This report provides a focused assessment of recent progress in the Synth subnet, where miners forecast BTC’s return distribution every day. Over the past month, the subnet has seen a major shift: miners prioritizing volatility accuracy have rapidly climbed the leaderboards, outperforming more complex models that previously held top positions.

We analyze the performance of these top miners across multiple metrics, volatility, intraday variation, and kurtosis, using data from June 15 to July 27, 2025. While recent gains in volatility prediction mark a significant milestone for Synth, they also reveal a new competitive phase: can miners now combine precision in volatility with deeper modeling of the return distribution? This report highlights where things stand, what’s improved, and where the next edge may lie.

Analysis Setup

The primary focus of this analysis is on volatility forecasts, which are the core of the subnet’s design. We evaluate forecasts produced by the Top 10 miners by leaderboard as well as the Top 10 miners by meta-leaderboard (i.e., total rewards accumulated over the past two weeks). Since the composition of these top miners changes over time, the aim is to track how forecast quality has evolved week by week.

We concentrate on 24-hour volatility forecasts (computed using all 5-minute return predictions), comparing them against realized BTC volatility over the same period. We also analyze shorter-term volatility forecasts (1-hour, 3-hour, 6-hour, 12-hour, and 18-hour) to assess consistency across time horizons.

Recognizing that volatility is not constant throughout the day, we also explore how miners vary their short-term volatility estimates relative to the 24-hour horizon. Additionally, we assess kurtosis, a metric reflecting the probability miners assign to extreme or unexpected outcomes.

To provide a more detailed picture, we supplement aggregated Top 10 statistics with per-axon data, helping identify which axons may be driving changes in the overall forecast quality.

The analysis covers the period from June 15th, 2025 to July 27th, 2025.

Volatility Analysis By Top Miners

Figure 1 shows the 24-hour volatility forecasts from the Top 10 miners by both leaderboard and meta-leaderboard, alongside realized BTC volatility.

From the figure, we observe that the gap between forecasted and realized BTC volatility has narrowed over time, with an almost perfect match in the final weeks of the period. Additionally, whereas the two groups of miners diverged in late June, they now appear to be in strong agreement.

Figures 2 and 3 report the absolute and relative differences between forecasted and realized volatility, along with 7-day moving averages.

These figures confirm the trend: After a temporary increase in errors from late June to early July, both absolute and relative differences have dropped sharply. In the last two weeks of the sample, the relative difference fell below 20%—to 18.47% (leaderboard) and 17.04% (meta-leaderboard)—more than halving the values from the prior two weeks (43.13% and 46.48%, respectively). This improvement represents a major milestone for the Synth subnet.

Table 1 breaks down volatility forecast performance across multiple horizons and periods.

Key takeaways from the table:

• Shorter-term volatility forecasts tend to be less accurate, with relative differences decreasing as the time horizon increases.

• Both top miner groups showed significant improvements across all horizons in the last two weeks, following a dip in late June and early July.

Volatility Analysis By Axons

To better understand these improvements, we analyzed the individual axons to which the top miners belong. Figures 4 and 5 show how miners from each axon moved in and out of the Top 10 (by total rewards and by leaderboard).

We identify two main groups of axons:

• Previously leading axons (through end of June): 35.77.6.189, 18.183.47.137, 3.112.97.164, 186.233.184.223

• Newly leading axons (late July): 116.202.53.142, 95.216.99.113

Figure 6 shows the average forecasted volatility from each group.

A clear pattern emerges: Previously leading axons consistently forecasted higher volatility, which initially seemed advantageous. However, as BTC volatility declined, this strategy failed to adapt, causing these axons to fall in the rankings.

By contrast, miners from the newly leading axons forecasted lower volatility, more aligned with realized BTC movements. Their apparent strategy of focusing on accurate volatility estimation has clearly paid off in leaderboard performance.

Intraday Volatility Analysis

Intraday volatility variation is a critical aspect in probabilistic forecasting, particularly in financial markets where volatility often fluctuates between day and night.

To evaluate whether the new top miners are adapting to intraday dynamics, or simply improving average 24-hour volatility forecasts—we conducted the following:

• Calculated normalized short-term volatilities (1h, 3h, 6h, 12h, 18h) and averaged them.

• Compared these to the normalized 24-hour volatility.

• Computed the ratio of the two metrics: A ratio near 1 suggests constant volatility; a ratio far from 1 suggests adaptive intraday modeling.

Figure 7 shows this ratio over time.

Interestingly, while miners initially varied their intraday volatilities effectively, this behavior diminished over time. Ratios approached 1 in late July, indicating a shift toward constant-volatility paths.

Figure 8 examines these ratios by axon group.

The pattern is revealing: Previously dominant axons made significant efforts to model intraday variation. In contrast, newer top axons appear to be submitting forecasts with constant volatility across time. While this simpler strategy has improved long-term accuracy, it reveals a potential weakness, namely lack of intraday adaptability, which could be exploited by competitors in the future.

Kurtosis Analysis

Finally, we examine the kurtosis of forecast distributions, considered an important measure of tail risk and sensitivity to rare events.

For each miner, we averaged the kurtosis across 100 simulated paths. We omit BTC realized kurtosis from the plots, as it is a single-path estimate and not directly comparable to the multi-path averages from simulations. A comparison with BTC will be the subject of a future report.

Figure 9 displays average kurtosis over time for top miners by both leaderboards.

Similar to the intraday volatility findings, kurtosis estimates were higher and more dynamic in early weeks, but have since flattened. By the end of the sample, top miners' kurtosis estimates dropped near zero, particularly for those by leaderboard.

Figure 10 shows the same analysis split by axon.

Again, we find that newer leading axons produce flatter distributions with little to no kurtosis. Previously leading axons, despite declining in rankings, still make active efforts to estimate and adapt kurtosis.

Takeaways and Future Challenges

The past month on Synth has seen a notable shift in leaderboard dynamics. A new group of miners—likely relying on simpler models narrowly focused on volatility—has climbed to the top by steadily improving their accuracy on this one key metric. Their forecasts are now consistently closer to BTC’s realized volatility, especially over the past couple of weeks, with errors dropped below 20%—18.47% (leaderboard) and 17.04% (meta-leaderboard), less than half of the previous period’s values, marking a major milestone for Synth.

Meanwhile, the previously dominant miners, who likely used more complex models, initially succeeded due to their broader probabilistic modeling, capturing volatility, intraday variation, and kurtosis. However, their failure to adapt to recent moves in BTC volatility caused them to fall behind.

This shift suggests that the Synth competition structure is driving real adaptation: simpler strategies have temporarily outpaced more complex models that were previously dominating. But the big questions are now coming into focus:

• Will the more sophisticated miners rise again, inspired to improve their models and regain the top spot?

• Will current leaders evolve to include better modeling of intraday variation and tail risk?

• Or will a new axon emerge that combines the strengths of both approaches—accurate volatility forecasts and rich distribution modeling?

The coming weeks will be an important test for the subnet. There’s still clear room for improvement, and the miners who manage to balance precision in volatility with attention to deeper characteristics of BTC’s return distribution are likely to be the ones who rise next. Whether it’s the current leaders who evolve, previous top performers making a comeback, or a new group altogether, the competition is far from over.

In short: the subnet is improving, but there’s still plenty of value to be captured, both in leaderboard positions and in rewards. That makes now a very good time for skilled new quants to enter the game.

Synth Rewards

With the current value of Synth Alpha, Synth paid 20,694 Alpha tokens to miners last week, equivalent to $59,599. Since March 2025, a total of 448,640 Alpha tokens have been distributed to miners, worth $1,292,083. On a monthly average since March 2025, Synth has paid $258,417 to miners, exceeding Numerai’s average monthly payout of $197,554.

BTC, ETH, and XAU Volatility

Figure 1 illustrates hourly log returns and volatility for Bitcoin (BTC), Ethereum (ETH), and Gold (XAU) during the week.

Of the three assets, XAU was the least volatile, ETH the most, with BTC falling in between. The most volatile period occurred mid-week, from July 23 to July 25.

Miners Performance

This week, we monitored the performance of six key miners:

• Top Performers: Miners 85, 169, and 61

• Dropped Out of Top Three: Miners 153, 65, and 154

Axon affiliations:

• Miner 61, 85, 154 → axon 95.216.99.113

• Miners 65, 153, 169 → axon 116.202.53.142

Performance was evaluated using two main metrics:

• CRPS (Continuous Ranked Probability Score) for BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS). Lower CRPS values indicate better predictive accuracy.

• For BTC, Miners 153 and 154 performed well early in the week. By Sunday, July 27, Miners 85, 169, and 61 were leading.

• For ETH, Miners 85, 169, and 61 excelled on Tuesday, July 22, while Miners 65 and 154 stood out mid-week. Overall, all miners except Miner 65 performed consistently well throughout the week.

For XAU, Miner 61 demonstrated consistent strong performance mid-week, with all miners performing similarly at the start and end of the week.

Leaderboard Scores. These are exponentially weighted averages of past CRPS values over a 10-day window. Lower scores indicate better performance.

Miners 85, 169, and 61 consistently improved their scores during the week—dropping from just over 150 to around 110. In contrast, Miners 153, 65, and 154 had stable scores, fluctuating around 120–130.

Miner 65 saw a slight deterioration in its leaderboard score due to poor weekend performance, increasing from ~135 at the start to ~145 by week's end.

Miners’ Strategies

This section analyzes summary statistics—volatility, kurtosis, and skewness—to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution—how likely extreme events are

• Skewness: Measures the asymmetry of return distributions (0 = symmetric)

Figure 4 presents these statistics for the four miners across the week.

Across all assets and metrics, miners showed highly similar distributional choices, likely due to shared axon affiliation. The only noticeable difference was in forecasted volatility for XAU, which was flatter and less variable for Miners 169 and 61 compared to others.

Miners’ Ranking vs. Summary Statistics: All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across ranking tiers.

Miners ranked 1–70 had very similar average weekly volatility and kurtosis across all assets. In contrast, lower-ranked miners exhibited much greater variability in these statistics, suggesting less consistency or riskier strategies.

Analysis by Axons

To assess performance at the group level, we analyzed:

• Axon size (number of miners)

• Average and total rewards per axon

• Axon-level summary statistics (BTC, ETH, and XAU forecasts)

Figure 6 shows the number of miners per axon at the end of the week.

• Axon 116.202.53.142 had the largest group (47 miners)

• Axon 95.216.99.113 increased from 20 to 24 miners

• Axon 138.201.62.165 remained at 21 miners

Figure 7 shows the evolution of average (per miner) and total daily rewards for the top 10 axons by average reward.

As far as the weekly rewards are concerned: Axon 116.202.53.142 earned the most, also due to its size in terms of number of miners Axon 95.216.99.113 had higher per-miner rewards Axon 35.77.6.189 started strong but declined mid-week All-Time Rewards (Since May 19, 2025)

1. Axon 116.202.53.142 - Total Alpha Tokens: 55,501 (= $ 159,842)

2. Axon 95.216.99.113 - Total Alpha Tokens: 30,253 (= $ 87,128)

3. Axon 35.77.6.189 - Total Alpha Tokens: 26,250 (= $ 75,600)

4. Axon 186.233.184.223 - Total Alpha Tokens: 20,834 (= $ 60,001)

5. Axon 18.183.47.137 - Total Alpha Tokens: 13,666 (= $ 39,357)

6. Axon 3.112.97.164 - Total Alpha Tokens: 13,435 (= $ 38,693)

7. Axon 95.111.205.93 - Total Alpha Tokens: 8,456 (= $ 24,352)

8. Axon 35.78.218.102 - Total Alpha Tokens: 7,275 (= $ 20,953)

9. Axon 103.88.234.233 - Total Alpha Tokens: 6,121 (= $ 17,629)

10. Axon 160.202.130.77 - Total Alpha Tokens: 4,347 (= $ 12,518)

Figure 8 compares the top 10 axons (by average rewards) based on BTC, ETH, and XAU modeling statistics: average intraday volatility, volatility variability, and average intraday kurtosis.

The two leading axons—116.202.53.142 and 95.216.99.113—shared similar modeling styles in terms of volatility and kurtosis. They successfully captured mid-week volatility spikes and the end-of-week drop.

While their volatility estimates reflected real-time market conditions effectively, both produced zero kurtosis forecasts throughout the week. This could indicate an area for strategic improvement that may offer an edge over other axons.

Takeaways

This week’s leaders were miners from axons 116.202.53.142 and 95.216.99.113, who demonstrated strong adaptability in capturing real-time asset volatility.

To further improve, these miners should consider incorporating non-zero kurtosis into their forecasts. Meanwhile, miners from other axons need to refine their volatility estimates to match the performance of the top groups.

Synth Rewards

With the current value of Synth Alpha, last week, Synth paid 20,692 Alpha tokens to miners, equivalent to $79,457. Since March 2025, a total of 427,944 Alpha tokens have been distributed to miners, worth $1,643,307. On a monthly average since March 2025, Synth has paid $345,959 to miners, more than Numerai’s average monthly payout of $205,606.

BTC, ETH, and XAU Volatility

Last week marked the first full week where Synth miners were tasked with forecasting gold (XAU). Therefore, this commentary also evaluates miners’ performance on gold predictions.

Figure 1 illustrates hourly log returns and volatility for Bitcoin (BTC), Ethereum (ETH), and Gold (XAU) during the week.

We observe large differences in volatility among the three assets. Ethereum was the most volatile, while gold was the least—though it did momentarily reach BTC-level volatility during a spike on July 17. Bitcoin's volatility generally fell between that of Ethereum and gold.

Miners Performance

This week, we monitored the performance of six key miners:

• Top Performers: Miners 153, 65, and 154

• Dropped Out of Top Three: Miners 221, 32, and 129

Axon affiliations:

• Miner 221 → axon 35.77.6.189

• Miner 32 → axon 3.112.97.164

• Miner 154 → axon 95.216.99.113

• Miners 65, 129, 153 → axon 116.202.53.142

Performance was evaluated using two main metrics:

• CRPS (Continuous Ranked Probability Score) for BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS). Lower CRPS values indicate better predictive accuracy.

Across all three assets, CRPS scores were tightly clustered between the two groups. However, ETH scores show that Miners 65 and 153 gained an edge between July 18–19 and during the weekend. In contrast, Miners 32 and 221 were hampered by poor gold forecasts early in the week and higher ETH CRPS scores over the weekend.

Notably, the CRPS score magnitude varied by asset, which is expected due to differences in each asset’s volatility.

Leaderboard Scores. These are exponentially weighted averages of past CRPS values over a 10-day window. Lower scores indicate better performance.

Poor gold performance at the beginning of the week negatively impacted Miners 32 and 221, causing them to drop from the top 2 by Monday, July 14—replaced by Miners 65 and 154. Miner 129 held the top spot for most of the week but dropped out of the top 3 on Sunday, July 21, likely due to weaker ETH performance.

Miners’ Strategies

This section analyzes summary statistics—volatility, kurtosis, and skewness—to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution—how likely extreme events are

• Skewness: Measures the asymmetry of return distributions (0 = symmetric)

Figure 4 presents these statistics for the four miners across the week.

Miners from axon 116.202.53.142 (65, 129, 153) and Miner 154 exhibited consistent modeling: similar volatility, kurtosis near 0, and stable, symmetric skewness.

In contrast, Miners 32 and 221 produced more volatile and fat-tailed forecasts—particularly for gold—with greater skewness variability.

Miners’ Ranking vs. Summary Statistics: All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across ranking tiers.

We observed that lower-ranked miners are beginning to match the volatility levels of top-ranked miners for BTC and ETH. For gold, differences are more pronounced.

In terms of kurtosis, miners ranked 10–30 are approaching top-tier performance for BTC, but disparities remain for ETH and XAU.

Analysis by Axons

To assess performance at the group level, we analyzed:

• Axon size (number of miners)

• Average and total rewards per axon

• Axon-level summary statistics (BTC, ETH, and XAU forecasts)

Figure 6 shows the number of miners per axon at the end of the week.

Axon 116.202.53.142 had the most miners, followed by 138.201.62.165 and 95.216.99.113.

Figure 7 shows the evolution of average (per miner) and total daily rewards for the top 10 axons by average reward.

Axon 116.202.53.142 earned the highest total rewards, benefitting from both miner quantity and strong individual performance.

On a per-miner basis:

• Early week: Axon 185.150.117.101 led

• Midweek: Axon 3.135.57.163

• Weekend: Axon 35.77.6.189

All-Time Rewards (Since May 19, 2025)

1. 116.202.53.142 – 44,143 Alpha ($169,510)

2. 35.77.6.189 – 26,007 Alpha ($99,867)

3. 95.216.99.113 – 23,192 Alpha ($89,058)

4. 186.233.184.223 – 20,834 Alpha ($80,001)

5. 18.183.47.137 – 13,597 Alpha ($52,211)

6. 3.112.97.164 – 13,308 Alpha ($51,104)

7. 95.111.205.93 – 8,456 Alpha ($32,469)

8. 35.78.218.102 – 7,244 Alpha ($27,816)

9. 103.88.234.233 – 6,121 Alpha ($23,505)

10. 160.202.130.77 – 4,347 Alpha ($16,691)

Figure 8 compares the top 10 axons (by average rewards) based on BTC, ETH, and XAU modeling statistics: average intraday volatility, volatility variability, and average intraday kurtosis.

Forecasting styles are converging, particularly for BTC and ETH volatility. Some axons like 185.150.117.10 use low-volatility models, while others like 35.78.218.102 and 170.64.173.44 are more aggressive.

Axon 185.150.117.10 showed significant intraday volatility for BTC and ETH but used flat volatility for gold. It modeled kurtosis for BTC/ETH but ignored it for gold. Axon 116.202.53.142 maintained consistent zero kurtosis across all assets.

Takeaways

This week marked the beginning of performance evaluation for gold forecasts. As observed last week, the gap between top and mid-ranked miners is narrowing, with lower-ranked miners adopting increasingly similar modeling approaches—particularly in volatility.

Synth Rewards

With the current value of Synth Alpha, last week, Synth paid 20,716 Alpha tokens to miners, equivalent to $85,766. Since March 2025, a total of 410,208 Alpha tokens have been distributed to miners, worth $1,698,261. On a monthly average since Synth’s dtao launch in February 2025, Synth has paid $377,391 to miners, more than double Numerai’s average monthly payout of $186,058.

BTC and ETH Volatility

Figure 1 illustrates hourly log returns and volatility for the week, highlighting the relationship between BTC and ETH. While overall volatility remained relatively stable, two notable spikes occurred midweek (October 9th and 10th). These were instances of positive volatility, driven primarily by strong upward price movements.

Miners Performance

This week, we monitored the performance of four key miners:

• Miners 221, 32, and 129 – Finished the week in the top three on the Synth leaderboard

• Miner 200 – Dropped out of the top three during the week

Miners 200 and 221 are part of axon 35.77.6.189, Miner 32 belongs to axon 3.112.97.164, and Miner 129 is part of axon 116.202.53.142. The data suggests increasing competition between axons, indicating that models from different clusters are becoming increasingly competitive—benefiting the overall health and quality of the subnet.

As usual, performance was evaluated using two key metrics:

• CRPS (Continuous Ranked Probability Score) for BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS). Lower CRPS values indicate better predictive accuracy.

Throughout the week, CRPS values for BTC showed intense competition. Notably, Miner 129—previously the leaderboard leader—lost ground toward the weekend, particularly on Sunday the 13th, likely due to overestimating weekend volatility.Miner 129 also experienced a modest drop in ETH performance over the weekend, despite strong results earlier in the week. Conversely, Miner 200 underperformed midweek but posted much-improved ETH CRPS values during the low-volatility weekend. Miners 32 and 221 showed performance patterns similar to Miner 200.

Leaderboard Scores. These are exponentially weighted averages of past CRPS values over a 10-day window. Lower scores indicate better performance.

This week, ETH performance and weekend scores had a significant influence on leaderboard positions. Miner 129, despite a strong midweek comeback, lost the top spot by the weekend but remained in the top three. Miner 200’s fluctuations closely mirrored their ETH CRPS trajectory, with midweek underperformance and weekend recovery. Miners 32 and 221, with slightly better overall scores, secured their places in the top three.

Miners’ Strategies

This section analyzes summary statistics—volatility, kurtosis, and skewness—to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution—how likely extreme events are

• Skewness: Measures the asymmetry of return distributions (0 = symmetric)

Figure 4 presents these statistics for the four miners across the week.

Miners 221, 32, and 200 displayed very similar statistical profiles, with closely aligned volatility, kurtosis, and skewness, both in level and variation. In contrast, Miner 129 appeared to downplay kurtosis and skewness, focusing instead on maintaining lower volatility. This approach did not pay off in the long term, as evidenced by the loss of the leaderboard top position over the weekend.

Miners’ Ranking vs. Summary Statistics: All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across ranking tiers.

This week, even lower-tier miners appeared to place greater emphasis on volatility modeling. For both BTC and ETH, their forecasted volatility levels approached those of the top-tier miners—another sign of rising quality across the subnet. Kurtosis modeling is also gaining traction, with second-tier miners (ranks 10–20) increasingly emulating the modeling patterns of the top 10.

Analysis by Axons

To assess performance at the group level, we analyzed:

• Axon size (number of miners)

• Average and total rewards per axon

• Axon-level summary statistics (BTC forecasts)

Figure 6 shows the number of miners per axon at the end of the week.

As in previous weeks, axon 116.202.53.142 had the highest number of registered miners, followed by 138.201.62.165 and 95.216.99.113.

Figure 7 shows the evolution of average (per miner) and total daily rewards for the top 10 axons by average reward.

The axon-level dynamics mirrored those of individual miners. Axon 116.202.53.142 earned the highest total weekly rewards. However, its total rewards decreased toward the end of the week, and its average daily rewards were in line with others—suggesting its performance heavily depended on Miner 129. Meanwhile, axons 35.77.6.189 and 3.112.97.164 saw average daily rewards increase over the weekend, reflecting stronger miner performance compared to Miner 129.

All-Time Rewards (Since May 19th 2025) Per Axon

1. Axon 116.202.53.142 - Total Alpha Tokens: 33,847 (= $ 140,126)

2. Axon 35.77.6.189 - Total Alpha Tokens: 25,375 (= $ 105,053)

3. Axon 186.233.184.223 - Total Alpha Tokens: 21,309 (= $ 88,218)

4. Axon 95.216.99.113 - Total Alpha Tokens: 19,307 (= $ 79,929)

5. Axon 18.183.47.137 - Total Alpha Tokens: 13,499 (= $ 55,886)

6. Axon 3.112.97.164 - Total Alpha Tokens: 13,073 (= $ 54,120)

7. Axon 95.111.205.93 - Total Alpha Tokens: 8,478 (= $ 35,100)

8. Axon 35.78.218.102 - Total Alpha Tokens: 7,202 (= $ 29,815)

9. Axon 103.88.234.233 - Total Alpha Tokens: 6,318 (= $ 26,155)

10. Axon 185.141.218.212 - Total Alpha Tokens: 4,902 (= $ 20,295)

Figure 8 compares the top 10 axons (by average rewards) based on BTC modeling statistics: average intraday volatility, volatility variability, and average intraday kurtosis.

This figure sheds light on Miner 129’s weekend decline. Miners from axon 116.202.53.142 generally anticipated higher volatility compared to others. While most axons reduced volatility forecasts over the weekend, this axon increased them. Additionally, its miners had some of the lowest intraday volatility and kurtosis values, possibly limiting their ability to adapt to real-time market changes. In contrast, other axons displayed higher and more variable intraday volatility and kurtosis—potentially enabling better responsiveness.

Takeaways

This week’s analysis reveals several noteworthy trends:

• ETH performance is becoming more influential in overall miner rankings

• Competition among axons is intensifying, with three different axons battling for top spots throughout the week

• Lower-tier miners are improving, particularly in volatility modeling, and are starting to adopt strategies used by top performers

These developments point to a healthier and more competitive subnet, with greater emphasis on modeling distributional nuances—especially for ETH, which is playing a larger role in leaderboard scoring.

Starting last week, Gold has been added to the set of assets modeled by Synth miners. We expect miners who successfully capture meaningful dynamics in this asset’s distribution to gain a competitive edge. Beginning next week, we will start reporting miner scores for Gold forecasts.

BTC and ETH Volatility

Figure 1 shows the hourly log returns and volatility observed throughout the week, highlighting the relationship between BTC and ETH. Volatility remained relatively stable during the weekdays and declined over the weekend (July 6–7), before a final spike; particularly noticeable in Ethereum, at the end of the week.

Miners Performance

This week, we tracked the performance of six miners:

• Miners 221, 32, and 200 – consistently among the top-ranked throughout the week

• Miners 253, 213, and 76 – each from a different axon, showing strong performance

As usual, performance was evaluated using two key metrics:

• CRPS (Continuous Ranked Probability Score) for BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS). Lower CRPS values indicate better predictive accuracy.

The weekly CRPS values reflect a highly competitive environment among top-tier miners. While there were moments when either group (221/32/200 vs. 253/213/76) outperformed the other, no miner group consistently dominated across all days and volatility regimes. This signals a healthy competitive dynamic in the subnet, one that fosters continuous model improvement and innovation.

Leaderboard Scores. These are exponentially weighted averages of past CRPS values over a 10-day window. Lower scores indicate better performance.

Miners 32, 200, and 221 maintained top-three positions throughout the week. However, the performance gap between them and Miners 76, 213, and 253 narrowed, decreasing from ~40 points to less than ~20 points by week’s end.

Miners’ Strategies

This section analyzes summary statistics, including volatility, kurtosis, and skewness to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution, how likely extreme events are

• Skewness: Measures the asymmetry of return distributions (0 = symmetric)

Figure 4 presents these statistics for the four miners across the week.

The figure reveals distinct modeling styles. Miners 32, 200, and 221 tended to produce forecasts with higher volatility and kurtosis, and they showed greater variation in distribution moments (including skewness). Conversely, Miners 253, 213, and 76 used less volatile and more stable distributions.

Interestingly, the latter group closed the performance gap with the frontrunners in leaderboard scores, suggesting that under the specific volatility regime of the week, their more stable approach paid off.

Miners’ Ranking vs. Summary Statistics: All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across ranking tiers.

Consistent with previous weeks, miners in the top 10 reward tier produced forecasts that were, on average, more volatile and had more stable kurtosis than those ranked lower.

Analysis by Axons

To assess performance at the group level, we analyzed:

• Axon size (number of miners)

• Average and total rewards per axon

• Axon-level summary statistics (BTC forecasts)

Figure 6 shows the number of miners per axon at the end of the week.

As in the previous weeks, axon 116.202.53.142 had the highest number of registered miners, followed by 138.201.62.165 and 95.216.99.113.

Figure 7 shows the evolution of average (per miner) and total daily rewards for the top 10 axons by average reward.

At the beginning of the week, axon 35.77.6.189 received most of the rewards. However, starting over the weekend, likely due to exploiting lower-volatility periods, miners from axons 95.111.205.93 and 116.202.53.142 began to earn significantly more rewards on average.

Figure 8 compares the top 10 axons (by average rewards) based on key BTC modeling characteristics: average intraday volatility, volatility variability, and average intraday kurtosis.

Except for axon 95.216.99.113 and axon 95.216.2.50, who produced less volatile, less "fat-tailed" forecasts, most axons showed consistent forecasting behavior, with similar volatility and kurtosis levels, both in magnitude and direction.

Takeaways

This week’s results highlight an increasingly dynamic and competitive environment among miners, with no single frontrunner consistently outperforming across all conditions. The shrinking gap in leaderboard scores between the frontrunners and their closest challengers suggests that competition will intensify further in the coming weeks.

BTC and ETH Volatility

Figure 1 presents the hourly log returns and volatility observed throughout the week, highlighting the relationship between BTC and ETH. Volatility generally declined as the week progressed, reaching its lowest levels on Sunday, June 29, before a final spike at the end of the day.

Miners Performance

This week, we analyzed six miners: Miners 100, 236, and 238—who started the week at the top of the leaderboard—and Miners 32, 200, and 221, who ended the week in the leading positions.

As usual, performance was evaluated using two key metrics:

• CRPS (Continuous Ranked Probability Score) for BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS). Lower CRPS values indicate better predictive accuracy.

In the early part of the week, miners submitting “equivalent paths” performed best, benefiting from the high-volatility environment. However, as volatility declined, Miners 32, 200, and 221 made a strong comeback, consistently outperforming Miners 100, 236, and 238—especially over the weekend.

Leaderboard Scores. These are exponentially weighted averages of past CRPS values over a 10-day window. Lower scores indicate better performance.

The leaderboard trends clearly illustrate the decline in performance for Miners 100, 236, and 238. Their use of equivalent paths and lack of adaptability to changes in the volatility regime negatively impacted their results. Conversely, miners able to adapt to changing conditions quickly regained top positions—even after temporarily falling behind.

Miners’ Strategies

This section analyzes summary statistics—volatility, kurtosis, and skewness—to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution—how likely extreme events are

• Skewness: Measures the asymmetry of return distributions (0 = symmetric)

Figure 4 presents these statistics for the four miners across the week.

The results confirm that Miners 32, 200, and 221 employed a more dynamic approach to modeling volatility, with their estimates varying significantly more than those of Miners 100, 236, and 238. They also produced more symmetrical return distributions, while the others tended to show more negative skewness. In terms of kurtosis, the top-performing miners modeled extreme events more cautiously, resulting in lower kurtosis values overall.

Miners’ Ranking vs. Summary Statistics: All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across ranking tiers.

During the high-volatility early part of the week, miners that consistently forecasted high volatility—without adjusting—captured most of the rewards. Their lack of adaptation also resulted in low variability in kurtosis estimates.

Analysis by Axons

To assess performance at the group level, we analyzed:

• Axon size (number of miners)

• Average and total rewards per axon

• Axon-level summary statistics (BTC forecasts)

Figure 6 shows the number of miners per axon at the end of the week.

As in the previous week, axon 116.202.53.142 had the highest number of registered miners.

Figure 7 shows the evolution of average (per miner) and total daily rewards for the top 10 axons by average reward.

As previously noted, miners generating equivalent high-volatility paths received most of the rewards until the weekend. These miners mainly belonged to axon 186.233.184.223. However, as the volatility regime shifted, their performance declined, and they lost ground to axons with miners who better adapted to changing conditions.

Figure 8 compares the top 10 axons (by average rewards) based on key BTC modeling characteristics: average intraday volatility, volatility variability, and average intraday kurtosis.

Axons 186.233.184.223 and 160.202.130.77 (which likely belong to the same group and overlap visually in the chart) maintained stable volatility estimates throughout the week, showing nearly zero variation. In contrast, miners from other axons adjusted their volatility estimates downward, following the trends observed in BTC and ETH. Regarding kurtosis, axon 95.111.205.93 stood out for its extremely high values—indicating a modeling approach that overestimated the likelihood of extreme events.

Takeaways

This week once again demonstrated that miners who fail to adapt their forecasts to evolving market and volatility conditions are unlikely to remain on top for long. While static, high-volatility forecasts may perform well in turbulent conditions, they lose effectiveness as volatility moderates. In contrast, adaptive miners—those who adjust their forecasts to reflect changing market realities—are better positioned to sustain high performance in the longer term.

BTC and ETH Volatility

Figure 1 presents the hourly log returns and volatility observed throughout the week, highlighting the relationship between BTC and ETH under the current market regime.

Miners Performance

This week, we analyze the performance of six miners: three that reached the top of the leaderboard (Miners 46, 109, and 236), and three that lost their leading positions (Miners 170, 200, and 165).

As usual, performance is evaluated using two key metrics:

• CRPS (Continuous Ranked Probability Score) for BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS) Lower CRPS values indicate better predictive accuracy.

Similar to previous weeks, the competition for BTC and ETH forecasts was tight through the beginning of the week, with relatively stable volatility. Toward the end of the week—particularly between Saturday the 22nd and Sunday the 23rd—a clear performance gap emerged between Miners 46, 109, and 236, and the others. This shift coincided with a spike in unexpected weekend volatility. Figure 2 shows that these three miners belonged to the same group and submitted identical forecast paths throughout the week, resulting in identical scores.

Leaderboard Scores. These are exponentially weighted averages of past CRPS values over a 10-day window. Lower scores indicate better performance.

These scores confirm the weekend struggles of Miners 165, 170, and 200, who lost their lead and saw a significant decline in performance. Conversely, the group of miners submitting identical paths capitalized on the change in volatility during the weekend.

Miners’ Strategies

This section analyzes summary statistics—volatility, kurtosis, and skewness—to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution—how likely extreme events are

• Skewness: Measures the asymmetry of return distributions (0 = symmetric)

Figure 4 presents these statistics for the four miners across the week.

The results show that the miners who ended the week at the top of the leaderboard benefited from distributions with higher kurtosis and lower skewness. In terms of volatility, their BTC forecasts were on par with (or slightly less volatile than) other miners but were more consistent. In contrast, Miners 46, 109, and 236 displayed more variable volatility, suggesting a strategy that relies on consistently large volatility, which does not adapt well to changing conditions but pays off during high-volatility periods. This pattern is also evident in ETH forecasts, where the leading miners clearly projected higher volatilities than their peers.

Miners’ Ranking vs. Summary Statistics: All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across ranking tiers.

As in previous weeks, top-ranked miners (across the full week, not just the weekend) demonstrated better volatility adaptation than lower-ranked miners. However, some lower-ranked miners appear to be experimenting with modeling kurtosis for both BTC and ETH. While this has not yet translated into higher rewards, this kind of innovation is essential for long-term success in the Synth Subnet.

Analysis by Axons

To assess performance at the group level, we analyzed:

• Axon size (number of miners)

• Average and total rewards per axon

• Axon-level summary statistics (BTC forecasts)

Figure 6 shows the number of miners per axon at the end of the week.

As in the previous week, axons 116.202.53.142 had  the largest memberships with 49 miners. Axons 95.216.99.113, on the other hand, saw a reduction in the number of registered miners, from 33 to 25.

Figure 7 shows the evolution of average (per miner) and total daily rewards for the top 10 axons by average rewards.

Axons 186.233.184.223 and 160.202.130.77 benefited the most from weekend volatility. The top-performing miners during the weekend belonged to Axon 186.233.184.223.

Figure 8 compares the top 10 axons (by average rewards) based on BTC modeling characteristics: intraday average volatility, volatility variability, and average intraday kurtosis.

The data confirms that the top axons employed more stable volatility strategies, which proved advantageous over the weekend. Moreover, the modeling outputs of Axons 186.233.184.223 and 160.202.130.77 were nearly identical, suggesting they likely belong to the same underlying group.

Takeaways

The unexpected weekend volatility disrupted the leaderboard, displacing previous frontrunners and elevating miners who used strategies built on consistently large volatility. While this shift temporarily favored miners submitting identical paths, we expect upcoming regime changes and better model adaptability to return more sophisticated models and miners to the top.

As noted in previous weeks, miners are encouraged to focus on developing high-quality models rather than registering multiple miners. Larger groups do not show a clear advantage in terms of average or total rewards.

BTC and ETH Volatility

Figure 1 shows the hourly log returns and volatility observed throughout the week. Higher ETH volatility and the BTC/ETH correlation are the most notable patterns in the two graphs.

Miners Performance

This week’s analysis focuses on four miners: 32, 170, 200, and 234. Two of these miners were in the top 3 both at the beginning and end of the week. Miner 170 climbed into the top rankings during the week—ultimately finishing first—while Miner 32, initially in the top 3, dropped out by the end.

As usual, we evaluate performance using two key metrics:

• CRPS (Continuous Ranked Probability Score) for BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS). Lower CRPS values indicate better predictive accuracy.

As observed last week, CRPS values for both BTC and ETH were very close among the top miners, indicating a tightening competition. This suggests that success increasingly depends on accurately modeling price nuances.

Leaderboard Scores

These are exponentially weighted averages of past CRPS values over a 10-day window. Lower scores indicate better performance.

The leaderboard continues to reflect close competition among top miners. Miner 170 entered the top 3 on Friday, June 13th, and reached the top position over the weekend.

Miners’ Strategies

This section analyzes summary statistics—volatility, kurtosis, and skewness—to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution—how likely extreme events are

• Skewness: Measures the asymmetry of return distributions (0 = symmetric)

Figure 4 presents these statistics across the week for the four miners.

Modeling strategies appear broadly similar across top miners, with only slight differences—particularly in how Miners 234 and 200 modeled BTC kurtosis compared to the others.

Miners’ Ranking vs. Summary Statistics: All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across different ranking tiers.

As seen last week, lower-ranked miners struggled to match top-tier performance, especially in volatility and kurtosis for both BTC and ETH. Still, it's encouraging that some lower-tier miners attempted to model return distribution kurtosis, even if results fell short. Such efforts are likely to be rewarded over time as model quality improves.

Analysis by Axons

To assess performance at the group level, we analyzed:

• Axon size (number of miners)

• Average and total rewards per axon

• Axon-level summary statistics (BTC forecasts)

Figure 6 shows the number of miners per axon at the end of the week.

As in the previous week, axons 116.202.53.142 and 95.216.99.113 had the largest memberships, with 49 and 33 miners, respectively.

Figure 7 shows the evolution of average (per miner) and total daily rewards for the top 10 axons by average rewards.

Axons 35.77.6.189, 3.112.97.164, 35.78.218.102, and 18.183.47.137 achieved the highest average and total daily rewards. Notably, the two largest axons were not among the top earners—suggesting that a larger group size does not guarantee greater rewards.

Furthermore, the reward system's updated parameters now encourage forecast diversification. Miners from the same axon who submit identical path sets are penalized, as identical paths receive identical scores—and poor-performing models will impact all such miners equally.

Figure 8 compares the top 10 axons (by average rewards) based on their BTC modeling characteristics: intraday average volatility, its variability, and average intraday kurtosis.

There is a clear difference in how the top four axons model volatility. These axons tend to produce forecasts with greater and more adaptive volatility. They also place more emphasis on modeling kurtosis. In contrast, lower-ranked axons often produce less volatile forecasts and seem to neglect kurtosis modeling. One exception was axon 186.233.184.223, which made some effort in kurtosis modeling but did not adapt forecast volatility as much as the top axons.

Takeaways

This week continued the trend of tightening competition. Top-ranking miners submitted high-quality forecasts with very close scores. A sharper distinction emerged between high-reward axons and others, largely due to superior modeling strategies among the top axons.

While the current market and volatility regime appears to favor these top-performing axons, lower-tier axons and miners are encouraged to keep improving their models. Leaderboard shifts are not uncommon, as seen in previous weeks.

Finally, the recent change in EMA calculation parameters seems to have had two effects:

1. Large axons do not necessarily yield larger total rewards unless their models are competitive.

2. Lack of diversification within axons is penalized. Submitting identical paths results in shared outcomes, and when the model underperforms, all miners using it suffer lower rewards.

In this week’s issue, in addition to reviewing individual miners’ performance and modeling strategies via forecast summary statistics, we also examine axon-level performance among subnet participants to assess how groups of miners perform collectively.

Important Update: Starting on June 4, 2025, at 4:00 PM UTC, the parameters used for calculating EMA scores were updated:

• Window length increased from 7 to 10 days

• Half-life increased from 3.5 to 5 days

• Beta parameter decreased from -0.0475 to -0.1

The axon analysis offers an initial overview of the impact of these changes on the rewards distribution.

BTC and ETH Volatility

Figure 1 shows the hourly log returns and volatility observed throughout the week. It’s clear that ETH exhibited higher volatility than BTC, and that the two assets are strongly correlated, with volatility spikes occurring at similar times.

Miners Performance

This week’s analysis focuses on six miners:

• The top three at the start of the week: Miners 32, 234, and 200

• The top three at the end of the week: Miners 170, 221, and 90

As usual, we evaluate their performance using two key metrics:

• CRPS (Continuous Ranked Probability Score) for both BTC and ETH forecasts

• Leaderboard Scores

Continuous Ranked Probability Score (CRPS)

Lower CRPS values indicate better predictive accuracy.

The competition was extremely tight this week, as reflected in the CRPS values for both BTC and ETH. The scores were very close, making it hard to clearly distinguish between miners.

Leaderboard Scores

Leaderboard scores are an exponentially weighted average of past CRPS values over a 10-day window (half-life: 5 days). Lower scores indicate better performance.

The leaderboard also reflected the week’s close competition, with frequent leadership shifts until the ranking began stabilizing from Friday, June 6, through the weekend. Notably, the six miners finished the week within a narrow range of smoothed scores, between 91 and 97.

Miners’ Strategies

This section analyzes summary statistics—volatility, kurtosis, and skewness—to detect potential strategic differences among top miners for both BTC and ETH.

Summary Statistic Definitions:

• Volatility: Variability in forecasted returns

• Kurtosis: Measures the "tailedness" of the distribution—how likely extreme events are

• Skewness: Indicates the asymmetry of return distributions (0 = symmetric)

Figure 4 shows these distributions across the week for the six miners.

The modeling strategies appear to be quite similar, with no significant differences in weekly volatility, kurtosis, or skewness. This helps explain the tight CRPS values.

Miners Ranking vs. Summary Statistics

All miners who submitted forecasts were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across different ranking tiers.

Compared to the previous week, lower-ranked miners struggled more to keep up with top miners, particularly regarding BTC volatility modeling. As in past weeks, top miners showed a clear edge in modeling ETH volatility and kurtosis for both assets.

Analysis by Axons

To evaluate miners’ performance in groupings (axons), we analyzed weekly data on:

• Axon size over time

• Average and total rewards per axon

• Axon-level summary statistics (for BTC forecasts)

Figure 6 shows axon size evolution throughout the week, with a vertical red dashed line indicating when the EMA parameter update occurred.

Notably, axon 116.202.53.142 and axon 95.216.99.113 had the most miners (49 and 33 at the end of the week, respectively). But does a larger size guarantee more rewards?

Figure 7 explores this by showing average and total daily rewards for the top 10 axons by average rewards. The EMA parameter change is again marked by a vertical line.

Key insights:

1. The EMA update steepened the rewards curve (as intended), making performance differences more pronounced (as noticeable from the top graph)

2. Some large axons (e.g., 116.202.53.142 and 95.216.99.113) saw declining total rewards post-update, despite their size

3. Smaller axons with better modeling strategies (e.g., axon 35.77.6.189) significantly outperformed larger ones in total rewards

Figure 8 compares the top 10 axons (by average rewards) on their BTC modeling characteristics—intraday average volatility and its variability, as well as intraday average kurtosis.

There is a clear distinction in modeling quality between top- and lower-ranked axons. High-reward axons were able to model BTC volatility effectively and adjust it throughout the week. Similarly, strong performance in modeling kurtosis contributed to their rewards. By contrast, some axons appear to neglect one or both of these aspects altogether.

Takeaways

This was a highly competitive week, with tight CRPS values and similar modeling strategies across top miners. This is a positive sign for the subnet, suggesting that miners are refining their models and focusing on nuanced aspects of BTC and ETH distributions.

The axon-level analysis reveals that the EMA parameter update improved reward fairness: axons producing higher-quality models were more strongly rewarded, while large axons relying solely on size saw diminished returns. This encourages skill-based clustering rather than strategy based purely on scale.

Starting this week, Synth has begun scoring miners on their Ethereum (ETH) predictions in addition to their Bitcoin (BTC) forecasts. This report presents a combined analysis of miner performance across both assets.

BTC and ETH Volatility

Figure 1 displays the hourly log-returns and volatility observed throughout the week. It clearly shows that ETH experienced higher volatility compared to BTC. Furthermore, both assets exhibited volatility spikes at similar times, reflecting their correlation.

Miners Performance

This week’s analysis focuses on six miners: the top three on the leaderboard at the beginning of the week (Miners 46, 241, and 250), and the top three at the end of the week (Miners 33, 103, and 252). We assess their performance using two key metrics: CRPS (based on forecasts for both BTC and ETH) and Leaderboard Scores.

Continuous Ranked Probability Score (CRPS)

Lower CRPS values indicate better predictive accuracy.

As shown in Figure 2, Miners 46, 241, and 250 — the early frontrunners — received identical CRPS values throughout the week. Notably, they completely failed to predict ETH outcomes in the initial prompts. This suggests they were producing identical forecasts, likely due to shared model architectures or coordinated strategies. The Synth team suspects that this group may include additional miners and is currently refining the scoring system to address this issue.

In contrast, Miners 33, 103, and 252 generally performed better in the middle and latter parts of the week, especially in BTC predictions. Their models seem to have adapted more effectively to the changing volatility regime. For ETH, performance between the two groups varied throughout the week, but over the weekend, Miners 33, 103, and 252 clearly outperformed.

Leaderboard Scores

Leaderboard scores are an exponentially weighted average of past CRPS performance over a 7-day window, with a half-life of 3.5 days. Lower scores are better.

Thanks to their consistent CRPS performance, Miners 33, 103, and 252 took the lead on May 20, surpassing the previous top scorers. While all miners saw slight score increases over the week, the new top trio experienced only a modest rise (from ~100 to ~140), whereas Miners 46, 241, and 250 saw a sharp increase (from ~60 to over 220).

Miners’ Strategies

This section analyzes summary statistics, volatility, kurtosis, and skewness to identify potential strategic advantages among the top miners. These metrics were collected and assessed for both BTC and ETH.

Top Miners Statistics Distribution

• Volatility indicates variability in forecasted returns.

• Kurtosis: Kurtosis reflects the thickness of the distribution tails, indicating the degree to which miners expect extreme events

• Skewness measures the asymmetry of return distributions (a skewness of 0 implies a balanced distribution around its average).

Figure 4 shows these distributions across the week for the six miners.

The data supports the hypothesis that Miners 46, 241, and 250 share a common model, given their identical distributions for both assets. In contrast, Miners 33, 103, and 252 displayed:

• Higher and more variable volatility: indicating more dynamic modeling.

• Lower kurtosis: suggesting reduced expectations of extreme outcomes.

• More symmetrical skewness: reflecting more balanced forecasts.

This highlights the risks of using undifferentiated models, especially for organizations managing multiple miners, since shared models can underperform collectively during market shifts.

The observed diversity in forecast behavior highlights the drawbacks of a lack of model differentiation. For entities managing multiple miners, homogeneous forecasts increase vulnerability to performance degradation during regime shifts.

Miners Ranking vs. Summary Statistics

All miners who submitted forecasts during the week were ranked by total weekly rewards. Figure 5 shows how summary statistics vary across different ranking tiers.

Key Insights

• BTC: Volatility modeling is improving across the board. Compared to past weeks, lower-ranked miners are making noticeable efforts to align their volatility modeling more closely with that of top-tier performers.

• ETH: Modeling remains less developed. As competition is still in early stages, miners need more time to optimize their approaches.

• Kurtosis: Still largely overlooked by lower-ranked miners. Many exhibit flat kurtosis near zero, indicating a lack of attention to tail risks—likely contributing to poorer results.

Takeaways

• Identical Forecast Clusters: Some miners are submitting indistinguishable forecasts, likely due to shared models. While this may offer short-term advantages, these miners are more vulnerable when model performance degrades.

• Value of Diversification: Entities managing multiple miners should diversify their forecasting models. Relying on a single model can expose all associated miners to simultaneous underperformance.

• Volatility Modeling Progress: Lower-ranked miners are making gains in modeling BTC volatility. ETH modeling, however, still lags as miners continue to adjust their strategies in this early stage of competition.

• Neglected Kurtosis: Despite progress in volatility modeling, kurtosis remains underutilized among lower-ranked miners.

Ongoing Improvements: The Synth team is addressing the issue of forecast copying. One potential solution under consideration is increasing the steepness of the reward curve. Further updates will follow.

Bitcoin Price Action

During the week of May 5 to May 11, 2025, Bitcoin opened at $94,300 and closed at $104,150, with a weekly low of $93,400 and a high of $105,000. The week saw several volatility spikes, primarily driven by positive market sentiment and rumors of new tariff deals between the U.S. and other countries, notably the U.K. A significant volatility spike occurred around May 7, between 00:30 and 01:00 UTC, aligning with a sharp price increase as market optimism grew, pushing Bitcoin toward its weekly high of $105,000. Volatility remained elevated through the mid-week before gradually stabilizing as the price settled around $104,150 by the week's end.

Miners Performance

At the start of the week, Miners 32, 165, and 103 led the leaderboard. By the end of the week, Miners 136, 133, and 195 had taken the top spots. We analyze their performance using two key metrics: CRPS and Leaderboard Scores.

Continuous Ranked Probability Score (CRPS). Lower CRPS values indicate better predictive accuracy.

The CRPS plot reveals a tight competition among the top miners, with values generally close throughout the week. Very early on, Miners 32, 165, and 103 produced more accurate forecasts, reflected in their lower CRPS values. However, mid-week volatility spikes, particularly around May 7, proved decisive. Miners 136, 133, and 195 adapted more effectively to these market shifts, achieving lower CRPS values between May 8 and 9. By the end of the week, the CRPS scores of all miners under examination stabilized around 2,500 to 3,000.

Leaderboard Scores

Leaderboard scores are an exponentially weighted average of past CRPS performance over a 7-day window, with a half-life of 3.5 days. Lower scores are better.

At the week's start, Miners 32, 165, and 103 held the top ranks with scores around 70 to 90, while Miners 136, 133, and 195 started with scores between 120 and 150. The mid-week volatility spikes caused a gradual shift. Miners 136, 133, and 195 consistently improved their forecasts, with their leaderboard scores steadily converging toward those of the initial leaders. By May 10 to 11, they overtook the top spots, closing the week with scores around 90 to 110, while Miners 32, 165, and 103 saw their scores rise to around 100 to 110 due to their worse performance during the volatile periods.

Top Miners’ Strategies

In this section, we examine summary statistics derived from the forecasts of all miners during the week to identify what might have given an edge to the top performers over the last 7 days.

To achieve this, we first created an "overall" weekly ranking by summing the reward weights earned by each miner across the week. Next, for each set of forecasts provided by each miner during the week, we calculated the following summary statistics: (a) variability in volatility (i.e., the standard deviation of the standard deviations) of the returns; (b) standard deviations in the hourly standard deviation of the returns; and (c) the average distribution’s kurtosis. While (a) helped gauge how each miner's model captured long-term volatility, (b) was used to observe whether miners adjusted their models to variations across different times of the day; and (c) served to check if miners prepared for "unexpected" events by assigning more weight to the tails of their distributions (which kurtosis measures). After obtaining these three statistics, we divided the miners’ rankings into different bins and analyzed how these statistics correlated with each bin. Results are shown in Figure 4.

As observed, all three plots reveal clear differences in how the top 10 miners modeled Bitcoin price returns compared to all other miners:

• Top miners adapted their volatility estimates more frequently to different market conditions, as shown in the left-most plot;

• Top miners also varied their intra-day volatility more, adjusting their models to different times of the day (an example of how this can be done is detailed in our article "Bitcoin’s Volatility Hack: How Savvy Synth Models Crush the Market");

• Top miners consistently produced distributions with fatter tails on average compared to other miners in the competition (some lower-ranked miners also showed this, but it appears to be more noise than a deliberate attempt to model distribution tails).

Takeaways

This week underscores the importance of adaptability of Synth miners in volatile markets and weeks. The tariff deal rumors and a general sentiment shift triggered significant volatility, challenging miners’ predictive models. Miners 136, 133, and 195 demonstrated superior flexibility, adjusting their forecasts effectively during key volatility spikes, which ultimately propelled them to the top of the leaderboard. The consistent performance of these miners, as reflected in their improving leaderboard scores, highlights the value of robust, adaptive algorithms capable of handling both turbulent and stable market conditions. Miners aiming for top rankings must focus on refining their models to navigate such dynamic environments successfully.

Additionally, the analysis of top miners’ strategies reveals that frequent volatility adjustments, intra-day modeling, and fatter-tailed distributions were key to their success, emphasizing the need for comprehensive modeling approaches in the Synth competition.

Introduction

This report analyzes miner performance within the Synth Subnet over April. We examine:

1. Leaderboard dynamics: consistency in the top ranks, week‑to‑week turnover, and the ability of miners to return to leading positions;

2. Performance drivers: how forecast distribution characteristics—volatility and tail fatness (kurtosis)—distinguish top miners from the rest.

Our aim is twofold: to identify whether dominant players exist and can maintain or regain leading status, and to highlight model attributes that correlate with success.

1. Leaderboard Dynamics

We analyze three aspects of Top 10 behavior:

1. Week‑to‑week flows into and out of the Top 10 based on average prompt scores.

2. Total time spent in the Top 10 (in minutes) during April, using smoothed leaderboard rewards.

3. Ranking trajectories of the longest‑standing Top 10 miners over the month.

Participant Overview.  Over the course of April, a remarkable 798 distinct miners entered and exited the Synth Subnet leaderboard. Of these:

• 85 were active both before April 1 and after April 30,

• 245 began before April 1 but dropped out before month’s end,

• 317 joined after April 1 and remained past April 30,

• 151 joined and were deregistered within April.

This level of engagement underscores the broad interest in our forecasting contest—and against such a large field, securing a spot among the Top 10 is a true mark of excellence.

1.1 Weekly Top-10 Turnover

To gauge the depth of competition, we gathered every miner’s prompt scores from April 1 to April 30 and computed each miner’s weekly average (excluding the final four-day partial week). We then ordered miners by ascending average prompt score—an approach that isolates true shifts in ranking, independent of the smoothed leaderboard metrics. In total, 798 miners participated at various points in April (including 85 who were active throughout the entire month), so cracking—and holding—a spot in the Top 10 reflects an exceptionally strong forecasting engine.

Our week-to-week breakdown (Figures 1–4) reveals how a handful of standout miners consistently outpaced the field:

Week 1 → Week 2: Every Week 1 Top 10 remained within the Top 30, underscoring the solidity of their models. Notably, Miners 240, 91, and 196 preserved their Top 10 rankings in Week 2.

Week 2 → Week 3: Only Miner 208 managed to stay in the Top 10, while many of its peers slipped past the 100-rank mark—illustrating the high caliber required to stay at the forefront of the competition.

Week 3 → Week 4: While Week 3’s leaders faced challenges, with many dropping below the Top 200, their ability to recalibrate models paved the way for a strong recovery. Impressively, six of the Week 4 Top 10 had previously ranked in Week 1’s Top 10, underscoring that elite miners can swiftly reclaim leadership with refined strategies.

Key insight: The Synth Subnet’s Top 10 is a showcase of the truly best miners, where only the most sophisticated forecasting models thrive. While the leaderboard sees dynamic movement, reflecting the competitive vigor of nearly 800 miners, the ability of top performers to repeatedly secure Top 10 positions—sometimes across multiple weeks—signals their far superior models. This vibrant contest is successfully sourcing the highest-quality forecasters, driving innovation and excellence in the subnet.

1.2 Miners’ Duration In Top 10

We measured each miner’s total time in the Top 10 by ranking all participants according to their smoothed reward each time rewards were updated, then summing the minutes spent in the Top 10 over April. Figure 5 shows that Miners 33, 39, and 103 each exceeded a total of two weeks inTop 10 , while another cohort—including Miners 196, 91, 240, and 200—spent more than 10 days in the Top 10. These results highlight a group of consistently high‑quality forecasters who sustain leadership for extended periods.

1.3 Miners Ranking Evolution

Focusing on the 20 miners with the longest Top 10 tenures, Figure 6 charts their daily rankings through April. Once they break into the Top 10, many maintain positions for days or even weeks. Though temporary declines—driven by unexpected market conditions or model drift—do occur, miners frequently rebound with timely model adjustments.

Key insight: Sustained Top 10 presence is achievable, and setbacks are reversible through focused refinements.

2. Miners’ Performance Drivers

To uncover what separates the most successful miners from the rest, we examined two fundamental features of each miner’s forecast distribution over April’s one hundred 24‑hour price paths: the way they estimate price volatility and the weight they assign to extreme events (captured by kurtosis).

Price volatility reflects the degree of market uncertainty—a higher volatility implies wider swings in Bitcoin’s returns, while lower volatility suggests more stable behavior. A model that adapts its volatility estimates to changing market conditions can better anticipate periods of calm and turbulence. We measured this adaptation in two complementary ways: first, by calculating the empirical standard deviation of each forecast’s 24‑hour returns and then assessing how those daily volatilities fluctuate across all submitted prompts; second, by reconstructing an intra‑day volatility profile for each forecast—estimating the volatility in hourly increments—and analyzing the dispersion of these profiles over time. The result is a clear view of whether a miner treats volatility as a static parameter or actively refines it to capture both day‑ahead and intra‑day market dynamics.

Kurtosis, on the other hand, quantifies the “fatness” of a distribution’s tails. A Gaussian (normal) distribution has a kurtosis of zero, implying that extreme deviations are exceedingly rare. Positive excess kurtosis indicates heavier tails, meaning the model anticipates and assigns higher probability to large, infrequent price moves. By averaging the excess kurtosis of returns across all forecasts for each miner, we gauge how much each forecaster prepares for tail risk—an essential trait when sudden market events can trigger outsized losses or gains.

Putting these two metrics together produces a nuanced picture of modeling sophistication. In Figures 7 and 8, boxplots of volatility variation reveal that the miners who ranked in the Top 20 by April rewards consistently display higher—and more tightly clustered—volatility adjustments than their lower‑ranked peers. Those top performers not only update their day‑ahead volatility but also intricately shape intra‑day patterns, suggesting a deep responsiveness to Bitcoin’s intraday rhythms.

For a deeper dive into how leading miners fine-tune their volatility forecasts across day vs. night and weekday vs. weekend periods, see our article “Bitcoin’s Volatility Hack: How Savvy Synth Models Crush the Market”, which walks through the exact intraday and weekend dynamics they’re capturing.

Figure 9 then plots each miner’s average kurtosis against their final ranking. The miners who secured the most rewards tend to cluster at much higher kurtosis values, demonstrating a deliberate emphasis on tail risk. Conversely, miners with near‑zero kurtosis mimic Gaussian assumptions and underplay the likelihood of extreme moves—an approach that correlates with lower leaderboard performance.

Together, these analyses illustrate that mastering both volatility adaptation and tail‑risk modeling is not optional but essential for success in the Synth Subnet. Models that treat volatility as dynamic and embrace heavy‑tailed forecasts consistently outperform those that do not.

Takeaways

April’s data reveals three core lessons for Synth miners:

• First, the Top 10 leaderboard experiences substantial weekly turnover, reflecting the dynamic nature of Bitcoin markets and the competitive environment within the subnet. Yet within this churn, a resilient core of miners repeatedly reclaims leading positions by fine-tuning their approaches.

• Second, dropping out of the Top 10 does not preclude a comeback. Whether due to model drift, temporary market shocks, or new entrants, miners that focus on targeted refinements—particularly in volatility calibration and tail-risk modeling—can regain top slots, even if they were previously unranked.

• Third, the most consistent high performers distinguish themselves through distributional mastery. They treat volatility as a dynamic parameter, actively adjusting both day-ahead and intra-day estimates, and they build in heavy tails to anticipate extreme market moves. This combination proves essential for capturing rewards.

By emphasizing these three principles—adaptability, resilience, and statistical rigor—miners can sharpen their models for the days ahead. May these insights inspire your next breakthroughs on the Synth Subnet leaderboard.

Next Steps

Building on our findings that miners have mastered the Phase 1 forecasting challenge, we’re now elevating the contest’s demands and refining our incentive structure to drive ever-greater innovation:

• Expanded Modeling Scope: Participants will be required in Phase 2 to forecast additional crypto assets alongside Bitcoin. By diversifying the asset set, we encourage miners to develop multi-asset forecasting frameworks that seek to perform across varied market regimes.

• Sharper Reward Curve: With the subnet’s proven ability to identify top-tier forecasters, we’ll steepen the reward curve so that the very best miners capture a larger share of incentives, while minimizing rewards of average performers. This adjustment ensures that high-fidelity models receive proportional recognition and funding, further intensifying competition at the leading edge.

These enhancements are designed to push miners to continually raise the bar for forecasting excellence and to ensure that the Synth Subnet remains the premier environment for next-generation market models.

Bitcoin Price Action

During the week of April 21 to April 27, 2025, Bitcoin opened at $85,200 and closed at $93,700, with a weekly low of $85,150 and a high of $95,700. The week experienced significant price movement, driven by a major volatility spike on April 22 between 21:30 and 22:00 UTC. During this period, rumors of a potential tariff deal between China and the U.S. triggered a sharp 2.3% price surge within 30 minutes, pushing Bitcoin from around $91,600 to $93,700. Volatility remained elevated for the next few days, with smaller fluctuations, before stabilizing toward the end of the week as the price settled around $93,700.

Miners Performance

At the start of the week, the top miners by leaderboard score were Miners 4, 44, and 100. By the end of the week, Miners 32, 247, and 33 had taken the lead. Below, we analyze their performance using two key metrics: CRPS and Leaderboard Scores.

Continuous Ranked Probability Score (CRPS)

Lower CRPS values indicate better predictive accuracy.

Early in the week, sudden volatility spikes caused the CRPS values of all miners under analysis to peak around 45,000. Miners 4 and 100, who started with higher scores, saw their CRPS values soar, reflecting their difficulty in adapting to the rapid market shift. Meanwhile, Miners 32, 247, and 33 adjusted more effectively, producing forecasts with lower CRPS values between April 23 and 24. The miners who ended the week at the top of the leaderboard consistently outperformed Miners 44, 4, and 100, especially over the weekend when no significant volatility spikes occurred, leading overall scores to stabilize between 15,000 and 20,000.

Leaderboard Scores

Leaderboard scores reflect an exponentially weighted average of past CRPS performance over a 7-day window, with a half-life of 3.5 days. Lower scores are better.

At the start of the week, Miners 4 and 100 led with scores around 600, while Miners 32, 247, and 33 had scores ranging from 1,200 to 1,400. The volatility spike on April 22 caused a significant reshuffling. Miners 4 and 100 saw their scores rise sharply to around 2,000 by April 23 due to their weaker predictive accuracy during the tariff rumor event. In contrast, Miners 32, 247, and 33 improved steadily, with their scores dropping to around 1,000 by April 27, reflecting their ability to adapt to market conditions.

Takeaways

This week’s performance once again highlights the critical need for miners to adapt swiftly to unexpected market events. The tariff deal rumors on April 22 created a challenging environment, and miners who could recalibrate their models quickly—such as Miners 32, 247, and 33—gained a clear advantage. The leaderboard’s emphasis on consistent, long-term performance underscores the importance of robust, adaptive algorithms that can effectively handle both volatile spikes and stable periods. Miners aiming to maintain or achieve top rankings must prioritise flexibility and continuous model refinement to navigate such dynamic market conditions.

Bitcoin Price Action

During the week of April 14 to April 20, Bitcoin opened at $83,700 and closed at approximately $85,200, fluctuating between a low of $83,100 and a high of $86,400. Overall, volatility was notably lower than in previous weeks. However, brief episodes of heightened price movement persisted.

One of the most notable events occurred on Monday, April 14, between 2 PM and 3 PM UTC, when Bitcoin jumped 1.4% within an hour—from $84,500 to $85,700. This spike coincided with the U.S. stock market opening and the announcement of tariff exemptions for tech companies.

Other volatility events followed—such as Fed Chair Jerome Powell’s speech at the Chicago Economic Club on April 16—but became increasingly muted as the week progressed. The second half of the week was relatively stable, until a final sharp jump on Sunday, April 20, saw the price rise from about $84,000 to over $85,000.

Miners Performance

At the start of the week, the top 3 miners by leaderboard score were Miners 91, 196, and 3. By week’s end, leadership had shifted to Miners 44, 4, and 100. Below, we analyze their performance through the lens of two key metrics: CRPS and Leaderboard Scores.

Continuous Ranked Probability Score (CRPS)

Lower CRPS values indicate better predictive accuracy.

In the more volatile first half of the week, all miners produced relatively similar BTC forecasts, with CRPS values clustering around 35,000 early in the week and improving to about 20,000 by Friday, April 18. During April 15–16, Miners 44, 4, and 100 began to show slightly stronger, more consistent performance than their peers.

As volatility tapered off by April 18, these three miners further distinguished themselves by better adapting to changing market conditions. From April 18–19, their models clearly outperformed others. By April 20, Miners 91, 196, and 3 appeared to have recalibrated their models, regaining performance parity with the top miners.

Leaderboard Scores

Leaderboard scores reflect an exponentially weighted average of past CRPS performance over a 7-day window, with a half-life of 3.5 days. Lower scores are better.

This smoothed scoring system rewards consistent performance over time rather than short bursts of excellence. Initially, Miners 44, 4, and 100 trailed behind, with scores above 1,800 compared to the 1,600–1,800 range of Miners 91, 196, and 3. However, the consistent accuracy of their models allowed them to quickly rise in the rankings. By April 16, they had taken the lead with scores around 1,400.

Their strong performance continued through the week, pushing their scores down to approximately 600. Meanwhile, the early leaders—Miners 91, 196, and 3—struggled to recover, finishing the week with leaderboard scores more than double those of the top three.

Takeaways

This week’s performance evolution offers valuable lessons for subnet participants:

• Miners must strike a delicate balance between quickly adapting to market shocks and maintaining steady, reliable performance over time. The leaderboard’s extended evaluation window means that occasional spikes in accuracy aren’t enough—continuous excellence across varying conditions is essential to earn and hold a top spot.

• The rapid leaderboard shifts observed this week underscore the importance of continual model refinement. As the subnet increasingly rewards long-term, high-quality forecasting, miners who want to break into or stay atop the leaderboard must invest in robust, adaptive algorithms capable of navigating diverse market environments without compromising consistency.

• For Miners: Stay up-to-date on miners' performances with our new Miners Tooling Dashboard! Besides live CRPS and leaderboard values, the dashboard also contains the most recent information about Bitcoin price action and volatility.

• For Traders: Powered by the Synth data produced by the subnet's miners, futures leveraged traders can now make more informed decisions about their liquidation risks! This tool is available in our new Liquidation Probabilities insight page.

Data From 2025-04-07 To 2025-04-13

Bitcoin Price Action

Between April 7 and April 13, Bitcoin opened at approximately $78,300 and closed at around $83,700, marking a significant 6.9% increase over the week. The price exhibited extreme volatility, particularly during the early part of the week following continued uncertainty around the new U.S. tariffs against China announced the previous week, which triggered substantial market fluctuations.

Bitcoin reached its lowest point of approximately $74,500 on April 7, during a volatility spike directly attributed to fears surrounding the escalating trade war between the U.S. and China. From this low, the market initiated a strong recovery, reaching a peak of approximately $85,900 on April 13 before moderating slightly to close the week.

The most dramatic price movement occurred on Monday, April 7, when Bitcoin experienced a substantial intraday volatility spike, with the 5-minute rolling volatility reaching nearly 0.9%. Besides this event, this week saw multiple significant spikes in the 5-minute rolling volatility throughout the period. Notable secondary volatility peaks occurred on April 9  and 10,   reaching approximately 0.8% and 0.4% respectively, suggesting persistent market uncertainty as traders continued to digest the implications of the ongoing trade tensions.

By the weekend, although price momentum had turned strongly positive, volatility remained elevated compared to historical weekend patterns, with another distinct spike observable on April 13 as the market approached its weekly high.

Miners Performance

Four miners were analyzed this week: the previous week's top performers (Miners 91, 196, and 234) and one new contender (Miner 3). We are going to comment on these miners' performance based on the aggregated Continuous Ranked Probability Score (CRPS) and leaderboard values, with particular attention to their ability to manage the multiple volatility spikes observed throughout the week.

Continuous Ranked Probability Score (CRPS)

Lower CRPS scores indicate better predictive performance. The data reveals distinct patterns in miner performance throughout the week. All miners struggled significantly during April 7-8, with CRPS values spiking dramatically to around 80,000. This represented the highest CRPS values observed during the week, reflecting the extreme difficulty of generating accurate predictions during this period of unprecedented volatility.

As market conditions began to stabilize mid-week, CRPS values improved substantially across all miners, with values generally trending downward from April 9 through April 13. The improvement was particularly notable from April 11 onward, with CRPS values dropping below 30,000 by April 12-13, indicating significantly enhanced predictive accuracy as miners adapted their models to the evolving market conditions.

A secondary spike in CRPS values occurred around April 10, coinciding with another volatility event, though this impact was less severe than the initial spike. During this second volatility event, Miner 234 demonstrated the poorest adaptation, with CRPS values consistently higher than its peers.

Leaderboard Scores

Lower leaderboard scores indicate better performance. The leaderboard scores reflect the exponentially weighted average impact of the CRPS values over the preceding 7 days, with a half-life of 3.5 days. This extended evaluation window reveals interesting patterns in miner performance sustainability.

Miner 3, which was not among the top performers in the previous week, demonstrated remarkable improvement throughout this week. Starting with a relatively high score around 2,500 (indicating initial disadvantage), Miner 3 showed consistent improvement, particularly in the latter half of the week, ultimately achieving the best score of approximately 1,750 by week's end.

In contrast, Miners 91 and 196, which were top performers in the previous week, maintained their strong positions throughout most of this week as well. Both miners started with scores around 1,300-1,400 and continued to improve, ending the week with scores around 1,750, further cementing their positions among the top performers.

Miner 234, despite being a top performer last week, struggled to maintain its competitive edge under the new parameters of the leaderboard scoring system. Its score ultimately settled around 1,830, dropping it out of the top three positions by week's end.

The extended EWMA time window with its 7-day cutoff and 3.5-day half-life appears to be effectively identifying miners with sustained performance excellence, as evidenced by the continued presence of Miners 91 and 196 in the top three, while also allowing for the emergence of new top performers like Miner 3 that demonstrate consistent improvement over time.

Takeaways

• Multiple Volatility Events: Unlike previous weeks where market volatility was primarily triggered by a single major event, this week featured multiple significant volatility spikes throughout the period. Miners that successfully adapted their models to these recurring fluctuations performed best, highlighting the importance of developing predictive frameworks capable of rapidly assimilating new market conditions.

• Leaderboard Stability and Turnover: The new leaderboard parameters (7-day EWMA window with 3.5-day half-life) are clearly rewarding sustained excellence. Two of the previous week's top performers (Miners 91 and 196) maintained their positions, while Miner 3 entered the top three at the expense of Miner 234. This pattern suggests that the Subnet scoring system is effectively balancing continuity with the opportunity for new entrants to rise based on merit.

• Adaptation vs. Consistency: The week's data reinforces that miners need to strike a delicate balance between rapid adaptation to market shocks and maintaining consistent performance over time. With the extended evaluation window, sporadic high-performance periods are insufficient; miners must demonstrate sustained excellence across varying market conditions to secure and maintain top leaderboard positions.

• Model Improvement Imperative: The leaderboard dynamics observed this week underscore the critical importance of continuous model improvement. As the Subnet now rewards quality models that deliver consistent results over extended periods, miners seeking to enter or remain at the top of the leaderboard must invest in developing more sophisticated predictive algorithms capable of handling diverse market scenarios while maintaining reliability across time.

Bitcoin Price Action

Between March 31 and April 6, Bitcoin opened at approximately $82,400 and closed at around $78,300, marking a 4.9% decrease over the week. The price experienced significant volatility, particularly following the U.S. announcement of new tariffs on April 2 at 8:00 PM UTC, which triggered an immediate drop. Bitcoin reached a peak of $88,400 on April 2 before declining sharply to a low of $78,100 on April 6. Notably, the end of the week saw another surge in volatility, with 5-minute rolling volatility spiking to about 0.05% on April 6, driven by market reactions to the tariff news and ongoing global economic uncertainty.

The most pronounced price movement occurred right after the tariff announcement, with Bitcoin dropping $6,100 (7.4%) within hours. This decline persisted through the week, with additional volatility spikes at the end of the week, particularly on April 6, as the market continued to react to the tariff implications. Unlike previous weeks, where weekends often brought stability, this weekend maintained high volatility levels, reflecting the broader market's uncertainty.

Miners Performance

Six miners were analyzed: the previous week's top three performers (Miners 54, 148, and 232) and three new contenders (Miners 91, 196, and 234). We are going to comment on these miners’ performance based on the aggregated Continuous Ranked Probability Score (CRPS) and leaderboard values, with a particular focus on their ability to handle the significant volatility observed at the end of the week.

Continuous Ranked Probability Score (CRPS)

Lower CRPS values indicate better predictive performance. The data reveals distinct patterns in miner performance throughout the week. All miners experienced a sharp increase in CRPS values (worsened performance) following the U.S. new tariffs announcement on April 3, with values peaking around 50,000, reflecting the difficulty of predicting during this high-volatility period. Miner 148 performed the worst during this spike, reaching a CRPS value close to 49,500.

Performance stabilized for a short period on April 4 as market conditions temporarily calmed, with CRPS values dropping to around 30,000 before rising again following another period of increased volatility. However, the significant volatility at the end of the week, particularly on April 6, caused another spike in CRPS values, reaching almost 35,000.

Miners 234 and 54 adapted particularly well to the lower-volatility between April 5 and April 6, showing the most consistent improvement during this period, with CRPS values dropping back close to 15,000 by April 6.

Leaderboard Scores

Lower leaderboard scores indicate better performance. The leaderboard scores reflect the (exponentially weighted) average impact of the CRPS values throughout the previous 7 days. Miners 54, 148, and 232 started the week with values between 800 and 900, but they saw their scores increase considerably during the week, ending with scores around 1,400. On the other hand, Miners 91, 196, and 234.

Miners 91, 196, and 234, on the other hand, showed stable and consistent performance throughout the week, even demonstrating slight improvement as their scores decreased from around 1,200 to below 1,000. This suggests that the subnet leaderboard rewards sustained performance rather than sporadic high scores from individual prompts.

Takeaways

• Impact of Geopolitical Events: The U.S. new tariffs announcement on April 3 significantly impacted Bitcoin’s price and volatility, challenging miners’ predictive models. The significant volatility at the end of the week further tested miners’ adaptability, with those who recalibrated quickly (e.g., Miner 234) performing best.

• End-of-Week Volatility: Unlike previous weeks where weekends often brought lower volatility, the significant volatility spike on April 6 highlights the need for miners to develop models that can handle unexpected market reactions, particularly during weekends.

• Scoring System Effectiveness: The extended EWMA time window (from 4 to 7 days, with a half-life increase from 2 to 3.5 days) continues to reward consistent performance, as seen in the relatively narrow range of leaderboard scores by the week's end, despite the significant volatility challenges.

Bitcoin Price Action

Between March 24 and March 30, Bitcoin opened at approximately $86,800 and closed at around $82,300, representing a significant 5.2% decrease over the week. The price exhibited notable volatility, reaching a peak of approximately $88,300 on March 25 before experiencing a sharp decline following the release of the PCE Inflation Data on Friday, March 28 at 11:00 AM UTC.

The most dramatic price movement occurred immediately after the PCE inflation report, when Bitcoin dropped nearly $3,000 (3.5%) within hours. This decline continued through the weekend, with prices stabilizing around the $82,000 mark by week's end.

Throughout the week, the 5-minute rolling volatility showed several spikes, with the most pronounced occurring during the PCE data release. Interestingly, unlike the previous week where weekend volatility was significantly lower, this weekend maintained moderate volatility levels as the market continued to digest the inflation data.

Miners Performance

Six miners were analyzed: the previous week's top three performers (Miners 154, 92, and 57) and three new contenders (Miners 54, 148, and 232). Performance was assessed using the Continuous Ranked Probability Score (CRPS) and leaderboard values.

We are going to comment on these miners’ performance based on the aggregated Continuous Ranked Probability Score (CRPS) and leaderboard values.

Continuous Ranked Probability Score (CRPS)

Lower CRPS scores indicate better predictive performance.

The data reveals distinct patterns in miner performance throughout the week. All miners struggled significantly during the early part of the week (March 24-25), with CRPS values spiking to around 30,000, indicating poor model predictions during this period of high volatility. Miner 57 performed the worst, with CRPS values reaching 31,200.

Performance improved substantially mid-week (March 26-28) as market conditions stabilized, with CRPS values generally staying below 27,000. However, the PCE inflation data release on March 28 triggered another performance deterioration across all miners, though not as severe as earlier in the week.

The weekend period (March 29-30) showed the best performance for all miners, with CRPS values consistently trending downward to their lowest levels of the week, reaching values below 20,000. Miners 148 and 54 adapted particularly well to weekend conditions, showing the most consistent improvement during this period.

Leaderboard Scores

Lower leaderboard scores indicate better performance.

The leaderboard scores reflect the cumulative impact of the CRPS values throughout the week. Miner 232 started the week with the highest score (worst performance among the miners under examination) at around 3,000 points, but showed consistent improvement, eventually joining the top performers with a score around 900.

Miners 154 and 57, which were top performers in the previous week, maintained their strong positions throughout most of this week as well, particularly after the mid-week stabilization period. Miner 57 showed more volatility in performance, with scores fluctuating considerably.

By week's end, all six miners had converged to relatively close leaderboard scores between 800-1300, with Miners 148 and 54 showing the most dramatic improvements, dropping from mid-range scores to among the lowest (best) by Sunday.

Takeaways

• Economic Data Impact: The PCE inflation release demonstrated how significant economic indicators can create immediate volatility in crypto markets. Miners that quickly adapted their predictions before/after the data release performed best.

• Convergent Performance: The narrowing gap between top and bottom performers by week's end suggests that the extended EWMA time window (recently increased to four days) is having its intended effect of rewarding consistent performance over longer periods.

• Adaptation Speed: Miners showing the greatest improvement (particularly 148 and 54) demonstrated superior ability to recalibrate their models following the PCE-induced volatility, highlighting the importance of adaptive algorithms in changing market conditions.

• Weekend Resilience: Unlike the previous week where weekends showed significantly lower volatility, this weekend maintained moderate activity levels, reinforcing that miners should develop models that can adapt to changing market dynamics.

Bitcoin Volatility

One of the key challenges in the Synth subnet is forecasting volatility. Having models that can distinguish between periods of high and low volatility provides miners with a significant edge, particularly when predicting assets prone to frequent price swings, such as cryptocurrencies. In this phase of the subnet, the focus is specifically on Bitcoin.

The baseline model initially recommended for forecasting Bitcoin price movements is the Geometric Brownian Motion (GBM) with fixed volatility. Volatility is estimated from historical data, and miners must determine the optimal time window for this estimation, as well as how frequently to update it.

However, miners who dynamically adjust their volatility models to changing market conditions are more likely to outperform others in the subnet. For instance, consider Bitcoin price movements from December 1, 2024, to March 25, 2025, along with its log-returns and daily volatility:

The figure illustrates frequent peaks of high volatility and periods of sustained volatility. The average daily log-returns volatility during this period was approximately 0.07%. Using this as a fixed input for the GBM model would likely result in forecasts that fail to capture significant price swings, placing miners at a disadvantage over time.

A clear example of structural volatility variation emerges when analyzing data by days of the week (Monday to Sunday) and hours of the day (0–23). The following heatmap displays the standard deviation of 1-minute net returns for Bitcoin over the same period, aggregated by weekday and hour:

The heatmap reveals how volatility fluctuates throughout the week. Monday, particularly Monday afternoon, exhibits the highest volatility. Weekday afternoons tend to be more volatile than mornings, with elevated volatility persisting into the night (until around 1–2 AM), after which it declines to the more stable morning hours. Saturdays, by contrast, appear relatively stable compared to other days. The low volatility observed on Saturdays continues through Sunday morning and early afternoon. However, by late Sunday afternoon and evening, volatility starts rising again, seemingly "preparing" for the upcoming Monday.

Miners’ Volatility Modelling: Two Case Studies

Do miners adjust their forecasts based on changing volatility patterns, or do they rely on a fixed volatility value when predicting Bitcoin prices? To investigate this, we analyzed the forecasts of two miners as follows:

• We examined the period from March 15 (when the V2 scoring system was fully implemented) to March 25.

• We aggregated the total rewards earned by miners during this period and selected the top-ranked miner and the miner ranked 100th.

• For these two miners, we analyzed the first hour of all their provided forecasts within the selected period.

• We calculated net returns from these price forecasts, grouped them by weekday and hour, and generated heatmaps.

The heatmap of the top-ranked miner demonstrates an awareness of Bitcoin’s volatility dynamics. Similar to the actual Bitcoin price, Monday exhibits higher volatility, and weekday afternoons are more volatile than mornings. Additionally, this miner effectively captures the low volatility observed on Saturdays and early Sundays (except for an anomaly on Saturday at 3 PM, possibly due to outliers).

In contrast, the heatmap of the 100th-ranked miner shows relatively fixed volatility values across all days and hours. This miner likely relies on the baseline GBM model with fixed volatility. The lack of variation in forecasted volatility places this miner at a disadvantage, reflected in their lower ranking.

Takeaways

While the GBM model with fixed volatility serves as a solid starting point, miners gain a true competitive edge by effectively modeling Bitcoin’s dynamic volatility. This report highlights one possible approach: accounting for volatility variations across different periods of the week.

However, miners can enhance their models further by incorporating additional factors such as:

• Seasonal and Cyclical Trends – Monthly patterns and broader seasonal influences.

• Volatility Clustering – Periods of high volatility followed by further high volatility, and vice versa.

• Market Regimes and Sentiment – External market conditions and investor behavior.

The core principle of the Synth subnet is the ability of miners to integrate these factors into their forecasts. Achieving this requires careful analysis and robust modeling techniques, but the rewards are substantial.

Our findings demonstrate that top-performing miners utilize more sophisticated models than the baseline GBM. The Synth subnet benefits from miners who continually refine their models, as this leads to higher-quality data for stakeholders such as liquidity providers and Automated Market Makers (AMMs). The liquidity probabilities and expected impermanent losses derived from the forecasts of top miners in the Synth subnet are likely to be more accurate than those produced by in-house models.

By striving for more precise volatility modelling, miners not only improve their own performance but also contribute to the overall robustness of the ecosystem.

Bitcoin Price Action

Between March 17 and March 23, Bitcoin opened at approximately $82,500 and closed at around $86,080. The price fluctuated between a low of $81,177 on March 18 and a high of $87,450 on March 20. Throughout the week, volatility surged, with rolling log-return volatility increasing fourfold from about 0.05% to 0.2%. These fluctuations were likely driven by geopolitical events and general market uncertainty.

Notably, the FOMC (FED) meeting on March 19 triggered a volatility spike to approximately 0.3% around 6 PM, leading to a 2.6% increase in Bitcoin's price within an hour. Over the weekend, volatility significantly decreased, averaging around 0.02%, compared to 0.05% on weekdays.

Miners Performance

Six miners were analyzed: the top three based on leaderboard scores at the start (midnight, March 17) and the top three at the end (11:59 PM, March 23). These were:

• Top miners at the start: Miners 90, 200, and 91.

• Top miners at the end: Miners 154, 92, and 57.

Performance was assessed using the Continuous Ranked Probability Score (CRPS) and leaderboard values. Given the Synth subnet's new scoring system, final emission rewards are now a transformation of leaderboard scores, making additional discussion of those values redundant.

Continuous Ranked Probability Score (CRPS)

CRPS scores indicate model performance, with lower scores reflecting better results.

Volatility spikes, such as after the FOMC meeting, led to increased CRPS values (worsened performance), while stable periods, such as the weekend, improved scores. Miners 90, 91, and 200 performed well early in the week but struggled after the FED meeting volatility spike. Conversely, Miners 154, 92, and 57 performed better in the latter half, benefiting from both post-FED volatility and the calmer weekend conditions. Miner 154, in particular, started with poor CRPS values but showed significant improvement, securing the top leaderboard position by week’s end.

Leaderboard Scores

Leaderboard scores, now a moving average of CRPS values (and thus lower values indicate better performance) , reinforce the observed trends.

Miners 154 and 57 had high initial scores (above 5,000) but improved significantly, ending with scores near or below 1,000. Their performance surged as volatility shifted mid-week, helping them climb the leaderboard. Meanwhile, Miners 90, 91, and 200 saw gradual score deterioration from March 19 onward, ultimately losing their top spots.

Takeaways

• Miners who adapted to Bitcoin’s changing volatility performed best, especially under the new scoring system, which rewards sustained success over short-term luck.

• Bitcoin volatility differs significantly between weekdays and weekends (and between day and night hours). Incorporating these patterns into models can provide a competitive advantage.

• Recent increases in subnet registration costs aim to discourage multiple miner registrations for lottery-like rewards.

• To further emphasize long-term performance, the Synth team plans to extend the EWMA time window (currently set at four days) used in leaderboard calculations.