A short excursion into TradFi-metrics…

Okay,... we are not here to glorify the existing financial system. It is broken in many ways, and we need to build a better one! But…some of these folks came up with pretty good models and metrics to explain how this whole system works, and they did this a long time ago. Many have withstood the test of time. One of these modelers was Wiliam F. Sharpe who came up with the eponymous Sharpe ratio in 1966! The Sharpe ratio describes the risk-adjusted returns of an asset compared to a risk free asset. Generally speaking, riskier assets should yield higher returns to compensate for the additional risk, and the Sharpe Ratio is a way to measure that. Pretty cool, huh?

Let's dive into the math...

…on second thought, let’s not!

If you’re not a math geek, the Credmark Terminal will do this work for you and just show you the results. You only need to select the tokens of your choice and you will be able to compare their returns against each other. It is that simple. All you have to do is go to the Terminal and navigate to the Sharpe Ratio page.

...but I still don't get it!

Ok…yeah there are some weird graphs and numbers, but how will that help you? To recap, Sharpe Ratio is a number that allows comparison of different assets and portfolios in terms of risk-adjusted returns.

To put it even more simple:

• Number high - 👍

• Number low or negative - 👎

As a rule of thumb, Sharpe ratio numbers above 2.0 are considered very good, anything between 1.0 and 2.0 is acceptable and anything below 1.0 is considered bad. Thus, the Terminal gives you an easy option of selecting the tokens you like and comparing them over a period of time.

If you hold some ETH, SHIB and AAVE Token, you can easily track their numbers (Spoiler alert: they all performed pretty bad recently) but keep in mind that we look at very short timeframes of 30, 60 or 90 days here.

How to move on from there

Insight from the Sharpe Ratio can be implemented to evaluate if tokens are worth the risk. Do they perform dramatically better then some so-called risk-free assets like 10y Treasuries or are they even worse? This is a very important question, especially if you manage a big portfolio of different assets and tokens. The Sharpe ratio might be a useful indicator to identify tokens that are underperforming, especially if you combine them with other metrics like volatility. But we will leave this for you to decide.

Again…the Terminal proposes a use case for Credmarks data. It will not be perfect, it is just a demonstration of what you could do. It is not the ultimate truth you should base any financial decision upon but it should trigger your creativity and broaden the entrance into the DeFi rabbit hole.

This article was first published here:

https://credmark.com/blog/visualizing-on-chain-data-part-2

So if you are looking to understand this massive amount of data, you need to use tools. And one of the most efficient tools to understand whopping amounts of data is to visualize it and look for patterns.

At Credmark we organize and produce data, but we don’t build visualization tools. Nonetheless, we wanted to see it ourselves, and we wanted to show others an example of the data we produce. We wanted to connect all the dots to show what we have to offer. Ideally, we wanted to inspire people to think even further about the possibilities that lay beyond the raw data.

For us, it was important to demonstrate the blend between on-chain data and useful risk analysis for DeFi. So we built a showcase, the Credmark Terminal.

Not Another Dashboard!

Sure, there are dashboards out there….there are thousands of them. So why release yet another one?

Well, simple. Ours is different. And we don’t only mean the nice design and the interactivity. Most importantly, the data that drives this dashboard is different. This data that has not been seen in any other dashboard. This data is the product of financial risk analysis models that use on-chain data as their source. These models were built and deployed on the Credmark Platform.

As we are very proud of what is possible with the Terminal, we want to get a step further and dive into some “real world” examples, to show how you can visualize our data from the Terminal to gain insights into what is going on in the markets. This data, of course, is also directly available on our platform or through one of our APIs.

We will now highlight some of the metrics we have in the Terminal. The first one will be the Balance Ratio.

Looking at the FTX drama on-chain

2022 is a weird time to be in Crypto. Some heavy macro events pulled us down. We had the Terra / Luna debacle as well as the fallout of 3AC, BlockFi, Voyager, Celsius….but the biggest 💩 of all was set to come in the form of SBF and his Alameda / FTX empire.

While FTX was a centralized exchange (CEX), they (and Alameda even more) had their claws in DeFi as well. And as soon as they leave their centralized orderbooks and OTC trading desks, and enter the main stage on the blockchain, they leave traces. And we are able to pick them up.

Liquidity is king... but for which tokens?

So with every crisis there comes the liquidity crunch. Degens flew out from risky assets into some safe havens like USDC or DAI. But what about other so-called stablecoins? Remember that Alameda was a major user of Abracadabra, and they used FTT (FTX´s own utility token) to mint Magic Internet Money (MIM). And they minted A LOT. As on November 3rd, more than 35% of the outstanding MIM supply was backed by FTT¹.

As the rumors about Alameda surfaced and CZ officially announced his selloff plan for FTT, users flew from MIM. And you could follow it live on the biggest decentralized stablecoin exchange of all….Curve.

Here we see some direct insights that the Terminal offers us. We see the trading volume and total value locked (TVL) in the 3Crv/MIM pool but, most importantly, we see the Balance Ratio on the top left.

In short, the Balance Ratio shows how unbalanced the pool is. A number close to 1 means perfectly balanced, so if there are two assets in the pool, each one makes up roughly 50%. When there are 3 assets in the pool and each individual one is about 33%, you would consider this pool also perfectly balanced and the Balance Ratio would show a number close to 1.

The Balance Ratio of the 3Crv/MIM pool reached a low of 0.3535 on Nov 8. The Balance Ratio recovered recently, but the Curve pool is still very unbalanced towards MIM with over 86%.

So what does this mean?

People were selling MIM against the other stablecoins of the pool, thus causing sell-pressure against the peg of MIM. This finally caused MIM to depeg to $0.95 on Nov 8th but soon restored the peg again.

But not only was MIM under pressure, so was USDT, which raised concerns worth looking at. The 3Pool on Curve (DAI/USDC/USDT) shows some significant imbalance as well, with a share of 13.9% DAI, 12.5% USDC, and 73.5% USDT, instead of the ideal 33% split.

You will see the change in the Balance Ratio chart for the 3pool in the Terminal as well as rise in the trading volume. The Balance Ratio tanked from nearly 1 to 0.3634 while the Trading Volume spiked from below $100m daily volume to over $2b.

You can draw your own conclusions from these charts, but in general, investors seemed to limit their exposure to USDT and once again favor USDC and DAI instead.

Balance Ratio Conclusions

Imbalances in Curve pools should always be watched closely as they are indicators that something is going on in the markets. So every savvy investor should pay attention to this very important metric and Credmark´s data, available via our APsI or the Terminal, should be your go-to source.

Footnote

¹ Insights from the FTX implosion (Bankless Substack)

² Data retrieved from Curve.fi/#/ethereum/pools/mim/deposit

³ Chart based on price data from Credmark Token API

⁴ https://curve.fi/#/ethereum/pools/3pool/deposit

This article was first published here:

https://credmark.com/blog/visualizing-on-chain-data-part-1

Artificial intelligence (AI) is no longer a distant dream or a concept confined to the realms of science fiction. It's a reality that's transforming every facet of our lives and businesses. One area where AI is making significant strides is in the world of finance, specifically in the burgeoning field of cryptocurrencies. The crypto market, with its volatility and complexity, presents a unique challenge and an equally unique opportunity for AI.

The Power of Crypto Data

The beauty of the crypto world lies in its transparency and the vast amount of data available on-chain. Every transaction, every movement of funds, every interaction with a smart contract is recorded on the blockchain, providing a treasure trove of data for AI to learn from. However, the sheer volume of this data can be overwhelming. It's like trying to find a needle in a haystack. You need to know what you're looking for, and you need the right tools to sift through the noise and find the valuable insights hidden in the data and be able to train AI and machine learning (ML) models.

The Challenges of Training AI Models

Training an AI model is not a straightforward task; it's a complex process that requires a careful balance of various elements. Here are some of the key challenges and necessities:

1. Quality and Quantity of Data

The foundation of any AI model is the data it's trained on. The quality and quantity of this data can significantly impact the model's performance. In the crypto world, there's no shortage of data, thanks to the transparency of blockchain technology. However, the challenge lies in filtering out the noise and identifying the relevant data points. This is where platforms like Credmark come into play, offering curated data sets that can help train more accurate models.

2. Labeled Data

Understanding the meaning of data can be important when training a model. We need to be able to distinguish between a mint and a burn event, for example. Naively decoded smart contract data doesn’t provide these labels. When possible, Credmark always provides this critical context.

3. Feature Selection

In the context of AI, features are individual measurable properties or characteristics of the phenomena being observed. Selecting the right features is crucial for training an effective model. In the crypto space, these features could be transaction amounts, wallet addresses, timestamps, or smart contract interactions. The challenge is identifying which features are most predictive of the outcomes you're interested in.

4. Overfitting and Underfitting

Overfitting occurs when a model is too complex and starts to learn the noise in the data, resulting in poor performance on unseen data. Underfitting, on the other hand, happens when the model is too simple to capture the underlying structure of the data. Striking the right balance is crucial for creating a model that generalizes well to new data.

5. Model Interpretability

While complex models like deep neural networks can provide high accuracy, they often act as "black boxes," making it hard to understand how they're making predictions. This lack of interpretability can be a significant issue, especially in a field like finance where transparency and understanding are crucial.

6. Adapting to Market Dynamics

The crypto market is highly volatile and can change rapidly. An AI model trained on past data might not perform well if the market dynamics change significantly. Therefore, it's essential to continuously update and retrain the models to keep up with the changing market.

Navigating the complex landscape of AI model training in the crypto space can be daunting. The challenges are numerous, from ensuring the quality and relevance of data to selecting the right features, avoiding overfitting and underfitting, ensuring model interpretability, adapting to market dynamics, and complying with regulatory and ethical standards. However, these challenges are not insurmountable, especially with the right tools and resources at your disposal.

Enter Credmark: Providing Crypto Data for AI

Credmark offers a comprehensive solution for harnessing the power of crypto data to train AI models. Their platform allows developers to create DeFi (Decentralized Finance) risk models faster than ever before. The integration of curated data sets, Python, and a library of pre-built data structures and models provide an incredible springboard for AI development.

Credmark's Data Offering

Credmark's platform enables developers to focus on product iteration instead of building their own web3 data stack or creating custom integrations to wrangle data. This means more time can be spent on refining the AI models and less on dealing with the intricacies of data management.

Token API

This API provides access to granular token data made for deep analysis and decentralized applications (dApps). It includes current and historical token prices, decentralized exchange (DEX) liquidity, holdings, and token metadata. This data can be used to train machine learning models to predict token prices, understand liquidity patterns, and more.

Portfolio API

This API allows you to trace all wallet activity and manage your portfolio with integrated risk models. It provides token price and balance, wallet activity, and risk metrics. This can be particularly useful for building AI models that predict portfolio risk or track wallet activity patterns.

DeFi API

This API is a gateway to DeFi, purpose-built for quants, modelers, and web3 builders. It provides access to over 200 financial models with one endpoint and allows you to create your own models with web3's most robust modeling platform. This API can be used to build sophisticated AI models that leverage financial data from the DeFi space.

Raw Data

Credmark also provides raw blockchain data as well as derived data, which can be used to monitor or trace any activity from the genesis block. This data can be used to train AI models on a variety of tasks, such as detecting fraudulent transactions or understanding blockchain activity patterns.

How Can Credmark Help?

Credmark, with its comprehensive suite of tools and services, is uniquely positioned to help both seasoned and budding AI practitioners overcome these challenges and unlock the full potential of AI in the crypto space. One of the challenges with other data integrations is the difficulty in accessing and processing blockchain data. Credmark simplifies this process by providing decoded, indexed data that can be easily accessed from a Python environment. This means that data scientists and developers can use familiar Python tooling to work with the data, which can significantly speed up the model development process.

Moreover, Credmark's platform integrates with all the necessary data science libraries and supports Jupyter Notebook, making it an excellent tool for developing and training machine learning models. The platform also allows developers to easily write their own models and query historical data sets via APIs using an open-source framework, which can save time and resources that would otherwise be spent on building a custom data stack or creating custom integrations.

Final Thoughts

In conclusion, the data that Credmark offers is a powerful tool for training AI models in the crypto world. By providing a rich, curated, and accessible source of crypto data, Credmark can help to drive the next wave of AI innovation in the crypto space. The future of AI and crypto is bright, and platforms like Credmark are at the forefront of this exciting intersection of technologies.

About Credmark

Credmark runs a financial modeling platform powered by reliable on-chain data. We curate and manages DeFi data making it available via API and thearound the globe and across industries.

Our community of quants, developers, and modelers actively build models for the DeFi community by leveraging our data API and tools. Join the growing community and together we will advance the next-generation financial system.

This article was initially published here:

https://credmark.com/blog/harnessing-the-power-of-crypto-data-for-ai