This research presents a comprehensive behavioral analysis of DEX wallets on Uniswap, combining both liquidity provisioning and token swapping activities to derive unified wallet-level behavioral scores. Across a sample of 20,000+ randomly selected addresses, we evaluate user behavior through two key lenses:

• LP scores reflecting engagement, holding patterns, and category coverage across stable-stable, stable-volatile, and volatile-volatile pools, and

• Swap scores capturing volume, frequency, and holding intent in token trades.

By analyzing score distributions and behavior patterns, the study identifies distinct user archetypes—from passive one-time participants to high-conviction, protocol-aligned actors. The findings underscore how jointly modeling LP and swap behavior enables a robust proxy for wallet-level commitment, risk exposure, and contribution quality within decentralized exchange ecosystems.

Dust LPs: Are like “hit-and-run” participants — they provide tiny amounts of liquidity for extremely short durations, often just to test or exploit price movements. They rarely stick around and contribute little to long-term protocol stability.

Anchors: Behave like long-term investors — they provide depth, consistency, and rarely withdraw. Anchors are essential to the protocol’s liquidity health and act as its most reliable economic contributors.

• Total Sample size: 20,825 Wallets

• JSON-Formatted Source Data

• LP zScore Range: 0-1000

• Swap zScore Range: 0-300

• Protocol: Uniswap

• Latest Swap

  • Timestamp: 1747076253, May 12, 2025 at 06:57:33 PM UTC

  • Wallet: 0x9db8c22fbeef0352ee5c0a3a3aceb39002e8a161

• Latest Deposit

  • Timestamp: 1747061247, Wed, May 12, 2025 at 09:27:27 AM (UTC)

  • Wallet: 0x611f6479ca9370b163c815699163572597b49e50

• Latest Withdraw:

  • Timestamp: 1747072129, Wed, May 12, 2025 at 12:08:49 PM (UTC)

  • Wallet: 0xae5cae800e3e2d4b59a34a023ee89012c140c168

• Mean LP zScore: 390

• Median LP zScore: 364

• Mode LP zScore: 520

Most LP wallets show a clear preference for operating within a single category (e.g., only stable–volatile). As score increases, a small subset of wallets begins to diversify across two or all three LP categories, indicating more strategic and risk-aware behavior. [[[To see the graphs.]]]

1. Stable–Stable: e.g., USDC–USDT — lowest volatility, conservative exposure

2. Stable–Volatile: e.g., USDC–ETH — moderate risk with some price exposure

3. Volatile–Volatile: e.g., ETH–WBTC — high volatility, higher potential yield

• Single-category wallets dominate the population, especially in the 100–600 score range. Most seen activity is covered in the Stable-Volatile category.

• Two-category wallets become visible starting from the 200–300 range and peak between 300–500. Most common combinations were Stable–Volatile + Volatile–Volatile, indicating moderate diversification.

• Only a handful of wallets (fewer than 100 total) engage with all three LP categories, and these are largely concentrated between scores 300–600. Their rarity emphasizes how behavioral completeness (depth + breadth) is difficult to achieve and rare even among high scorers.

In total, over 92% of all wallets remain confined to a single LP category, while only less than 1% demonstrate full category diversification. This pattern reinforces how the LP score not only reflects depth of activity, but also breadth of strategy.

• Deposit frequency peaks in the mid-tiers

  • Wallets scoring 300–500 show the highest deposit rates (~2.3–2.5 deposits/month), reflecting strategic, layered provisioning. Both lower (0–200) and higher (>600) tiers show reduced frequency, suggesting either casual or highly selective engagement.

• Withdrawals are frequent early, then fade

  • LPs in the 200–400 range average up to 8 withdrawals/month, often exiting quickly or reacting to volatility. After 500, withdrawal rates plummet — top scorers rarely withdraw at all.

• Churn dominates low scores; retention dominates high

  • The withdrawal frequency ratio is >1.0 for score ranges below 400, meaning these wallets withdraw more often than they deposit. Ratios fall below 0.3 after 500, indicating a shift to long-term commitment.

• Holding duration scales with score — and trust

  • Holding periods increase from ~1.8 days (0–100) to over 131 days (900–1000). This clearly shows a growing level of patience and alignment with protocol stability among high scorers.

• Liquidity retention defines elite behavior

  • LPs with scores above 600 retain

• 0–100: Dust LPs

  • One-time deposit + immediate full withdrawal.

  • Zero liquidity remaining; avg hold = minutes.

  • Purely test-driven behavior with no commitment.

• 100–200: Hesitant Testers

  • Slightly more holding (up to a few days), still full exits.

    • No diversification, minimal re-entry.

    • Mostly exploring, not investing.

• 200–300: Early Explorers

  • Holding time increases (~5–20 days), some multi-deposit activity.

  • Still single-pool, full exits common.

  • Shows curiosity, but not consistency.

• 300–400: Transitional LPs

  • Some wallets hold >60 days, begin partial exits.

  • More deposits and staggered withdrawals.

  • Signs of strategy, but still inconsistent.

• 400–500: Strategy Seeds

  • Frequent deposits, phased exits begin.

  • Holding up to months, retention rising (10–20%).

  • Commitment building, but narrow (still single-pool).

• 500–600: Retention Takes Hold

• zScore is not just a number — it’s a behavioral fingerprint.

  • Our model captures how users provide liquidity, how long they stay, how often they churn, and how much they commit. This goes far beyond volume-based metrics — it decodes intent.

• Higher scores reflect elite LP behavior — long-term, disciplined, and protocol-aligned.

  • Users in top score ranges demonstrate patient capital, minimal withdrawals, layered deposits, and liquidity retention exceeding 95%. These are not speculators — they behave like Dex-native portfolio managers.

• Lower scores expose reactive or opportunistic users.

  • Quick exits, short holding durations (often < 2 days), and high withdrawal frequency (ratios > 1.0) are signatures of churn-prone wallets. The model surfaces them with precision.

• Mid-range score ranges reveal evolution in action.

  • Wallets here are not yet optimal, but they’re improving — reducing churn, increasing retention, and experimenting with longer holds. The score doesn’t just classify — it tracks behavioral growth.

• The scoring curve is consistent, interpretable, and scalable.

  • From 0 to 1000, every score jump corresponds to observable on-chain behavior shifts. This makes the model not only fair, but explainable — a key trait for trust and adoption in credit systems.

• Strategic behaviors are rewarded, not just volume.

Just-In-Time (JIT) liquidity providers are wallets that add liquidity right before a trade and remove it shortly after, aiming to extract fees while minimizing exposure. In our current framework, these LPs are not assigned a numerical score. Instead, they are classified separately based on deposit/withdrawal timing and low capital commitment. This allows us to flag opportunistic behavior without distorting score distributions meant to reflect long-term, protocol-aligned liquidity.

As the framework matures, future iterations may incorporate dedicated scoring or penalties to better account for JIT patterns.

• 72.75% of wallets didn't have any swap transactions.

• Total Wallets: 20825

• Missing swap transactions: 15150

• Wallets which show swap transactions: 5675

Below is the score distribution of those wallets.

• Mean Swap zScore: 56

• Median Swap zScore: 33

• Mode Swap zScore: 12

• Most users are low-frequency, low-risk swappers

  • Wallets in the 0–50 score bin average just 2 swaps and ~$890 total volume. They typically use only 1–2 tokens and avoid volatile pairs. These are likely occasional or test users, not strategic traders.

• Mid-tier users show structured, cautious engagement

  • Users scoring 100–200 execute 10–20 swaps and grow volume to $170K–770K. Token diversity also rises to 3–4. They're more risk-aware, mixing stable→volatile and volatile→stable activity without going full degen.

• Swap volume scales exponentially with score

  • From ~$1.9M in 200–250 to $96M+ in 300–350, high scorers transact massive volumes — reflecting institutional or bot-level behavior. Swaps are not just frequent but strategically timed and large.

• Volatile trading defines high performers

  • In the 300–350 bin, users average 933 volatile↔volatile swaps, compared to <1 in the lowest bin. These wallets embrace volatility — likely for yield farming, arbitrage, or leverage-based plays.

• Token diversity rises gradually, but meaningfully

  • Low scorers stick to ~1–2 assets, while high scorers use 5–6 unique tokens. This suggests growing portfolio complexity and protocol exploration as score increases.

• Top scorers are rare, active, and risk-optimized

  • Only a sliver of users land in the 250+ range, but they dominate in every behavioral metric: swap count, volume, token variety, and volatile exposure. These users are likely whales, bots, or highly engaged DeFi veterans.

• 0–50: Dust Swappers

  • Minimal activity — likely one-time or test swaps

  • No meaningful risk exposure or strategy

  • Behavior suggests casual curiosity, not real intent

• 50–100: Cautious Tinkers

  • Begin exploring volatile assets, but still lightly

  • Token use is narrow, swaps are reactive

  • Behavior indicates early discovery phase without follow-through

• 100–150: Experimental Traders

  • More varied swap types, including volatile pairs

  • Token diversity begins to show

  • Behavior reflects deliberate testing of market conditions

• 150–200: Developing Strategists

  • Swap patterns show deliberate sequencing across types

  • Consistent use of volatile assets appears

  • Strategy is forming, though still moderate in volume and scope

• 200–250: Tactical Executors

  • Behavior reflects planning — swaps are diversified and balanced

  • Token usage expands, swaps target varied risk/reward paths

  • Likely active DeFi users with moderate conviction

• The score reveals trading intent — not just activity.

  • Our model tracks how wallets engage with volatility, asset diversity, and swap structure. It doesn’t simply tally trades — it uncovers how users approach markets: are they probing, rotating, or executing with conviction?

• High scores reflect strategic, risk-calibrated execution.

  • Top scorers aren’t just active — they’re purposeful. They maintain diverse token positions, navigate between stable and volatile assets with intent, and often favor volatile↔volatile swaps that signal market-making, arbitrage, or high-yield strategies.

• Low scores surface passive, casual, or test-driven behavior.

  • Single-swap wallets with minimal token use and no risk rotation are easily identified. These are either casual users, bot tests, or early-stage wallets not yet committed to any DeFi thesis.

• Mid-tier scores capture behavioral progression.

  • Users in the middle ranges are developing patterns — diversifying tokens, experimenting with swap types, and scaling volume. They may not be experts, but they’re learning. The score reflects this growth arc.

• The scoring is volume-aware but behavior-first.

  • Large volume alone won’t yield a high score — it must be paired with intentional risk-taking, diversification, and frequency. This prevents whales or bots from gaming the model through raw size.

• Swap score generalizes across user types and intents.

  • Whether a wallet prefers slow, diversified stable swaps or aggressive volatile trading, the score adapts. It doesn’t favor one strategy — it evaluates how committed, thoughtful, and engaged the behavior is across the swap spectrum.

This study offers a data-driven framework for understanding Uniswap user behavior, unifying liquidity provisioning and token swapping into a comprehensive wallet-level scoring system. By analyzing over 20,000 wallets across both LP and Swap dimensions, we demonstrate how decentralized exchange (DEX) activity can be transformed into structured, interpretable behavioral signals.

The LP zScore captures a wallet’s relationship with liquidity — measuring not just deposits and withdrawals, but the consistency, patience, and retention that define long-term alignment with protocol needs. In parallel, the Swap zScore reflects how wallets interact with volatility, risk, and asset diversity — distinguishing between casual testers and strategic, high-conviction traders.

Together, these scores form the zScore: a behavioral fingerprint that quantifies how users engage, evolve, and contribute to protocol stability. From dust swappers and hesitant LPs to whale-caliber traders and disciplined liquidity managers, the zScore enables precise segmentation of user archetypes — with clear implications for creditworthiness, capital reliability, and on-chain reputation.

While lower-tier wallets remain the numerical majority, it is the top 5–10% of users — those with high behavioral scores — who represent the backbone of protocol-aligned activity. These users don’t just transact; they participate with purpose, structure, and persistence.

As decentralized finance moves toward more personalized incentives and trust-aware mechanisms, tools like the LP and Swap zScore will be essential. They help protocols move beyond raw activity metrics — to identify not just who showed up, but who stayed, who adapted, and who mattered.