How we analyzed 43,618 hyperliquid addresses and found the most solid 12 ones' strategies
Source: https://x.com/i/article/2094385134783131648
Over the past week, we studied
43,618 Hyperliquid addresses to answer one question: among accounts that actually made money, what trading strategies showed up most often?
The answer was not just as simple as “buy the right coin early.” In the strongest 12 accounts we inspected, the dominant pattern was much more operational. Eight behaved like high-turnover, two-sided execution systems. Three looked like active intraday traders or scalpers. Only one looked like a concentrated directional trader.
The useful signal is how the address moved through the market: cadence, buy/sell balance, turnover, position concentration, market selection, and the amount of profit generated per unit of risk recycled.
The screen: from 43,618 addresses to 1,681 profitable candidates
The screen had five stages: 1. Read the full leaderboard CSV
We loaded all 43,618 rows, including each address, rank, nickname, account value, and 1D, 1W, 1M, and all-time PnL, ROI, and volume. This gave us the full snapshot rather than a hand-picked sample.
2. Remove low-quality samples
We kept accounts with at least $10,000 in value, positive all-time PnL and ROI, at least $1 million in all-time volume, and at least $100,000 in monthly volume. This removed tiny high-ROI accounts, inactive wallets, and one-off winners.
3. Check whether performance continued recently
We compared PnL and ROI across the 1D, 1W, 1M, and all-time windows, paying particular attention to weekly and monthly results. A strong historical trade was not enough if the address had stopped making money.
4. Build a robust composite score
We scored each address using all-time PnL, all-time ROI, account value, all-time volume, monthly volume, and the number of recent profitable windows. Combining these measures reduced the distortions created by ranking accounts on any one metric.
5. Attribute strategy from fills
For the top addresses, we inspected Hyperliquid public fills to see which assets they traded, how often they traded, whether they bought and sold both ways, how much notional they turned over, and how concentrated their activity was. This moved the analysis from who made money to how they appeared to make it.
That screen produced 1,681 profitable candidates. From there, we selected 12 high-scoring addresses for closer fill analysis.
The method is still a proxy. The CSV gives 1D, 1W, 1M, and all-time snapshots, not a complete daily equity curve. We cannot prove that an address compounded smoothly for 365 days. What we can do is use a stricter filter than leaderboard rank alone: positive PnL, positive ROI, sufficient account value, meaningful volume, recent continuation, and observable trading behavior.
The strategy split was 8-3-1. More importantly, the three groups were not simply trading at different speeds. They appeared to rely on different sources of profit.
Type 1: high-turnover, two-sided execution
The eight addresses produced 16,000 returned fills and $51.37 million in gross fill notional. Their combined order flow was almost perfectly balanced: 49.4% buys and 50.6% sells.
Across the returned slices, six of the eight addresses had positive net closed PnL. Together they generated $156,464 in net closed PnL, equal to about 30.5 basis points of gross notional.
Those numbers describe a strategy built around throughput. The observed profit was small relative to turnover, but the addresses repeated the process thousands of times. They did not need a large market move. They needed a small execution advantage to persist across many trades.
The public fills cannot prove that these addresses were market makers. We do not have their resting orders, maker/taker status, full inventory, or hedges on other venues. But the observable behavior is consistent with several related methods:
- quoting or trading on both sides of the market;
- capturing small spreads or temporary price dislocations;
- reducing inventory after short directional moves;
- short-horizon mean reversion;
- cross-market or cross-asset hedging.
Three addresses show different versions of this operating model.
0xe4c6ae25959d7fc66cf2dd5965fb78c5e09c4048 traded $24.92 million in gross notional across its returned 2,000 fills. The slice produced $113,079 in net closed PnL, or 45.4 basis points of gross notional. BTC represented 69.2% of its observed notional, and buys accounted for 56.7% of fills. This looked like a large-ticket execution strategy concentrated in the deepest market.
0x523852be2db1a76a0e088ecbff32e849544054e5, known as perpfumbler, spread the same broad behavior across BTC, xyz:SP500, HYPE, xyz:XYZ100, and xyz:TSLA. Its returned fills covered $12.27 million in gross notional and produced $28,864 in net closed PnL, or 23.5 basis points. Buys represented 43.4% of fills. The address was not expressing one asset thesis; it was applying a high-turnover method across markets.
0x399965e15d4e61ec3529cc98b7f7ebb93b733336 was the fastest example. The API returned 2,000 fills covering about 20 minutes, with a median gap of 0.21 seconds. Its flow was 49.85% buys and 50.15% sells. The returned slice generated $2,187 in net closed PnL on $1.44 million in gross notional, or 15.2 basis points.
The three addresses differed in asset mix and trade size, but their economic model looked similar: find a small edge, keep exposure controlled enough to trade both ways, and repeat.
Type 2: active intraday trading and scalping
Three addresses traded frequently but failed the balance test for the two-sided group. Their order flow leaned more heavily in one direction, making their PnL more dependent on short-term price movement.
Together, these three addresses produced 5,071 returned fills and $15.35 million in gross notional. Their weighted buy share was 70.1%. All three returned slices showed negative net closed PnL, even though the three accounts had earned $61.43 million in all-time leaderboard PnL.
That difference is important. These are profitable addresses, but their recent returned fills did not show the same small, positive execution edge as the two-sided group. Their profitability appears more episodic. A directional intraday trader can lose through one sequence and recover through a later move; a market-making-like strategy is expected to distribute results more evenly across turnover.
The three addresses also used different forms of intraday concentration.
0x8c625ff57d8a4374784c7eff585dfdc42ccec974 focused heavily on DOGE. DOGE represented 93.0% of returned notional. The API returned 1,071 fills across 23.35 days, with 30.1% buys and 69.9% sells. This looks less like neutral liquidity provision and more like repeatedly trading one directional campaign within a single volatile market.
0x77375a8c9d13bf79afb2a87f1b0ac1dfd5f5bf66 traded ETH, SOL, PUMP, and BTC, but 94.7% of its returned fills were buys. The slice contained $14.09 million in gross notional and negative $142,648 in net closed PnL. The account still showed $47.06 million in all-time profit. The observable sequence could represent aggressive accumulation, a hedge against exposure elsewhere, or a losing leg inside a broader profitable strategy.
0xc926ddba8b7617dbc65712f20cf8e1b58b8598d3 traded much smaller tickets, averaging about $161 per fill. Buys represented 67.1% of its 2,000 returned fills. Its behavior fits a rapid scalping or short-term directional system, but the returned slice lost $14,572.
The shared feature of this group is not simply high frequency. It is high frequency plus directional imbalance. Their method needs price movement as well as execution.
Type 3: concentrated directional trading
Only one of the 12 addresses fit the concentrated directional category.
0x862dd8e68f30693e3d3c9daa42a440bc6d2a1f0c had just 14 returned fills across 12.05 days. Every fill was a buy, and every fill was in
@107. The observed gross notional was $18,504, with no realized PnL in the returned slice.
The account had $6.45 million in all-time leaderboard PnL, but the public record does not connect that lifetime profit to these 14 fills. The observable behavior only supports a narrower statement: during the returned period, the address was building a concentrated position rather than trading around a short-term inventory.
This is what a directional strategy looks like from fills: low cadence, one-sided execution, high asset concentration, and profit that depends on a subsequent price move rather than immediate turnover.
It was also rare among the 12 strongest accounts. Concentrated directional trading can produce large gains, but it did not appear to be the typical method among the addresses that ranked highest for profitability, consistency, scale, and activity.
Asset choice follows the strategy
The addresses did not share a single favorite asset. Their market choices followed the kind of edge they were trying to capture.
BTC, ETH, and SOL dominated observed notional. These markets support larger trade sizes, deep liquidity, and frequent entry and exit. HYPE, PUMP, xyz:SKHY, and xyz:CRCL generated many more fills relative to their notional, suggesting smaller tickets and faster recycling.
The difference is substantial. The average observed BTC fill was roughly eight times the size of the average HYPE fill. xyz:SP500 had the largest average fill in the table, while xyz:CRCL had one of the smallest.
For the high-turnover two-sided group, BTC accounted for $21.66 million in returned notional. ETH, xyz:SP500, HYPE, xyz:SKHY, and xyz:CRCL formed the next tier. These addresses appeared to move between deep markets for size and smaller markets for denser execution opportunities.
The intraday group concentrated more heavily in ETH, SOL, PUMP, BTC, and DOGE. That mix is consistent with a strategy seeking short-term directional movement and volatility rather than only depth.
The useful lesson is not that profitable addresses prefer BTC. It is that asset selection is part of strategy design. A throughput strategy needs liquidity. A scalping strategy needs movement. A concentrated directional strategy needs conviction in a specific market.
What the profitable addresses actually have in common
The 12 addresses did not agree on direction. They did not trade the same assets. They did not even operate on the same timescale.
What separated the dominant group was the way profit was produced. The two-sided addresses turned over $51.37 million in the returned slices to realize roughly 30.5 basis points. Their edge looked small, repeated, and operational. The intraday group accepted more directional imbalance and much noisier local outcomes. The final address concentrated risk in one market and waited.
So the most common strategy among these profitable addresses was not simply "high frequency." It was high-turnover, two-sided execution designed to earn a small edge many times while controlling inventory.
That is also why copying the latest position reveals so little. The position is one output of the system. The strategy lives in the cadence, balance, sizing, market selection, turnover, and risk recycling around it.
Follow Minara for more: > minara.ai > strategy.minara.ai