30 strategies2024-08-15 → 2026-09-29
Do crypto trading bots work?
Do crypto trading bots work? We backtested 30 distinct trading strategies, from trend following and momentum to mean reversion and funding carry, on real Hyperliquid data from 2024-08-15 to 2026-09-29, net of fees, slippage and funding. 12 cleared a Sharpe of 1.0, 17 cleared 0.8, and 23 of 30 made money, while 7 lost, one of them about 86%. Some work, most are mediocre and a few are traps, so test a bot's rules before you trust them.
By Keel Research Team · Data as of 2026-09-29 · how these backtests were produced
Cleared Sharpe 1.0
12 / 30
Made money
23 / 30
Lost money
7
Median Sharpe
0.95
All 30 strategies, ranked
30 distinct strategies: trend following, momentum, mean reversion, funding carry and more. Some are in our strategy library, others we tested only for this study. Ranked by net Sharpe; the library ones link to their full backtest. Strategies that effectively blew up show n/a for Sharpe, so judge those by return instead.
| Strategy | Net Sharpe | Return | Max DD | Trades |
|---|---|---|---|---|
| Momentum Funding | 1.90 | +124.2% | −16.6% | 7,355 |
| MACD Crossover | 1.87 | +408.8% | −33.3% | 2,746 |
| Confluence | 1.78 | +262.1% | −21.1% | 1,554 |
| Supertrend Bollinger | 1.73 | +385.1% | −28.5% | 460 |
| Stochastic | 1.69 | +128.2% | −14.4% | 74 |
| Funding Carry | 1.39 | +49.2% | −10.5% | 17,680 |
| Regime Gated Crypto | 1.20 | +196.8% | −42.5% | 394 |
| Momentum | 1.17 | +151.0% | −38.8% | 1,243 |
| AI Trading Bot | 1.15 | +213.4% | −50.2% | 1,141 |
| Long Hype Short Alts | 1.08 | +123.7% | −42.3% | 252 |
| MA Crossover | 1.05 | +130.8% | −30.4% | 54 |
| RSI | 1.02 | +33.7% | −9.6% | 32 |
| Supertrend | 0.99 | +105.0% | −31.6% | 679 |
| Regime Rotation | 0.98 | +102.4% | −30.4% | 91 |
| Breakout | 0.96 | +127.0% | −40.4% | 4,034 |
| Ichimoku Cloud | 0.94 | +115.4% | −45.5% | 825 |
| ADX Trend | 0.85 | +81.6% | −39.7% | 1,403 |
| Vol Targetingtested | 0.78 | +85.6% | −58.6% | 339 |
| Supertrend Crypto Btested | 0.58 | +30.7% | −34.6% | 379 |
| RSI Trend Pullbacktested | 0.48 | +19.4% | −18.9% | 429 |
| Rebalancing Indextested | 0.33 | -4.8% | −75.6% | 66 |
| Mean Reversion | 0.23 | +4.5% | −37.0% | 3,838 |
| Pairs Tradingtested | 0.22 | +3.9% | −33.6% | 383 |
| Gridtested | 0.16 | +2.3% | −34.5% | 173 |
| Bollinger Bandstested | 0.11 | -2.3% | −37.2% | 472 |
| Short Squeezetested | 0.09 | -21.4% | −62.0% | 7,122 |
| Ttm Squeezetested | -0.28 | -17.2% | −28.6% | 2,432 |
| Delta Neutraltested | -1.18 | -28.6% | −28.6% | 13,555 |
| Vwap Reversiontested | n/a | -86.0% | −89.1% | 8,636 |
| Mean Reversion Regimetested | -2.46 | -18.2% | −18.8% | 804 |
AI trading bots, tested the same way
An AI trading bot still trades a set of rules, whether a model wrote them or a person did. What decides whether it works is those rules, measured after fees, slippage and funding, which is exactly what this study does for every strategy in the table.
So the useful question about an AI bot is what rules it trades and how they did. Keel's AI builds a strategy from the same kind of blocks as the strategies above, and you can read and backtest what it builds before anything trades.
How profitable trading bots were after costs
23 of the 30 strategies ended with more money than they started with, after every fee, slippage charge and funding payment. Making money is a low bar, though. Only 12 earned a Sharpe of 1.0 or better, which is where a return starts to look worth the swings it took to get there.
Momentum Funding came out on top by Sharpe. At the other end, 7 lost money, and the worst lost about 86% of the account. A bot that trades often pays fees and slippage every time, so a rule needs a real edge just to break even.
What the results say about algo trading
The median strategy in the study had a Sharpe of 0.95. Split by market, 13 of the 30 made money on the days Bitcoin was in a downtrend, below its long-term moving average, so a falling market is not the end of a good rule.
Algo trading can be profitable, but only for rules that hold up after costs, over many trades, across more than one stretch of the market. That is a test you can run before a bot touches your money.
If you are comparing bot platforms, see the best crypto trading bots, or how an AI trading bot on Keel is built and tested.
Run one that held up
12 of 30 cleared the bar. Momentum Funding was the strongest. Start it on your own account, or fork it and make it yours first.
Free to start. Keel trades from your own Hyperliquid account through an agent wallet that can place trades but never withdraw. Every strategy draws down, so size accordingly.
What this is: a same-engine, same-window, same-cost backtest of 30 systematic crypto strategies on real Hyperliquid data. When a strategy blows up, a Sharpe ratio stops being meaningful, so we judge those on total return and mark the Sharpe n/a. A backtest is evidence, not a promise: every strategy here can lose, and past performance does not predict future results. This is not investment advice.