A bot for your goal
No universal best bot. Each one is built for a specific market situation and your goal. This guide shows which to pick if you're accumulating a position, trading sideways, combining approaches, or hedging your main portfolio.
Choose by goal
You want to accumulate a position on dips
You see a long-term uptrend, but want to enter gradually, buying on dips. DCA-bot buys on a grid when price falls, lowers your average entry cost, closes on bounce. Keeps leverage controlled (max 2×), accounts for funding rates.
DCA-bot →Sideways — price oscillates in a range
Market is choppy, no real trend. Grid-bot works like an exchanger: buys in range lows, sells in highs, catching every bounce. Many small profits on volatility. No leverage or minimal — this is sideways mechanics. Yield depends on bounce frequency and range width.
Grid-bot →Mixed approach — part of position on trend, part on sideways
You want to catch trends AND earn on volatility simultaneously. Combo-bot runs DCA and Grid at the same time on one pair or in parallel on different ones. DCA takes long-term direction, Grid feeds on short swings. Harder to tune, but gives more diversification within one position.
Combo-bot →Insurance for your main portfolio
You already have a main position on exchange or off-chain. You want to insure it against drops without closing. Hedge-DCA and Hedge-Combo run in opposite direction on Hyperliquid perps: if your portfolio falls, hedge profit offsets the loss. Mechanics are like regular DCA/Combo, but logic is flipped.
Hedge-bots →How to compare strategies
Sharpe ratio
Income relative to volatility. High Sharpe (> 1.5) — stable equity curve, doesn't jump around. Low (< 0.5) — profit is unstable, sharp drawdowns. Main metric for comparison.
Max drawdown
Capital drop from peak to trough. If peak is $10k, trough is $8k — drawdown is 20%. Lower drawdown at same profit is better. Shows your worst-case loss.
Profit factor
Ratio of all gains to all losses. If you earned 1000, lost 500 — profit factor 2.0. Minimum for viable strategy — 1.2. It's not magic: even 1.1 can be good over many trades.
Win rate
Share of winning trades out of total. 60% wins is good, 70% is very good. BUT it's a trap: 80% wins with losses 5× bigger than wins = negative expectancy. Look together with profit factor.
Trade count
How many markets and timeframes the strategy trades. One trade a month — untestable. 50+ trades in backtest — that's statistics. In paper mode, trade count shows how many real signals the strategy gave on live market.
Honesty about limits
• Past performance is not a guarantee of future results. A strategy that made 30% in backtest on 2023–2024 history may return −10% on a different market or period. Backtest shows ideal execution, no slippage or delays. Reality is worse.
• Curve fitting (overfitting). Parameters selected on historical data often don't work forward. That's why AI Traders requires paper trading before live — it's a signal: does actual trading match the backtest? If not, rethink your parameters.
• Funding rates and fees eat into profits. On Hyperliquid, fee is 0.02% on longs, −0.01% (sometimes longs earn) on shorts. Funding varies. All included in backtest, but if funding spikes (market panic), profit drops.
• Liquidity can run out. Trading a rare altcoin, your big order slips 1–2%. Hyperliquid is deep, but depth isn't forever.