Every trading bot page shows you a number. Usually a green one, usually large, usually denominated in percent. That number is the least informative thing on the page. Whether you are choosing a platform, tuning your own bot, or reviewing a month of results, the questions that matter are answered by five quieter metrics — and none of them is the headline return.
Beyond the headline number
Headline returns are cherry-picked by construction: the best period, the best market, before fees, sometimes before the losing half of the portfolio. Useful evaluation starts by refusing that frame and asking quality questions instead: how often does it win, how much does it make when right versus lose when wrong, how bad does it get, and does it work outside the period it was sold on?
Win rate: necessary, not sufficient
Win rate is the share of closed trades that were profitable. It is the most quoted and most misleading metric in the industry, because it says nothing about magnitude. A bot can win 90% of trades and still lose money catastrophically if the 10% of losers are ten times the size of the winners.
- High-frequency, small-edge strategies (grids, market-making) typically run 60–80% win rates with tiny wins and occasional large losses.
- Trend strategies typically run 30–45% win rates with rare large wins that pay for many small losses.
- Rule: never interpret win rate without profit factor. Ever.
Profit factor: the quality ratio
Profit factor divides gross profits by gross losses. It is the single best one-number summary of a strategy’s quality:
| Profit factor | Reading |
|---|---|
| < 1.0 | Loses money. Stop. |
| 1.0 – 1.3 | Marginally alive — one bad regime from trouble |
| 1.3 – 1.8 | Healthy, realistic retail range |
| 1.8 – 2.5 | Strong — verify it survives fees and slippage |
| > 3.0 | Almost always overfit, mis-measured, or lucky period |
If a platform shows you returns but not profit factor, ask why. Computing it requires the full trade history — the same history you should demand to see anyway.
Max drawdown: the number that hurts
Max drawdown is the largest peak-to-trough decline in the equity curve — the worst continuous losing stretch, in money or percent. It matters more than any return figure for one reason: it is the number you will actually feel, and it is the number that determines whether you can keep the bot running.
- A 40% drawdown requires a 67% gain to recover; a 60% drawdown requires 150%. Recovery math is asymmetric and unforgiving.
- Duration matters as much as depth. A −15% drawdown that lasts four months breaks more users than a −25% one that lasts a week.
- Compare drawdown to the return over the same period. A strategy earning 30%/year with 35% max drawdown is a worse deal than one earning 22% with 12%.
Consistency across regimes
A bot tuned on one market condition is a snapshot, not a strategy. Before trusting any performance figure, segment it:
- By time: does the edge hold across at least one full bull, one full bear, and one boring chop period?
- By asset: is it one lucky pair like SOL, or does it work across the pairs you actually intend to run?
- By size: grids and market-making degrade with size. Double the capital and the same bot often earns less than half the rate.
Consistency is also the difference between backtest and reality — the topics in our backtesting guide apply directly here.
Fees and slippage: the silent majority
On a busy grid or scalping bot, fees can consume 30–60% of gross profit. Any evaluation that ignores them is fiction. Practical rules:
- Always recompute headline returns after the venue’s real fee tier, not the advertised one.
- Add a slippage assumption (0.05–0.2% depending on the pair’s liquidity) to every fill.
- Watch the fee-to-profit ratio over time: if it trends above ~50%, the strategy is working for the exchange.
A simple scorecard
| Metric | Green zone | Red flag |
|---|---|---|
| Profit factor | 1.3 – 2.0 | > 3.0 or < 1.2 |
| Max drawdown | < 25% of yearly return, multiple | > 50% at any point |
| Fee share of gross | < 40% | > 60% |
| Regime coverage | ≥ 2 complete market phases | One phase only |
| Win rate (context) | Read with profit factor | Quoted alone |
The bottom line
Returns tell you what happened; profit factor, drawdown, consistency and fee-load tell you whether it will happen again. Demand the full trade history, compute the five metrics yourself, and treat any platform that shows you only the green number as marketing, not measurement. On NexoBot every simulated trade is itemized in your demo dashboard — fees included — so you can run exactly this analysis.
Konten edukasi saja — bukan nasihat keuangan, pajak, atau hukum. Kinerja masa lalu, nyata atau historis, tidak menjamin hasil di masa depan.
Praktikkan ini
Uji strategi ini di akun demo — 10.000 USDT virtual pada data pasar live.