How the rule works
RSI measures how hard a price has been falling or rising over the last 14 candles, on a scale from 0 to 100. Below 30 is usually read as oversold. This rule doesn’t buy while the price is still falling. It waits until RSI climbs back above 30, which means the selling has eased, and then buys.
It sells when RSI crosses above 65, or earlier if the price falls 3% below what the bot paid.
How it did
Every test used a 1,000 USDT balance, 1-hour candles, 0.1% fees and 0.05% slippage, on Binance prices from 20 July to 30 October 2026. Prices rose over those 102 days.
| Coin | This rule | Holding | Trades won | Worst drop | Holding’s worst drop |
|---|---|---|---|---|---|
| BTC | +0.3% | +19.7% | 10 of 16 | −8.8% | −12.1% |
| ETH | +11.6% | +19.8% | 13 of 19 | −6.9% | −21.2% |
| SOL | −2.0% | +25.7% | 14 of 26 | −17.2% | −24.5% |
| BNB | +7.1% | +23.6% | 15 of 20 | −8.6% | −13.5% |
It won most of its trades and still trailed holding. The bot held a position about a third of the time. In a rising market, the other two thirds is where holding made its money.
It fell less. On every coin its worst drop was smaller than holding’s. That’s the trade this rule offers: less of the rise, less of the fall.
The stop loss mattered most on SOL. 12 of its 26 trades ended at the stop, which is how a rule that wins more than half its trades can still lose money.
When it tends to work, and when it doesn’t
It suits a market that moves sideways in a range, where dips are followed by recoveries. It struggles in two cases. In a strong rise there are few dips to buy, so it sits out. In a long fall each dip is followed by a lower one, and the stop loss gets hit again and again.
What to change first
- The RSI levels. Buying at 25 instead of 30 means fewer, deeper dips. Selling at 70 instead of 65 holds winners longer.
- The stop loss. A wider stop is hit less often and costs more each time it is.
- The coin. The table shows how differently one rule behaves across four coins.
Change one thing at a time and test again, or the result won’t tell you which change did it.
Common questions
Is an RSI bot profitable?
Sometimes. In this test it made money on three coins out of four and lost on one, and holding beat it every time. A profit in a test is a reason to keep looking, not a forecast.
What RSI settings should a trading bot use?
There’s no best setting. 14 candles with levels at 30 and 70 is the textbook default. Settings tuned until the past looks perfect often fail later, so prefer round numbers that work across several coins.
Can I run this without coding?
Yes. In Tradenos it’s a template: pick it, choose the coin and the amount, and test it before it trades.
What this doesn’t tell you
This is one period of 102 days in a market that went up. The same rule can do better or worse in a different market, and it says nothing about what happens next. The past doesn’t predict the future.
You can run this test yourself on the testing page, then change the rule, the coin and the stop loss and watch the result move.