Strategy Recipes
RSI Mean Reversion Strategy: Backtested on Crypto
How to express an RSI mean reversion strategy as no-code rules, plus the honest backtest result from Traseq's RSI demo template.
Traseq··7 min read
How to express an RSI mean reversion strategy as no-code rules, plus the honest backtest result from Traseq's RSI demo template.
Start with a no-code crypto spot strategy, lock the version, run the backtest, and keep the result traceable for comparison.
An RSI mean reversion strategy assumes that price stretches too far in one direction and then snaps back. When the 14-bar Relative Strength Index drops into oversold territory, you buy; when it climbs into overbought territory, you sell. This guide turns that idea into a small set of no-code rules, then shows what it actually did on real BTC/USDT history — including the parts that are not flattering.
Traseq is a research workspace, not a live trading or exchange-execution platform. It does not place orders, connect to exchange accounts, or guarantee performance. Everything below is research on historical data, not a recommendation to trade.
The Relative Strength Index (RSI) measures the speed and size of recent price moves on a 0–100 scale. The default lookback is 14 bars. Two thresholds do the work:
A mean reversion strategy bets on the snap-back: buy weakness, sell strength. This is the opposite stance to a trend-following strategy, which buys strength and sells weakness. Neither is universally right; each suits a different kind of market. We cover that tension in trend following vs mean reversion.
Here is the exact RSI template the Traseq demo uses, expressed as plain rules you can build with Sentence mode, Canvas mode, or a template — no Pine Script or Python required:
A few things worth noting about how this evaluates:
For a primer on building rules without code, see no-code backtesting.
Traseq's Learn hub ships a no-signup interactive demo that runs three system templates on real BTC/USDT 1H candles from 2024-11-03 to 2024-12-31 ($10,000 starting balance, zero fees, 100% position sizing). That window was a sideways-to-down chop after a rally — deliberately unflattering. Here is what the RSI template produced over the full sample:
Read this carefully before getting excited. Of the three demo templates (an SMA(200) trend filter and a Donchian breakout were the other two), this was the only one that did not net-lose in that window. But "did not lose" is a low bar. A profit factor of 1.12 means gross profit was only 12% larger than gross loss. A Sharpe of 0.05 is statistically indistinguishable from zero. This is a thin edge, not a green light.
The core risk of mean reversion is structural, not a tuning detail. The strategy assumes price will revert. In a strong, sustained trend, it does not — it keeps going.
Picture a real downtrend. RSI drops below 30, the rule buys, and price keeps falling. RSI can stay oversold for a long time while a knife keeps dropping. Your 3% stop fires, you take the loss, RSI is still below 30, and the rule buys again into the same decline. Mean reversion fights the trend, so a trending market can hand it a string of losses precisely when it feels most "oversold."
This is why the choppy, range-bound demo window is the kinder environment for this strategy, and it still barely cleared breakeven. Drop the same rules into a clean downtrend and the stop-loss math gets ugly fast. It is also why a headline win rate tells you so little on its own: a 44.4% win rate with the wrong loss distribution would have sunk this, a point we unpack in high win rate can still lose money.
Even setting aside the trend risk, the bigger problem is the sample. Nine trades is not evidence. It is an anecdote with a spreadsheet attached.
With nine trades, a single lucky exit can flip the profit factor from below 1.0 to above it. Flip one winner to a loser and the +1.74% return likely turns negative. There is no way to tell skill from noise at that count. Statisticians would not draw a conclusion from nine data points, and neither should a trader.
The fix is not to optimize the thresholds until the nine trades look better — that is how you overfit to luck. The fix is more data: test the same rules across a longer date range, across different market regimes, and ideally on a separate slice you did not look at while building. That out-of-sample discipline is the whole point of in-sample vs out-of-sample testing. A rule that survives a 2-year window across trends, ranges, and crashes is worth far more attention than one that "won" over two choppy months.
You can run the RSI template — and the two that net-lost — yourself, with no signup, in the interactive demo on the backtesting basics page. Drag the position-sizing knob and watch return and drawdown scale together. Then build your own version with different thresholds and compare.
When you are ready to author your own rules from scratch, no-code backtesting walks through the workflow, and the Core Concepts guide covers how Traseq evaluates signals and keeps each result tied to a strategy version.
It is a strategy that buys when the Relative Strength Index signals oversold (typically below 30) and sells when it signals overbought (typically above 70), betting that price will revert toward its mean. It is the opposite of trend following, which buys strength rather than weakness.
The demo enters long when the 14-bar RSI crosses below 30 and exits when RSI rises above 70, with a 3% stop loss and a 5% take-profit. These are conventional defaults and starting points, not tuned or optimized values.
It returned +1.74% on real BTC/USDT 1H data from 2024-11-03 to 2024-12-31, with a 44.4% win rate, -9.0% max drawdown, and a profit factor of 1.12 over 9 trades. It was the only one of the three demo templates that did not net-lose, but a profit factor barely above 1.0 on nine trades is a thin edge, not proof the strategy works.
Mean reversion assumes price will snap back, but a strong trend keeps moving in one direction. Buying every oversold reading in a downtrend means repeatedly catching a falling knife: the stop loss fires, RSI stays oversold, and the rule buys into the same decline again.
No. Nine trades is far too small a sample to separate skill from luck — a single different outcome can flip the result from profit to loss. To trust a rule set, test it across a longer date range and multiple market regimes, and validate it on data you did not look at while building.
| Component | Rule |
|---|
| Entry | Enter long when 14-bar RSI crosses below 30 |
| Exit (signal) | Exit when 14-bar RSI rises above 70 |
| Stop loss | 3% below entry |
| Take-profit | 5% above entry |
| Metric | RSI Mean Reversion |
|---|
| Return | +1.74% |
| Win rate | 44.4% |
| Max drawdown | -9.0% |
| Trades | 9 |
| Profit factor | 1.12 |
| Sharpe | 0.05 |
May 29, 2026