What it is
The trend pullback to the 20 EMA is the standard "buy the dip in an uptrend" rule. A 20-period exponential moving average tracks the short-term trend; a 50-period EMA defines the longer one. When the fast average is above the slow one and price pulls back to touch the fast average and holds, the trade goes with the trend, with the stop under the pullback low. The idea is sound: in a trend, pullbacks to the average are where late buyers get filled. The question is whether a mechanical version of it has any edge on crypto, and the table says: barely, and only on some timeframes.
Rules
- Market / timeframe: BTC or ETH, 15-minute, 1-hour or 4-hour candles.
- Trend filter: EMA20 above EMA50 and close above EMA50 for longs; the mirror for shorts.
- Entry: a candle whose low touches or crosses the EMA20 but which closes back above it, and closes green (for a long). Enter at the close.
- Stop: the lowest low of the last three candles including the entry candle. Skip if the stop is closer than 0.2% of price.
- Target / exit: 2R, or 24 bars on 1h (18 on 4h, 48 on 15m).
- Position size: 1% of equity per trade in the table.
Chart example
Evidence
Our own backtest, Coinbase BTC-USD and ETH-USD candles, 2025-10-19 to 2026-10-19 (one year), no fees, no slippage, one position at a time. Buy-and-hold over the same year: BTC −23%, ETH −38%.
| Market / TF | Trades | Win rate | Avg R | Profit factor | Net R | Max DD (R) | Return @1% risk | Max DD @1% risk |
|---|---|---|---|---|---|---|---|---|
| BTC 1h | 397 | 37.5% | 0.01R | 1.02 | 5.6R | 22.9R | 2.2% | 21.9% |
| BTC 4h | 107 | 38.3% | 0.08R | 1.14 | 8.2R | 11.9R | 7.5% | 11.3% |
| BTC 15m | 1,181 | 35.2% | 0.02R | 1.02 | 17.8R | 53.4R | 6.8% | 43.2% |
| ETH 1h | 386 | 35.2% | −0.03R | 0.94 | −13.3R | 28.5R | −15.3% | 26.6% |
| ETH 4h | 104 | 35.6% | −0.02R | 0.97 | −1.8R | 11.3R | −2.7% | 11.3% |
| ETH 15m | 1,280 | 34.2% | −0.01R | 0.98 | −14.3R | 40.6R | −23.0% | 39.2% |
How to read it: every row sits within a few hundredths of an R of zero. That is what "no edge before costs" looks like. The one slightly positive row (BTC 4h) is 107 trades, too few to separate from luck. The 15-minute version fires over a thousand times a year and would be a reliable loser once fees are added.
When it fails
- It is not a strategy, it is an entry technique. The pullback tells you where to enter if you already have a reason to expect the trend to continue. On its own, the touch of an average has no predictive power; the table shows it.
- Ranges. A 20/50 EMA cross happens constantly in a sideways market, so the "trend" filter passes half the time while price is going nowhere. Add a higher-timeframe trend requirement or a minimum EMA slope.
- Stop placement. The three-candle low is often just under the EMA, where everyone else's stop is. Expect to be swept.
- Overtrading. 1,200 signals a year on 15m is a signal per 7 hours. Most of them are noise.
If you use pullbacks, use them to improve the entry on a trade you would take anyway. RektChat's paper desk lets you test the difference between "EMA touch" and "EMA touch plus a reason" without paying for the education.