What it is
The 3-day pullback is a mechanical short-term pullback rule: in a stock or index above its 50-day and 200-day averages, after three consecutive down closes (or three lower lows), buy the next day's open or the first break of the prior day's high, and exit within a few days. It is a compact way to trade the same idea as ema-pullback-trend with less discretion and a shorter hold, and it makes a good first system for someone learning to test rules honestly.
The logic
Three down days in an uptrend are usually short-term sellers (profit takers, day traders, stop-outs) rather than a change in the fundamental picture. The sellers are exhausted after a short run, dip buyers who have been waiting for a discount step in, and the trend's longer-horizon buyers were never selling in the first place. The setup profits from the imbalance between the exhausted short-term sellers and the patient buyers.
On the other side are traders selling because three down days "look like" a reversal, plus momentum shorts. In an uptrend they are more often wrong than right over the next 3 to 5 days; in a downtrend the same pattern is a continuation, which is why the trend filter matters more than the pattern.
Setup rules
- Market: liquid stocks and ETFs; index futures on daily bars.
- Timeframe: daily.
- Trend conditions: close above the 50-day and 200-day simple moving averages; the 50-day above the 200-day; the 20-day EMA rising.
- Pullback conditions: three consecutive lower closes, or three consecutive lower lows; the third day closes within 2 percent of the 20-day EMA or below it; the pullback has retraced less than 62 percent of the last up leg.
- Disqualifiers: the pullback includes a gap down of more than 3 percent; earnings within the next 7 days; the index has had three down days too and is below its 20-day EMA (market-wide selling is a different regime).
Entry, stop, target
Two entry styles: the aggressive version buys the next open; the confirmed version buys a break of the third day's high (which skips some winners but avoids fourth and fifth down days). Stop is 1 ATR below the third day's low. Exit on the first close above the entry day's high, or at the close of the fifth day after entry, whichever comes first.
| Item | Level | Notes |
|---|---|---|
| Last swing high | 55.00 | |
| Day 3 low | 51.20 | Close 51.50, at 20-day EMA |
| Entry (confirmed) | 52.10 | Break of day 3 high |
| Stop | 49.90 | 1 ATR (1.30) below day 3 low, risk 2.20 |
| Typical exit | 54.00 | Close above entry day's high on day 2 or 3, reward 1.90, 0.9R |
| Time exit | Day 5 close | Whatever it is |
The average winner is below 1R and the win rate is where the expectancy comes from; historically a rule like this shows win rates in the 55 to 65 percent range in strong uptrends and far worse elsewhere. That is a description of past backtests, not a forecast.
Position sizing and risk
Short holds with wide-ish stops mean small positions. Size at /tools/position-size with 0.5 percent risk. Because signals come in clusters (many stocks pull back on the same three days), treat the cluster as one correlated trade and cap the total in /learn/risk-management terms; five 3-day-pullback longs in the same week are one bet on the index bouncing.
What breaks it
- Fourth and fifth down days. The pattern's failure is a continued slide, and in a correction that is what happens. The confirmed entry style avoids some of this; the stop handles the rest.
- Regime. In a bear market the setup loses money reliably. The index filter is non-negotiable.
- Costs. The average win is small in percentage terms (2 to 4 percent), so commissions and spread are a bigger fraction of profit than in longer trades. Do not trade this in illiquid names.
- Edge decay. Short-term pullback effects in liquid US stocks have weakened since the 2000s. The rule still shows a positive expectancy in most tests, but with thinner margins and more dependence on the regime filter.
- Boredom and rule-breaking. The trader adds discretionary tweaks, holds losers past day 5 or skips signals after a losing streak. Rule-based setups fail through the rule-breaker.
How to test it
This is the ideal first system to code. Run it across at least 500 liquid stocks and 20 years, survivorship-free, with 0.1 percent per side costs. Report win rate, average win and loss in R, expectancy, profit-factor and the distribution of consecutive losses. Test both entry styles and both exit styles. Then test it in a bear-market subsample; if you still see a positive expectancy there, look for a bug. Minimum 1,000 trades, which the rules will produce. Trade it on paper for three months to verify you follow the rules, then live with minimum size for three more.
Variations
- RSI(2) or connors-style entries that replace the three down days with an oscillator reading; see range-mean-reversion-20ma.
- Weekly version: three down weeks into a rising 30-week average, held for 2 to 4 weeks.
- Index-only version traded on a single broad ETF, which reduces selection effects and makes the sample cleaner.
Further reading
pullback, trend, moving-average, atr, win-rate, expectancy, profit-factor, correlation, backtesting, buy-the-dip.
Related playbooks: ema-pullback-trend, range-mean-reversion-20ma, abcd-measured-move, weekly-inside-bar-breakout