Everyone knows the phrase "the trend is your friend." But there is a second, equally researched truth that works on a shorter clock: when price snaps too far, too fast, it tends to bounce back a little. Traders call it short-term reversal or mean reversion, and it has been tested in top journals for over 30 years.
Who found it, and where (the who / what / when)
The landmark paper is "Evidence of Predictable Behavior of Security Returns" by Narasimhan Jegadeesh, published in The Journal of Finance in 1990. A companion study by Bruce Lehmann appeared the same year. Jegadeesh is a famous finance professor (now at Emory) — the same researcher who later helped prove momentum works over longer windows. So the same scientist documented both: prices trend over months, but reverse over days and weeks.
What they actually did (the how they studied it)
Jegadeesh measured whether last week's or last month's biggest winners kept winning or gave it back. He sorted assets by their very recent return, then watched what happened next. The pattern was consistent: the recent big winners tended to underperform, and the recent big losers tended to bounce. Buying the beaten-down names and fading the stretched ones produced returns you could not explain by luck.
What they found (the data points)
At the one-month horizon, returns showed reliable negative autocorrelation — a technical way of saying "a strong move up is more likely to be followed by a pullback than by another equal move up." The contrarian strategy earned statistically significant profits before costs. Later research confirmed the same short-horizon snap-back shows up in futures and currencies too, though costs eat into it because it trades often.
The math — explained like you are 12
You need a way to say "this move is unusually big." Two simple tools do it.
1) The z-score. A z-score answers "how many normal steps away from average is price right now?"
z = (Price − Average price) ÷ (Typical wiggle)
If price is 2.5 normal steps below its recent average, z = −2.5 — that is very stretched, and reversal says a bounce is likely. A z near 0 means "nothing unusual, stand aside."
2) RSI(2). RSI is a 0–100 gauge of how one-sided recent moves have been. Set its lookback to just 2 bars and it becomes a fast "too far, too fast" detector: below 5–10 means violently oversold (a buy candidate), above 90–95 means violently overbought (a short candidate). It is the two-line version of the same idea as the z-score.
The rules (the what to actually do)
A clean, tested-style reversal setup looks like this:
- Trend filter first: only buy dips when price is above a long moving average (like the 200-period), so you are buying pullbacks in an uptrend — not catching a falling knife.
- Entry: buy when RSI(2) drops below 5 (or z-score below −2). Mirror it for shorts below the long average.
- Exit: take the snap-back — exit when price reverts to its average (RSI back above 50), plus a hard stop for the times the bounce never comes.
Why it should work (the why)
Short-term reversal is the market paying you to provide liquidity. When a fund is forced to dump a position fast, or a news pop triggers a herd, price overshoots what the information was really worth. Calm buyers who step in against that panic get a small, repeatable edge as price drifts back to fair value. It is the mirror image of momentum: momentum is slow information spreading; reversal is fast overreaction correcting.
Does it still hold — honestly?
With caveats you must respect. Reversal and momentum live on different clocks — reversal over days, momentum over months — so never blend the two horizons or they cancel out. It trades a lot, so costs and slippage can erase a paper edge; you must test after realistic fees. And the tail risk is real: sometimes an oversold market keeps crashing (a trend day), which is exactly why the stop and the trend filter are not optional. This is a strategy that needs honest testing more than most.
Build and test it in TapeScript, step by step
- Create it in plain English. Type: "Build a mean-reversion strategy on 5-minute MES: only long when price is above the 200-EMA, enter when RSI(2) is below 5, exit when RSI(2) crosses back above 50, hard stop at 1.5x the recent ATR. Mirror the rules for shorts below the 200-EMA."
- Classify and baseline. Type: "Classify this and run the baseline backtest with realistic costs." Reversal only counts if it survives fees.
- Fit the stop and target with real data. Type: "Run the MAE/MFE analysis and tell me where my stop and target really belong." This is gold for reversal, where a too-tight stop kills the bounce and a too-wide stop invites the falling knife.
- Find when it works. Type: "Run a session test — is the snap-back stronger in some sessions than others?" Reversal often concentrates in specific hours.
- Sweep the trigger. Type: "Sweep the RSI(2) entry level from 2 to 15 and show the parameter heatmap." A smooth plateau means a real edge; a lonely spike means overfitting.
- Prove it and export. Type: "Run walk-forward and Monte-Carlo, spend the holdout, then export the Pine and verify it matches the backtest." TapeScript has a faithful RSI-based Pine exporter, so what you test is what you can trade on TradingView.
The bottom line
Snap-backs are not a myth or a Reddit trick — they are a 30-year-old, journal-published edge that shows up on futures and forex when you measure it correctly. The catch is that a reversal strategy lies to you more easily than any other if you skip the costs, the stop test, and the out-of-sample check. TapeScript runs all three automatically and tells you the truth. Test a mean-reversion strategy the honest way →
Citation: Jegadeesh, N. (1990). "Evidence of Predictable Behavior of Security Returns." The Journal of Finance, 45(3), 881–898. See also Lehmann, B. (1990), Quarterly Journal of Economics. Many free versions are posted on Google Scholar.