Here's a rule almost every breakout trader takes as gospel: when the market is in a correction, get to cash and wait. It's in every O'Neil summary. I believed it. Then I tested it against 20 years of my own system's trades — and the data said something I didn't expect.
The danger isn't the correction. It's a different market state entirely — one most people never even name.
The setup
I took every trade my locked system generated across a 20-year walk-forward backtest (2006–2026, US market) — 1,535 trades. For each one, I looked up what market "regime" the index was in on the day I entered, using the IBD market-condition state machine. Three possible states:
- Confirmed Uptrend — the market is trending up, distribution days low
- Uptrend Under Pressure — the market is wobbling, distribution days piling up
- Market in Correction — the market has rolled over
Then I asked the simple question nobody asks: what did trades entered in each regime actually return?
The result
Average R-multiple per trade, by the regime on entry day:
| Regime on entry | Avg result |
|---|---|
| Confirmed Uptrend | +0.69R |
| Uptrend Under Pressure | −0.00R |
| Market in Correction | +0.49R |
Read that again. Trades entered while the market was in a correction averaged +0.49R — solidly positive. The real dead zone was "Uptrend Under Pressure" — trades there made essentially nothing, just chop and whipsaw.
And this wasn't a fluke of one strategy. I run two independent sleeves. The pattern held in both: in each one, Under Pressure was the flat dead zone, and corrections were strongly positive.
Why corrections aren't the enemy
Once you see it, it makes sense. My entry filters demand a stock be a genuine leader — relative strength in the top 20% of the market, rising relative-strength line, above its own 200-day, in a Stage 2 advance.
A stock that is breaking out to new highs while the broad index is in a correction is showing extraordinary relative strength. It's swimming up a waterfall. Those are exactly the stocks that lead the next advance. O'Neil knew this — the leaders of every new cycle are forged in the prior correction. My data just put a number on it: +0.49R.
The "Under Pressure" state is different. The index isn't trending up and it isn't washed out — it's grinding sideways, indecisive, distribution days accumulating. Breakouts in that chop get faked out. That's where the money quietly bleeds.
What I'd actually do with this
I tested a sizing change: keep full size in Confirmed Uptrends and in Corrections, but cut position size to zero during "Under Pressure." The result:
- Total return: identical to trading every regime at full size (because the Under Pressure trades netted ~zero anyway — cutting them costs nothing)
- Maximum drawdown: down 39%
- Return-per-unit-of-drawdown: up 65%
You remove a pure source of pain — trades that add volatility without adding return — at no cost to the bottom line. That's about as close to a free lunch as backtesting ever offers.
The honest caveats
This is US data. My system trades Thai stocks too, and Thai market regimes behave differently — I won't apply this to SET until I've run the same test on Thai data (it's next on my list). It's also a sizing overlay, not a change to which stocks I buy. And "Under Pressure" is the smallest of the three buckets, so its exact average carries some sampling noise — but the decision (cut it, lose nothing, gain stability) is robust.
The bigger point
I publish things like this because the willingness to test your own gospel — and report when it's wrong — is the whole method. "Don't trade in corrections" sounds wise. For a system built on genuine market leaders, it turned out to be backwards. The next leaders are born in the correction. The place to be careful is the chop nobody names.
Track the pattern. Study what the leaders are actually doing. Wait for your setup — and know which market state actually hurts you. It isn't the one everyone warns you about.
Personal research. Not investment advice. Backtest on US data 2006–2026 with transaction-cost assumptions; results are a sizing study on a fixed trade set, not a forecast. Past results do not guarantee future performance.
[Get the free 36-year research brief →](https://moeasymmetry.netlify.app/the-36-year-test.html)