Breakout Trading Strategy: Same Rules, 480 Backtests on Nasdaq, S&P 500 and Dow

TL;DR: the answer box

• 441 of 480 breakout variants made money. 1 of 480 cleared its random control's 95th percentile. Luck alone would clear about 24.
• The typical breakout scored below random. The average score was 40.5, where 50 means "no better than random". Only 156 of 480 had an average trade above the random average.
• From 2007 to 2026, the Nasdaq was not a better breakout market against random entries. It was a stronger market for buy and hold and for random 5-bar entries alike. A random 5-bar QQQ entry earned 1.71x a random SPY entry and 2.00x a random DIA entry.
• Buy and hold won almost everywhere. QQQ buy and hold made $500,263.82. The best QQQ breakout made $111,444.09. 1 of 480 variants out-earned holding its own market.
• The "best rule per index" moves around and means little. 0 of 20 classic rule and market pairs beat the luck bar at a 5-bar hold.

How we tested

Most traders assume a breakout that works on one index works on the others. The Nasdaq, the S&P 500 and the Dow move together, after all. Some go further: the Nasdaq "trends harder", so breakouts belong there.

I tested both beliefs the boring way. Same rules. Same days. Five markets. And a luck test on every single variant.

The markets. SPY for the S&P 500, QQQ for the Nasdaq 100 and DIA for the Dow. Plus the futures I trade: ES (E-mini S&P 500, $50 a point) and NQ (E-mini Nasdaq 100, $20 a point), day session, one contract. The data holds no day-session Dow futures file, so the Dow is DIA only.

The window. Entries from 2007-07-02 to 2026-05-14 on all five. That is 4,748 possible entry days on SPY.

The rules. Long only, 16 entries, all simple. They come from five price lines. ATR, the average true range, is the average daily range of the last 10 days, gaps included.

The four "classic" rules in this test are: close above the prior close, high above the prior high, and the two 0.25 ATR stops. Every entry is held 1, 2, 3, 5, 10 or 20 bars, then sold at that bar's open. No stop loss, no target, no filter. That is 96 variants per market, 480 in all.

The execution. Execution rule v2. The signal is read at the close. The trade buys the next bar's open, or fills at its stop inside the next bar. If the market opens above the stop, it fills at the open. A 5-bar hold sells at the open five bars after entry. One position at a time. A $35,000 account, whole shares on the ETFs, no compounding, no costs.

Here is one real trade, start to finish.

The luck test. This is the part most backtests skip. For every variant I took its trade count, picked that many random entry days in the same window, bought the open, held the same number of bars and sized it the same way. I did that 100 times with fixed seeds.

The score is how many of those 100 random runs the breakout's average trade beat. A score of 50 means no better than random. A score of 95 or more clears the luck bar.

Test 480 things and luck alone hands you about 24 winners. So you should count the winners, not admire one.

Is breakout trading a good strategy?

Most of the 480 variants made money. Only 1 cleared the luck bar. Chance alone would clear about 24.

That is the whole story in one picture. From 2007 to 2026, a mostly rising window, 441 of the 480 breakout variants I tested made money. Random entries made money too.

If breakout timing added nothing, the scores would spread evenly from 0 to 100. About 48 variants would land in each bin of ten. They do not. The bars lean left, toward "worse than random".

The typical breakout scored below a plain random entry.

The one winner stood alone

The one variant that cleared the luck bar was the NQ buy stop at the prior high + 0.25 ATR, held 1 bar. It scored 95.0 over 1,273 trades.

Now look at the same rule at every other hold.

Four of its five other holds sit below random. Ask yourself: would a real edge vanish when you hold the trade one more day? One hit in 480, when luck predicts about 24, with weak neighbours: that is one variant clearing the bar, not a pattern across holds.

When I rebuilt my trading on the 2014 Turtle Traders lesson, it was simple systems plus proper risk management, not a better trigger. These 16 entries showed no broad timing edge over random entry.

Does the same breakout work differently on the Nasdaq, S&P 500 and Dow?

In dollars, yes. In edge, no.

The dollar gap is real. At a 5-bar hold the four classic rules made $63,161.57 to $80,472.94 on QQQ. On SPY they made $41,223.47 to $49,277.40, and on DIA $26,133.39 to $59,219.59. Stop there and the Nasdaq "fits breakouts best".

Now do what you should do with any backtest: score every rule on every market against its own random runs. This is the pivot chart. Blue beats random. Orange loses to it.

0 of 80 cells reach the luck bar at a 5-bar hold. The best four sit on DIA, topped by 87.0 for the prior high + 0.25 ATR stop.

Widen it to every hold and rank the markets by their typical score.

By this measure the Nasdaq is not the best breakout market. The Dow, through DIA, sits highest at 45.0. And it is still below random.

The "best rule" depends on where you look

Rank the four classic rules by Ret/DD, short for return to drawdown: net profit divided by the worst dollar drop. Then watch the order change from market to market.

Part of the popular story survives. On NQ futures the plain close breakout ranks first, with a Ret/DD of 2.35. On ES the close + 0.25 ATR stop ranks first at 4.15.

Part of it does not. On QQQ, the ETF for the same index as NQ, the plain close breakout ranks 4 of 4. Same index, opposite rank. A ranking that flips between two ways of trading the same index is not something to build on.

If you build breakout systems, this is the check that matters. A tool like AlgoChef exists to put a strategy's results next to what chance and the market would have handed it anyway, before you trade it.

Why do breakouts look better on the Nasdaq?

Random entries did better there too.

Think of a moving walkway at the airport. Everyone on it arrives faster, whether they walk, stroll or stand still. Measure only the walkers and you will credit their stride for the walkway's work.

Your breakout on the Nasdaq rode the faster walkway. Over this window QQQ buy and hold made $500,263.82. That is 3.67x SPY's $136,493.28 and 5.08x DIA's $98,514.00.

So random entries on QQQ earned more as well. A random 5-bar QQQ entry earned 1.71x a random SPY entry and 2.00x a random DIA entry, on the same $35,000 sizing. Here is what the breakouts earned against that walkway.

On four of the five markets the median breakout earned less per trade than a random entry. Only DIA tipped the other way: $62.18 against $55.91.

I am not claiming why the Nasdaq rose more. Tech weight, higher volatility, bigger swings: those are stories this data cannot test. What it shows is simpler. The median QQQ and NQ breakouts earned less than their random baselines.

Do breakouts beat buy and hold?

Almost never. Buy and hold out-earned 479 of 480 variants on net profit.

The one exception was an ES variant at $283,956.25 against $269,250.00 for holding.

A breakout sits in cash much of the time, so that comparison is a little unfair. Give it a fair shot: scale its profit up as if it were in the market every day. The close breakout on SPY was in a trade 41.56% of the time. Scaled up, it would have made $118,569.30, which is 86.9% of buy and hold. On QQQ, 0 of 96 variants beat holding even after that boost.

Did the time out of the market at least buy a smaller drawdown, the biggest drop from a high point?

Not reliably. On SPY the four classic rules fell 43.32% to 57.71% at their worst, against 56.24% for buy and hold. On QQQ the range is 24.12% to 67.70% against 54.09%. Some rules fell less than holding. The plain close breakout on QQQ fell more. Less time in the market did not mean less pain for you.

Does holding longer, or a stop entry, fix it?

Neither.

Holding longer. At a 1-bar hold 1 of 80 variants cleared the luck bar. At 2, 3, 5, 10 and 20 bars, 0 of 80 did.

Does it hold over time? I split the window at the start of 2017 and scored each half against its own random runs. 1 of 20 classic rule and market pairs sat above random in both halves. The prior high + 0.25 ATR stop, the best classic rule on SPY, QQQ and DIA, scored lower against random in the second half on all five markets.

A stop entry. A buy stop above the prior close sounds like it waits for proof. In practice the market often jumps over it at the open, and the order fills right there.

On SPY, 48.0% of the close + 0.25 ATR fills came at a gap open. The fill was the open: the same price any next-open rule would have paid. Stops changed which rule won the ranking. They did not change the verdict.

The verdict, and the honest limits

441 of the 480 simple daily breakout variants made money from 2007 to 2026. Only 156 had an average trade above their random control's. 1 of 480 beat the luck bar.

Where the popular view is right: the same rules do give different results on the Nasdaq, the S&P 500 and the Dow. The best rule changed from market to market, and the Nasdaq paid more dollars.

Where it is wrong: this test found no broad breakout timing edge behind those dollars. Random QQQ entries earned 1.71x random SPY entries, so the Nasdaq's extra breakout dollars came with a faster walkway, not a better walker. And buy and hold out-earned 479 of 480.

The limits, in plain words:

• No costs. No commission or slippage. The busiest variant trades 2,144 times, so costs would hit the short holds hardest. Their effect was not measured.
• Long only, daily bars, time exits only. No stop loss, no profit target and no trend filter were tested. A breakout with a filter or a trailing exit is a different system.
• One window, mostly rising. 2007 to 2026 rewards being long. No flat or falling decade was tested.
• The Dow is DIA only. There is no day-session Dow futures file, so there is no YM view.
• Futures are leveraged. One ES contract on $35,000 is heavy leverage: buy and hold fell 123.53% of the account at its worst, which means wiped out. For ES and NQ I show the worst drop as a share of the account, but profits only in dollars, never as a return percentage.
• The random control matches trade count and hold, not exact days. Random entries that land on an open position are skipped, so the realised random count can run lower.

What this means for you

1. Test every breakout against random entry, not against zero. 441 of 480 variants made money here, and 1 of 480 cleared its luck bar, where chance alone would clear about 24. "It's profitable" did not show timing skill.
2. Don't pick your market by its breakout backtest. QQQ had bigger breakout dollars, and random QQQ entries earned 1.71x SPY's. Compare each market to its own walkway.
3. Don't trust a "best rule per index". The close breakout ranked first on NQ and last on QQQ, the same index.
4. In this window, holding beat almost everything. Buy and hold out-earned 479 of the 480 variants on net profit.
5. Treat exits, filters and sizing as untested here. These 16 entries with time exits showed no broad timing edge. An edge from exits, filters or sizing still has to be proven the same way.

Reference tables

For serious readers, the full numbers behind the charts.

The four classic rules at a 5-bar hold. Score is the share of 100 random-entry runs whose average trade the rule beat. Ret/DD is net profit divided by the largest dollar drawdown. Futures are one contract.

Every variant against buy and hold of its own market. "Scaled to full time" divides net profit by the share of days in the market, the most generous way to read a part-time strategy.

Entries are the cheap part. Proof is the expensive part.

This is one study in an ongoing series. If you want the next one, the same kind of large-scale backtest with a luck test on every variant, join The Overfit newsletter at StatOasis.com/Overfit.

Further reading: the Donchian channel definition, and the average true range the stop entries are built on. Brock, Lakonishok and LeBaron ran a classic test of trading-range breakouts on the Dow. Bailey, Borwein, López de Prado and Zhu explain why testing many variants makes a lucky winner the expected outcome, in "Pseudo-Mathematics and Financial Charlatanism".

Methodology: 480 variants (16 long-only entries x 6 holds x 5 markets), daily bars, entries 2007-07-02 to 2026-05-14, Execution rule v2 (signal at the close, next-open or stop fill, time exits sell at the open), flat-only, $35,000 account, no compounding, frictionless. Random control: 100 seeded runs per variant. Educational research, not investment advice. By Ali Casey, StatOasis.