Breakouts are the easiest strategy to believe in. You scroll a chart, find the spot where price cleared a long-standing high and ran, and the rule writes itself: buy the break.
This guide walks through backtesting one properly. The example is a real test we ran on Trade Manager on 20 August 2026, and it lost half its capital. That is the useful part.
What a breakout backtest actually tests
Looking at a chart and seeing breakouts that worked is not a test. Your eye finds the ones that ran because those are the ones that are visually obvious. The failures look like noise, so you skip past them without counting them.
A backtest is the counting. It answers three questions your eye cannot:
- How often does the rule trigger at all? Ten times a year and 81 trades across the index are very different strategies to live with.
- What happens after every trigger — including the ones you would never have screenshotted.
- Does the sequence survive? A strategy that is flat overall but drew down 59% along the way is not tradeable, whatever the final number says.
For more on the mechanics of setting a test up, see how to backtest a trading strategy without coding.
The strategy we tested
Named "S&P Breakout + Volume". S&P 500, daily bars.
| Trigger | Breakout, 50-bar lookback, price at least 0.5% above the level, evaluated on the close |
|---|---|
| Confirmation | Volume between 1.2x and 2x the 20-candle average |
| Take profit | +6.0% |
| Stop loss | -3.0% (a 1:2 reward-to-risk) |
Nothing exotic. This is close to the strategy most people describe when they say they trade breakouts.
Choosing the lookback
The lookback is how far back the strategy looks to find the high that price has to clear. We used 50 bars — on daily data, roughly a quarter of a trading year.
- 20 bars triggers far more often. Most of those highs are minor swing points inside a range, so you get a large sample of low-significance breaks.
- 100 bars triggers rarely, and each break means more. The cost is sample size: you can run a long test and end up with too few trades to conclude anything.
The 0.5% buffer matters as much as the number. Without it, a break of one cent counts, and you take a position every time price brushes the level.
Why the confirmation filter didn't rescue it
The volume band was there to insist that a real crowd showed up: at least 1.2x the 20-candle average, but under 2x, on the theory that an enormous spike is often the end of a move rather than the start.
It is a reasonable idea. It did not save the strategy. A filter can only remove trades — it cannot improve the ones that survive it unless the thing it measures actually predicts follow-through. Here it removed a lot of triggers and left the remaining ones failing at almost the same rate.
That is the common outcome. When a trigger has no edge, stacking filters on top of it mostly buys you a smaller sample of the same losing trade.
The results
| Win rate | 25.6% |
|---|---|
| Total profit | -50% |
| Max drawdown | -59% |
| Sharpe | -3.14 |
| Closed trades | 81 |
| Average hold | 5 bars |
| Coverage | All 503/503 tickers, up to 170 calendar days |
The break-even maths — the real lesson
Every strategy with a fixed take-profit and stop-loss has a win rate it must beat simply to end flat. The formula is short:
Break-even win rate = risk / (risk + reward)
| Reward-to-risk | Win rate needed to break even |
|---|---|
| 1:1 | 50% |
| 1:2 | 33.3% |
| 1:3 | 25% |
Ours risked 3% to make 6% — 1:2 — so it needed roughly 34% of trades to win. It got 25.6%. The gap between those two numbers is the whole result. Every trade was, on average, a small negative expectation, and 81 of them compounded into -50% and a 59% drawdown.
Read the rules again at the top of this page. Nothing in them tells you the win rate would be 25.6%. A 1:2 ratio even sounds prudent. The only way to find out was to replay the rules against real bars and count.
The caveats — and they cut the wrong way
Two things about this test flatter the result, not the opposite:
- Survivorship bias. It uses the index membership as it stands today. Companies that were dropped from the S&P 500 over the period are absent, and they are disproportionately the ones that fell hardest.
- Idealised fills. Exits assume you got exactly the take-profit or stop-loss price, with no slippage and no commissions. Breakouts are fast-moving by definition, which is precisely where real fills drift.
The honest reading: the real number is probably worse than -50%.
What to do with a result like this
Change one thing and re-run, rather than rebuilding the whole strategy. Widen the take-profit so the break-even bar drops. Lengthen the lookback so only significant levels count. Drop the volume band entirely to see whether it was helping at all. Try the same rules on a different market — for example, the NASDAQ 100 instead of the S&P 500. One variable at a time, or you will not know which change did what.
And when a version finally clears its break-even bar with a decent number of trades, activate it and let your strategy watch the market so you don't have to — it runs against your chosen market around the clock and messages you when the rules line up.
This is what backtesting is for
A 25.6% win rate on a 1:2 strategy is a bad strategy. Finding that out on history cost nothing but the minutes the test took to run. Finding it out live costs 50% — and by the time the drawdown told you, you would have been 59% down and convincing yourself to hold on.
Related reading: setting an alert with multiple conditions and best free stock screeners.