Confirmation bias in backtesting is the habit of studying the trades that prove your strategy works more closely than the ones that contradict it. It isn't a flaw in your tool or in the historical data. It's a well-documented mental shortcut, and it kicks in after the backtest is over, when you sit down to read your results.
What confirmation bias is
Psychologists have described it for decades: the tendency to look for, interpret and remember information in ways that support what you already believe (see the full definition on Wikipedia). In trading, it means that if you believe your strategy is good, your brain will keep finding reasons to go on believing it, even when the evidence is mixed.
It isn't unique to backtesting. It shows up anywhere people judge their own work. But backtesting gives it extra room to operate: you generated the data yourself, marking every trade by hand, so it's much easier to explain away an uncomfortable result than if someone else had produced it.
How it shows up when you review your backtest
You finish a session of 40 trades. As you go back over it, you linger on every winner: "the breakout was crystal clear here," "that entry was perfect." You skim past the losers with a generic excuse: "bad luck," "the market was acting strange." You walk away with a mental picture of your strategy that's far stronger than the numbers actually show.
Losing trades aren't the problem. Losses are part of any profitable strategy. The problem is that if you don't review your losers as carefully as your winners, you'll never spot the real pattern behind them. It might be a recurring execution mistake, a market condition your strategy doesn't handle well, or just normal statistical noise. Each of those three causes calls for a different fix.
Reviewing your whole journal, winners and losers, is the first step to knowing what actually needs fixing.
See plansHow it differs from hindsight bias
The two are easy to confuse because both distort your backtesting, but they strike at different points in the process. Our complete guide to backtesting already covers hindsight bias as the sneakiest mistake in the process. This article covers the other one, the one that comes afterward.
| Criterion | Hindsight bias | Confirmation bias |
|---|---|---|
| When it happens | During the replay, as you decide each trade | Afterward, as you review closed results |
| What it distorts | Your entry, stop and exit decisions | Your explanation of why it worked or failed |
| How to fight it | A tool that forces you forward one candle at a time | An audit routine that treats both groups the same |
| End result | Artificially inflated backtest results | Inflated confidence in a mediocre strategy |
A 4-step audit routine
You don't need special software. You need a fixed order that doesn't depend on your mood when you sit down to review. Always apply it to the full sample, never to a hand-picked selection:
- Split the sample into two columns: winners and losers. Don't mix them, and don't sort them by how much they made or lost.
- Review the losers first. Flipping the usual order weakens the anchoring effect, the warm glow that reviewing winners first leaves behind and that colors everything you look at next.
- Label the cause of every loss using one of three categories: execution mistake, a market condition your strategy didn't account for, or normal statistical noise.
- Count how many fall into each category. If most are execution mistakes, the problem is your discipline, not your strategy. If the same market conditions keep showing up, the problem is your strategy.
A journal that records the reason for every entry is what makes this routine possible. Without it, you're labeling causes from memory.
See plansA worked example
A 40-trade journal ends with 17 winners and 23 losers. Read with bias, the usual conclusion is: "the strategy needs minor tweaks, and the 17 winners show the idea is sound." Run the routine above on the 23 losers, though, and something else appears: 14 of them share the same cause. They were all opened in the first 15 minutes of the London session, when the spread (the gap between the buy and sell price) is still wide and unstable.
Confirmation bias had hidden that completely. It isn't a problem with the strategy as a whole. It's a problem with the time of day you're trading it. The real fix isn't "tweak the settings," which is what the biased reading would have led to. It's filtering out that specific time window.
Why this matters more than it seems
A backtest skewed by confirmation bias doesn't fail in the data. It fails in the conclusion, and the conclusion is the only thing you carry into a live account or a prop firm challenge. You can have the most complete journal in the world and still trade live on a wrong reading of your own strategy if your final review wasn't neutral.
Frequently asked questions
- Is confirmation bias the same as hindsight bias?
- No. Hindsight bias happens while you're backtesting, when your eyes jump ahead of the current candle. Confirmation bias happens afterward, when you review your closed results: you pay more attention to the trades that prove your strategy works than to the ones that contradict it.
- How do I know if confirmation bias is creeping into my backtesting?
- Here's a clear sign: you can explain in detail why each winning trade worked, but you only remember your losing trades vaguely ("bad luck," "the market was weird that day"). If that sounds familiar, you're reviewing with bias.
- Why do we review winning trades more than losing ones?
- Because winners confirm what we already want to believe: that the strategy works and that we executed it well. Reviewing a loser forces you to consider that the strategy is flawed or that your execution was poor, and most of us avoid that without even noticing.
- Does confirmation bias invalidate an entire backtest?
- It doesn't invalidate the data itself, since the trades are already recorded. It does invalidate the conclusions you draw from that data if you only reviewed the favorable half of the sample. The fix isn't to rerun the backtest. It's to audit your review.
- Does this bias affect automated backtesting too?
- Yes, just in a different way. It doesn't show up while the system runs, because that part is objective. It shows up when you interpret the results: for example, keeping the set of settings you like best instead of the one the overall statistics support most.
- Does Trade & Repeat help you avoid confirmation bias?
- It automatically records every trade, winners and losers alike, in the same journal, so the full sample is always there to audit. Avoiding the bias when you review it is still a habit that's up to you.
Keep reading: Manual vs. automated backtesting: differences, pros and cons · How to read your backtesting statistics · How to keep a backtesting journal without falling into hindsight bias · Cherry-picking without realizing it: why your backtesting date range isn't neutral
Educational content only; it is not financial advice. Trading financial markets involves the risk of losing capital.