Overconfidence and Recency Bias: Why Your Best Days Predict Your Worst Ones

The most dangerous moment in a trading career is right after a winning streak. You feel sharp. You feel like you've "figured it out." You size up. The market, indifferent to your recent feelings, then delivers a normal losing trade, which now hurts much more because you sized for the version of you that thought they were invincible. This dynamic is overconfidence fueled by recency bias, and it's how the best weeks of trading set up the worst ones.

Recency bias, the cognitive driver

Recency bias is the tendency to weight recent events more heavily than older ones in your mental model of reality. A strategy that worked the last 5 trades feels more robust than the same strategy's 100-trade backtest suggests. A losing day feels like the strategy is broken; a winning day feels like the strategy is "really clicking."

In trading specifically:

• Recent wins → strategy must be working really well → size up
• Recent losses → strategy is failing → reduce size or change
• Recent volatility → next move will be calm (or vice versa, depending on your prior)

The bias is asymmetric in damage. Sizing up on a hot streak exposes you to larger losses when the streak inevitably ends. Sizing down after a losing streak underweights you when the strategy reverts to working. Both errors compound across cycles.

How overconfidence develops from a streak

The mechanism:

1. Three or four good trades in a row. Statistically normal, most strategies have streaks like this multiple times per month.
2. Pattern recognition (false positive). Your brain registers "I'm in a hot streak" rather than "this is normal variation."
3. Attribution to skill. The wins feel like proof of skill rather than luck. Confirmation bias amplifies this , you remember the specific reasons you took each trade and they all "worked."
4. Confidence inflation. You feel sharper than usual. You start trusting your reads more. You take entries that would have been "watch only" yesterday.
5. The size-up. Often subtle: 1.2R instead of 1R, then 1.5R, then 2R. Sometimes explicit: "I'm clearly in a groove, let me run with it."
6. The reversion. A normal losing trade arrives. With inflated size, the loss is 2-3x what it would have been in your normal regime.
7. The cascade. The loss feels disproportionate (because it is, relative to your normal). You're thrown off. The next few trades are made from emotional disequilibrium, often producing further losses.

The end state is often that the streak's gains are entirely given back over the next few sessions, sometimes more. The "hot hand" became a cold one without warning.

Why your win rate doesn't justify the size up

Suppose your strategy has a 50% win rate with average winner +1.5R and average loser -1R. Expectancy is 0.25R per trade.

Streak math: at 50% win rate, you'll see at least one 4-trade winning streak roughly every 32 trades. That doesn't mean your strategy got better, it means you're seeing a normal random sequence.

If you size up after each winning trade, the inevitable losing trade hits you at the inflated size. Across many streaks, your average outcome gets worse, not better, because the increased size during streaks pairs with reverted size right when you take the proportionally larger loss.

Mathematically, the optimal sizing strategy is constant risk per trade, not adjusted for recent results. (Variants like Kelly sizing exist but don't increase or decrease based on recent streaks, they're set by long-run win rate and edge, which doesn't change between trades.)

The mirror: under-confidence after losses

The same recency bias works in reverse. After a losing streak, you feel uncertain about every setup. You skip valid trades because "the strategy isn't working right now." You size down on the trades you do take. Then the strategy reverts to its normal expectancy and the wins you do capture are at smaller size than usual.

This is also negative-EV across cycles. You participate in the losing streaks at full size and the winning streaks at reduced size. Net: you systematically underperform your strategy's true expectancy.

The asymmetry is structural: the streak ends regardless of your sizing choice, but your altered sizing during the streak captures only the losing portion at full weight.

A common mistake: sizing by "feel"

A trader doesn't formally size up after wins. But they "feel" comfortable taking a slightly bigger position because "things are working." They don't track this, it's just intuition. The cumulative drift is invisible to them, but real in their PnL.

The defense: pre-define sizing rules in writing and don't deviate. "1% risk per trade, regardless of recent performance." When you find yourself wanting to size larger on a "high-confidence" setup that happens to come after wins, recognize the pattern. Confidence isn't generally a function of the setup, it's a function of recent results that your brain is misattributing to the setup.

A common mistake: confusing learning with overconfidence

Genuine learning happens in trading, over months and hundreds of trades. You really do get better. The question is how you can tell genuine improvement from a temporary streak.

Genuine improvement signals:

• Your average R-multiple has improved measurably across rolling 50-trade windows
• Your max drawdown has shrunk as a proportion of expected return
• You can articulate what you do differently now vs 6 months ago
• The improvement is consistent across multiple market conditions, not just trending markets

Overconfidence signals:

• Recent results dramatic but small sample size (less than 20-30 trades)
• "I just understand it now" without specifics
• Improvement only visible in current market regime
• Increased trading frequency or position size

If your confidence is growing but you can't point to specific process improvements, it's probably recency bias talking. Wait for sample size to confirm.

A common mistake: changing strategy mid-streak

You're winning. You start adding refinements: tighter stops, wider targets, new entry rules. Each change feels like "continuing to improve." But you're now trading a different strategy than the one that produced the wins, with no backtested basis. The strategy that won the last 5 trades isn't the one you're running anymore.

The defense: lock the strategy. Mid-streak is the worst time to change strategy because your judgment is contaminated by recency. Make changes only at scheduled review intervals (e.g., monthly), with explicit data and explicit hypotheses about why the change improves expectancy.

A common mistake: trusting external validation

You post a great trade on twitter. People cheer. The cheering adds to your sense of being right. Two weeks later, you have a losing trade. The same audience disappears or pivots to mocking you. The validation was correlated with recent results, not with strategy quality.

The defense: external validation is noise. The only signal that matters is your trade journal across many trades. Build the habit of caring about your data, not your timeline.

Mental model, overconfidence as the streak amplifier

Imagine you're flipping a fair coin and betting on the outcomes. You bet $10 per flip. Sometimes you'll get 5 heads in a row by chance. If you start betting $20 after each head ("I've figured out the pattern"), the next tails (which is 50% likely) costs you $20. Across many such streaks, your average outcome is worse than $10/flip betting.

Trading is the same dynamic with one difference: in trading, the underlying probabilities (your strategy's edge) are real, but the streaks are still mostly random variation within that edge. Size up during streaks and you're betting $20 per flip while ignoring that the coin's probabilities haven't changed.

The right behavior: bet $10 per flip regardless of the streak. Boring. Effective.

Why this matters for trading

Streaks are normal. The temptation to alter behavior in response to them is also normal. The discipline to not alter behavior is what separates traders who compound consistently from traders whose accounts oscillate. Hex37's journal page tracks rolling expectancy and position-sizing trends so you can see whether you're drifting larger or smaller in response to recent results. The data is the counter to the bias.

Takeaway

Recency bias makes recent events feel more representative than they are. Overconfidence emerges when winning streaks feel like proof of skill rather than normal variation. The typical pattern: streak → confidence inflation → size up → normal losing trade hits at inflated size → cascade. The defense is structural: fixed risk per trade, no mid-streak strategy changes, attention to long-run rolling expectancy rather than recent PnL. The boring discipline of constant sizing through streaks (in both directions) is what makes expected value actually compound.