Recency Bias: Why the Last Few Trades Overwhelm Your Overall Judgment
Ask someone with 200 trades how their strategy is doing, and they often discuss only the last 5–10 trades. The other 190 have almost disappeared from memory. Recency bias gives the newest information more weight than it deserves and influences decisions to change trading rules.
The latest 10 trades are only 5% of the record
Recency bias emphasizes recent information and discounts older information. This can be cognitively efficient, but distorts trading outcomes governed by probabilities. See trading cognitive biases for its relationship with other biases.
Full record:
200 trades; 104 wins and 96 losses.
Actual win rate: 52%.
Latest 10 trades:
3 wins and 7 losses.
Perceived win rate: 30%.
─────────────
Share represented by the latest 10:
10 ÷ 200 = 5%.
Only 5% of the information
occupies most of the judgment.
The discrepancy is straightforward: 5% of the sample can receive almost 100% of the mental weight. See win rate for interpretation and trade sample size for when a number becomes more reliable.
Losing streaks can be normal
The feeling of repeatedly losing may be accurate, but does not by itself mean the strategy has failed. Around a 50% win rate, losing streaks are to be expected.
Loss probability per independent trade: 0.48.
Five losses: 0.485 ≈ 2.5%.
Six losses: 0.486 ≈ 1.2%.
Eight losses: 0.488 ≈ 0.28%.
─────────────
The guide approximates about 195 overlapping windows within 200 trades.
Expected five-loss windows:
195 × 0.025 ≈ 4.9, roughly five or six.
Expected eight-loss windows:
195 × 0.0028 ≈ 0.5, or roughly one over 400 trades.
These are approximate expectations, not guaranteed occurrences.
A five-loss streak is therefore an event to plan for, rather than necessarily an accident. Recency bias can misread a normal stretch as the end of a strategy's useful life. See losing-streak probability for calculations by length.
The opposite error is believing that enough losses mean a win is now due. That does not follow for independent trades and is the gambler's fallacy. Both biases reach opposite conclusions from the same data while overweighting the latest few observations.
What happens when size is cut after losses?
A common reaction is reducing position size immediately after a losing streak. Intended to reduce losses, it can worsen the result if the following period recovers.
Rule: Risk 1% of initial capital per trade.
Capital $2,000 → $20 risk per trade.
Trades 1–20, including losses: −$180 total.
Trades 21–40, recovery under unchanged sizing: +$260.
─────────────
A. Keep the rule:
−180 + 260 = +$80.
B. Halve size after trade 20:
−180 + (260 × 0.5) = −$50.
All the losses were already taken;
only the recovery was halved.
The issue is the timing of the reduction: full size during the loss and reduced size during the recovery. Recency bias bases the change on the period that has already passed. See position sizing for using predefined formulas instead of emotion.
A strategy can, of course, actually deteriorate. The distinction should follow a threshold defined beforehand, rather than how painful recent trades felt. One illustrative rule reduces size when equity falls below its 30-trade moving average; see equity-curve trading.
Increasing size after wins can cost more
Expansion after a winning streak can damage the account more than losses alone. Recent success lets confidence outrun the rule while size quietly grows.
Capital: $2,000.
Original 1% risk: $20.
Four wins at +$30 each:
accumulated +$120.
“Conditions are good” → 3% risk: $60.
Four subsequent losses:
−$60 × 4 = −$240.
─────────────
Result: 120 − 240 = −$120.
Without increasing size:
120 − (20 × 4) = +$40.
The same wins and losses
produce an opposite sign.
The same eight trades move from +$40 to −$120 solely because of when size increased: after recent results looked best. The opposite trap of increasing size after losses appears in martingale.
Judging the market from only the last few days
The bias affects market assessment as well as personal performance. Quiet recent days encourage expectations of continued calm; volatile recent days encourage expectations of continued turbulence.
Average ATR over the last five days: 1.2%.
→ Choose a narrow 1.5% stop.
Average ATR over the last 60 days: 2.4%.
When volatility returns to its usual level:
normal movement 2.4% > stop distance 1.5%.
→ You may be stopped first despite the correct direction.
─────────────
The reverse mistake:
After a sharp move, leave a 4% stop.
Once the market calms, it can permit unnecessarily large losses.
Also examine a longer window, rather than only the last five days. See ATR for volatility measurement. Stops too narrow for normal variation become a stop-placement problem.
Three ways to reduce the bias
Willpower alone rarely removes recency bias. A promise not to be influenced can disappear after the next three losses. Instead, make the relevant decisions beforehand.
“Do not change the rule before 50 trades.”
→ Define the review point in advance.
② Write numerical change conditions
✗ “It seems less effective lately.”
○ “Pause if PF over the latest 50 trades is below 1.0.”
→ A condition, not a feeling, is the trigger.
③ Keep the full sample visible
Latest 10 trades: 30% win rate.
All 200 trades: 52%.
→ Write them side by side to reduce
the illusion that 5% represents everything.
The third measure is inexpensive and useful. Displaying “latest 10” and “all trades” together continually restores context. See the trading journal for record formats.
Profit factor and expectancy and R multiples can express review conditions numerically. Predefine the latest N trades window. Shortening it every time results worsen reintroduces recency bias.
A mechanism that physically interrupts a losing day can also help. A daily loss limit ends trading before emotion dictates the next action.
Do not evaluate the process only by the result
Outcome bias often accompanies recency bias: retrospectively calling a decision good because it won or bad because it lost.
Follow the rule → win:
“The rule is correct.”
Ignore the rule → win:
“My intuition was good.” ← Danger.
Follow the rule → lose:
“The rule does not work.” ← Danger.
─────────────
Evaluate whether the rule was followed.
Assess wins and losses after a sufficient sample accumulates.
Luck materially affects individual trades. Judging decision quality from each result teaches the wrong lessons. See outcome bias, and revenge trading and overtrading for how post-loss emotion spreads to the next trade.
Recap
② That 5% can dominate most judgment.
③ At a 52% win rate, the guide estimates roughly five or six overlapping five-loss windows in 200 trades.
④ Losing streaks are events to anticipate.
⑤ Cutting size after losses can halve only the subsequent recovery.
⑥ Tripling size after wins changes +$40 to −$120 in the example.
⑦ Stops based only on five quiet days can trigger prematurely.
⑧ Use predefined rules, not promises of willpower.
⑨ Choose the sample size first.
⑩ Express change conditions numerically.
⑪ Compare recent and full records side by side.
⑫ Evaluate rule adherence separately from individual wins and losses.
Recency bias makes 5% of a sample feel like the whole record. Address it by defining review timing and change conditions before the results arrive.
Caution
Trade counts, win rates, streak probabilities, capital and PnL figures are hypothetical examples, not observed account or strategy results. Streak calculations assume independent trades and a constant win rate; actual correlations and changing regimes alter them. No particular rule or sample size guarantees future returns or loss avoidance. Leverage can lose all principal, and decisions remain your responsibility.
NOONOO TRADING invites you to follow live trading in our free chat.
Start in the bot📈 OKX trading fee discount for new registrations
Register for the OKX Fee Discount →