How many losing days should I accept before changing my strategy?

The number that your own history shows to be unusual, decided in advance. If the backtest and the live record contain runs of five losing days, a run of four is not evidence of a change.

Is recency bias always wrong?

Recent information matters when conditions really change, such as a new regime or a structural change in the market. The bias is giving recent results more weight than their sample size supports, without evidence of such a change.

Recency bias

Recency bias is the tendency to give recent events more weight than they deserve relative to the longer history. Traders show it when they abandon a tested strategy after a few losing days, or increase size after a few winning ones, although the recent sample is too small to change what the longer record says.

Senzoukria · Glossary · Updated September 2026

How it distorts decisions

Abandoning a strategy after three losing days that its history shows to be common.
Increasing size after a winning week, just before the normal reversion of results.
Rewriting a playbook around the last few trades instead of the full sample.
Reading today's market through the lens of yesterday's move, for instance expecting another trend day because the last one was memorable.

The antidote is the base rate

Every strategy has a normal range of bad periods. With a 50% win rate and independent trades, a run of six or more consecutive losses appears in more than half of all 100-trade samples, and losing days come in runs as well. Before reacting to recent results, compare them with the range the long record produced: the longest losing streak, the deepest drawdown and the longest time under water. If the recent period sits inside that range, the long record still describes the strategy; if it falls well outside, it is evidence worth investigating.

Decision rules that help

Fix in advance the conditions that would suspend a strategy, in number of trades and depth of drawdown.
Change size only at scheduled reviews, on the full sample, never after a streak.
Keep the long-term statistics visible next to the recent ones.
Label thin samples as thin rather than judging them.

In Senzoukria

Several displays keep recent results in proportion. The journal dashboard's Day of Week Track Record requires at least three sessions before calling a weekday good or bad and reports it as thin below that. The Performance panel lists every past replay and backtest run so the latest one can be compared with earlier ones, and shows a rolling expectancy over 30 trades. The Gauntlet's rolling Sharpe over 20 sessions and its reshuffled drawdown distribution show what range of results the same trades could have produced in another order.

In the same section

This page in other languages

Français

Frequently asked questions

How many losing days should I accept before changing my strategy?: The number that your own history shows to be unusual, decided in advance. If the backtest and the live record contain runs of five losing days, a run of four is not evidence of a change.
Is recency bias always wrong?: Recent information matters when conditions really change, such as a new regime or a structural change in the market. The bias is giving recent results more weight than their sample size supports, without evidence of such a change.

Keep reading

Strategy researchTest a trading strategy, from observation to challenge rulesDefine an observation, question it with the copilot, write a rule and explore costs, drawdown and fictional challenge limits in an interactive example.Quant · Start hereQuant trading for beginners: from idea to backtestDefine a testable rule, choose suitable data and evaluate later periods after costs.Order flow · HubOrder flow trading: the complete guideThe hub. What order flow is, what it shows — aggression, delta, imbalance, absorption, liquidity — and the tools that read it, each linked to a full guide.