Cognitive biases in trading

It is 10:20 ET (22:20 ICT, winter). You short NQ at 18,060 on a rejection at yesterday’s value area high, stop 18,082, written before the entry. Ten minutes later price grinds to 18,075 against you. Nothing in the plan has changed, yet you catch yourself re-reading the bearish case, remembering that 18,075 was yesterday’s top, and reaching for the stop. Which of those thoughts are analysis, and which are biases running the trade? This lesson names six of them and gives each one a defense that does not depend on noticing it in the moment.

Key term

A cognitive bias (thiên kiến nhận thức) is a systematic error in judgment: a mental shortcut that pushes many people’s decisions the same wrong way in the same situations Definition. Biases are not rare lapses; they show up in large samples of ordinary people and ordinary investors MeasuredOther market.

Knowing a bias’s name does little to stop it PlausibleOther market. In Fischhoff’s hindsight experiments, people who were told an outcome rated it as more predictable than people who were not, even when asked to answer as if they did not know it. That is why every defense below is structural: a resting order, a number written before the event, a record you compare against. These are house rules House rule; whether they reduce bias costs for traders is not yet measured.

Six biases and their defenses

Each bias below is well documented in general psychology or in investor records MeasuredOther market. How strongly each one shows up in NQ day traders has not been measured; the trading versions are plausible extrapolations PlausibleOther market.

1. Confirmation bias (thiên kiến xác nhận): looking for evidence that supports your position and discounting evidence against it, such as re-reading the bullish case while price trades through your level. Defense: the four pre-trade questions (M0.06) make you write the invalidation price before entry; the journal’s deviation tags (M7.03) catch the rest.

2. Recency bias and overconfidence (thiên kiến gần đây, tự tin thái quá): giving the last few trades too much weight. Three winners lead to bigger size and more trades; three losers lead to dropping the method. In brokerage records, men (the study’s proxy for overconfidence) traded 45% more than women, and the extra trading cost them more in net returns MeasuredOther market. Defense: size comes from the written plan, not from the last result, and the method is judged on its whole sample with its interval (M7.10). A short streak is mostly noise: at a 40% win rate, two or fewer wins in ten trades happens about 17% of the time (M7.07).

3. Anchoring (neo giá): fixing on a number that no longer matters, such as “I should have gotten 18,100” or “18,075 was yesterday’s top”, and trading the anchor instead of current structure MeasuredOther market. Defense: every entry and stop is re-derived from the levels in today’s written plan (M6.01–M6.05); “should” is not a level.

4. Loss aversion and the disposition effect (né tránh thua lỗ, hiệu ứng bán lời giữ lỗ): a loss feels worse than an equal gain feels good. In laboratory gambles, losses weighed about twice as much as equal gains (Tversky and Kahneman estimated 2.25); later estimates vary with the method MeasuredOther market. In trading it shows up as the disposition effect: taking winners early and holding losers, including widening a stop MeasuredOther market. Defense: the stop rests as a working order from entry and is never widened (M6.08); targets and exits follow the plan (M6.11). Planning in R (risk units) instead of dollars is a house habit meant to blunt the asymmetry House rule; its effect is not yet measured.

5. Outcome bias: judging a decision by its result MeasuredOther market. A rule violation that wins “proves” the violation works; a correct stop-out “proves” the setup is bad. Defense: the adherence score (M7.03) grades process separately from P&L on every trade.

6. Hindsight bias: “it was obvious” after the fact MeasuredOther market. Grading yesterday’s decisions with today’s knowledge corrupts the review. Defense: the morning sheet is written before the open (M7.01), and the review compares each decision with what was written at the time, not with memory.

What the data says

Worked example: three biases in one trade

The opening short, step by step (synthetic prices, one NQ contract at $20 a point). Entry 18,060, stop 18,082: risk 22 points, $440, which is 1R.

1. At 10:30 price reaches 18,075, still below the stop. Unrealised loss: 15 points, $300.
2. The trader re-reads the bearish case three times (confirmation), recalls that 18,075 was yesterday’s top (anchoring), and moves the stop to 18,095 “to give it room”, because taking the loss now hurts more than the plan’s logic says it should (loss aversion).
3. At 10:35 price spikes to 18,096.50. The moved stop fills at 18,096: 36 points, $720, about −1.6R before costs. The planned stop would have filled near 18,082 for −1R.
4. By 10:45 price has fallen to 18,020.

Post-mortem by process: the planned trade lost 1R, which is what a stop is for. The bias cost was the extra 0.6R from moving it. What price did after the exit is not evidence about the setup: one trade cannot show whether a rejection short at value area high has an edge, and judging it by the drop to 18,020 is outcome bias. The structural fix written into the plan: “The stop rests as a working order from entry. It may move only to reduce risk (M6.11); widening it does not exist in my plan.”

The mirror and the failure

Mirror. On a long the same chain runs downward: price sags toward the stop, the trader recalls a “strong low” a few points lower, and slides the stop under it. Same biases, same fix.

Failure. The resting stop is not free. On this day it took the trade out at 18,082, and price then fell 62 points without you. Sometimes a stop-out is followed by the move you expected. That is the cost of a defined loss, and it can only be judged over many trades (M6.08, M7.10), never on the one where it hurts.

Prop-firm note. On evaluation accounts, a daily loss limit is a firm rule, and an extra 0.6R per bad trade reaches it sooner; the PROP course shows each rule with its dated card. Put the stop in as a bracket at entry. Evaluation fills are simulated, so a stop’s fill there is not a live fill (MKT2.03), and firms treat strategies that exploit simulated fills as prohibited conduct (PROP1.04).

Common mistake: Treating awareness as protection. The trader in the example knew about loss aversion and moved the stop anyway. The test is not whether you can name the bias; it is whether your plan makes the biased action impossible: a resting stop, a pre-written size, a written invalidation.

Common mistake: Reading the investor studies as proof about your own NQ trading, or the reverse, dismissing them because they are about stocks. They are strong evidence that ordinary people trade this way, and a reason to look for the same pattern in your own journal. Your journal, not the papers, tells you which bias costs you money.

Common mistake: Grading a stop-out by what price did afterwards. “It went my way after I was stopped, so I was right” and “it went my way, so the stop was wrong” are both outcome bias. Grade the decision with what was known at the time.

Drill: price your biases

Take your last 20 closed trades from the journal (M7.03).

1. For each trade, write the planned stop and target, and the actual exit, all in R.
2. Mark every trade where the exit differs from the plan, and tag the bias that best explains the change (confirmation, recency or overconfidence, anchoring, loss aversion, outcome, hindsight).
3. Add up the R lost (or gained) by deviating from the plan, per bias tag.

What we don’t know yet

• How often each bias appears in NQ futures day traders: not yet measured; the strongest records are from stock investors and from Treasury futures floor traders of the late 1990s.
• Whether resting stops, R-based planning and written invalidations reduce the cost of biases for traders: not yet measured. The supporting evidence (checklists, if-then plans) comes from other domains.

Recap

• Six biases: confirmation, recency and overconfidence, anchoring, loss aversion (the disposition effect in trading), outcome, hindsight, each with a concrete trading behavior.
• Investor records measure the disposition effect and overconfidence; for futures day traders the evidence is plausible, not measured on NQ.
• Knowing a bias’s name does little; every defense here is structural (a resting stop, a pre-written size, a written invalidation) and is a house rule.
• Grade a trade by its process. What price did after the exit says nothing about the decision.
• The journal’s deviation tags are your bias detector: patterns in the tags show which bias is costing you R.

Further reading

• Odean 1998, “Are Investors Reluctant to Realize Their Losses?”, Journal of Finance: the disposition effect in 10,000 brokerage accounts.
• Barber & Odean 2001, “Boys Will Be Boys: Gender, Overconfidence, and Common Stock Investment”, Quarterly Journal of Economics: overconfidence, trading volume and net returns.
• Coval & Shumway 2005, “Do Behavioral Biases Affect Prices?”, Journal of Finance: loss aversion in futures proprietary traders.
• Fischhoff 1975, “Hindsight ≠ foresight”, Journal of Experimental Psychology: Human Perception and Performance: why knowing the outcome rewrites what seemed predictable.
• Baron & Hershey 1988, “Outcome bias in decision evaluation”, Journal of Personality and Social Psychology.
• Our lab: our evidence review grades the process research (if-then plans, checklists, illusion of control in traders) as measured in other domains and plausible for trading; none of it has been tested on NQ day traders.

Connects to: M7.04 (the plan that cages these biases) → M7.06 (the emotional cycle that rides on top of them).

Educational content only — not financial advice. Futures trading involves substantial risk of loss.