Psychology as a Systems Problem

The standard treatment of trading psychology is exhortation: be disciplined, control your emotions, stick to the plan. This is useless advice, in the same way that "write fewer bugs" is useless advice.

The engineering framing is better. You are an unreliable component in your own system. Design around yourself rather than attempting to fix yourself.

The specific failure modes

Each of these has a structural mitigation, which is the only kind that works:

Overriding the system. You see a trade the system took and you disagree, so you skip it or exit early. This is the most common and most expensive failure. Note that your override might genuinely have positive expectancy - but you have no way to know, because you have no sample of your overrides, and your memory of them will be selective. Mitigation: full automation. Not partial. If a human decision point exists, it will eventually be used badly.

Revenge trading. After a painful loss, you take a trade the system didn't signal, sized larger. Mitigation: a hard daily loss limit that halts the system, plus physical separation from the terminal after it triggers.

Hesitation after a losing streak. You skip the next signal because the last five lost. Chapter 11 explains why this is catastrophic: for a positive-skew strategy, a small number of trades produce all the profit, and skipping trades is a lottery over which ones you miss. Mitigation: automation, plus internalising the drawdown simulation from Chapter 4 so streaks are expected rather than alarming.

Premature abandonment. You stop a working strategy during a normal drawdown. Mitigation: written kill criteria, set in advance - Chapter 20.

Size creep. After a good run you increase risk, right before the mean reversion in your results. Mitigation: position sizing derived from a formula, in code, not adjustable without a config change and a cooling-off period.

The core problem you cannot solve by thinking harder

Your emotional response is calibrated to a sample size of one. You feel each trade individually. Your strategy's edge exists only across hundreds.

This mismatch is not fixable through willpower or experience, because it isn't a character flaw - it's a mismatch between the timescale of your feelings and the timescale of the statistics. The only reliable fix is to remove yourself from the loop at the timescale where your feelings operate.

Practical expressions:

Automate execution fully. The gap between signal and order is where discretion leaks in.
Do not watch P&L intraday. There is no decision it informs. If it informs one, that decision should be in the code.
Review on a fixed cadence - weekly or monthly - with a fixed agenda, not whenever you feel uneasy. Feeling uneasy is not evidence.
Separate research time from execution time. Never modify a strategy on a day it lost money. The impulse to fix is strongest exactly when your judgement is worst.
Size small enough to be boring. If a normal drawdown would be distressing, your position size is wrong regardless of what the maths said. The maths assumed you'd still be trading afterwards.

Write the drawdown plan while calm

Before deploying, write down: what a normal drawdown looks like for this strategy (from the simulation), what you will do during one (nothing), and what specific evidence would change that.

Then, during a drawdown, read the document rather than re-deriving the decision. The version of you who wrote it had better information than the version reading it, because it had the statistics and no pain.

The distinction that matters, and the reason the document must exist:

"I should stop because the edge is gone" and "I want to stop because this hurts" feel identical from the inside, and you cannot tell them apart while it hurts.

That's precisely why the criteria have to be written in advance, when you can.