Thinking in a Sample of Trades, Not One
The book's famous recommendation: judge a system across a defined batch of trades, never off any single one.
Douglas recommends a specific practical exercise: commit in advance to executing a defined system, without deviation, across a pre-set number of trades (commonly cited as around 20) before drawing any conclusion about whether the system works. This directly operationalizes the casino-style thinking from earlier in this course — a large enough sample is what actually reveals an edge; any individual trade within it is close to noise.
The specific commitment matters as much as the general idea. Simply "trying to think in samples" is easy to abandon quietly under pressure, the same way a vague intention to "stay disciplined" tends to erode; committing to an exact, pre-chosen number of trades, decided before any of them are taken, gives the trader a concrete finish line to hold to rather than an open-ended, constantly-renegotiable standard.
This chapter functions as the practical bridge between the statistical thinking covered earlier in this course and the discipline-building techniques of the previous chapter — it's the single specific exercise that puts both into practice at once, and Douglas presents it as the most direct way to actually feel, rather than just intellectually accept, that an edge lives in the sample and not in any individual trade.
| Trade-by-trade judgment | Pre-committed sample of ~20 trades | |
|---|---|---|
| When a conclusion gets drawn | After every single trade, win or lose | Only after the full, pre-chosen sample is complete |
| Effect of a losing streak mid-sample | Often triggers abandoning the system immediately | Expected — normal variance the sample size was chosen to absorb |
| What actually gets evaluated | The emotional sting or relief of the most recent result | The system's real win rate and risk/reward across the full batch |
A system with a genuine, real edge can still lose several trades in a row purely from normal statistical variation — exactly the "random distribution between wins and losses" from the Five Fundamental Truths chapter. A trader without a pre-committed sample size will often abandon a perfectly good system after a losing streak that was, statistically, entirely expected and unremarkable — destroying a real edge purely from impatience with normal variance.
A system with a genuine 55% win rate and a favorable risk/reward ratio can still lose 5 or 6 trades in a row at a completely normal frequency, purely from statistical variance — not because anything about the market or the system changed. A trader who judges the system after each individual trade, rather than committing to a full sample first, is likely to abandon a genuinely profitable system during one of these entirely expected losing streaks, right before it would have reverted to its actual long-run performance.
Douglas doesn't present 20 as a rigorously derived statistical minimum — the right number varies with a system's actual win rate and the trader's own psychological tolerance for a run of losses. The point of naming a specific figure is practical rather than mathematical: a concrete, pre-committed number is something a trader can actually hold themselves to, where "a while" or "long enough" is vague enough to be reinterpreted downward the moment a losing streak becomes uncomfortable.
A system with a lower win rate or a wider range of possible outcomes per trade genuinely needs a larger sample before its results say much of anything reliable — the specific number matters less than the underlying principle that the required sample size scales with how much variance the system produces. What stays constant across any system is the discipline of picking a number in advance and holding to it, rather than quietly shrinking the sample the moment the running results turn uncomfortable.
- A pre-committed sample size (a defined number of trades, decided before starting) is a specific, practical technique for actually living out the statistical thinking covered earlier in this course, not just agreeing with it in the abstract.
- A losing streak within a real edge's expected variance is not evidence the edge stopped working — the whole point of committing to a sample size in advance is removing the temptation to draw that conclusion too early.
- The specific number (commonly cited as around 20) matters less as a statistical minimum than as a concrete, hard-to-renegotiate commitment made before pressure sets in.
- This exercise is presented as the practical bridge between the statistical thinking of earlier chapters and the discipline-building techniques of the previous one — it's where both get put into practice together.
- This chapter's technique is the direct practical bridge into the final chapter's concept: "the zone" as a state built from genuinely trusting the process across a sample, not from getting any single trade right.