Process and Outcome
In any domain with substantial randomness, a good decision can produce a bad result and vice versa. Judging decisions by their outcomes teaches the wrong lesson at exactly the wrong moment.
MadStockAlerts Research · Updated August 28, 2026
What to take away
- Outcome and decision quality are only loosely connected over small samples.
- Resulting is judging a decision by how it turned out.
- A profitable mistake is more dangerous than an unprofitable good decision.
- Only a process can be controlled, so only a process can be improved.
- Over a large sample, outcomes do become the evidence.
MAD Academy Training Video · 0:45
Four Boxes, Not Two
A good decision can lose and a bad one can win, so judging yourself on outcomes teaches exactly the wrong lessons.
This lesson is part of a Stock Alerts + Tools plan.
The four cases
| Decision | Outcome | The lesson usually taken | The correct lesson |
|---|---|---|---|
| Good | Profit | The method works | One observation, consistent with the method |
| Good | Loss | The method failed | Variance. Nothing to change |
| Bad | Profit | The rule was too strict | The most dangerous cell on this table |
| Bad | Loss | Bad luck | The one case where something should change |
The third row
A rule broken profitably is the most expensive event on this table. It is rewarded immediately, which reinforces the behaviour, and the cost arrives later and is attributed to something else.
The second row is the one that ends methods prematurely. A sound approach producing a run of losses is indistinguishable, in the moment, from a broken one, and abandoning it is a decision made with the least information and the most discomfort.
Resulting
The habit of judging a decision by its outcome has a name in poker, where the same structure applies: resulting. It is a reasonable heuristic in domains where results follow decisions closely, and it is actively misleading in domains with heavy randomness.
Markets are firmly in the second category over short horizons. A single trade carries almost no information about whether the decision behind it was sound, which is the same sample-size point the expectancy article makes arithmetically.
What can actually be controlled
- Whether a situation met the written criteria.
- Where the exit was placed and why.
- How position size was derived.
- Whether the plan was followed.
- Whether the trade was recorded and reviewed.
Not on that list: whether the trade made money. Grading against the controllable items is what makes improvement possible, because those are the only inputs that can be changed deliberately.
Where outcomes do matter
Over a large enough sample, outcomes are the evidence. A hundred trades that met a written plan and lost money say something real about the method, and continuing to run it on the grounds that the process was sound is its own error.
The distinction is between a single result, which is noise, and an accumulated record, which is data. Both errors are available: abandoning a good method on three losses, and defending a bad one for a hundred.
Sample size is what makes the distinction usable
The claim that process matters more than outcome is only true over enough repetitions for outcomes to average out. Over a handful of trades the outcome carries almost no information about the process, which is the whole point. Over several hundred, the outcome is the main evidence about the process there is.
That leaves a practical problem: at the sample sizes most individuals accumulate, the outcome is too noisy to judge the process and the process cannot be judged any other way. The conventional resolution is to evaluate the process on whether it was followed and on whether its logic survives inspection, and to hold judgement on whether it works until the sample is large enough to say.
| Trades | What the record supports |
|---|---|
| Under 10 | Nothing about the method. It is noise in both directions |
| 10 to 30 | A first look at whether the rules are followable in practice |
| 30 to 100 | The rough shape of the win rate and the size distribution |
| 100+ | A hit rate that describes the method rather than the sample |
The dangerous region is the middle. A method that has worked twenty times feels established, and twenty is small enough that a run of that length happens by chance regularly. Most abandoned methods and most overconfident ones are abandoned or adopted in that band.
Resulting, and how it operates
Resulting is the habit of judging a decision by how it turned out. In a domain with a large random component it is a systematic error, because a good decision can produce a bad outcome and a reckless one can produce an excellent result, and both happen constantly.
| Good outcome | Bad outcome | |
|---|---|---|
| Good process | Deserved. Learn nothing new | Expected sometimes. Change nothing |
| Bad process | The dangerous cell. It teaches the wrong lesson | The lesson is available, and easy to attribute to bad luck |
The top-right and bottom-left cells are where the damage is done. A good decision with a bad outcome invites abandoning a process that was working; a bad decision with a good outcome reinforces one that was not, and the second is more dangerous because nothing prompts a review.
The only defence is separating the two evaluations in the record. Whether the rules were followed and whether the trade made money are two different columns, and keeping them apart is what makes it possible to notice the bottom-left cell at all.