Base Rate Neglect: Why Your Setup Feels Better Than the Data Says

Your setup wins forty-one percent of the time. You know this, because you counted, once, a while ago.

It does not feel like forty-one percent. It feels like a good setup. It feels, if you are honest, like it works more often than not, and the reason it feels that way is that the last three times you took it, it worked.

Those two beliefs are describing the same setup. Only one of them is being used to size the trade.

What the gap costs

Take a setup with a genuine 41% win rate and 1.5R winners. Modest. Real.

The trader who believes 55% expects ninety-four R a year. He gets six.

Now charge him for the privilege. Spread, commission, and slippage of five hundredths of an R per trade, which is generous for most retail execution.

Note what has happened. He does not have a discipline problem, a psychology problem, or an execution problem. He has a number problem, and it is fourteen percentage points wide.

Why three wins prove almost nothing

Suppose you are open-minded. You give this setup a twenty percent prior chance of being genuinely good, meaning a 55% win rate, and an eighty percent chance of being ordinary at 41%.

You take it three times. You win three times.

What should you now believe?

Thirty-seven ¤6 percent. Still more likely than not that you are looking at an ordinary setup having an ordinary run.

And to reach ninety percent confidence, starting from that twenty percent prior, you need thirteen consecutive wins.

Thirteen. You have never had thirteen. Nobody has. Which means no winning streak you have ever experienced was sufficient evidence to promote a setup in your mind, and yet every winning streak you have ever experienced did exactly that.

The mechanism

Kahneman’s term is representativeness. You judge how likely something is by how much it resembles your mental image of the thing, and not at all by how common the thing actually is.

The setup in front of you looks like the picture in your head of a good setup. The picture in your head was assembled from your most vivid memories, and your most vivid memories are the winners, and they are vivid precisely because they were unusual.

The base rate has none of these properties. It is a number in a spreadsheet, from a period you barely remember, computed over trades that have blurred together. It cannot compete with a recent, specific, emotionally charged image of the same pattern paying you three times.

Four sentences that mean you have abandoned the base rate

“This one’s different.” It is not. If your setup definition permits you to notice that this instance is special, then your setup definition is not a definition. The base rate already contains every instance that was special in that way.

“It’s been working well lately.” A statement about a sample too small to distinguish from noise, being used to override a sample large enough to be informative. Backwards, in both directions.

“I’ll size up here.” Sizing is a function of the base rate and your account. If the base rate has not changed, the size cannot have.

“The chart looks great.” The chart always looks great. That is what a setup is. If the chart looked bad you would not be here.

The compound problem: base rates plus survivorship

Where did your 41% come from?

Almost certainly not from a count. Almost certainly from an impression formed while trading, in which the trades you remember are the ones that resolved dramatically, and the trades you skipped, cut early, or forgot to log are absent.

Which means the number you are neglecting was probably already wrong, in the same direction, for a different reason. You are not choosing between a warm feeling and a hard number. You are choosing between a warm feeling and a slightly cooler feeling that you have written down.

Bulkowski’s entire body of work exists to solve this. Not to tell you a pattern works, but to tell you how often it does not, and what the failure looks like. It is the least glamorous trading book ever written, and it is the only one that supplies a denominator.

What to do

Count, once, properly. Fifty instances of your setup, tagged, with the outcome in R. Not from memory. From the chart, or from your journal if it is honest. This is a weekend, and it will be the most valuable weekend of your trading year.

Write the number where you will see it at the moment of entry. On the chart. In the order ticket comment. Anywhere the story is about to be told.

Update slowly, in public. If you want to believe your setup has improved, say how many trades it will take to convince you, before the trades happen. If the answer is fewer than thirteen, you are not updating. You are being persuaded.

Treat “this one’s different” as an alarm, not an insight. The sentence has never once, in the history of retail trading, preceded a good decision.