The Share That Is Real

Six habits, already priced

A habit here is not a feeling. It is a decision rule with a number attached, and it earns a place on this page only because some earlier page measured what it does. Six of them have been priced, one at a time, and the pages that priced them were about other things.

Read down the middle column and notice what is not there. Not one of these was measured by a psychologist, and not one of them needs a theory of why people do it. Every figure came out of the same arithmetic the rest of the course runs on. They are in six different units and nothing here adds them up, but two of them can be put through the same test, which is the only reason the second half of this page exists.

Read across and the six share a shape. Looking at more settings, checking more often, abandoning sooner and adding a filter are all the same decision made at different moments: to consult the record more than the record can support. The other two are not about the record at all. They are about not doing what the rule book already says: sending the order it produced, and stopping at the count the limit fixed.

Almost nobody has written their own six down with the numbers beside them. Every figure is already on a page you have read, and the exercise is to admit which of the six you did last month.

Module 11’s list of the degrees of freedom this course did not have gains its third item, and it is the one every page has assumed away: the person who has to sit through all of it.

Now put two of them together, which nothing in this course has done. Lesson 65’s test, run over 156 trades against lesson 67’s tenth of an R, has a power of 0.3461: it accepts a system that genuinely works about a third of the time. Take a hundred candidate rules of which some share are alive, run them through the test twice — once with the discipline lesson 65 asked for and once with two of the habits — and ask what fraction of the rules you end up accepting are real.

The last column is the finding and it points the wrong way for comfort. The habits cost most where the good rules are rarest, which is exactly the situation a person who has not found one yet is in. At a base rate of one in two they cost a factor of 2.38, and at one in fifty they cost 10.52. The reader who most needs the discipline is the reader for whom breaking it is most expensive, and nothing about that is psychological. It is Bayes.

A hundred candidates, twice

Take the middle row and follow the hundred rules through both runs, because the counts are more use than the percentage.

Ten are alive. Judged once, the test accepts 3.46 of them and accepts 4.55 of the ninety dead ones, so you accept 8.0 rules and 43.2 per cent of what you accepted is real.

Now peek. The ninety dead ones pass 24.25 per cent of the time rather than 5.05, so 21.83 of them get through.

Now abandon at eight R. That line removes 59.5 per cent of the live ones, so the 3.46 becomes 1.40.

You now accept 23.2 rules, of which 1.40 are real: 6.0 per cent.

So the two habits together got you 2.90 times as many accepted rules and 40.5 per cent as many true ones. That is the shape of it, and it is worth saying slowly: you did not end up with fewer systems. You ended up with nearly three times as many, and almost none of them work.

The habits do not cost you money. They cost you the meaning of the word yes.

Which is why the remedy is not resolve. Every one of the six is defeated by writing a number down before you start: the horizon, from lesson 65; the depth you will sit through, from lesson 67; the count of settings you searched, from lesson 64; the position limit, from lesson 75. None of those costs anything, none of them requires you to feel differently on the day, and each of them is a decision made once instead of a decision made every time you look.

So write your own six down, put the number beside each, and mark the ones you broke last month.

What this does not settle

That these are the six habits that matter. They are the six this course has priced. The ones it has not priced are the ones that do not reduce to a number, and inventing figures for them is exactly what a page like this is usually made of. What is here is arithmetic that was already on the pages the table names, and the reader who wants a seventh row has to find a page that measured one.

That the two habits are independent. They are treated as independent in the second table and they plainly are not: the person who checks the record every day is the same person who abandons it at eight R, and the checking is often what produces the abandoning. Any positive relationship between them makes the compound worse than 6.0 rather than better, so the table is the optimistic version of its own finding.

That 0.3461 is the power of the test. It is the power of lesson 65’s bar over 156 trades against lesson 67’s tenth of an R, and that edge is a hypothesis this course has carried since lesson 65 rather than a measurement. A larger edge raises the power, which raises both columns of the second table and shrinks every ratio in it. The finding survives the change and the numbers do not.

That the base rate is knowable. It is the one input in the second table you cannot measure, which is why every row is printed and none of them is recommended. The nearest thing to a hint this course has anywhere is the result in lesson 63, that 52.5 per cent of series with no time structure in them produce a best-of-253 rule beating holding by at least as much as the real one did, which is a reason to read the top rows rather than the bottom ones.

And the concession that costs most: this page has priced a person by listing decision rules with numbers on them, and that is a model of a person with the part that makes habits habitual left out. It says nothing about what it is like to be ten years into lesson 79’s unanswered question, or why the fourth losing month is the one where the eight-R line gets crossed rather than the first. That is the missing half of every page in this module, and this course does not have it. Lesson 85 is the last page, and it goes back to the one number every page since 67 has held perfectly still: how far a trade travels, set at one R and never moved.