Writing a Trading Plan You'll Actually Follow

A trading plan is decisions made in advance. Everyone says it. Nobody says what a decision made in the moment actually costs.

So I tried to measure that, and the measurement went somewhere I did not expect. By the end you'll be able to write a one-page plan and check how often you followed it.

Here is the setup. One signal — a touch of a level — and three decisions: where to enter, where the stop goes, how to exit.

Two entries, three stops, four exits. Twenty four combinations, eight hundred and eight touches, six instruments.

A written plan is one of those twenty four, chosen in advance and not changed. An unwritten decision means you take some version of it each time, and which one depends on your mood.

So the cost of not writing a line down should be the width of the spread across that decision. That was the idea.

First, the whole grid. The worst of the twenty four combinations returns minus zero point two one of an R per trade. The best one still loses: minus zero point zero one.

A spread of zero point two of an R. That looks like the price of having no plan at all, and I was ready to end the video there.

Then I asked the obvious next question: which of the three decisions is the expensive one?

Fix two, leave the third open, measure the spread. Across the full window: the exit gives zero point zero six one, the stop zero point zero five eight, the entry zero point zero four one.

Exit first, stop second, entry last. A clean order to write your plan in. So I checked it on the two halves of the window, as this channel checks everything.

On the first half the order was entry, stop, exit. On the second half it was entry, exit, stop.

Three orderings from the same data. Which means the ranking is not a ranking — it is noise wearing one.

And that sent me back to check whether any of these numbers are measured at all.

Here is the thing I had been doing wrong. The variants of one decision run on the same trades, not on separate ones.

Comparing two averages of the same trades as if they were independent samples throws away the pairing and inflates the uncertainty enormously.

The right way is to take the difference trade by trade and look at the average of those differences, with its own error bar. The error bar is how far that average would typically move if you ran the same test on a fresh batch of trades. Four times the trades, half the error bar.

Done properly, the stop is the clearest result in the video, and the result is: nothing.

The clearest pair of stops, one range against one and a half, differs by zero point zero four of an R, with an error of zero point zero three, on three hundred and eighty five trades.

That is not a difference. It is one point two error bars from zero. A difference counts as measured only when it is at least two error bars away from zero.

The exit does clear the bar, barely — and it is the extreme pair of its four options, so I hold it loosely. A two-stop target against a three-stop target: zero point zero five of an R, error zero point zero two five.

Measured, real, and small enough that it would take about three hundred and fifty trades, each checked both ways, before it showed in your own record. And thousands, if you only ever trade one of the two.

And the entry, on shared trades the biggest of the three paired numbers — plus zero point three five of an R — is the one you cannot use.

Because the confirmation rule does not only change your entry price. It changes which trades you take at all.

Comparing the two on their shared trades means asking: on the touches that later closed back, was it better to enter at the touch? Yes, by a lot.

But at the moment of the touch you do not know whether the bar will close back. That comparison is only available afterwards, and afterwards is not a place you can trade from.

So of three decisions, one is unmeasurable in principle by this method, one is unmeasurable in practice at this sample size, and one is probably real and worth about zero point zero five of an R.

Which leaves the spread of zero point two from the beginning of the video looking rather different.

It is a real spread between the best and worst combinations. It is not a prize you collect by choosing the best one.

Because choosing the best one does not survive. On this data, the best combination on the first half returned plus zero point one eight there and minus zero point one eight on the second half.

The same thing happened in four of the five setup videos that picked a best combination on one half. And the spread itself stays: on the second half the grid is still zero point one eight wide.

So the spread is durable and the choice inside it is not. Both of those are true at once, and together they say something specific about plans.

Writing the plan does not earn you the spread. Nothing in this data says that your particular line makes money.

What writing it down does is something else, and the statistics video, two lessons from now, is where it pays off.

If you decide each trade in the moment, your last fifty trades are not fifty trades of one strategy. They are fifty samples from twenty four different strategies.

The statistics video shows how much those numbers move even for one fixed rule. Every statistic you compute on a mixed record is measuring the mixture, not the method.

Your average loss stops describing your stop. Your win rate stops describing your exit. Your expectancy, as that video will show, is the least stable number you have even for one fixed rule — and on a mixed record it is not even about one thing.

That is the argument for a written plan, and it is not the argument usually given.

It is not that the rule is right. On this data, mostly, nobody can tell whether the rule is right.

It is that a fixed rule is the only thing that makes your own record readable. Without it you cannot learn from your trading, because there is no single thing to learn about.

Which also tells you what to write down, and the order is the opposite of the intuitive one.

Write down first the decisions that change WHICH trades you take. Entry conditions, filters, what makes a setup valid at all.

Not because they are worth more R — I could not show that — but because they are the ones that change what your sample is made of.

Write down second the mechanical ones: stop distance, exit rule. They change how each trade ends, not which trades you take, and this data cannot say which of the two matters more.

And do not write down anything you cannot check later. A plan that says "enter with confidence" cannot be violated, so it cannot be tested, so it is decoration.

Three limits. One: twenty four combinations of one setup. A different setup might have a decision that genuinely dominates, and I would have missed it here.

Two: three hundred and eighty five paired trades. The exit result just clears the bar at that size; a smaller effect would not be, and most of the effects here are smaller.

Three: all of it before costs. Every variant pays one round trip, so costs cannot reorder the exits. They do charge a narrower stop more in R, because the same dollars are a bigger share of a smaller stop.

So, what to go and do. It takes about twenty minutes and produces one page.

Write down your rules in three blocks: what makes a trade valid, where the stop goes, how you get out.

Then go through your last twenty trades and mark, for each one, whether you followed all three blocks or not. No trades yet? Take twenty setups from a chart six months back and mark whether your one page would have let you take each of them.

If the number of trades where you followed everything is smaller than you expected, that number, not the rules, is what to work on first.

And when you next want to change a rule, change one block, keep the rest fixed, and give it fifty trades before judging.

Fifty will not be enough to prove the change helped. It will be enough to tell whether you can follow it.

Which is the honest order of business: follow it first, measure it second, improve it third, and expect the third step to take longer than anyone tells you.

A plan does not buy you the spread between the best and the worst rule. It buys you a record you can read. Educational content only. Nothing here is financial advice.