The Cost of the Trades You Didn't Take

Open your account history and you are looking at a list of trades you took. Every row is a decision that went one way. What is not in the file, anywhere, is the other set of decisions — the calls you read and did not act on, the setups you saw and passed, the entries your own rule rejected before you had a chance to want them.

That set is usually bigger than the one on the screen. For most people it is much bigger. And it is the set their process actually operates on: a filter's whole job is to reject, so the more a rule works, the less of its work is visible.

Your account is a record of half your decisions. The other half is the half you were trying to improve.

This is not a motivational point about missed opportunity. It is a measurement problem, and it has a specific consequence: the part of your trading that does the most work is the part you have no evidence about, so it can drift for months in either direction and nothing in your records will contradict it.

The feedback is one-sided, so the drift has a direction

Consider what each kind of mistake leaves behind. A bad trade you took leaves a row, a number, and a feeling. A good trade you skipped leaves nothing at all — no row, no number, and usually no memory, because you never found out what happened to it.

So the two errors are not treated equally by your own record-keeping, and over time the rules move toward the error that does not show. Tightening a filter reliably removes losses from the statement. It also removes wins, but those come off a page nobody is keeping. Every quarter the rule gets a little stricter, every quarter the statement looks a little calmer, and nothing anywhere suggests the trade-off went too far.

The opposite drift is just as real and comes from the same cause. Someone who does notice their skips usually only notices the painful ones — the call that was passed and then ran to target while they watched. That single memory is vivid and unopposed, because the skips that went on to lose leave no impression whatsoever. One bad miss loosens the rule that produced it, and the rule stays loosened.

Both drifts are the same failure: a decision with no record attached. Which one you get depends on your temperament, not on your method.

Three ways not to take a trade, and only one of them is testable

Before keeping any record it is worth separating things that look identical in hindsight and are not the same event.

A rule rejected it

The ratio at the price you could get was below your floor. The entry was further away than your chase allowance. The driver was already carried by two open positions. Whatever the clause, something you wrote down in advance did the rejecting, and you can name which one.

This is the only category that can be evaluated, because it is the only one where a specific, stated rule made a specific, checkable decision. Everything useful in this article is about these.

You decided in the moment

Nothing rejected it. You did not like the look of the chart, or you were already down for the day, or it felt too easy, or you had been wrong twice that morning and did not trust yourself. These are real reasons and some of them are good ones, but none of them is a rule, and a record of them cannot tell you whether to keep doing it — a discretionary judgement made differently every time has nothing stable to test.

The useful move is not to log these and hope a pattern appears. It is to notice how many there are. If most of your skips are in this category, your process is discretionary and the filters you believe you run are decoration. That is worth knowing on its own, and the repair is to turn the reasons that recur into clauses so they become the first category.

You never saw it

Asleep, at work, the notification buried. This is not a decision at all, and filing it with the others is the fastest way to make the whole record useless — it puts entries in the column that no rule produced and no rule change could affect.

Keep these, but keep them apart. They measure coverage, not judgement, and they answer a different question: which hours of the day your participation is real, and whether the answer to a run of them is an alert, a different session, or accepting that a part of the feed is not for you.

The skip that is not a skip

There is a fourth case, it belongs to none of the above, and it is almost certainly the most expensive item on this page.

You read the call. You decided not to take it. Twenty minutes later you took it, at a worse price.

In your account this is a trade, indistinguishable from any other. In your process it is a failed rejection — the rule fired, the rule was overridden, and the override was paid for in entry price. Nothing in your records will ever surface it, because the only trace of the first decision was a thought.

It deserves its own mark for one reason: it is the single clearest evidence that a rule is not binding. A rule you overrule is not a filter, it is an opinion you had earlier. Counting these is the cheapest audit available, and if the count is anything other than very small, no other finding in your log means much, because the log is describing rules that do not actually govern anything.

What to write down, and when

The timing matters more than the content. Write the entry at the moment of the skip, before you know what happened — the same reason a journal is written at entry rather than at exit. A skip recorded after the outcome is known is not a record of a decision; it is a record of your reaction to an outcome.

• The instrument, the direction and the time. Enough to find the call again.
• The price you could actually have had. Not the published entry — the one on the side of the book you would have crossed.
• The clause that rejected it, named. "Ratio below floor", "outside chase allowance", "third position on the same driver". If you cannot name a clause, write "discretion" and be honest about it; the discretion count is one of the two numbers this log produces.
• One line of state, only if it would change the reading — already at the day's loss limit, a position open in something correlated, out of the session you trade.

Then one more field, filled in later: what the trade went on to do. This is the half that makes the exercise work, and it is the half that requires the call to have a published outcome. A skip against a source that never says what happened cannot be completed. You will have a column of things you declined and no way to find out whether declining them was right — which is a list of regrets, not a record.

Why you should not add the money up

The first thing everyone wants from this column is a total. What did my rules cost me this month? Resist it, because the number that comes out is wrong in a way that is not fixable by being more careful with the arithmetic.

A skipped trade did not merely fail to happen. Had you taken it, it would have been taken in some size, at some point in a sequence, alongside whatever else was open at the time — and each of those would have changed what came next. The position would have occupied a share of the exposure you had agreed to carry, so at least one later trade would not have been available. If it lost, the balance every subsequent size was calculated from would have been smaller. If it won, the same in reverse.

The trades you skipped are not missing from your account. They are missing from an account you do not have.

So the counterfactual is not additive, and a total pretends it is. Worse, it is a number that will be produced during a bad stretch, read as evidence that the rules are the problem, and used to justify loosening them at the exact point in the sequence where that costs the most.

Read it by clause instead

The column answers a different question well: which of my rules is earning its place, and what is each one charging me for it?

Group the skips by the clause that rejected them and look at each group separately. What you are looking for is not whether the rejected trades won or lost — it is the balance of the two within each clause, read against what the clause is for.

• A clause whose rejections mostly went on to lose is doing its job, and the group is evidence you can point at the next time you are tempted to override it.
• A clause whose rejections mostly went on to win is not automatically wrong — a filter that removes volatility is allowed to cost something — but it is charging a price, and now you know roughly what it is. Decide whether you want to keep paying it, on a day when nothing is moving.
• A clause with no rejections at all is not protecting you from anything. Either the condition never occurs, in which case delete it and stop believing you have a safeguard, or it is being overridden silently.
• A clause that rejects almost everything reaching it is not a filter, it is an off switch with a complicated description. That may be what you want; it should at least be what you meant.

Notice that none of those readings requires a count of trades or a percentage. Each of them is a shape, and a shape shows up in a small column long before a statistic would.

The cost that is not a missed winner

There is one more cost in this column, and it is the one people never look for because it does not feel like a loss.

Your attention and your risk budget are both finite and both spent in order. When a rule rejects a trade, the slot it would have occupied does not stay empty — something else goes into it, later that day, often something you would not have taken if the first one had filled. A skip is rarely a trade minus a trade. It is usually a trade swapped for a different one.

Which means the honest comparison, when a rule rejects something that goes on to work, is not against zero. It is against whatever you did instead with that share of the day. Sometimes the substitute was better. Sometimes the rule rejected a good trade and you spent the freed capacity on a worse one an hour later, and only the first half of that ever gets discussed.

You can see this in the column without doing anything clever: put the skip and the trades you did take on the same line of the same day. If a pattern exists, it will be visible as a pattern, and it will be about your day rather than about any single call.

Common questions

How do I work out what my rules have cost me overall?

You cannot, and we would rather say that than hand you a method that produces a confident wrong number. The obstacle is not effort — it is that every skipped trade would have changed the size, the sequence and the availability of the trades that came after it, so the account in which you took them all is not your account with some amounts added. Read the column by clause, look at the balance of outcomes within each group, and accept that the question "how much" has no answer while "which rule, and is it worth it" has a good one.

A rule just made me miss a big winner. Should I loosen it?

Not on that one observation, and the reason is the whole argument of this piece: that trade is memorable because it hurt, and the trades the same clause saved you from are unmemorable because nothing happened. You are comparing one vivid event against a set you never experienced. Wait until the group has more than one entry in it, read the group, and if you do change something, change a clause — a specific threshold, by a specific amount, written down with the date. "Be less strict" is not a change you can evaluate later.

Do I log the calls I was asleep for?

Yes, in a separate place. They are not judgement and they should never be mixed with the skips a rule produced. What they tell you is how much of the feed you are actually present for, and that has its own decisions attached — whether an alert is worth setting, whether a resting order makes sense for the sessions you miss, or whether a portion of what gets published is simply not available to you and your expectations should say so.

Won't keeping this just make me feel worse?

It can, and if it does, the log has been built wrong. A column that contains only the ones that got away is a gallery of regrets and you will stop keeping it within a fortnight, correctly. Two things prevent that. Record the clause, so what you are reviewing is a rule rather than a memory. And record every skip, including the ones that went on to lose — which will be many of them, and which is exactly the half your recollection deletes. Read by group and the exercise stops being about what you missed and starts being about which of your rules you can now defend.

How long before this tells me anything?

Longer than you want and less long than a statistic would need, because you are not computing an average — you are looking for a group whose entries point the same way. One clause will usually declare itself early, simply because it fires most often. The clauses that fire rarely will take a long time and that is not a defect; a rule that rejects one trade a month is also only costing you one trade a month. The first thing the log tells you is normally not about a clause at all. It is the discretion count and the overridden-skip count, and both of those are readable within a couple of weeks.

Why we published this

We publish free calls on Telegram — entry, stop and target, with what happened to each one posted afterwards, losers included — and we sell a paid room and say so openly. Our interest is in you taking the trades we post, so an article encouraging you to keep a careful record of the ones you declined is not obviously in our favour. Some of those records will say the call was not worth taking.

We are publishing it anyway because of one thing it implies about sources, and the implication is not subtle. A skip column has a field that can only be filled in by the outcome of the call. Against a channel that posts winners and quietly drops the rest, that field stays empty forever — you can record what you declined and never learn whether declining was right. The filter you built cannot be tested, and you are back to drifting.

So the test this article hands you is one you can run on us. Take a week of our calls, write down the ones your own rules reject and why, and fill in the outcome column from what we publish. Whatever the column says about your filters, the fact that it can be completed at all is the thing to check first — in a source, and in everything else you take calls from.