Which rule breaks first — your clocks, not the comparison table

An account is not one rule. It is a small set of clauses, each with its own trigger, and each of them is a clock: given a trader's numbers, there is an expected number of trades before it first fires. The account is settled by whichever clock runs out first, and that ordering is nowhere in the rule book, because the ordering depends on numbers the rule book does not have.

So the useful question is not what are the rules. It is in what order do they arrive. That is a ranking problem with a definite answer, and the answer changes from trader to trader while the rules stay exactly where they are.

Take a worked profile. A trader wins 44% of the time, makes 2.2R on a winner and gives back 1R on a loser, risks 0.5% of the account per position and takes six trades a day. That is an expectancy of 0.408R per trade — a real edge, not a marginal one. The account allows a 7% loss from its high-water mark, a 2.5% loss in any one day, requires 9% of profit to pass, and caps any single day at 40% of total profit.

The clause in the offer's headline — the loss allowance, the number every comparison table sorts on — is last by an order of magnitude. It sits 18.1 times further out than the clause immediately behind the leader, and across the whole horizon it is touched at all in only 54.7% of runs.

For this trader the finish line is the nearest clause, which is another way of saying the account is survivable. Nothing about that was visible in the rule set until the four clocks were put on the same axis.

The four clauses split into two families, and the families need different mathematics.

The clauses that live inside a day

A daily loss limit and a consistency cap are both settled inside one trading day and then forget it. Whether today breaches has nothing to do with yesterday, so the waiting time is geometric over days and the whole clause collapses to a single per-day probability. That probability is computed exactly: every ordering of a day's trades is walked, the running total is checked against the limit at each step, and the orderings that breach are added up. With six trades in a day there are two to the power of six of them, which is a small enough number to enumerate outright rather than estimate.

The clauses that remember everything

A loss allowance and a profit target are different animals. Both are first-passage problems: the account carries a running state, and the clause fires the first time that state touches a line. Neither has a closed form for a general win and loss pair, so both are solved on a lattice — the distribution of the state is pushed forward one trade at a time and the surviving mass is added up, which is exactly the expected number of trades before the line is touched.

The state for the allowance is the distance below the account's high-water mark. A win reduces it and cannot push it below zero; a loss deepens it. Written that way it is a random walk reflected at its running maximum, and the recursion is the one Lindley wrote down for a queue with the sign of the increment reversed. That is not a metaphor; it is the same line of algebra.

Why every clock is censored

A clause that a trader may never meet has no finite expected waiting time, and a readout that says "infinity" tells nobody anything. So every clock here is censored at a horizon of 1,200 trades — about 200 trading days at the profile's pace — and reported as the expected number of trades to fire or the horizon, whichever comes first. A clock sitting at the horizon is not a prediction that the clause fires there. It is the statement that it does not arrive inside the horizon at all, which is why the probability of ever reaching it is printed beside it.

The one clause that does not end anything

The consistency cap is the odd member. It does not close an account; it moves the finish line. The moment a single day closes above the cap's share of the posted target, the posted target stops being the real one, because total profit must now grow until that day is a small enough slice of it. Where that threshold sits and what a consistency cap does to a posted target is already worked out on this site. Here it is used as given: the clause fires the first time a day closes above 3.6%, and the clock counts how long that takes.

Two of the four clauses are memoryless and two are path-dependent. Ranking them means putting a geometric waiting time and a first-passage time on the same axis — legitimate, because both are measured in trades, and necessary, because the account does not care which kind of clause reaches it first.

Below is the whole method as an instrument. Set the five numbers you bring and the four the account publishes; the ladder re-sorts itself and the clause at the top is the one that decides the account. Nothing is precomputed and nothing is looked up.

Read the lead before the order. The lead is how many times further out the runner-up sits, and it is the only honest measure of how much the ranking is worth. A lead near one means two clauses arrive at about the same time and the order between them is noise; a lead of three or more means one clause genuinely owns the account. Expected arrival is also not the same thing as probability of arriving first — the clock with the shortest mean usually wins that race, but two clocks with the same mean and different spreads do not.

The last readout is the one worth watching. It is the chance the loss allowance is reached at all inside the horizon. On a great many settings it is the smallest number on the page, attached to the largest number in the offer.

Hold the rule set completely still and vary only the trader. The map below is the same four clauses, computed 90 times across position size and win rate, with each cell lettered by whichever clock is shortest there.

Four different clauses lead in four different regions of one rule set. The profit target leads where size is small and the edge is real; the consistency cap takes over through the middle; the daily limit owns the low-win-rate band, where losing days come often enough to matter. The loss allowance leads in one cell of 90, in the corner where the smallest size meets the weakest edge — the only place on the map where nothing else gets there first.

The same exercise, one input at a time, gives the second figure. Each ribbon is a single input swept from its floor to its ceiling with everything else held at the declared profile, divided into the stretches over which one clause leads.

The account's headline number changes the answer one time across its entire published range. The five numbers the trader supplies change it twenty times between them.

The position-size ribbon is worth stopping on, because its steps are not gradual. The daily clause is decided by how many whole losses fit inside the daily limit, and a count of losses is an integer, so its clock does not bend as size rises — it drops in flights. At 0.4% risk a full day of losers costs 2.4%, which the limit will not notice: the clause cannot fire at all. Add 0.05% to the size and the worst day becomes 2.7%, the limit becomes reachable, and the clock lands at 194.2 trades. Across the whole size range that one clock takes nine distinct values, not a curve.

A change in position size too small to notice on a ticket can move a clause from impossible to live. That is a property of integers, not of markets, and it is invisible to anyone reading the rule set rather than computing against it.

Every clock above rests on assumptions worth naming, because each of them can move the answer.

• Trades are independent and identically distributed. Real trading is neither. Streaks, regime changes and a trader's own behaviour after a loss all break it, and every clock here would move if they were modelled.
• A trade has two outcomes, not a distribution. The model gives every winner the same size and every loser the same size. That is hardest on the two per-day clauses, which depend on the shape of a day's tail rather than its average, and it will understate both very good and very bad days.
• The loss allowance is measured from the account's high-water mark. If it is measured from the starting balance instead — the difference between a floor that moves and one that does not — the clause fires no sooner, so the clock lengthens and the allowance falls even further down the board.
• The daily limit is treated as touched intraday and as ending the evaluation. Where an account only locks the day, the same clock still tells you how often a day is lost, but the ranking against terminal clauses is no longer like for like. What a daily limit is measured on, and when a day begins, is a separate question this page does not answer; the site's glossary states the usual case and stops there.
• Clauses are computed independently. Each clock assumes it is the only clause in the account. Comparing them is what identifies the leader, but an account killed by one clause never goes on to meet another, so the clocks behind the leader are counterfactual.
• Expected arrival is not probability of arriving first. Ranking by mean is the right first cut and it is not the whole answer; where two clocks are close the lead readout says so rather than resolving it.
• The horizon is a modelling choice. Lengthen it and clauses that read at the horizon move; the ordering of the clauses that fire well inside it does not.

None of that makes the ranking useless. It makes it a ranking of a model, which is the only kind of ranking anyone has, and the honest use of it is comparative: which clause leads, by how much, and what has to change before a different one does.

The clocks above run on one fixed rule set. Real rule books do not agree on which clocks exist. Three evaluations — an intraday-trailing account and an end-of-day account from Apex, and Topstep's Trading Combine — carry the same $100,000 size, the same $6,000 profit target (6%) and the same $3,000 loss allowance (3%). On the headline row of any comparison table they are identical. They are not the same account.

Read down the columns and the four-clock model thins out. The intraday account runs only two clauses that can end it, and its allowance trails the account's highest equity inside the day, so the same $3,000 sits closer than a figure fixed once at the close. The end-of-day account adds a daily limit, but breaching it pauses the session rather than failing the account — a clock that costs a day, not the evaluation. The Combine carries every clause, yet its daily limit is optional, and its consistency cap, when the best day runs long, lengthens the target clock rather than firing one of its own.

So the bottom row is the same on all three: a target and a trailing allowance are the only clauses that end any of them. That is still the wrong place to stop. The two surviving clocks are shaped differently on each account — an intraday trail is not an end-of-day trail — and where they rank against each other is decided by the same five numbers as before: win rate, average win, average loss, risk per trade and trades per day. A comparison table cannot rank these three accounts. The instrument can, on your numbers.

Rank the clocks on your own numbers — free, in the browser →

Is this the same thing as risk of ruin?

No, and the difference matters. Risk of ruin is a probability: the chance a trader is eventually eliminated. A clock is a waiting time: how long until a particular clause first fires. The two can point opposite ways, because a clause with a short expected arrival may be almost certain to fire while a clause with a long one may hardly ever fire at all, and it is the pair together — arrival and reach — that describes an account.

Why is the profit target treated as one of the clauses?

Because it ends the evaluation too. An account stops when any of its lines is crossed, and one of those lines is the one you are aiming at. Leaving the target out of the ranking would mean asking which rule kills you while ignoring the possibility that you finish first, which is precisely the case a competent trader is trying to be in.

What if my drawdown allowance is measured from the starting balance rather than the high-water mark?

Then the clause arrives later, never sooner. A loss measured from the start can only reach the allowance when the peak is at the start, and the peak never goes down, so every path that breaches a fixed allowance also breaches a trailing one at the same trade or earlier. The effect on the ranking is to push the allowance further down the board, not up it.

Does the ranking change if the only thing I change is position size?

Repeatedly. Size scales every clause at once but not at the same rate, and one of them does not scale smoothly at all: the daily clause depends on how many whole losses fit inside the daily limit, which is a count, so it steps. A change in size too small to matter on a single ticket can take that clause from unreachable to the shortest clock on the ladder.

What does it mean when a clock reads at the horizon?

That the clause does not arrive within the horizon, not that it arrives there. The clocks are censored so that a clause a trader may never meet still produces a finite number, and the probability of ever reaching it is reported beside it. Two clauses both sitting at the horizon are tied only in the sense that neither is the account's problem.