When Exits Run Late, the Break-Even Win Rate Moves Before the Trader Does

A position held past the point where it was supposed to be closed is usually filed under discipline. That label is not wrong, but it is close to useless, because it describes the trader rather than the position. A more tractable description is arithmetic. A trade closed later than planned is a trade whose realised loss is larger than the loss the position was sized for, and the win rate needed to break even moves by a specific, computable amount when that happens. The amount can be worked out before the order is entered, from two numbers that are already on the ticket.

The structural claim here is narrower than the psychological one and easier to check. An exit condition that was not written before the order existed cannot be written while the position is open. What is available mid-trade is a decision, not a rule. A decision made against a live mark is a different object from a condition checked against a live mark, and only one of the two can be specified in advance. The practical consequence is that a trade carrying only two exit conditions, a stop and a target, has no defined behaviour in the region between them, and that region is where most of the holding time is spent.

Why the Exit Rule Cannot Be Written While the Position Is Open

Writing a rule means specifying a condition over states that have not happened yet. The specification has to be complete enough that checking it later requires no further judgment. Three things are true in the window before the order exists that stop being true the moment it fills: there is no mark moving against capital, there is no deadline attached to the next tick, and there is no exposure already committed. Those three absences are what make the pre-entry window the only place a condition can be authored.

Once the position is live, the same question looks like classification rather than specification. The trader is no longer asking what should happen if price does X; price is doing something, and the task has become deciding whether the thing it is doing counts. That task is performed under time pressure, against a number that is changing, with capital already at risk. It produces answers, but it does not produce rules, and the answers it produces are not reproducible across two similar trades.

This is why an intention such as "close it if the setup stops working" does not survive contact. The phrase is not evaluable without a position in front of it, which means it can only be evaluated in the window where evaluation is least reliable. A rule that can only be checked in the worst available conditions is not a rule; it is a plan to improvise later.

The Break-Even Win Rate Has a Closed Form

Let the planned stop distance be one unit of risk, and let R be the target distance measured in the same unit. Let L be the loss realised on losing trades, also in that unit. Expected value per trade is p*R - (1-p)*L. Setting that to zero and solving gives the break-even win rate:

p* = L / (R + L)

When every loss is taken at the planned stop, L = 1 and the expression collapses to p* = 1 / (1 + R). The four values below are exact, not rounded estimates of a simulation:

Target size (R)	Break-even win rate, loss taken at plan
1.0R	50.00%
1.5R	40.00%
2.0R	33.33%
3.0R	25.00%

Nothing in that table says which R is attainable. The frequency with which a given target is reached is an empirical property of a setup and an instrument, and this identity has no opinion about it. What the identity does provide is a fixed reference point: for any pair of stop and target, there is one win rate below which the arithmetic is negative regardless of how the trades felt.

What a Late Exit Costs, in Percentage Points

Now let the loss run past the planned stop by a factor k, while the target is left where it was. The break-even win rate becomes p* = k / (R + k). The table below is the same four target sizes under three loss outcomes.

Target size	Loss at plan (k=1.0)	Loss 1.5x plan	Loss 2.0x plan
1.0R	50.00%	60.00%	66.67%
1.5R	40.00%	50.00%	57.14%
2.0R	33.33%	42.86%	50.00%
3.0R	25.00%	33.33%	40.00%

Expressed as movement rather than level, a loss taken at 1.5 times the planned stop raises the required win rate by 8.33 to 10.00 percentage points across those four target sizes. A loss taken at twice the planned stop raises it by 15.00 to 17.14 percentage points. At a 2.0R target specifically, break-even moves from 33.33% to 42.86% at 1.5x and to 50.00% at 2.0x.

The second row is worth reading slowly. A 1.5R target with losses running to twice the stop requires a 57.14% win rate simply to avoid losing money. Win rates in that range are not impossible, but a strategy that was designed around a 40% hit rate and now needs 57% has not been degraded at the margin. It has been moved into a different category.

A Minority of Late Exits Is Enough

The tables above assume every loss runs long, which is not what happens. A more realistic construction is that a fraction f of losing trades exit late while the rest exit at plan. The mean loss is then L = (1-f) + f*k, and the same closed form applies to that mean.

Take one losing trade in five running to twice the planned stop, so f = 0.2 and k = 2. Mean loss is 1.20. At a 2.0R target the break-even win rate moves from 33.33% to 37.50% — a shift of 4.17 percentage points, which looks survivable.

It is not, and the reason is that expectancy is a small difference between two larger quantities. Hold the win rate fixed at 40% with a 2.0R target. With every loss taken at plan, expectancy is 0.40*2 - 0.60*1 = +0.200R per trade. With one loss in five running to twice the stop, expectancy is 0.40*2 - 0.60*1.20 = +0.080R per trade. Sixty per cent of the per-trade edge is gone, and the win rate never changed.

Over a hundred trades at those parameters, the difference is 40 wins at 2R against 60 losses. All exits at plan gives 80R - 60R = +20R. Twelve of those 60 losses running to 2R gives 80R - (48*1R + 12*2R) = 80R - 72R = +8R. The margin above break-even falls from 6.67 percentage points to 2.50. Twelve trades out of a hundred did that.

The Third Exit Condition Is a Bar Count, Not a Price

A stop covers the adverse case. A target covers the favourable case. Between them sits every trade that does neither, and for that population there is no condition at all — which is exactly the population that gets closed by judgment, late, at whatever the mark happens to be when attention runs out. A time stop is the third condition, and its defining property is that it is checked against a counter rather than a price.

On a tick-based chart the counter is bars, not minutes, and that distinction matters. A 144-tick bar closes after 144 trades have printed, so a bar count measures traded activity rather than elapsed clock time. A limit of thirty bars is a shorter wall-clock leash when the tape is busy and a longer one when it is thin. A clock-based limit behaves in the opposite direction. Neither is inherently better; they are different measurements, and a plan that mixes them without saying which is being used has an ambiguity in it.

A time stop is not free. It truncates the loss distribution at something smaller than the full stop, which lowers mean loss and pulls the break-even rate down. It also closes trades that would have reached the target later, which lowers the win rate and pushes the break-even rate back up. Both effects run through the same identity, p* = L / (R + L), so the question of whether a given bar limit helps is empirical and specific to one trader's own record. It is answerable by counting, and it is not answerable by argument.

Settlement and Capital Sit on a Separate Clock

Closing a position and having the capital back are two events, not one. Under 17 CFR 240.15c6-1(a), a broker or dealer may not enter into a contract for the purchase or sale of a security that provides for payment of funds and delivery of securities later than the first business day after the date of the contract. That is the T+1 standard, effective for applicable transactions on and after 28 May 2024, replacing the previous T+2 cycle. Paragraph (c) of the same rule carves out securities priced after 4:30 p.m. Eastern in a registered firm commitment offering, which settle on the second business day.

The relevant point for exit planning is that a settlement cycle is a hard exogenous parameter, unaffected by how good or bad the exit was. A trade closed late does not settle late; it settles on the same schedule as one closed on time. Where lateness interacts with capital is upstream of settlement, in whatever intraday equity constraint the account operates under.

That constraint is currently changing. FINRA Regulatory Notice 26-10 adopted new intraday margin standards that replace the day trading margin requirements in their entirety, including the day trade count used to designate a customer a "pattern day trader" — four or more day trades within five business days — and the $25,000 pattern day trader minimum equity requirement. The effective date is 4 June 2026, forty-five days from publication of the Notice, and members needing more time may phase in implementation over eighteen months, until 20 October 2027. Any exit plan built around counting day trades against a five-day window is reading a rule that is being withdrawn, and during the phase-in two firms may be operating under different regimes on the same date.

What Would Invalidate This

• Variable position size. The identity assumes one risk unit per trade. If size varies trade to trade, R and L are not commensurable across trades and the tables do not apply; the same calculation has to be redone in currency rather than in R.
• An unexecutable stop. The whole argument treats k greater than one as a consequence of a missing condition. Gaps, halts and thin books produce the same k with no decision involved. In that case a late exit is not the cause of the fat tail, and adding a time stop does not remove it.
• Give-back rather than run-past. If the exit was late because the target was reached and then surrendered, the outcome is not a loss of k times the stop and none of these tables describe it. That case needs a separate computation on the distribution of give-back, which is a different measurement.
• Correlated trades. Per-trade expectancy assumes outcomes are independent. Repeated entries into one setup during a single regime are not, and expectancy computed this way will understate drawdown even where it correctly states the mean.
• Costs are excluded throughout. Commissions, financing and slippage raise every break-even figure in both tables. The direction is certain; the size is specific to the instrument and the account.
• Instrument scope. Rule 15c6-1 governs securities transactions effected by brokers and dealers. Futures, spot currency and other markets settle under other regimes entirely, and nothing in that paragraph transfers to them.

Concrete Framework

1. Write three numbers before the order exists, not two. Stop distance, target distance, and maximum bar count. All three go somewhere readable without looking at the chart.
2. Convert stop and target into R and compute the break-even rate. p* = 1 / (1 + R). Record it on the same line as the trade, before the fill.
3. Record realised loss in units of the planned stop. One number per losing trade: the value of k. This single column is the entire measurement the rest of the framework runs on.
4. At the end of a sample, compute mean loss. Average k across losing trades to get L, then recompute p* = L / (R + L).
5. Compare the two break-even numbers against the observed win rate. The distance between the planned p* and the realised p* is the cost of late exits, denominated in the same units as the edge itself.
6. Count the trades closed by the bar limit. If the count is zero, the limit is set high enough to be doing nothing and is not a condition. If it is closing most trades, the limit is substituting for an entry filter and the problem is upstream.
7. Re-read the two external constraints before the next change in size. The settlement cycle applicable to the instrument, and the account's own implementation date for the intraday margin standards during the phase-in window that runs to 20 October 2027.

None of this makes a late exit feel different in the moment. It makes it countable afterwards, which is the only property that lets it be reduced. The trade that should have been closed an hour earlier is not evidence of a character flaw; it is evidence that the position had two exit conditions where it needed three, and the cost of the missing one is a number, not a mood.