What does a break-even stop actually do?
A break-even stop moves the stop loss to the entry price, sometimes plus a small buffer for costs, once the trade has moved a stated distance in its favour, and it converts every trade that reaches that distance and then returns to the entry into a scratch rather than a loss or a win.
The trigger distance is the parameter. A trigger at half an R moves the stop early and catches many trades; a trigger at two R moves it late and catches few. Whatever the trigger, the rule partitions the trades that reached it into three groups: those that went on to the target, those that came back to the entry and were scratched, and those that would have come back to the entry and then gone on to the target had the stop not been moved.
The first group is unaffected. The second group are the losses the rule is meant to prevent, and it does prevent them; those trades would have retraced through the entry to the original stop. The third group is the cost: winners that the original rule would have taken and the break-even rule gave away.
The whole question is the size of the second group against the third, and it is not obvious in advance which is larger.
Why does it usually cost expectancy?
A break-even stop usually costs expectancy because a return to the entry price after an initial favourable move is one of the most common paths a winning trade takes, so the stop at entry is placed exactly where the market is most likely to visit before continuing.
The intuition that fails is that a trade which has gone in your favour and come back is a trade that has failed. Very often it is a trade that is pulling back inside a move that will resume, and the entry price has no significance to the market. Placing the stop there is placing it at a level chosen for the trader's comfort rather than for any feature of price, and levels chosen that way are hit by noise.
The arithmetic follows. Suppose the original rule wins forty percent of the time at two R. Move the stop to entry at one R, and some fraction of those winners are scratched on the way. If a third of the eventual winners pull back through the entry before reaching the target, the win rate falls to about twenty-seven percent, and the scratches from prevented losses have to be worth more than that to make up the difference. On many rules they are not.
The rule can also interact badly with the stop distance. A break-even trigger inside the noise band of the instrument is reached and then reversed through constantly, and the strategy becomes a machine for producing scratches that each pay the spread.
When can it help?
A break-even stop can help when the rule's losers are fat-tailed, so that the losses it prevents are large, when the trigger is tied to a structural level rather than to a round number of R, and when an external constraint such as a prop-firm drawdown rule makes avoiding a full loss worth more than its expectancy cost.
The fat-tail case is real. A rule whose losing trades occasionally run to several times the intended stop, because of gaps or slippage, loses a great deal on those trades, and a break-even move that happened to be in place before the adverse move prevents some of them. Whether it prevents enough to pay for the scratched winners is again a count.
The structural case changes the rule. Moving the stop to just beyond a swing that has formed since entry, which may or may not be near break-even, is a stop placed at a level price has already respected. That is a different rule from moving to the entry price, and it tends to test better because the level means something.
The constraint case is legitimate and it is not about expectancy. A trader whose account is subject to a daily loss limit may rationally accept a lower expectancy for a lower probability of a full-sized loss, and the test for that trader is against the constraint, not against the unmanaged rule.
How do you test a break-even rule?
Test it by running the same entries with and without the break-even move across a range of trigger distances, grading in R after costs, resolving the trigger bar and the stop-out bar pessimistically, and reading the scratch rate alongside the expectancy.
The scratch rate is the diagnostic. A rule whose scratches are mostly prevented losses is doing what it claims; a rule whose scratches are mostly given-away winners is costing the strategy its tail. The way to tell them apart is to look at what the unmanaged version of each scratched trade would have done, which the comparison on the same entries provides directly.
The intrabar problem applies here as it does to trails. On the bar where the trigger is reached, a tester that moves the stop off the bar's favourable extreme and then tests it against the adverse extreme can book a scratch on a trade that the adverse extreme had already stopped out at the original level. The pessimistic ordering, adverse extreme first, is the honest one.
The trigger distance is a parameter surface like any other, and the useful reading is whether there is any region of it in which the managed rule beats the unmanaged one on the out-of-sample segment. Often there is not, and that is a finding.
How does QuantParadox grade break-even rules?
QuantParadox grades a break-even move as a trigger stated in R, applies the pessimistic intrabar rule on the trigger bar so that a favourable move inside the bar cannot rescue a trade the adverse move had already ended, and records each trade's maximum favourable and adverse excursion so the scratch rate can be read against what the unmanaged trade would have done.
Running the same rule with and without the move is two runs on the same history, and the variant ledger counts them, so the better of the two is graded as a selection rather than as a single result.
The limitation is the one that applies to all managed exits on the platform: the stop moves once per bar on the entry timeframe unless finer data is consulted, so a break-even move that in live trading would have been placed mid-bar is placed at the bar's pessimistic point. That makes the platform's break-even results conservative relative to a tick simulation.
The Reconciliation view shows the scratched trades as a group, which is the quickest way to see whether they were prevented losses or given-away winners for a specific rule.