Break-Even Trades: Win, Loss, or Ignore? What Each One Does to Your Win Rate

A break-even trade moves no money. It is the one result in your journal with no consequence for your account, which is exactly why nobody thinks hard about it.

Then you have to put it somewhere. Wins, losses, or neither. And depending on which box you choose, the same 100 trades report a 40% win rate or a 50% win rate, a reward-to-risk of 1.92 or 2.88, and a current streak of six losses or one win.

Nothing about your trading changed. Only the bookkeeping did.

This post works all three conventions on one sample, shows which statistics move and which cannot, and then makes the case that the argument is usually about an almost-empty bucket.

The Sample

One hundred trades from one setup, risking $250 a trade:

• 40 winners, averaging +$600, for $24,000 of gross profit
• 50 losers, averaging -$250, for $12,500 of gross loss
• 10 scratches at exactly $0

Net result: +$11,500. That figure is the anchor for everything below. It does not move, cannot move, and is not a matter of opinion. Every disagreement in this post is a disagreement about how to describe $11,500, not about whether it is there.

The Three Win Rates

Ten points of spread. That is the number everyone quotes when this comes up, and it is real. It is also the least interesting thing the convention does.

Notice that the three modes change the fraction in two different ways. Ignoring scratches shrinks the denominator and leaves the numerator alone. The other two keep the denominator at 100 and move trades in or out of the numerator. That distinction is what causes the second-order effects nobody checks.

What Else Moves, and Why Nobody Notices

Here is the part that almost no article about break-even trades covers.

If you count a scratch as a win, it does not sit quietly in the numerator. It joins the pool of winners you average. And it is a zero, so it drags the average winner down.

Counting ten scratches as wins spreads $24,000 of profit across 50 winners instead of 40, so the average winner reads $480. Counting them as losses spreads $12,500 of loss across 60 losers instead of 50, so the average loser reads $208.33 and your reward-to-risk looks better.

So the two most-quoted numbers in retail trading, win rate and reward-to-risk, both move, and they move in opposite directions:

• Counting scratches as wins: win rate up 5.6 points, reward-to-risk down 0.48.
• Counting scratches as losses: win rate down 4.4 points, reward-to-risk up 0.48.

Which means you cannot improve both by choosing a convention, and anyone quoting a flattering pair of numbers has picked one of them and hoped you would not check the other. If you are working out what your target is really doing to your hit rate, the win rate post runs the same trade-off on exits rather than bookkeeping.

What Cannot Move

Three numbers are immune, and understanding why is the whole argument.

Total P&L. $11,500 in every mode. A zero adds zero.

Profit factor. Gross profit is $24,000 and gross loss is $12,500 whichever box the scratches sit in, because a trade at exactly zero adds nothing to either sum. Profit factor is therefore identical under all three conventions while the win rate swings ten points, which is precisely what makes it the tiebreaker when two win rates disagree. The profit factor post works that property out in full.

Dollar expectancy, as long as the scratches stay in the denominator:

• Ignore: (0.4444 x $600) - (0.5556 x $250) = $266.67 - $138.89 = $127.78
• As wins: (0.50 x $480) - (0.50 x $250) = $240.00 - $125.00 = $115.00
• As losses: (0.40 x $600) - (0.60 x $208.33) = $240.00 - $125.00 = $115.00

The two conventions that argue hardest with each other produce exactly the same expectancy. That is not a coincidence in this sample, it is an identity. Expectancy is total P&L divided by the number of trades in the denominator, and both of those modes keep the denominator at 100. Check it: $115.00 x 100 = $11,500.

Ignoring scratches gives $127.78, and that is also exactly right for the question it answers: $11,500 across 90 resolved trades. Multiply it out and you get the same $11,500. It is only wrong if you then use it to forecast the next 100 trades, because 10 of those will be scratches and $127.78 x 100 is $1,278 of profit that will not arrive.

That is the honest summary of the whole debate: the convention cannot change what you made. It changes what you predict you will make, and only through the denominator. The expectancy post has the formula and its failure modes.

The Streak, Which Moves More Than Anything Else

Streak counters read backwards from your most recent trade, and a scratch interrupts the read.

Take your last six trades, most recent first: a scratch, then five consecutive losses.

• Ignore: the scratch is skipped, the counter reads straight through it. Five losses.
• Count as losses: the scratch is a loss, and the run extends. Six losses.
• Count as wins: the most recent trade is now a win, so the losing run has already ended. One-trade winning streak.

You are six trades into a drawdown and the counter says you are on a winning streak.

This is the strongest practical argument against the "count them as wins" convention, and it has nothing to do with vanity. A streak counter exists to tell you when to size down or stop. Setting it so that every stop-to-break-even resets the warning defeats the only job it has. If you move stops to break-even under pressure, the convention that flatters you is also the one that switches off the alarm at exactly the moment it should be loudest.

The Cumulative R Trap, In Our Own Field

An honest disclosure about how this journal stores R, because it is the one place where the convention can invent a number that never existed.

When you type a trade into the form with an entry, a stop and an exit price, the R-multiple is derived from what actually happened: the distance from entry to exit divided by the distance from entry to stop. A trade that exits at its entry price gives a distance of zero, so it carries 0.00R. Count that scratch as a win and cumulative R gains nothing. Correct.

But if you pasted the trade in from the TradingView indicator, the R-multiple field holds the planned ratio, entry to target over entry to stop, and the form deliberately stops overwriting it. That is the right behaviour for a planned figure. It means a scratch logged that way can carry 2.00R while having returned $0. Count it as a win, and cumulative R gains 2.00R for a trade that paid nothing.

So the rule is narrower than "pick a convention": if your R column mixes planned and realised values, counting scratches as wins converts that mixture into phantom R. The R-multiples post covers the planned-versus-realised problem in general, and this is its sharpest special case.

Now the Part That Refuses the Finding

Everything above assumes you have a meaningful pile of break-even trades. You probably do not.

The test for a scratch is exact equality with zero. Not "close to zero". Not "within a few dollars". A trade at plus 40 cents is a winner and a trade at minus 40 cents is a loser. So the bucket only fills when your exit lands precisely on your entry and your P&L column is gross.

Two things push trades out of it:

Costs. A trade that exits at your entry price is flat before costs and negative after them. At 1.2 pips of round-turn cost against a 20-pip stop, that is 0.06R per trade, or $15 on $250 of risk. Ten scratches cost $150 that the journal records as $0. If your P&L is net, those ten trades are not scratches at all. They are ten small losses, and your "scratch convention" question was moot.

Partial fills and near misses. A stop moved to one pip above entry, a trade closed manually a few seconds before the level, a swap charge on an overnight hold. All land near zero, none land on it.

Which leads to the test worth running instead of the argument worth having:

Sort your trades by absolute P&L, ascending. Count how many sit within 0.1R of zero in either direction. That band, not the bucket labelled break-even, is the size of your break-even-stop habit.

If that band holds 30 trades and your exact-zero bucket holds 3, then the convention debate was about 3 trades and the thing you actually wanted to measure was sitting next to it the whole time, filed as ordinary small wins and small losses.

And measuring that band is a different question from counting it. A cluster of near-zero exits is a habit with a cause, usually a stop moved to entry because the trade went against you for a moment. That belongs in the rules half of the journal rather than the statistics half. Why you break your own trading rules sorts a repeatedly broken rule into four causes, only one of which is discipline.

So Which One Should You Pick

Derived from the above rather than asserted:

Default to ignoring scratches when you are comparing setups, symbols or sessions against each other. It answers the cleanest question, how often the trade paid when it resolved, and it keeps the denominator honest across slices that scratch at different rates.

Switch to counting them as losses when you are auditing whether your break-even-stop habit is costing you. It is the conservative reading, it matches what your account felt, and it stops the habit from hiding. Prop firm traders working against a drawdown limit should live here, because a scratch consumed a day and paid nothing.

Avoid counting them as wins as a standing default. It is the only convention that simultaneously inflates your win rate, deflates your average winner, resets your losing-streak counter and, if your R column carries planned values, invents R. There is one narrow case for it, which is measuring capital preservation specifically: how often did I get out without damage. Set it deliberately for that question, then set it back.

And whichever you choose, the rule that matters more than the choice: write it down and do not change it mid-analysis. Switching conventions between two comparisons is how a trader accidentally proves whatever they were hoping for.

One more thing worth doing this week. Open your journal and find out which convention it already applied to the win rate you have been quoting for the last six months. Most journals do not document it, and searching their help pages in September 2026 does not turn up an answer. The 60-second test: note your win rate, then delete or temporarily flag one break-even trade and watch the number. If it moves up, scratches were being counted as losses. If it moves down, they were counted as wins. If the win rate holds and only the trade count drops, they were being ignored.

How This Works in TradingSFX

The break-even convention is a toggle in the dashboard toolbar rather than a hidden assumption. It cycles through Ignore, Win and Loss, it remembers your choice, and every statistic on the page recomputes: win rate, average win, average loss, reward-to-risk, expectancy, profit factor and the streak counter. You can watch the ten-point swing happen and pick a convention on purpose instead of inheriting one. That is on the free plan.

Two behaviours worth knowing because they follow directly from what this post worked through. Profit factor holds steady across all three positions of the toggle, which is the property that makes it useful when two win rates disagree. And recent-trade lists always show your break-even trades whatever the toggle says, because the setting governs the arithmetic rather than what you are allowed to see.

For the near-zero band, the trades tab sorts and filters on P&L, so you can isolate everything inside a tight range around zero and read the notes on those trades together. On Pro and above, the same question can be asked per symbol and per strategy, which is where a break-even-stop habit usually turns out to be concentrated rather than spread evenly.

Basic is free forever at 10 trades a month with no card. Pro is $19.99/mo or $215.99/yr, Premium is $29.99/mo or $323.99/yr, both with a 7-day trial.

Bottom Line

The convention cannot change what you made. It changes what you report, and it changes more of the report than you think: win rate, average win, average loss, reward-to-risk and your streak counter all move, while total P&L, profit factor and dollar expectancy do not.

Ignore scratches for comparing, count them as losses for auditing, and do not leave the setting on "win" while judging an exit rule.

Then go and check the band within 0.1R of zero. That is where the trades you were arguing about have been sitting all along.

Published September 23, 2026. Every figure in this post is arithmetic worked openly from a stated illustrative 100-trade sample. No competitor pricing, feature claims, survey data or study results are quoted. TradingSFX plan details and the break-even calculation behaviour were verified against the application code on 23 September 2026.