Sixty trades into the quarter, your account is up $900 and you feel like a trader with an edge. Maybe you are. But a lucky streak on a losing system looks identical from the inside, and so does an unlucky streak on a winning one. Feelings cannot separate the cases. One formula can, and it runs on numbers already sitting in your journal.
The formula, worked
Expectancy is the average value of one trade over the long run: win rate times average win, minus loss rate times average loss. Take a real-shaped sample of 60 trades: 27 winners averaging $170, 33 losers averaging $90. Win rate 45 percent, loss rate 55 percent. Expectancy = 0.45 × $170 minus 0.55 × $90 = $76.50 minus $49.50 = $27 per trade. Every trade you take under this system is worth $27 on average, including the losers. Sixty trades should deliver around $1,620, and the $900 quarter above suddenly has context: positive system, slightly unlucky sample, keep going. A negative number flips the verdict: every additional trade is a purchase of loss, and more effort makes things worse until the system changes.
Do it in R
Dollar expectancy drifts whenever your size changes, so the durable version uses R, the risk per trade. Same sample: average winner 1.7R, average loser 0.9R (some losers cut early). Expectancy = 0.45 × 1.7 minus 0.55 × 0.9 = 0.27R per trade. On a $10,000 account risking $100, that is the same $27, but the R figure survives account growth, drawdown rebasing, and comparisons between instruments. As a working benchmark, a sustained expectancy of 0.2R or better after costs is a genuinely tradeable intraday edge. Plenty of profitable careers run on less with high trade frequency, but 0.2R gives variance enough cushion to be forgiving. Keep costs inside the numbers by logging every result net of spread and commission, since an edge that exists only before costs is not an edge.
Sample size, or why 20 trades lie
Expectancy computed on 20 trades is a rumor. With a true 45 percent win rate, 20-trade samples routinely print anywhere from 25 to 65 percent winners through pure chance, and the calculated expectancy swings from strongly negative to fantasy-positive along with them. At 50 trades the picture is still blurry. Around 100 trades per setup, the numbers begin to mean something, and the streaks and outliers have had room to show up. Two practical rules follow. First, do not redesign a system on the evidence of a bad week; twenty trades cannot convict it. Second, compute expectancy on a rolling window of the last 100 trades rather than a lifetime average, so the number reflects the trader and market of today, not last year. If a setup is rare, three instances a week, accept that judging it will take most of a year, and weight your trading toward setups that generate evidence faster.
Read it per setup, not per account
A single account-level expectancy hides the most useful information you own. Tag every journal entry with its setup and session, then split the calculation. A common discovery looks like this: London pullback longs, 80 trades, +0.5R expectancy; New York breakout trades, 60 trades, minus 0.15R. The account total looks mediocre, but the truth is one excellent system subsidizing one quiet leak. Dropping the breakout trades raises expectancy, lowers stress, and costs nothing. This is the cheapest performance improvement in trading: no new indicator, no new market, just arithmetic on records you already keep, pointed at the question of which of your trades deserve to exist. Frequency then completes the picture, because expectancy is per trade and income is expectancy times trade count: 0.27R across 40 trades in a month is about $1,080 on this account, while the same edge traded 15 times is $405. Treat that multiplication as a warning label as much as a promise. Forcing extra trades to feed it dilutes the very number being multiplied, so grow frequency only by finding more valid instances of setups that already test positive, never by loosening the definition of valid.