Why an Expectancy Tile in Percent Understates a Sized Book

The expectancy tile on the Performance overview reads -1.20% against a total trades tile reading 87. That is a per-trade average, and per-trade averages have one property that matters more than anything else about them. Every trade counts once, regardless of how much money was on it.

If you always trade the same size, that is fine and the percentage is a faithful summary. Almost nobody always trades the same size. Sizes drift with conviction, with account balance, with what else is open, and with whether the last trade went well. The moment sizes vary, the percentage tile and the currency reality separate, and the direction they separate in depends entirely on whether your big trades were your good ones.

What the tile is averaging

Expectancy in percent is the mean per-trade return expressed as a percentage. Add up the percentage outcome of every closed trade in scope, divide by the count. The 87 in the total trades tile is that count, and the venue table lower down accounts for all of it, with Hyperliquid at 70 trades, Alpaca at 16 and Manual at 1.

That construction has a specific consequence. A trade risking two hundred dollars that lost 3% contributes the same -3 to the average as a trade risking twenty thousand dollars that lost 3%. In the account, one of those cost six dollars and the other cost six hundred. In the tile, they are identical.

The screenshot makes the mismatch visible in one glance. The header strip is a set of count-based and percentage-based summaries. The chart beneath it is denominated in dollars and moves in steps of tens of thousands. Nothing in the strip can tell you which trades were responsible for which step, because the strip has thrown size away before it starts averaging.

The same trades under two weightings

Here is a made-up four-trade book to show the mechanism cleanly. These numbers are an illustration, not readings from any account.

The percentage expectancy is the mean of plus two, minus one, plus one and minus three, which is -0.25% per trade. That reads as a slightly leaky strategy that needs a small improvement in edge or costs.

The currency expectancy is the mean of plus ten, minus five, plus four and minus six hundred, which is -$147.75 per trade. The book lost $591 on four trades. Nothing about that is slightly leaky.

Note also that you cannot recover the currency figure by multiplying the percentage by the average size. Average size here is $5,350, and -0.25% of that is about -$13, which is off by more than a factor of ten. The two numbers are not convertible, because the information that connects them, which is the pairing of each result with its own size, was discarded at the first step.

Recomputing expectancy in currency from the same trade list

The export button sits at the top right of the page and this is what it is for. Five steps, and the whole thing fits in one sitting.

1. Fix the scope and the window before exporting, using the account row and the timeframe row at the top of the page, so your recomputation covers the same population the tile covered.
2. Drop open positions. Expectancy is a closed-trade statistic and mixing open marks in makes the result untestable.
3. Take the realised currency P&L for each trade. Mean of that column is your currency expectancy. Sum of it is what actually happened to the account from trading.
4. Sort the currency column and look at the top three and the bottom three rows. If the sum of the bottom three is a large fraction of the total, your expectancy is a statement about three decisions, not about a strategy.
5. If you use stops, compute the R-multiple version as well, which is each trade's currency result divided by the currency you had at risk on it. The Distributions tab offers R-multiple as a metric alongside PnL percent for exactly this comparison.

Keep the two numbers side by side afterwards rather than replacing one with the other. The percentage version tells you about the quality of the signal. The currency version tells you about the quality of your sizing. They are different problems with different fixes.

What the currency number changes about your next decision

A negative percentage expectancy points you at the entry. It says the average trade does not work and invites you to improve the signal, tighten the filter, or trade less often.

A currency expectancy that is far worse than the percentage version points somewhere else entirely. It says your losses are concentrated in your largest positions, which means the sizing rule is the thing that is broken, and improving the signal will not help until it is fixed. That is a much more tractable problem, because a maximum position size is a rule you can write down tonight and follow tomorrow, whereas a better signal is a project.

The reverse case is worth knowing too. If the currency expectancy is better than the percentage version, your big trades were your winners, and you are size-selecting well. That is a real skill and it argues for keeping the discretion in your sizing rather than flattening it. The point of doing the arithmetic is that these two situations look identical on the tile.

The other numbers on this page with the same blind spot

Once you have seen it in expectancy, you will see it in half the page. Win rate reads 25.29% and is a count, so a run of tiny scalps can carry it while the account bleeds. The direction table reads 73 long trades at a 28.8% win rate and 14 short trades at 7.1%, both unweighted. The venue table has Manual at 1 trade and a 100.0% win rate, which is a percentage computed from a single observation and belongs in no argument.

There is one more thing on this page worth handling carefully. The header tile reads -1.20% while the distribution panel on the Distributions tab reports mean -0.53 with median 0.00 across the same 87 trades. I am not going to assert what accounts for the difference, because I cannot verify it from a screenshot. Two averages over the same trades can legitimately differ when the population is filtered differently, when per-trade return is defined against different denominators such as notional, equity or risk, when a position filled in several parts is counted as one trade or several, or when the two panels are reading different windows. The way to find out which applies to your account is the same export you already need for the currency version, and by the time you have built that column you will know which of the two definitions matches the question you were asking.