The Performance page opens on a strip of eight tiles. Account equity, win rate, Sharpe, Sortino, max drawdown, profit factor, expectancy, total trades. It is the part people screenshot and the part people quote.
Directly above it are three separate control rows, and every tile in that strip is downstream of all three. Until you know which combination was selected, the eight numbers are not a track record. They are one of several dozen possible readings of one account, and you have no way to tell which.
Three controls, not one
The first row is a pair of buttons, all accounts real and Blockcircle only. That is a choice about which venues are in the population, and the word real is doing the work.
The second row is six buttons, all, autopilot, MTE, paper, manual and wallet. That is a choice about how the trades got placed. Autopilot for automated signal execution, MTE for the momentum engine, manual for what you clicked yourself, wallet for on-chain activity, paper for simulated.
The third is a dropdown that on this capture reads all accounts cumulative, with Alpaca listed at 16 and Secondary at 11 as the individual options. That is a choice about which connected account.
Off to the right is a fourth control that people also forget, the timeframe row running 1D, 7D, 14D, 1M, 3M, 1Y, ALL and CUSTOM. Four independent choices, multiplied together, and the header strip reflects whatever combination happens to be active.
The page itself flags the most important boundary in small italics under the title. All reflects real-money trades, and paper is to be selected separately to view simulated performance. That is the one distinction the page draws attention to unprompted, which tells you it is the one that gets confused most often.
What each of the eight tiles means under a narrower scope
The tiles do not degrade equally as you narrow the scope. Three of them stay meaningful, three become fragile, and two stop meaning what you think they mean.
- Account equity is a balance for the accounts currently selected. Under the cumulative option it is a sum across connected accounts. Under a single account it is that account's balance. It is not a strategy result under any scope, because a balance also moves on transfers.
- Total trades and win rate are counts, so they degrade with sample size in the obvious way. The venue table further down this page makes the point better than any argument. Hyperliquid reads 70 trades at a 21.4% win rate, Alpaca reads 16 at 37.5%, Manual reads 1 at 100.0%. Same account, three win rates, all correctly computed, and the third is a percentage derived from one trade.
- Profit factor and expectancy follow the same trade population as total trades, so they inherit the same fragility. A scope with a dozen closed trades produces a profit factor that will move substantially on the next two.
- Sharpe and Sortino are ratios over a return series that is itself rebuilt when the scope changes. They are the tiles most sensitive to the selection and the least obviously so.
- Max drawdown is path-dependent and does not decompose. The drawdown of a subset is not a component of the whole's drawdown, and you cannot add two scopes' drawdowns together or average them. On this capture the header reads -91.31% and the risk panel further down the page reads -38.79%, which is a reminder that even on one screen the same statistic can come from more than one construction.
The rule that falls out of this is that narrow scopes are for diagnosis and wide scopes are for judgement. Narrow to answer where the losses came from. Widen to answer whether you are making money.
The paper boundary changes the claim entirely
Every other scope selection changes the sample. The paper button changes what kind of claim the page is making, from something that happened to something that was simulated.
That distinction matters more than the usual warnings about slippage suggest, because it changes what you are entitled to conclude. Real fills carry your actual execution, your actual latency, your actual fee tier and your actual behaviour under stress. Simulated fills carry none of that, and the last one is the biggest.
There is a practical wrinkle worth knowing in the other direction. On the Distributions tab, the maximum favourable and adverse excursion histograms are empty on this real-money record, with a note that excursions are only recorded on paper trades. So the paper scope is not just a lesser version of the real one. It carries data the real one does not, and if you want to study how far your trades ran in your favour before you closed them, the paper scope is where that analysis is available.
The failure to avoid is mixing. Judging a strategy on paper statistics and then sizing it on the real balance is the mistake, and it happens because the two live behind adjacent buttons on the same page and the tiles look identical either way.
Reading a scope somebody else chose
The page has a share button next to the export button, and performance screenshots circulate constantly. When one arrives, the control band is the first thing to look for and it is usually the first thing cropped.
Three questions cover it. Which of the two account buttons was lit, real or Blockcircle only. Which execution type was lit, and specifically whether it says paper. And which timeframe, because a 1M window on a nine-month account is a selection of the best or worst month if somebody wanted it to be.
If the image does not answer all three, the numbers in it are unusable and the polite request is for the uncropped version rather than for an explanation. This applies to your own screenshots from three months ago as much as to anyone else's.
The habit that makes this page usable in five minutes
Write the scope down as a triple before you read a single tile. Something like all accounts real, MTE, 3M. It takes four seconds and it makes the reading reproducible next week.
Then fix two canonical scopes and stop improvising. One for judging the account, which should be the widest real scope over the longest window with data. One for judging a strategy, which should be the execution type that strategy runs under, over a window long enough to contain enough closed trades to matter. Everything else is a diagnostic view you open for a specific question and close afterwards.
The reason to be rigid about this is that the alternative is not chaos, it is bias. When the tiles are a click away from changing, you will click until they look better, and you will not notice you are doing it. A written scope removes the option before the temptation arrives, and it costs nothing except the four seconds.