Scroll the Performance overview from top to bottom and you will pass a Sharpe twice. The header strip prints 1.23. The Risk and Statistics block further down prints a Sharpe ratio of -4.246. Same account, same page, same session, opposite signs.
The first instinct is to decide one of them is broken. Resist it. Two figures with the same name can both be computed correctly and still disagree, because the name is not a definition. A Sharpe is a recipe with at least four free parameters in it, and a page that shows you two of them is showing you two different recipes rather than one recipe and one error. Your job is not to catch a bug. It is to work out which recipe answers your question, and that is a job you can finish in about fifteen minutes.
The two numbers and where they sit
The header tile sits in the strip of eight that runs across the top, next to account equity, win rate, Sortino, max drawdown, profit factor, expectancy and total trades. That strip responds to the account scope buttons and the timeframe buttons above it.
The second figure sits inside a panel headed Risk and Statistics, which describes itself as risk metrics computed over your equity curve plus benchmark comparison. That panel has its own controls, a benchmark selector and a risk-free percentage field, and its own footer, which on this capture reads Observations 86, Days span 96, Risk-free 4.5%.
So the two figures do not merely differ in value. They sit in blocks with different stated inputs and different exposed controls. That is the shape of the problem, and it is not unique to this page. Every serious analytics screen has it somewhere.
It is not only the Sharpe. Sortino reads 9.84 in the header and -2.422 in the risk panel. Max drawdown reads -91.31% in the header and -38.79% in the panel. Whatever is different between the two computations, it is systematic rather than a one-off on a single tile.
Five reasons two honest Sharpe figures disagree
This list is not speculation about this specific product. It is the standard set of choices that anyone computing a Sharpe has to make, and any two implementations that make them differently will print different numbers.
- The input series. A ratio computed from a series of per-trade returns is a different object from one computed from a series of daily equity marks. The same account produces both, and 87 closed trades and 86 daily observations are not the same sample even when they cover the same calendar.
- The window. The timeframe buttons on this page run 1D, 7D, 14D, 1M, 3M, 1Y, ALL and CUSTOM. A figure computed over the last year and a figure computed over the span the account actually has data for will differ whenever those two are not the same, and the risk panel footer here says the span is 96 days.
- The observation frequency and the annualisation that follows from it. Daily, weekly and per-trade sampling scale by different factors. Two implementations that annualise from different frequencies will differ by a constant multiple even on identical data.
- The risk-free assumption. An excess-return ratio subtracts a rate before dividing. This page exposes a risk-free field in the risk panel and records 4.5% in the footer. A tile with no such control is not necessarily using the same assumption.
- The treatment of external flows. The equity panel here says it plots a cumulative balance summed across connected accounts. A balance moves on deposits and withdrawals, and returns computed from a raw balance differ from returns computed after those flows are stripped out.
There is a sixth that catches people out, which is what counts as an observation. Days with no trading, days with no price update, and days where the account sat in cash can be included as zero-return observations or excluded entirely, and that choice moves both the numerator and the denominator.
I am deliberately not telling you which of these explains the gap on this page. I cannot verify it from the outside and neither can you, and the confident-sounding explanation is exactly the thing that gets repeated until somebody sizes a position on it.
How to work out which one answers your question
Start from the decision rather than from the number. There are only three questions a retail trader actually asks this page, and each one points at different evidence.
If the question is whether your trading decisions are any good, you want the per-trade evidence, not a ratio at all. On this capture that reads a win rate of 25.29% over 87 trades, a profit factor of 0.08 and an expectancy of -1.20%, with the long bucket at 73 trades and a total P&L of -44.64% and the short bucket at 14 trades and -1.18%. A profit factor below one means gross losses exceeded gross profits, and 0.08 is not a marginal reading.
If the question is whether your account is growing, you want the balance and the flow record, and you want to know how much of the growth was money you added. The equity panel here reads Net +396.07% while the monthly returns table on the Distributions tab reads April -0.4%, May -20.5%, June -8.1% and July -16.8%. Those two facts sitting side by side is the thing to resolve.
If the question is whether you are taking too much risk for the return, that is where a ratio is the right tool, and then you have to pin down the recipe before you use it. Pick the block whose inputs you can state. The risk panel here tells you its input class, its risk-free rate, its observation count and its span. The header tile does not, so it is the harder of the two to defend to yourself.
Cross-checking with evidence that has no free parameters
The useful property of a disagreement like this is that the surrounding evidence usually does not disagree. Counts and sums have far fewer choices baked into them than ratios do.
On this page the venue table reads Hyperliquid 70 trades at a 21.4% win rate for -37.79%, Alpaca 16 trades at 37.5% for -8.50%, and Manual 1 trade at 100.0% for +0.47%. The asset class table reads crypto, 87 trades, 25.3%, -45.82%. None of those require an annualisation factor or a risk-free rate. They are counts and sums, and they point one way.
The Distributions tab adds the same kind of check for the ratio itself. Its rolling Sharpe panel plots 28 points between 11 April and 11 July, and the axis runs from -5.5782 at the low to 0.6361 at the high. Whatever single summary figure you settle on, that chart tells you a single figure is compressing a series that moved by six units of ratio inside three months. A summary number over a span like that is a weak claim regardless of which recipe produced it.
The rule I use before repeating a ratio anywhere
One line, and it has saved me from more embarrassment than any indicator. I do not repeat a Sharpe unless I can say, in the same breath, the window it covers, the series it was computed from, and the risk-free rate it assumed. If I cannot say all three, I quote the counts instead.
Making that operational on this page takes one session. Choose a scope from the account row and a window from the timeframe row and write both down. Export the CSV. Compute the ratio yourself once from the daily equity series and once from the per-trade series, at a rate you choose, and see what each produces. You are not trying to prove either published figure right or wrong. You are learning which recipe you have been reading, so that next month you know without repeating the exercise.
The version of this mistake that costs money is not misreading a Sharpe. It is sizing up because one tile looked good while the profit factor two tiles along read 0.08 and the monthly table was negative every month it had data for. When two figures on one page disagree, the safe assumption is that you do not yet know what either one means, and the cheapest response is to trade the smaller size until you do.