The Performance overview draws a panel called Account Equity and labels it, in its own words, the real cumulative balance summed across your connected accounts. That label is doing more work than it looks like it is doing. A balance is not a return. It moves when the book makes money and it moves when somebody wires money in, and the line itself does not distinguish the two.
On the capture I am working from, the header strip reads Net +396.07% on that series while the closed-trade statistics further down the same page read a profit factor of 0.08 and an expectancy of -1.20% across 87 trades. Both of those can be true of the same account on the same day. They answer different questions, and only one of them is built on something that qualifies as a return series.
What a summed balance actually measures
A cumulative balance answers the question a custodian asks, which is how much is there. A return series answers the question an allocator asks, which is what happened to each unit of capital while it was exposed. The gap between those two is external flow, and it is not a rounding problem. A single deposit on a small base changes the level permanently and changes every percentage computed off that level from that point forward.
The shape gives it away before the arithmetic does. Look at the equity panel in the screenshot below and note that the line does not drift, it steps. There are vertical moves into late January, a fall through February, and then a flat stretch from early March to the right edge. A step is not proof of a flow, because a single large marked position can produce one too, but it is the first thing you check against the cash record, and a flat stretch on an account that the monthly table says was losing money in April, May, June and July is the second.
The second reason to care is that the risk block on this page states its own input. The Risk and Statistics panel describes itself as computed over your equity curve. That tells you the input class. It does not tell you whether flows were removed before the computation, and you should not guess. Establish it, because everything in that block inherits whatever the answer is.
Building the flow register before you build the return series
The register is the deliverable that survives the analysis, so build it first and build it as a table, not as a mental note.
Every row needs six fields. The value date, the account, the direction, the amount, the currency, and a source reference you can point at in a review. Anything that changes the balance without a corresponding trade belongs in it. Deposits and withdrawals are obvious. The ones that get missed are transfers between two accounts that are both inside the composite, non-cash credits such as staking rewards or rebates, fee reimbursements, and any asset that arrived without being bought.
Internal transfers deserve their own treatment. Where the composite is a sum across connected accounts, and this page offers exactly that as a dropdown option alongside the individual accounts, a transfer between two members of the composite nets to zero at the top level and is a full-size flow at the account level. If you compute account-level returns from the same register without a flag on those rows, you will book the same capital as an inflow to one account and an outflow from another and produce two wrong series that add up to a right one.
Reconcile the register before proceeding. Beginning balance, plus net flows, plus realised and unrealised P&L, should equal the ending balance for every account and every period. A residual is a missing flow, not a rounding artefact.
The two chaining methods and when each is the honest one
With the register in place, cut the series at every date carrying a flow. For each sub-period, the return is the change in value less the net external flow over that sub-period, divided by the beginning value adjusted for the timing of the flow. Where you have daily valuations, and this page does draw a daily line, you rarely need the timing weight at all. Place every flow at the start or at the end of the day, pick one convention, write it down, and never change it mid-series.
Chaining the sub-period returns multiplicatively gives you a time-weighted return. That is the number that describes the strategy independently of when capital showed up, and it is the number to use when the question is whether the process works. Solving instead for the single rate that reconciles all the flows to the ending value gives you a money-weighted return. That is the number that describes what the capital actually earned, and it is the number the person who sent the capital cares about.
Both are legitimate. The failure is quoting one while the audience assumes the other, which happens most often when a manager reports time-weighted performance to a client whose experience was dominated by the timing of their own subscriptions.
One arithmetic warning. Sub-period returns computed on a small denominator explode. An account that is drawn down to a fraction of its peak and then receives a deposit will generate a sub-period return on the pre-deposit base that can be an order of magnitude larger than anything the book actually did. This is where extreme headline figures come from in practice, and it is why the denominator convention needs to be a written rule rather than a default.
What changes downstream once the series is clean
Everything in the risk block sits on top of whatever series you feed it. Volatility, drawdown, the excess-return ratios, the capture pair, the calendar table. Change the input and they all move together, which is the reason to fix the input once rather than argue about outputs repeatedly.
It is worth being precise about what this page can and cannot settle for you. It prints a Sharpe of 1.23 in the header strip and a Sharpe ratio of -4.246 in the Risk and Statistics block, and a max drawdown of -91.31% in one place and -38.79% in the other. I am not going to tell you which figure in each pair is the one you want, because I cannot see the inputs behind either from a screenshot, and neither can you.
What I can tell you is the short list of things that make two honest figures on one page diverge. A different input series, per-trade returns against equity-curve returns. A different window. A different observation frequency and therefore a different annualisation. A different risk-free assumption, and this page does expose a risk-free field and does print the rate in force in its footer. And a different treatment of external flows. The last of those is the one you control from the outside. Until it is controlled, you cannot even start the argument about the other four.
The way to settle it is not to reason about it. Export the series, recompute one period by hand under both treatments, and see which of the two published figures your raw version reproduces and which your flow-adjusted version reproduces. That takes an afternoon once and then never again.
The disclosure that has to travel with the number
A flow-adjusted return series is only defensible if the assumptions travel with it. Six fields, attached to the figure itself rather than to a footnote that gets cropped out of a slide.
- The window, with explicit start and end dates, not a period label.
- The observation count and the frequency. This page models the habit already, printing Observations 86 and Days span 96 under its calendar table.
- The flow convention, meaning start-of-day or end-of-day, and the treatment of internal transfers.
- Whether the figure is time-weighted or money-weighted.
- The risk-free rate used, if the figure is an excess-return ratio. The footer on this page records 4.5%.
- Whether the series is gross or net of fees and financing.
The reason to be strict about this is not tidiness. It is that a return series is a constructed object, and two competent people building one from the same account will produce different numbers if their conventions differ. When somebody in a review asks why your figure does not match the platform tile, the answer that ends the conversation is a stated convention and a reconciliation. The answer that extends it for an hour is a shrug.