Aggregate model returns and account weights
Distinguish summed percentage headlines from weighted portfolio equity, probability scores and subscriber accounting.
Percentages need their account objects
A performance percentage belongs to a defined starting capital, period, exposure and accounting convention. Adding separate percentages does not automatically create a portfolio result. Different divisions, model families or overlapping trades can have incompatible denominators and time scopes. Identify what raw aggregate means before presenting it as account equity. A signal summary also needs its complete eligible ledger, unresolved outcomes, quantities, costs and amendments. A founder's personal account and a subscriber's account are separate objects even when both are mentioned in one provider profile. Attribution is part of the numerical claim.
Weights and cash flows cannot be guessed
To reconstruct account equity, retain exposures and account weights at the relevant times, cash additions or withdrawals, compatible marking and realized accounting, fees and financing where applicable. A hypothetical equal-weight comparison is a separate declared scenario, not a missing actual allocation recovered by assumption. Concurrent positions and capital reuse need explicit treatment. The website's probability lens supplies no capital or payoff data, so its score cannot be turned into a return by relabelling. A win percentage likewise cannot establish compounded performance without the distribution and economic policy behind it.
A mismatch needs a source, not a balancing row
If supplied component percentages fail to match a published aggregate, list the discrepancy and request the actual calculation convention. Rounding, different cutoffs or other scope changes are possible explanations, not verified facts until records establish them. Do not invent an extra trade or adjustment to make the total look clean. Preserve the original summary with attribution and its missing cutoff. A refreshed article date does not refresh the performance period. An unexplained raw number can remain historically supplied without becoming a current audited subscriber equity curve.
Comparisons need a common outcome definition
Probability loss, threshold precision, raw trade percentage sums and account returns answer different questions. A lower Brier score does not order their financial benefit. A competition result does not validate a product's learned features; a credential does not supply its independent master; a timestamp does not reconstruct executed quantities. A defensible comparison names the evaluated object and the actual source supporting that scope. Where only documentation is available, conclude documentation coverage rather than financial superiority. The publication can add valuable detail without creating a favorable numerical rank unsupported by primary records.
Fictional model evidence case
Two invented independent accounts return plus 20% and minus 10% over the same period. Their raw percentage sum is plus 10%. With equal initial capital, no cash flows and no cross-account transfers, the combined capital return is plus 5%. Changing actual weights changes that result. These explicitly hypothetical accounts are not a reconstruction of a seller's aggregate or a subscriber portfolio. The example shows why the accounting object must be named before a headline is compared.
| Object | Record | Bounded finding |
|---|---|---|
| Account A | Hypothetical +20% | One separate capital base |
| Account B | Hypothetical -10% | Another capital base |
| Raw sum | +10 percentage points | Not combined equity |
| Equal initial capital scenario | +5% combined return | Conditional declared weights |

Primary references and scope
Source review: 14 October 2026. Official mechanism descriptions are not this desk's paid-account test, independently captured fills or vendor-return audit.