Import a broker statement and Edge Lab reconciles it against the broker’s own figures, hunts for patterns, holds each one to explicit statistical gates, and only promotes a rule after it has survived a forward test you pre-registered.
No credit card. Full Pro access.
Inside the app this is called your improvement loop: verified data in, a tested rule out. Every step stays connected to the one before it.
Every import is checked line by line against your broker’s own totals — realised P&L, balance movement, deal count and closing balance — before a single pattern is searched. A history that does not close is not analysed at all.
Dozens of detectors split your trades into an affected group and a comparison group and test the difference between them. Multiple-testing correction is applied within each family of related tests, so a pattern that only looks real because it was one of fifty comparisons is marked as such rather than reported as a discovery.
A finding describes what already happened. A forward test is the only way to find out whether it holds on trades that have not happened yet. The hypothesis, the rule, the metric and the required sample are frozen the moment you register it, and cannot be edited afterward.
Only a rule whose forward test came back supported is added to your playbook. From there, your process score tracks whether you actually followed the rule — kept as a measurement completely separate from whether any given trade won or lost.
Nothing is analysed until the history Edge Lab rebuilt from your statement matches the broker’s own totals. Realised P&L, the balance movement, the deal count and the closing balance are each compared against what the statement itself reports.
Commission, swap and fees are itemised because your statement itemises them. Spread is not measured separately — it is already inside the fill prices your broker reports. Slippage is not measured at all: a statement records the price you got, never the price you intended.
Edge Lab imports any broker CSV, plus dedicated formats for MetaTrader, cTrader and more, up to 10 MB per file. A format marked unconfirmed was built from the platform’s own documentation and has not yet been checked against a real export — it still imports and is still reconciled against your statement, and you are told so at import time.
A gate is a bar a finding has to clear before its significance is even considered: enough effective observations in both the affected and comparison groups, enough distinct trading days for each, and a difference large enough to matter rather than merely detectable. A finding that fails any gate is marked insufficient regardless of how small its p-value is, and none of the gates can be waived.
Findings that clear the gates still have to survive correction for how many patterns were tested, and to hold up on trading days that were deliberately held back from detection. What comes out the other side lands on one of three levels.
At least one evidence gate failed. Nothing about the p-value or q-value can change this.
All gates passed, but the corrected significance is not strong enough, or the effect’s sign is not stable across time. Shown, and explicitly not acted on.
All gates passed, correction cleared, the sign is stable across time, and the pattern was not contradicted on the trading days held back from detection.
When you filter the findings list, the corrected significance behind each one does not change. Re-running the correction over whatever subset you happened to select — after you have already seen which findings exist — would improve the surviving numbers by construction, and the false discovery rate the procedure claims to control would no longer be controlled. So the correction stays fixed at the full family that was actually tested. Filtering changes what is on your screen; it never changes what was tested.
A declined finding is not deleted. It stays on the list with the reason it did not qualify, so what was tested and what did not hold are both visible — not just the patterns that happened to clear every bar.
A finding is historical. Registering it as a forward test is the only way to find out whether it holds on trades that had not happened yet. At registration, the hypothesis, the rule, the metric, the baseline and the required sample size are all sealed and cannot be edited afterward — you cannot quietly move the goalposts once you see how the trades are going.
Results are withheld until the required sample is complete, and a test is only evaluated if you actually followed the registered rule on enough of the qualifying occasions — otherwise it is invalidated rather than reported, because it did not test anything.
The observed improvement clears zero with room to spare.
The improvement is positive, but the uncertainty around it still reaches slightly below zero.
The result comes back below zero.
The result straddles zero and the sample is spent. That is a real outcome, not a failure to compute.
The coach reads your own findings, stats and trades back to you in plain language. Every number it states has to trace to a value a tool actually returned in that conversation — it is not allowed to do arithmetic of its own on a figure, or state a number no tool printed.
Evaluation and funded accounts demand clarity about which setups carry the account. Edge Lab shows which ones survive scrutiny.
If your decisions vary by session, mood or market state, Edge Lab measures which of those variations actually change your results.
A manual journal records what happened. Edge Lab tests what it means, without you writing a single formula.
It stays visible, marked with the reason it failed, rather than being hidden. Edge Lab shows what was tested and what did not hold, not just the patterns that happened to clear every bar.
Because looking obvious is what a random streak also does. A pattern is measured against the whole family of patterns tested alongside it, checked for whether its direction holds across time rather than coming from one unusual month, and held to a minimum amount of history. A finding that fails any of those is reported as declined, with the gate that stopped it named.
Nothing stops you acting on your own history. What Edge Lab will not do is call a finding a rule until it has held on trades that had not happened when the finding was made — that is the only part of the loop the past cannot fake.
Edge Lab imports supported broker and platform statement formats, including MetaTrader and cTrader exports. Automatic sync is available for MetaTrader 4 and 5 through the local Bridge while your terminal is running. A read-only cTrader Open API connection is now in private beta; until account discovery and history sync pass live verification, cTrader trades still stay current through statement import.
It depends on the pattern. Each finding shows its sample size, and a finding that does not yet have enough data to test is shown with how many more trades it needs instead of being reported as a result.
No. Edge Lab analyses trades you have already taken and reports what the evidence in your own history supports. It does not generate trade signals, forecasts or recommendations.
The trial runs for 14 days with full Pro access and no credit card. If you do nothing at the end of it, the account simply stops having Pro access. Nothing is charged automatically, because no card was collected.
No credit card. Full Pro access.