Why every call is published, including the bad ones

Every trading tool claims accuracy. Almost none of them let you check, and the reason is structural: if you control which results get shown, you control the accuracy number. Screenshot the wins, quietly drop the losses, and any tool can look like an edge.

So mEEme publishes the whole ledger. Every non-demo call the engine has ever made appears on the track record, in order, with what happened next — wins, losses and the ones that went nowhere.

The grading rule lives in the codebase and is versioned in git. It was written before the calls it grades, and it runs automatically four hours after each one, so it cannot be retuned once the results are in.

The rule

A call is graded against what price actually did in the following four hours. What counts as right depends on what was claimed:

• Exit calls (EXIT IMMEDIATELY, SCALE OUT, NO TOUCH) are correct when price fell 10% or more by the horizon, and wrong when it ran 15% or more without you. Telling someone to leave before a run is a real cost, and it is counted as one.
• Entry calls (APEX ENTRY, SCALE IN) are correct when price rose 10% or more, wrong when it fell 10% or more.
• ARM EXIT is vindicated by the drawdown it warned about, even if price later recovered. A warning that was right about the risk was right, and that is what a warning is for.
• Anything smaller is neutral and excluded from accuracy entirely. Neutral calls are not counted as wins — the easiest way to inflate an accuracy figure is to quietly bank the noise.

Why exit calls are graded asymmetrically

A 10% fall makes an exit call right, but it takes a 15% run to make it wrong. That gap is deliberate and it is not flattering — it reflects that the two errors do not cost the same thing.

Being told to exit and watching a modest continuation costs you upside you never had. Holding through a collapse costs you capital you did. In a market where the median outcome is approximately zero, those are not symmetric, and grading them symmetrically would misrepresent what the tool is for.

It does mean the accuracy figure is measuring “did this protect capital” more than “did this maximise return”. That is the honest description of it.

What the track record cannot tell you

• It is not a forecast. It is a record of what already happened, published so the tool can be judged. Past performance says nothing about the next call.
• Four hours is one horizon. It suits how these positions actually behave, but a call that was right at four hours and wrong at twenty-four is recorded as right.
• It does not model slippage or fees. It grades price movement, not the fill you would have got. Real execution is worse than the chart.
• Early samples mean little. An accuracy figure over a few dozen graded calls is noise wearing a percentage sign. It gets meaningful with volume, and it is published from the start anyway.

None of that makes the ledger worthless — it makes it a measurement with stated limits, which is the only kind worth trusting. The alternative on offer everywhere else is a screenshot.