A track record is evidence. Most of it is read wrongly.
A published signal shows growth, win rate and drawdown. None of those tell you whether the system survives. That is decided by the shape of its losses, the width of its statistical margin, and whether its edge outlives your broker's execution.
Three questions a growth curve cannot answer
Are the losses bounded?
The ratio of the worst loss to the average loss reveals whether a hard stop is in force. When every loss lands on the same value, a fixed stop closed it. When the tail runs long, the largest loss has not happened yet.
Is the edge real, or just short?
A win rate measured over sixty trades carries a wide confidence interval. If its lower bound falls beneath the break-even rate the reward:risk demands, the record is consistent with a system that loses money.
Will it survive your broker?
An edge held for seconds belongs to the seller's server, feed and latency. The shorter the average hold and the thinner the edge against spread, the less of it arrives on a copied account.
Seven weighted dimensions, and the ceilings that override them
Capital preservation, edge robustness, structural integrity, statistical confidence, drawdown behaviour, execution transferability and consistency are each scored from the published record rather than from marketing claims. Certain findings cap the total outright: suspected martingale behaviour, an absent stop-loss, or an edge that fails its own confidence bounds will hold a score down regardless of how well it performs elsewhere.
MQL5 requires a login to expose individual trades, so this analysis works from the public aggregate. It can establish whether losses are bounded and whether the edge is statistically real. It cannot see entry logic, and it cannot audit a record the seller has hidden. Analysis of historical data — not investment advice, and not a prediction. A record that scores well can still lose money.