Using Custom Analysis to stress-test a signal before you copy it

The whole-record problem

Every statistic on a signal provider's page is an average over the entire record, and averages are where weaknesses go to hide. A signal with a 2.1 profit factor over three years might have a 0.9 profit factor over the last six months. A forex robot with a 64% win rate across nine pairs might have 44% on the one pair you will actually trade. Using Custom Analysis to stress-test a signal is how you stop reading the average and start reading the parts.

On ConfirmedTrades, if the owner allows it (it is on by default), a Custom Analysis button at the top of an account page lets any visitor filter the trades by date range, instrument, strategy or EA, buy or sell, deposits and withdrawals, weekday, hour, lot size, profit, duration, swaps and comment text. Every statistic on the page, from the Glance hub through the By symbol table to the Score-adjacent metrics, is recomputed over just those trades. It is read-only; nothing on the owner's page changes, and Reset returns you to the full record. The filtered view lives in the page link, so you can send a specific cut to a trading partner.

Filter 1: the last six months only

Set the date range to the most recent six months and compare the recomputed gain, profit factor and drawdown with the lifetime figures. This is the single most informative test for any automated trading system or copy-trading account, because edges decay and developers do not announce it. A signal whose lifetime profit factor is 1.9 and last-six-months profit factor is 1.1 is being sold on its past. Then try three months. If the recent window is negative and the provider is still advertising the lifetime number, you have learned what you needed to.

Do the reverse as well. A record whose last six months are far better than its lifetime average is either improving or running hotter risk; check Total lots in the filtered window against the lifetime figure to tell which.

Filter 2: one instrument at a time

The By symbol table already shows per-instrument totals, but Custom Analysis on a single symbol recomputes everything: drawdown, expectancy, average holding time, streaks. Filter to the instrument you intend to copy and see whether the strategy works there on its own. A common finding is that a 'multi-currency EA' is profitable on XAUUSD and a small net loser on everything else, with the gold trades carrying the drawdown too. If you only have the margin for two pairs, you want to know which two.

Then filter to the instrument that contributed the largest single loss. If removing one pair takes the permanent-style max drawdown from 28% to 11%, the provider's risk is concentrated, and so would yours be.

Filter 3: by EA or strategy

Where the owner runs several Expert Advisors on one account, the strategy filter (by magic number, or by the friendly EA name if the owner has set one) isolates each. This matters when a developer sells one EA but publishes an account that also runs two others. Filter to the EA you are buying and read its own expectancy, drawdown and trade count. Half the time the EA for sale is not the one producing the profit.

It also catches recovery systems. A main strategy with a small 'hedge' magic number that only trades during drawdowns is usually a martingale in a costume. Filter to the hedge magic and look at its lot sizes and trade durations.

Filter 4: hour and weekday

Filter to the hours you can actually trade, or to the hours your broker's spread is reasonable. A scalping signal provider whose edge lives between 22:00 and 02:00 UTC is trading the rollover window, where most retail spreads widen to three or four times normal. Recompute with those hours excluded and see what is left. I have seen signals go from +41% to −3% lifetime with a single hour filter.

Weekday filters find Friday risk and Monday gap risk. Filter to Friday, then to trades with duration over 48 hours, and look at the worst trade in the filtered set. That is your weekend exposure if you copy.

Filter 5: lot size

Filter to trades above the median lot size, then to trades below it. If the large trades have a much worse win rate or much deeper drawdown, the provider is sizing up after losses, which is the signature of averaging down even when the AI review has not flagged a full martingale. If the large trades are where all the profit is, the provider is sizing up after wins, which is fine until the streak turns.

A clean system shows roughly the same expectancy across size buckets. That is what you want to see, and it is rare.

Filter 6: profit and duration extremes

Filter out the five largest winning trades by using the profit filter, and see whether the record is still profitable. If removing a handful of outliers turns a +55% record into +4%, the edge is luck-shaped. Then filter to trades held over 72 hours: long-held trades on an intraday system are losers the provider refused to close, and their equity drawdown is the real drawdown. The Trade duration chart shows this as a cluster of long-duration losers, and the filter lets you measure it.

Finally, filter with swaps excluded and included. A carry-heavy signal can be profitable on paper and negative once financing is counted, and Total swap in Trade statistics will confirm the size of the effect.

Filter 7: comment text

Many EAs write their strategy name, version or signal type into the trade comment. Filtering by comment lets you isolate a version change ('v2.3' onwards) or a sub-strategy ('breakout' versus 'range'). If the developer says the losing period was 'before the update', this is how to check: filter to comments from the new version and see whether the improvement is real and whether the sample is large enough to mean anything. Owners can hide the comment column from the trade table, but if it is visible, use it.

What to do with the results

If a signal or trading bot survives the six-month window, the single-instrument test, the hour filter and the outlier removal with a positive expectancy and a drawdown you can size for, it has passed a harder test than its sales page ever set. Copy it at a lower risk than the provider runs, and re-run the six-month filter every month; the link you saved makes that a one-click check.

If the owner has switched visitor Custom Analysis off, that is their right, and sometimes there is a reason. But a provider asking for your money while declining to let you look at the parts has made a choice you are entitled to weigh. Rankings and Explore are free to view with no account, so there is no cost to moving on. Signing up free at /signup lets you follow candidates and compare up to five side by side in the dashboard once the filters have done their work. For the full vetting sequence, reading an EA track record before you buy is the companion to this guide.