Every audit here answers the question what was the worst day twice, and the two answers are not the same number. One adds up the profit and loss of the trades that closed on a day. The other follows account equity through the day, open positions included, and records how far it fell below the level the account carried into the day. In 19 of our 26 default audits the two agree to within 15 percent. In 6 they diverge, in both directions, and the widest pair differs by a factor of 358. In the remaining audit one of the two numbers does not exist at all, and the file says why.
The booked day. Our engine groups every paired trade by the date on which it closed and sums that day. A basket that stayed open for three weeks books its whole result on the day it is finally closed, and nothing on any of the days it was open.
The equity day. Our engine rebuilds account equity at every deal timestamp, anchors each day at the last equity value before that day began and keeps the lowest value the day reached. The drop between those two is the day figure, and the worst of them is what the audit prints.
The second number is a floor and not a measurement of the true low. Between two deals a tester report shows nothing, so an equity path that dipped and recovered without a deal in between leaves no trace. The audit states this in the section itself and names the tick based drawdown from the report head as the upper reference.
For 19 of the 26 default audits the pair is close enough that the distinction does not matter. Six of those agree to the cent.
The widest gap inside this group belongs to Gold Trade Pro MT5 at 322.10 USD booked against 369.23 USD of equity, which is 14.6 percent. Second widest is Market Anomalies EA at 18.67 against 21.19, or 13.5 percent.
One warning about the table before it is read across rows. Twenty two of the twenty six runs started from a 100,000 USD deposit in our own measurement frame. Four did not. Waka Waka, Quantum Emperor, Pulse Engine and Market Anomalies EA are older audits built from vendor reports of a different broker and started from 1,000 USD. A dollar figure from those four is a different kind of dollar. The comparison this article makes is always between the two numbers inside one audit, where the deposit cancels out.
The tempting story is that robots holding positions over midnight diverge and intraday robots do not. Our own data says no. Gold Atlas carries 2,188 trades that spanned a daily rollover and its two worst day figures are identical to the cent. Ultimate Breakout System carries 363 such trades and its pair differs by 4 cents. The Gold Reaper MT5 carries 334 and differs by 4.1 percent. Holding overnight is not the discriminator.
What the divergent runs have in common is something the audit already prints on its own line. All six carry our grid flag, and so does the seventh case below. None of the 19 agreeing audits carries it. That is an association across twenty six runs and not a law, but it is the only structural marker that sorts the two groups without an exception.
Here the equity number is the large one and the trade log stays quiet.
These three carry a second fingerprint that no other audit in the catalog shows. Each run reports a floating drawdown and a balance drawdown, and normally the two sit within a few percent of each other. For Quantum Queen they are 0.16 percent floating against 0.02 percent on balance, a factor of 8. For Quantum Athena X, 0.17 against 0.02. For Waka Waka, 14.32 percent against 0.30 percent, a factor of 48. The account went under water and came back up before anything was closed, so the closed trade record never learned about it.
Keep the size in view. For Quantum Queen the large factor sits between two small numbers. 158.15 USD against a 100,000 USD deposit is 0.16 percent of the account in its worst day. Waka Waka is the serious one of the three, because 96.89 USD against its 1,000 USD deposit is 9.7 percent, and its deepest recorded equity point stands at 900.11 USD against a peak of 1,050.51 USD.
Logan MT5 runs the other way and is the sharpest case in the catalog. Its worst booked day is 30,247.94 USD on 2016-12-09. Its second worst booked day in the entire run is 0.24 USD and the third is 0.20 USD. One day carries the whole event.
The equity series does not know that day as a catastrophe. Across 419 profiled days its worst single day drop is 6,867.38 USD. The deepest equity point of the run is dated 2021-01-11, five years later, at 68,317.63 USD against a peak of 100,636.82 USD. That hole is 32,319.19 USD deep, and since no single day cost more than 6,867.38 USD of equity, digging it took at least 5 days. The worst day the account lived through and the worst day the trade log records are two different events, five years apart.
Quantum Emperor is the mild version of the same shape, 21.13 USD booked against 11.92 USD of equity on a 1,000 USD deposit.
Every audit replays its run against three funded account profiles. One of them sets a maximum daily loss of 5 percent of the initial balance, which on our 100,000 USD frame is 5,000 USD, and our audit counts breaches on both bases at seven position sizes. For Logan MT5 the two rows disagree.
The rule these profiles copy is written on equity including open positions, which our own profile note records, so the equity row is the one that matches the rule and the booked row is merely the one that is easy to compute from a report. A vendor page that quotes a worst day from a trade log is quoting the number that is not being enforced.
Scalping Robot Pro MT5 books 1,802.78 USD on 2026-01-30 and has no equity figure to set against it. The audit prints the refusal instead of a number. The reconstruction failed its own validation, because the sampled floating drawdown of 53.3 percent exceeded the tick based report head drawdown of 29.2 percent, which means the implied contract size and pricing model does not fit that report. Both published runs of that robot carry the same refusal.
That is the seventh grid flagged run in the catalog, and it is the one where we would rather publish a printed reason than a plausible figure.
The equity day is a floor. It samples at deal timestamps, so a real low reached between two deals is invisible to it. Every figure in the first kind of divergence above could be deeper in reality and none of them can be shallower.
The day boundary is the broker server day. A funded account programme resets its daily loss at its own clock, and our own profile note records one such programme resetting at midnight central European time. A drop that straddles that boundary is split differently under the two clocks, and we do not model the second one.
A worst day is an order statistic. The number of profiled days in these audits runs from 159 to 5,533. The worst of five thousand days is drawn from a much larger sample than the worst of one hundred and fifty nine and will tend to be deeper for identical behaviour. Reading worst days across runs of different length compares sample sizes as much as robots.
What the pair is good for is the gap between them. When a robot shows a quiet trade log and a loud equity curve, the quiet part is what the seller can screenshot and the loud part is what the account rules measure.