The months that paid: what a monthly sweep would have paid out

2026-09-03. Every withdrawal figure quoted here is a field in the published audit JSON, and the audits are linked in place.

A backtest reinvests every dollar it makes. A person takes money out. Those are two different accounts, and only one of them appears in a tester report. So every audit we publish runs the observed daily balance path a second time under one plain rule. On the last trading day of every month, everything above the starting balance is swept out. A month without a surplus pays nothing, and nothing is ever paid back in. Across the 26 default audits in our catalog, 2 of them would have paid nothing in any month at all, and the spread between the rest is wider than any profit figure suggests.

What the replay does

The rule is deliberately simple, because a complicated rule cannot be checked by a reader.

Monthly sweep to the start balance. At each month boundary the account is measured. Anything above the start balance leaves the account and is counted as paid. Anything below it stays as a hole to be dug out.
A paid month is a month above the line. Surpluses under half a cent do not count as a payout, which keeps rounding noise out of the count.
The dry streak is the longest run of unpaid months. It is the figure that answers how long a working system can go without sending you anything.
The end balance is what is left. After the last sweep, the account either sits exactly at the start balance or below it. It can never sit above it, because everything above was paid out.
Two sizing modes, both printed. One keeps the observed daily profit and loss as measured. The other multiplies each day by the ratio of current balance to start balance, which is what an expert advisor does when its lot size follows the account.

The replay also reshuffles the same daily results a thousand times and runs the sweep again on every reshuffle, which is the part that shows how much of a payout record is the order of the days rather than the quality of the system.

Paying often and earning well are different questions

Put the long runs side by side and the payout frequency does not follow the profit ranking.

Gold House pays in 182 of 280 months, a total of $11,058.04, with a best month of $609.75 and a longest dry streak of 64 months. CryonX pays less in total, $3,498.74, but it pays in 176 months with a longest dry streak of only 13 months and a median paid month of $18.46. On the question of regularity, the smaller earner wins by a distance.

Lizard is the other end of the same window. It pays in 3 of 280 months. The three months that paid were good ones, a median of $60.90 and a best of $72.13, and the total ever withdrawn is $185.00 against a measured net profit of $183.99 over 23 years. The longest dry streak is 277 months. A system can be profitable, survive its whole test window and still send money three times in 23 years.

Quantum Athena X is the opposite extreme and the only 100,000 dollar run in the catalog with no dry month at all. It pays in 104 of 104 months, a total of $4,648.65, median paid month $23.05. The window is shorter, which is exactly why the dry streak is the number to read next to the count rather than instead of it.

Two more that look healthy in a profit column and thin in a payout column. Adaptive Gold Scalper pays 53 of 279 months with a median paid month of $6.74 and a dry streak of 126 months. TwisterPro Scalper pays 42 of 248 with a dry streak of 146 months. More than a decade of nothing, inside a run that ends in profit.

The end balance is the sentence nobody wants to read

The sweep is not a safety mechanism. It pays out on the way up and leaves the account carrying every loss that comes after.

Logan pays in 15 of 216 months, a total of $629.48, and the account ends at $75,028.53. That is a withdrawing account that banked $629.48 and finished $24,971.47 below its deposit, and its measured net result over the same run is minus $24,334.65.

Scalping Robot Pro pays in 0 of 280 months and ends at $46,766.04. Smart Gold Hunter also pays in 0 of 280, and ends at $98,581.50. Those are the two audits where the withdrawal replay has nothing to report, and the reason is visible in their net results of minus $53,234.68 and minus $1,418.55.

The sharpest version of the mechanism sits in an audit we hold unlisted. Pulse Engine, measured on a vendor report over 12 months on a 1,000 dollar deposit, pays out $2,030.42 across 8 months and ends at $580.85. Its measured net result on the same run is $1,598.18. You would have banked more than twice the deposit and would still be sitting on an account holding $580.85 of a $1,000 deposit. Nothing is wrong with either number. That is what monthly sweeping does to a volatile account.

The order of the days matters as much as the total

Take the same daily results, cut them into five day blocks, shuffle the blocks a thousand times and run the same sweep on each path. The calendar stays, the sequence changes. What comes back is a range instead of a single number.

Gold House observed $11,058.04 against a reshuffled range of $9,651.25 to $12,474.18 with a median of $10,991.75. The observed result is almost exactly the median, which is what a steady payer looks like.
Lizard observed $185.00 against a median of $419.53 and a range of $38.72 to $1,065.19. Its own order of days was worse for a withdrawing account than most rearrangements of the very same days.
Logan observed $629.48 against a median of $3,880.61 and a range of $289.89 to $6,589.72. The losses arriving when they did cost the withdrawing account most of what the same days could have paid.
Scalping Robot Pro pays nothing in 782 of 1,000 reshuffled paths, and even the ninety fifth percentile pays only $22.16. That is the difference between an unlucky sequence and a system with nothing to distribute.

This is the honest use of a resample. It does not predict the future. It separates the two sentences that get confused all the time, which are that a system paid badly and that a system cannot pay.

Two sizing modes, and the one that never applies

The audit prints the sweep under fixed lots and under balance scaled lots, and a separate detector says which one describes the system. Of the 26 default audits, 19 read as fixed lots and 7 read as mixed or unknown. None of them reads as balance scaled.

On a 100,000 dollar account the two modes barely differ. Gold Trade Pro withdraws $13,512.65 under fixed lots and $13,513.09 under balance scaled. On a small account they differ a lot. Pulse Engine withdraws $2,030.42 against $2,261.55 and ends at $580.85 against $549.30. The scaled mode is the optimistic one by construction, because it multiplies every day by a ratio while a real account cannot trade a fraction of a minimum lot. We print both and say which is which rather than choosing one quietly.

Run these checks yourself

Count paid months before you read the profit. Take any tester report, mark the balance at each month end and count how often it stood above the deposit. That single count tells you whether the system could have funded anything.
Ask what is left after the last payout. Profit that was withdrawn cannot repair a later loss. The end balance under a sweep is the number that survives the marketing.
Measure the dry streak against your own patience. A system with a 126 month dry streak is not a bad system. It is a system nobody holds long enough to collect from unless they know the figure in advance.

The withdrawal replay runs on every audit we publish, together with the resampled range and the capital what if table. The free browser check reads your own tester report in the browser.

The honest limits. All figures come from the replay block of the published audit JSON for each expert advisor, on our documented data world and the default preset of each system. The percent per year figures divide by trading days over 252 and not by calendar years, so they describe the tested span rather than a calendar return. The sweep is a model of an account and not an account. Transfer fees are not modeled, tax is not modeled unless a rate is supplied, and margin is not modeled, which means a small account may not be able to hold the positions at all. The daily path used by the replay starts at the close of the first trading day, so the result of that one day sits outside the sweep, which is why a total withdrawn can sit a few dollars away from the published net profit, Lizard $185.00 against $183.99 and Gold House $11,058.04 against $11,054.32. Waka Waka, Pulse Engine, Quantum Emperor and Market Anomalies EA are named and not linked, because those four audits still rest on vendor supplied reports with 1,000 dollar deposits and windows of eleven or twelve months, so their percent figures do not compare with the 100,000 dollar runs. Quantum Queen and Quantum Queen X carry identical replay numbers because the runs are deal identical, a republish confirmed in our own version census. Monthly sweeping to the deposit is one rule out of many. A quarterly rule or a rule that keeps a buffer would pay differently, which is why the rule is printed next to every figure it produces.