How long the hole lasts: 7 of 22 audits never repaired their deepest drawdown

2026-09-03, measured across every default audit that carries an episode table. The same tables sit on each audit page under the drawdown section, and those pages are free to read.

A drawdown figure says how deep an account went and nothing about how long it stayed there. Lizard has a deepest end of day drawdown of 2.55%, which reads like something anybody could sit through. The account set that high on 2003.05.13, bottomed out on 2007.11.08 and did not see the old level again until 2026.06.18. The counter between those two dates reads 6,027 days, and every one of them was spent holding an account worth less than it had already been.

What the counter actually counts

The episode table is built from one balance value per weekday. Our engine takes each closed trade, writes the account balance on the day that trade closed, then walks the calendar from the first closing day to the last, keeps Monday to Friday and carries the last known balance across the quiet days. An episode begins at a peak in that series, runs to the lowest point that follows and ends on the day the balance touches the old peak again. The deepest episode reproduces the maximum end of day drawdown of the same audit, which is the self consistency check the engine states inside the table.

Two properties decide how the duration should be read. The field is labelled in trading days and the counter runs on weekdays, so a public holiday counts and a week without a single trade counts. And when an episode never recovers the counter simply stops at the end of the run, which makes that number a floor rather than a total.

One more thing belongs in front of the figures. The balance series behind these episodes already carries our commission retrofit while the tester head does not, because the retrofit is applied to the deals before the daily series is built. Scalping Robot Pro shows the seam. Its deepest episode bottoms out at 46,765.32 USD and the retrofit line of the same audit reports a final balance of 46,765.32 USD.

Deep and long are separate questions

Twister Pro Scalper holds the second longest repair among the 22 deepest episodes, behind Lizard. Its own is worth 114.74 USD or 0.11% of the account, it started on 2012.07.16, reached bottom on 2015.05.22 and closed on 2024.10.08. That is 3,191 days under water for a hole of just over a tenth of a percent, across 113 trades of which 25 lost.

The two orderings come apart inside single audits as well. In 8 of the 22 the deepest of the three stored episodes is not the longest of them. Thundergold Scalper is the clearest case. Its deepest hole is 31.79 USD and closed after 346 days, while the episode ranked second at 30.41 USD opened on 2013.06.25, reached bottom on 2022.03.07 and did not close until 2024.04.12, which is 2,818 days for a difference in depth of 1.38 USD.

Logan is the opposite shape. Its deepest episode is a single step in the series, from a peak of 100,636.82 USD on 2016.12.08 to 70,388.88 USD on 2016.12.09, which is 30,247.94 USD or 30.06% of the account. Exactly 20 trades closed inside that step and all 20 of them lost. That run holds at most 20 long positions at the same time and adds against the move in 285 of its entries, so a single closing basket moves the whole balance line. The run then ended 2,524 days later without ever regaining the high, and its 1,372 trades leave a final balance of 75,665.35 USD after our commission retrofit.

The seven that never came back

In 7 of the 22 audits with an episode table the deepest episode carries no recovery date. Two of them are young rather than broken. Gold Atlas was 70 days into a hole of 501.15 USD when the window closed and Prop Firm Gold EA was 69 days into one of 683.08 USD. Both counters were still running on the last day of the test, so both figures are floors.

Two others are terminal. Scalping Robot Pro set its peak on 2003.05.19 at 99,999.28 USD and hit its lowest point on the last day its balance series has, 6,063 days later and 53,233.96 USD lower, with 302,797 trades closed inside that one episode. Smart Gold Hunter peaked on 2003.05.06 and stayed below that level for 6,071 days, which is the entire remainder of its run. The other three are Adaptive Gold Scalper at 2,720 days, Logan at 2,524 and OilVector X at 740.

The seven counts only the deepest episode of each audit. Widen it to the three episodes every audit stores and 10 of the 22 hold one that never recovered, which adds CryonX EA at 142 days, Gold House at 22 and Gold Trade Pro at 85. An episode can only stay open at the very end of a run, so each of those ten runs finished below a high it had already made. Ten is a floor too, because an audit stores its three deepest episodes and a shallow open one is simply not in the file.

One episode is a diagnosis

Three of the 22 runs carry exactly one episode over the whole test. That is not evidence of a calm account. An episode only ends when the balance reaches its old peak, so a run with a single episode is a run whose balance never returned to a level it had left. Logan, Scalping Robot Pro and Smart Gold Hunter are those three and all three also sit in the group of seven above. Their shapes differ completely. Smart Gold Hunter set its peak on 2003.05.06, the second day of a window that opens on 2003.05.05, and never beat it. Logan traded from that same opening date until 2016.12.08 before one basket took the account down. Scalping Robot Pro fell from its second week to the last day its series holds.

The count runs the other way as well. CryonX EA and Gold House each record 354 episodes, so the balance of those runs dipped below a running high 354 times. The deepest of the CryonX dips is worth 184.66 USD and was repaired in 339 days. A high episode count and a low one describe two different lives, and neither is a verdict on its own.

Run these checks yourself

Every audit page prints its episode table with the peak, the trough, the recovery and the trades closed in between. Your own tester report? The browser check is free.

The honest limits. These episodes live on a balance series, so a position that stays open leaves no mark until it closes and the floating drawdown between two day boundaries is invisible here, which the audit states in the table itself. Durations without a recovery date end with the run and are floors, not totals. The counter runs on weekdays and therefore includes days on which nothing traded. Each audit stores only its three deepest episodes while the episode count covers all of them, and because that selection is made by depth the longest episode of a run need not be in the file at all. Four default audits carry no episode table at all because they were built with an older engine version than the one that introduced the drill down, and those four are Waka Waka, Market Anomalies EA, Pulse Engine and Quantum Emperor. Quantum Queen and Quantum Queen X report the same episode of 11 days because our deal level census found the two tester runs identical. Every window and every data world differs by audit, so these durations rank nothing across EAs.