A backtest that runs from 2003 to 2026 has not necessarily been tested since 2003. The window is what the tester was told to cover. The first trade is when the EA actually did something. Across the 22 default audits that carry the measurement, the gap between those two dates runs from 1 day to 6,086 days. Eight of the 22 spent more than two years inside their own window without opening a single position.
The window comes from the tester report itself, the line that names symbol, timeframe and period. The first entry comes from the deals table, taking the earliest deal that is a buy or a sell with direction in, so closing legs and balance operations cannot move it. Both are exact, and the audit labels them that way. From them the engine derives two figures, the delay in whole days and the effective years, which is the span from the first trade to the end of the window.
The second figure is the one worth carrying around. It answers the question a reader actually has, which is how much history this result rests on.
Quantum Titan was tested from 2003-05-05 to 2026-08-23. Its first trade is dated 2020-01-02 15:19:54. That is 6,086 days of window with nothing in it, and the audit reports 6.64 effective years. For comparison, Gold House starts on the same 2003-05-05 and reports 23.28 effective years, because its first trade lands on 2003-05-06. Same start date on the period line, and two very different amounts of history behind the result.
Three further audits share the same first entry down to the second. Quantum Athena X, Quantum Queen X and the Quantum Queen document that republishes the same run all open at 2018-01-17 03:00:00, 5,371 days after the window starts, for 8.58 effective years. Athena X and Queen X are not the same run, they closed trades on 902 and 820 days respectively, and they still agree on the first entry down to the second. A start time two runs share that exactly is a property of the software rather than of the prices we fed it.
Three M15 runs begin trading in the same year. Twister Pro Scalper opens on 2006-01-06 after 977 days, Smart Gold Impulse on 2006-04-20 after 1,081 days and ThunderGold Scalper on 2006-05-04 after 1,095 days. All three windows open on 2003-05-05. Twister Pro Scalper and ThunderGold Scalper carry the same vendor name in our catalog and Smart Gold Impulse carries a different one, so at least one of the three reaches that year on its own.
The obvious objection is that the price history simply started later. It did not. Our audits carry a coverage figure for exactly this question, the reported bars converted into minute equivalents and held against the weekdays in the window. All three of these runs report 1.0005, so the bars were there and the EAs did not use them. The engine treats coverage below 0.50 as a missing data head and says so instead of blaming the EA, and none of the 22 audits falls into that class.
Most of the catalog starts trading almost immediately. Gold House and Smart Gold Hunter need 1 day, Gold Atlas and OilVector X need 3, Scalping Robot Pro needs 9 and The Gold Reaper needs 22. A short delay is what an indicator warmup looks like, and it is the reason the measurement exists as a scale rather than a yes or no. Eleven of the 22 audits carry an ok lamp here.
In the middle sit two runs on the daily timeframe. Gold Trade Pro waits 142 days and Ultimate Breakout System waits 151 days. On a daily chart a wait measured in months is a far smaller number of bars than the same wait on a minute chart, which is exactly the trap the next section is about.
Range Breakout EA with Range Filters starts trading on 2017-01-03 08:35:22, 2 days into its window, and still carries a caution lamp. That is not an error, and it is the reason this article quotes days rather than lamp colours. The run is on M1, where two calendar days are 1,895 bars of the tested timeframe, and the audit stores that figure next to the days. Our convention changed on 2026-08-31 from counting calendar days to counting bars of the tested period, because a warmup is a bar count and not a duration. Range Breakout is the only audit in the catalog rebuilt since that change, so the other 21 audits quoted here still carry lamps from the older calendar day rule while their delays themselves are unaffected. We would rather print that sentence than let a reader compare two lamps that were not computed the same way.
Every audited EA carries these dates on its audit page with the deal level evidence behind them. Your own tester report? The browser check is free.