Percent return and max drawdown are arithmetic on fallback deposit, fixed lots, no compounding. USD figures are divided by 1,000 USD, the fallback deposit (vendor names none). Run capital of 100,000 USD is the measurement frame, not the story basis.
Not measured here. live signal performance, individual risk sizing, alternative settings. Declared alternatives can be measured as separate matrix runs.
Sizing invariant Profit concentration, Active trading days, Trade count, Spread sensitivity per trade, Drawdown duration in trading days, Holding times, Win rate.
Sizing dependent CAGR, Drawdown in percent, Absolute PnL.
AI Algo Gold is sold by Artem Marukhov for a listed $99.00 on MQL5 Market.
The observed backtest (red) against 1000 resampled orderings of its own trades. Median dashed, bands P25 to P75 and P5 to P95. Hover or touch for exact values per trading day.
7 findings, 2 of them red.
9,782 trades, rebuilt deal by deal.
sampled at deal timestamps. Intrabar floating equity between deals is unobservable from a report, so the tester-head equity DD (tick-based) is the upper reference before added commissions; validation allows their total cost divided by initial capital, scaled by the native drawdown bound above 100%
The measurement record
| Trades reconstructed | 9,782 | exact |
| Pairing confidence | exact | |
| Balance reconstruction | exact | |
| History quality | 100% echter Ticks | exact |
| Real ticks from | XAUUSD: 2003.05.05 00:00:00 | |
| Generated tick minutes | n/a | |
| Execution delay | 10 ms | |
| Tester log provided | yes | |
| Tester log files | 20260913.log | |
| Randomizer prints | 0 | |
| Report SHA-256 | cff2d23ff66610bd... |
Reconstructed floating equity
| Floating drawdown (sampled) | 2.4% | reconstructed |
| Balance drawdown (deal-wise) | 2.4% | reconstructed |
| Report-head equity DD (tick-based) | 2.21% (2 217.23) | exact |
| Equity samples (deal timestamps) | 19,564 | |
| Worst sampled point | 2015-03-16 19:00:00 at 97,655.63 (from 100,009) | reconstructed |
| Equity between deals | unobservable in a tester report | unavailable |
Settings used in this run: ai-algo-gold.set (Version 1.0; 950 bytes). Measurement configuration of this audit, not a recommendation.
One strategy, one exit signature.
How trades actually ended
The symbol itself moved +1,151.2%.
approximated from entry prices, because a tester report carries no independent price series. HOLD is buying and holding the median traded lot size over the same span, before costs, using the contract value calibrated from this run's own paired deals. It is additive like the strategy PnL, so the three bars share one USD scale. Percent figures in the tooltips are arithmetic on the vendor deposit, no compounding.
Underlying table
| Symbol | Trades | Entry-price move | Traded span | Long/Short trades | Long PnL | Short PnL | Buy and hold |
|---|---|---|---|---|---|---|---|
| XAUUSD | 9,782 | +1151.2% | 2003-05-08 … 2026-08-13 | 5818/3964 | 5,936 (+593.6%) | 7,276 (+727.6%) | 4,036 (+403.6%) |
approximated from entry prices, because a tester report carries no independent price series. The buy-and-hold column is the entry-price move expressed in USD at the median traded lot size, before costs, using the contract value calibrated from this run's own paired deals. It is additive like the PnL columns. Percent figures in brackets are arithmetic on the vendor deposit, no compounding.
24 calendar years on record.
Year-by-year table
| Year | Trades | PnL | Win % | Long PnL | Short PnL |
|---|---|---|---|---|---|
| 2003 | 173 | -132 | 39.9% | -126 | -6 |
| 2004 | 317 | -145 | 43.5% | -146 | 0 |
| 2005 | 291 | -40 | 45.4% | -40 | 1 |
| 2006 | 422 | -357 | 44.1% | -313 | -44 |
| 2007 | 389 | 83 | 48.6% | 120 | -36 |
| 2008 | 436 | -932 | 46.8% | -920 | -13 |
| 2009 | 386 | 562 | 54.4% | 415 | 147 |
| 2010 | 393 | 26 | 47.3% | 19 | 7 |
| 2011 | 463 | -131 | 48.2% | -58 | -73 |
| 2012 | 485 | -160 | 50.1% | -238 | 78 |
| 2013 | 371 | -779 | 47.2% | -932 | 154 |
| 2014 | 392 | -80 | 48.2% | -159 | 79 |
| 2015 | 443 | -38 | 47.6% | -294 | 257 |
| 2016 | 370 | 563 | 48.9% | 316 | 248 |
| 2017 | 366 | 391 | 51.9% | 187 | 203 |
| 2018 | 375 | 475 | 53.9% | 179 | 296 |
| 2019 | 356 | 436 | 48.0% | 166 | 270 |
| 2020 | 418 | 1,020 | 49.0% | 642 | 378 |
| 2021 | 397 | 604 | 52.1% | 44 | 560 |
| 2022 | 460 | 752 | 48.5% | 334 | 418 |
| 2023 | 440 | 1,123 | 53.0% | 671 | 451 |
| 2024 | 414 | 1,871 | 50.7% | 1,481 | 390 |
| 2025 | 555 | 4,030 | 53.7% | 3,020 | 1,010 |
| 2026 | 670 | 4,069 | 54.5% | 1,568 | 2,501 |
Costs include the commission retrofit.
From gross to net
Commission retrofitted at 3.5 USD per lot per side on every deal, not tester-native. The balance chain was rebuilt deterministically.
Which Sharpe to trust
Cost and metric table
| Net profit | 13,211.24 | reconstructed |
| Commission | -782.56 | reconstructed |
| Commission model | commission retrofitted at 3.50/lot/side, not tester-native | |
| Swap | -6,203.58 | reconstructed |
| Gross before costs | 20,197.38 | reconstructed |
| Cost share of gross | 34.6% | reconstructed |
| Overnight trades | 4,744 | reconstructed |
| Tested history | First trade 182 bars into the window. The tested history effectively starts 2003-05-08 | |
| EOD Sharpe (annualized) | 1.01 | reconstructed |
| Report-head 'Sharpe' | 1.13 | exact |
| CAGR | 0.5% on 100,000 USD run capital | reconstructed |
| Max drawdown (EOD) | 235.2% of 1,000 USD fallback deposit | reconstructed |
80% of the profit happened on 65 of 3,215 active trading days (2.0%). That is 1.1% of the 23.4 year window.
That is 2.0% of 3215 trading days. Miss a handful of days and the edge is gone. Consistency rules at prop firms punish exactly this profile.
Best days
Worst days
The deepest drawdown recovered after 4,387 trading days.
Episode table
Top 3 of 120 episodes on the end-of-day balance curve.
| # | Peak | Trough | Depth | To trough | Underwater | Trades (losers) | Worst trade |
|---|---|---|---|---|---|---|---|
| 1 | 2003-06-03 (100,007) | 2015-03-16 (97,656) | 2,352 (2.4%) | 3074 d | 4387 d (recovered 2020-03-26) | 4603 (2422) | -75 on 2012-06-08 |
| 2 | 2026-04-06 (111,164) | 2026-04-29 (110,514) | 650 (0.6%) | 17 d | 36 d (recovered 2026-05-26) | 68 (42) | -73 on 2026-04-20 |
| 3 | 2025-12-25 (109,419) | 2026-01-02 (108,883) | 536 (0.5%) | 6 d | 14 d (recovered 2026-01-14) | 21 (18) | -52 on 2025-12-29 |
end-of-day balance curve, peak to recovery, the same series and formula as 'Max drawdown (EOD)', so the deepest episode reproduces that figure. trades listed are those CLOSED between the peak and trough day. Floating drawdown between day boundaries is invisible here, see the reconstructed-equity section where present.
34.2% of all start dates end in a dead account.
If you had started this EA fresh on any day of the tested history with a fresh 1,000 USD account (the standard fallback deposit. The run itself used 100,000 USD as measurement frame), 34.2% of all 19,564 start dates end in a dead account. The strip shows the measured fate of every possible start date. The curve below it shows why.
Death zones for a fresh account: 2003-05-08 to 2008-08-15, 2008-08-22 to 2008-09-11, 2008-09-17 to 2008-10-02, 2008-10-08 to 2008-10-10, 2009-02-20 to 2009-02-26, 2009-03-18 to 2009-03-20, 2009-03-20 to 2009-03-24, 2009-05-20, 2009-05-21 to 2009-06-08, 2009-10-15, 2009-10-20 to 2012-05-23, 2012-05-28 to 2012-06-08, 2012-06-14 to 2012-06-25, 2012-07-02 to 2012-07-10, 2012-07-10, 2012-07-11 to 2012-08-03, 2012-08-10 to 2012-08-15, 2012-08-20 to 2012-08-30, 2012-08-31 to 2013-02-12. The deepest episode sits 2,353 USD below the peak of 2003-05-20, trough on 2015-03-16.
Drawdown ribbon
The same history as one color band: how deep the account sat under its running peak in any week, measured against the 1,000 USD basis. Red weeks are the phases where a fresh account entering just before would have been wiped out.
Computed death chain. Restarting a fresh 1,000 USD account after every wipeout, the full history costs 2 account deaths (balance to zero on the closed curve). The mean run lifetime is 1848 days. The final run is alive for 4802 days since 2013-06-20.
Deposit ladder
| Deposit | Share of dying start dates | Computed chain deaths |
|---|---|---|
| 1,000 USD (fallback) | 34.2% | 2 |
| 2,000 USD | 10.0% | 1 |
| 4,000 USD | 0.0% | 0 |
Basis: closed balance curve, where floating drawdown would only be stricter. Fixed lot settings only (measured policy fixed). For compounding EAs this map is not transferable. Evaluation deposit 1,000 USD (fallback 1000 (vendor names no deposit)). Run capital 100,000 USD is the measurement frame. Small gap on each restart day of a chain is inherent to the tester workflow.
Only 0 days are guaranteed out of sample.
Net 0.00 USD at profit factor None out of sample, against 1.26 before release.
Product facts: released 2026-09-08, 0 recorded version entries, list price 99 USD.
The backtest runs the CURRENT version over the whole history. Every candle before the release was visible while the EA was built. Only the window after the last update is guaranteed out of sample, and it must be judged together with its length.
Price on record: 99 USD since 2026-09-11 (tracking baseline, append-only archive).
Window table
| Window | From | To | Days | Trades | Net USD | PF | Win rate |
|---|---|---|---|---|---|---|---|
| Before release | 2003-05-08 | 2026-08-13 | 8,499 | 9,782 | 13,211.24 | 1.26 | 49.5% |
| Release to last update | n/a | n/a | 0 | 0 | 0.00 | n/a | n/a |
| After last update (out of sample) | n/a | n/a | 0 | 0 | 0.00 | n/a | n/a |
The whole edge dies at +1.35 USD extra cost per trade.
Add that much cost to every trade (worse spread, slippage, commission) and the result is gone. The average position size is 0.01 lots.
Cost shock
Remove the best N trades
Sensitivity tables
| Extra cost per trade | Net profit | Profit factor | Win % |
|---|---|---|---|
| +0.50 USD | 8,320 | 1.153 | 44.9 |
| +1.00 USD | 3,429 | 1.06 | 41.1 |
| +2.00 USD | -6,353 | 0.899 | 35.0 |
| +5.00 USD | -35,699 | 0.574 | 25.4 |
| Best trades removed | Their PnL | Share of gross wins | Net without them |
|---|---|---|---|
| 1 | 73 | 0.1% | 13,138 |
| 5 | 354 | 0.5% | 12,858 |
| 10 | 634 | 1.0% | 12,577 |
| 20 | 1,161 | 1.8% | 12,050 |
flat USD shock per trade, not lot-scaled, so judge it against the average lot size. Leave-best-out removes the N most profitable trades from the observed set.
3 rule sets, simulated day by day.
IQ Capital Classic (funded)
Account deaths per sizing
2 breach years under the consistency rule
Rules used for this simulation (as of 2026-08-13)
| dd_mode | eod_trailing |
| dd_pct | 6.0 |
| max_position_loss_pct | 0.5 |
| consistency_pct | 30.0 |
| overnight_allowed | True |
| weekend_allowed | True |
Source: https://support.iqcapital.io
Worst days: -328 on 2025-10-21, -302 on 2026-04-28, -259 on 2026-01-02 · 4744 overnight trades · 1788 spanning weekends
FTMO Challenge
Account deaths per sizing
The intraday row samples equity against the day anchor (server-day boundary). The EOD row is shown for comparison.
Rules used for this simulation (as of 2026-08-13)
| dd_mode | static |
| dd_pct | 10.0 |
| daily_loss_pct | 5.0 |
| overnight_allowed | True |
| weekend_allowed | True |
Source: https://ftmo.com/en/how-it-works/ + academy/maximum-daily-loss + faq weekend
Worst days: -328 on 2025-10-21, -302 on 2026-04-28, -259 on 2026-01-02 · 4744 overnight trades · 1788 spanning weekends
Generic 6% EOD trailing
Account deaths per sizing
Rules used for this simulation (as of 2026-08-12)
| dd_mode | eod_trailing |
| dd_pct | 6.0 |
| overnight_allowed | True |
| weekend_allowed | True |
Source: generic, no firm source
Worst days: -328 on 2025-10-21, -302 on 2026-04-28, -259 on 2026-01-02 · 4744 overnight trades · 1788 spanning weekends
The same trades, a thousand other orderings.
1000 paths per method, seed 42, block length 5 trading days, additive resampling at the tested sizing. Drawdowns are expressed as % of start balance. Paths are not stopped at account death, so drawdowns beyond 100% mean repeated wipeouts at this sizing.
Resampled max-drawdown range vs. observed
Bars and observed marker use percent of start balance.
Bootstrap intervals
The share of resampled paths ending at or below zero net profit is 0.0%.
Streaks and recovery (block-bootstrap paths)
| Metric | Median | P90 | P95 | P99 |
|---|---|---|---|---|
| Max losing streak (days) | 6 | 7 | 8 | 9 |
| Time under water (days) | 621 | 1131 | 1327 | 1803 |
Ruin probability per profile and sizing
| Profile | 1/1 | 1/1.5 | 1/2 | 1/3 | 1/4 | 1/6 | 1/8 |
|---|---|---|---|---|---|---|---|
| iqcapital_classicfloor 6.0% | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| ftmo_challengefloor 10.0% | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| generic_6pct_trailingfloor 6.0% | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
Cell = probability that a resampled path breaches the profile floor at that sizing, in percent. Green ring = first sizing at or below the 10% target. This checks the floor only. The full rule set is stricter.
Challenge pass probability, ftmo_challenge
phase1: pass 89.5%, fail 0.0%, undecided 10.5%.
phase2: pass 100.0%, fail 0.0%, undecided 0.0%.
The probability of passing both phases (independent-resample approximation) is 89.5%.
Attempt economics. Passing phase 1 takes 1.1 expected attempts, a 90% chance of at least one pass needs 2 attempts, and the probability of 5 consecutive fails is 0.0%. multiply attempts by your challenge fee for the expected cost to fund.
Challenge detail tables
ftmo_challenge. Daily and floor checks are based on intraday-sampled equity vs. day anchor.
| Stage | Target | P(pass) | P(fail: floor) | P(fail: daily) | Undecided | Median days to pass |
|---|---|---|---|---|---|---|
| phase1 | 10% | 89.5% | 0.0% | 0.0% | 10.5% | 4377 |
| phase2 | 5% | 100.0% | 0.0% | 0.0% | 0.0% | 2214 |
Does reducing risk raise the pass chance? Phase 1 at each sizing.
| Sizing | 1/1 | 1/1.5 | 1/2 | 1/3 | 1/4 | 1/6 | 1/8 |
|---|---|---|---|---|---|---|---|
| P(pass) | 89.5% | 28.7% | 0.6% | 0.0% | 0.0% | 0.0% | 0.0% |
| Median days | 4377 | 5597 | 5901 | n/a | n/a | n/a | n/a |
- Monte Carlo resamples the SIMULATED trades of this backtest. It inherits every weakness of the simulation (in-sample bias, regime luck, execution gap) and cannot add information. All it does is reveal how fragile the observed path is.
- Same trades, different order. This answers only how path-dependent the drawdown is. Treats trades as exchangeable, which grid and martingale sequences are not (their order is caused by the market path), so read this as a diagnostic, not a forecast.
- Draws trades independently with replacement (IID assumption). Real EAs violate IID through clustering and position chains. Intervals are tighter than reality for such systems.
- Resamples 5-trading-day blocks of the end-of-day PnL series. The blocks are circular and wrap at the series end, so every day carries equal weight. Without wrapping, a crash on the final day would be undersampled. This preserves short-range clustering and losing streaks up to the block length. Longer regime shifts remain unmodeled.
- Floor breaches only, on block-bootstrap paths over a horizon equal to the tested history, profits above start swept (same convention as the prop-fit sweep). The firms' full rule sets (daily loss, consistency, position loss) are stricter, so treat every probability as a lower bound, and the safe-sizing answer as an upper bound on the truly safe size.
- The fan bands are pointwise percentiles across resampled paths. The envelope is not a single achievable path. Paths use the tested sizing and are not stopped at account death.
- Challenge-pass probabilities run each resampled path until profit target, max-loss floor or a daily-loss day is hit (checked in that day's order, daily first, then floor). Where the audit has reconstructed equity, daily-loss and floor checks use the intraday equity drop below the day anchor (sampled at deal timestamps, server-day boundary), which is still a lower bound on breaches, but far closer to the firms' equity-based rules than EOD deltas. Phases are treated as independent resamples for the combined figure. The sizing curve keeps the targets fixed while scaling the trading, which is why less risk can also mean more undecided paths.
The longest dry stretch lasted 201 months.
Lot policy detected: fixed. The monthly rule (monthly sweep to start balance) covers the active span 2003-05-08 … 2026-08-13.
Monthly PnL heatmap
Across 1000 resampled paths (fixed lots), total withdrawn spans 9,112 to 13,390 to 17,657 USD (P5/median/P95). The share of paths paying nothing at all is 0.0%.
Capital what-if (fixed lots)
Withdrawal and what-if tables
| Sizing model | Months paid | Total withdrawn | Withdrawn %/a | Median paid month | Best month | Dry streak (months) |
|---|---|---|---|---|---|---|
| fixed_lots | 57/280 | 13,212 | 0.55 | 172 | 954 | 201 |
| balance_scaled | 57/280 | 13,206 | 0.55 | 172 | 957 | 201 |
| Capital | MaxDD % | P(breach) iqcapital_classic | P(breach) ftmo_challenge | P(breach) generic_6pct_trailing |
|---|---|---|---|---|
| 1,000 (vendor basis) | 235.16 | 100.0% (hit) | 100.0% (hit) | 100.0% (hit) |
| 2,500 | 94.06 | 100.0% (hit) | 100.0% (hit) | 100.0% (hit) |
| 5,000 | 47.03 | 100.0% (hit) | 100.0% (hit) | 100.0% (hit) |
| 10,000 | 23.52 | 99.3% (hit) | 52.2% (hit) | 99.3% (hit) |
| 25,000 | 9.41 | 8.9% (hit) | 0.2% | 8.9% (hit) |
| 100,000 (tested) | 2.35 | 0.0% | 0.0% | 0.0% |
(hit) = the observed backtest itself breached this floor at that capital.
- Withdrawals are simulated monthly on the last trading day. Everything above the start balance is taken out. A month without surplus pays nothing. The dry-streak figure shows how long that can last. Paths are the OBSERVED backtest days. Replay deltas start after the first sampled weekday balance, excluding the move from the initial deposit to that sample. The withdrawal distribution adds resampled paths (same circular block bootstrap as the MC section, own seed).
- balance_scaled multiplies each day's PnL by balance/start. This is an approximation. Min-lot granularity and balance-coupled EA defaults do not scale linearly. The direction is optimistic for small accounts.
- Capital what-if keeps the observed USD path (fixed lots) and re-expresses risk against each capital. Margin requirements are NOT modeled. A small account may be unable to hold the positions at all. The direction is optimistic. Under balance_scaled sizing, percent metrics are unchanged by construction. Only the min-lot limit differs, which is why no table is shown.
Every term, in plain language.
Glossary
| Pairing confidence | How entry and exit deals were matched into trades. 'Exact' means taken from the tester log, 'validated' means FIFO/LIFO reproduced the report's holding-time figures, and 'heuristic' means an unconfirmed FIFO assumption. |
| EOD Sharpe (annualized) | Sharpe ratio computed from end-of-day balance returns, annualized with the square root of 252. Comparable across systems, unlike the report-head 'Sharpe', which is per-trade. |
| Cost share of gross | Commission plus swap as a share of gross profit before costs. High values mean the edge is eaten by fees and financing. |
| Max drawdown (EOD) | Largest peak-to-trough loss of the end-of-day balance curve. |
| Floating drawdown (sampled) | Largest drawdown of reconstructed equity (balance plus open-position value), sampled at deal timestamps. Between deals equity is unobservable from a report. The tester-head equity DD is tick-based. |
| Martingale | Position sizing that grows after losses. Looks smooth for months, then loses the account in one streak. |
| Consistency rule | Prop-firm rule capping the best day's share of total profit. It punishes concentrated profit profiles. |
| Account deaths | Number of times the simulated account breached the profile's drawdown floor over this history (account is then reset and the simulation continues). |
| Withdrawal %/a | Yearly withdrawal as percent of account size in the sweep simulation (profits above start are swept daily). |
| Z-Score | Serial correlation of the win/loss sequence. Strongly negative values often just reflect several sub-strategies interleaving, not necessarily a defect. |
| Undecided | A resampled challenge path reached neither its target nor a failure boundary within the sampled horizon. Each path spans the tested history's daily observations, not the duration of a real challenge. Undecided does not mean failed. |
| Monte-Carlo resampling | Re-arranging or re-drawing the audited trades many times to see the range of drawdowns and streaks the same trading could have produced. It cannot add information. It reveals path fragility, not future returns. |
| Block bootstrap | Bootstrap that draws whole multi-day blocks of the daily PnL series instead of single trades, preserving short-range clustering (losing weeks stay losing weeks). |
| Ruin probability (floor only) | Share of resampled paths that breach the profile's drawdown floor at least once, with profits above start swept. The firm's full rule set is stricter, so this is a lower bound. |
| Time under water | Longest stretch of trading days a path spends below its previous balance peak. |
| Withdrawal replay | Re-plays the backtest with a monthly payout. On the last trading day of each month, everything above the start balance is withdrawn. Shows what the strategy pays a trader who lives off it, instead of compounding like a backtest. |
| Dry streak | Longest run of consecutive months in which the monthly withdrawal was zero, the months a trader living off the account would have earned nothing. |
| Lot policy | Whether the EA trades fixed lot sizes or scales them with the balance, detected from the variation of lot sizes across the deal list. |
| Capital what-if | The observed USD path re-expressed against a different account size (fixed lots). Drawdown percentages and floor-breach risk change with capital even though the trades are identical. Margin limits are not modeled. |
| Underlying vs. strategy | The symbol's own price move over the traded span, approximated from entry prices. If most profit is long-side while the symbol itself rose strongly, part of the result is the market's tailwind, not the mechanics. The tester cannot separate the two. |
| Drawdown episodes | One episode runs from a balance peak through its deepest trough until the peak is regained (peak to recovery), on the end-of-day balance curve. The deepest episode is the max drawdown (EOD). The table shows which trades and sub-strategies dug each hole. |
| Take-home after tax | Total withdrawn multiplied by one minus a user-supplied flat rate. Pure arithmetic on the withdrawal table, no tax law modeled. It stays off unless a rate is provided. |
| Cost fragility | How quickly the result dies as per-trade costs rise. The break-even shock is the extra cost per trade (spread, slippage, commission) at which net profit reaches zero. High-frequency systems often die at cents. |
| Fresh start survival | With fixed lots the dollar swings that follow any date are identical for every account, so each start date has an exact, measured fate. A fresh account started there dies if the balance path later falls at least one deposit below its starting level. Death-chain runs concatenate into one continuous profit path. Computed from the closed balance curve, where a floating check would only be stricter. |
| Release timeline | The backtest runs the current EA version over history that mostly predates it. Candles before the release were visible during development, and candles up to the last update were visible to at least one refit. Only the window after the last update is guaranteed out of sample, and it must be judged together with its length. |
| Provenance labels | Where a number comes from. 'Exact' is taken as printed from the tester report or journal. 'Reconstructed' is recomputed by the engine from the deal list and journal, deterministic but a model of the run. 'Unavailable' cannot be derived from a report at all and is stated instead of silently approximated. |
The honest boundary.
A good backtest cannot prove an edge. This audit dissects the simulation you gave it. It can expose structural risks (martingale, grids, hidden concentration, cost fragility, rule conflicts), but it cannot tell you that the strategy will make money. Three limits are fundamental.
- In-sample bias. Commercial EAs are typically released and updated after most of the tested window. A strong report may replay what the vendor optimized against, not what the market will do.
- Regime dependence. A profitable window can reward almost any mechanic aligned with the prevailing trend. Year slices and long/short splits above hint at this, but cannot settle it.
- Simulation gap. Even with real ticks and execution delay, the tester does not model live spreads, slippage beyond the next tick, or broker-specific behavior.
This audit is software analysis, not investment advice, and contains no recommendation to buy or trade anything.
This audit is free, in full.
This audit is free. The list price of this EA is at or under 200 USD, so the complete audit is open. Every number, table and simulation on this page is the full deliverable.
Straight answers, measured where possible.
We measure it instead of guessing. The audit replays three prop rule sets (IQ Capital, FTMO, a generic 6% trailing profile) day by day over all 9,782 reconstructed trades, then runs the challenge as a Monte Carlo simulation at seven position sizings, with intraday equity resolution. The pass probability curve for AI Algo Gold MT5 is inside the audit.
The audit contains a withdrawal replay. The strategy is re run as if someone lived off it, with monthly payouts, months paid versus dry months and the longest dry streak, plus a safe sizing table answering how small you must trade to keep ruin risk under 10%. The numbers for AI Algo Gold MT5 are inside the audit.
We classify structure from the deal list, not from the equity curve. Split tickets, adds into adverse moves and size escalation are counted, not guessed. The audit states the classification for AI Algo Gold MT5 with the exact counts behind it.
Vendor live signals and social screenshots are outside our control, so we do not grade them. What we audit is a controlled strategy tester run under one documented protocol, real ticks, vendor defaults, tester log included. That evidence is reproducible and cannot be cherry picked after the fact. The audit shows exactly what that run supports for AI Algo Gold MT5.
The vendor's own recommended defaults, unchanged. Deposit, leverage and every input follow the product page or the vendor's published set, and the audit records the full input list from the tester run. Where a preset series exists, each preset is audited separately and the page names the one you are reading.
We do not answer that with an opinion. The audit answers it with measurements. The verdict lamps on this page are public, and behind them sit the exact numbers, the cost autopsy, the concentration test, the prop rule replay and the Monte Carlo stress for AI Algo Gold MT5. Read those and the answer becomes yours, not ours.
Audits of EAs listed at 200 USD or less are free. The risk of a cheap EA is not worth a paywall, and open audits show the depth of the method. Audits of more expensive EAs are paid.
Protocol: how we test · All audits · Audit your own report · Blog
Already running AI Algo Gold? Check your live account against this audit, free.
More grid audits: Quantum Queen X · The Gold Reaper · Scalping Robot Pro
Browse by category: Gold EAs · Martingale EAs · Grid EAs · Prop firm EAs · Free audits · Quantum family