Ilan 1.4 wins 74% of its trades and opens dozens a day. It reached +15% on a $10,000 account — then went to zero. Below is the full test: the conditions, the equity curve, the lot ladder, and the exact trades that killed it. I also backtested three ways to rescue it. None of them worked.
Balance (green) closes small profits all day. Equity (red) is what the account is actually worth.
This is the entire deception in one chart. The balance climbs smoothly to $11,537 because every basket eventually closes a few dollars in profit. The equity — the real account value with the open basket marked to market — collapsed to $2,085 with $9,465 floating in loss. The account blew up on 2014-01-06 01:02, 592 trades after it started.
Every 3 pips against it, Ilan adds a position 1.667× the size of the last.
| Level | This lot | Total lots | $ per pip against |
|---|---|---|---|
| L0 | 0.10 | 0.10 | $1 |
| L1 | 0.17 | 0.27 | $3 |
| L2 | 0.28 | 0.55 | $6 |
| L3 | 0.46 | 1.01 | $10 |
| L4 | 0.77 | 1.78 | $18 |
| L5 | 1.29 | 3.07 | $31 |
| L6 | 2.15 | 5.22 | $52 |
| L7 | 3.58 | 8.80 | $88 |
| L8 | 5.96 | 14.76 | $148 |
| L9 | 9.94 | 24.70 | $247 |
At level 10 the account holds 24.70 lots and every 1 pip against it costs $247. The loss is decided by this size, not by any signal. A $10,000 account cannot survive a few hundred pips of adverse move at that exposure.
This is what a martingale looks like five minutes from the edge.
On January 3 the EA added five levels in six minutes, going from 1.29 lots to 9.94 lots on a single basket. The entry price fell 12 pips while the position size grew 8×. There is no signal here — only the rule "keep adding until it turns around".
| Time | Type | Lots | Price | Level |
|---|---|---|---|---|
| 01-03 09:08 | buy | 9.94 | 1.36181 | L9 |
| 01-03 09:07 | buy | 5.96 | 1.36211 | L8 |
| 01-03 09:06 | buy | 3.58 | 1.36241 | L7 |
| 01-03 09:05 | buy | 2.15 | 1.36271 | L6 |
| 01-03 09:03 | buy | 1.29 | 1.36301 | L5 |
| 01-03 09:02 | buy | 0.77 | 1.36331 | L4 |
| 01-03 09:01 | buy | 0.46 | 1.36361 | L3 |
| 01-03 08:58 | buy | 0.28 | 1.36391 | L2 |
| 01-03 08:56 | buy | 0.17 | 1.36421 | L1 |
| 01-03 08:55 | buy | 0.10 | 1.36451 | L0 |
| 01-03 08:53 | buy | 0.10 | 1.36432 | L0 |
| 01-03 08:51 | buy | 0.10 | 1.36402 | L0 |
Hedging, hard stops, and a volatility filter. None of them beat the arithmetic.
A 0.5-lot hedge against a 24.7-lot basket covers 2% of the exposure. A fixed hedge cannot keep up with an exponentially growing basket. Blew up.
Hard stops at -$20/-$50/-$100 turn a fast blow-up into a slow bleed. The threshold is a knife-edge and no fixed number lands on the good side. Blew up.
A Kaufman efficiency filter ran Ilan only when the market looked like a range. Every threshold still blew up, because the death itself looks like a range.
The drawdown that kills a martingale is invisible until it is too late.
At the moment of death the efficiency ratio was 0.165 — a textbook range. The fatal damage happened on January 2–3, when EURUSD fell 170 pips and Ilan kept adding against it. By January 5–6 the price had flattened out and was oscillating gently — and that "safe range" was where the already-crippled account finished bleeding out. A trend does the damage; a quiet range delivers the killing blow.
A martingale cannot be saved by patching it. Its failure is not an event you can recognise early; it is an accumulation that only becomes visible in hindsight. Every rescue either catches the drawdown too late, or cuts so early that it destroys the one thing the system is good at.
If you insist on running a martingale, the honest options are not "hedge" or "filter". They are: trade it small enough that a blow-up is a survivable loss, cap the account at a hard equity stop so -100% becomes -20%, or replace it with an EA that has a real stop and a position cap built in. The win rate is seductive. The equity line is not.
Method note: Ilan closes at a 1-pip target, so a faithful replay needs tick data. This run used M1 bars; the intra-bar order shifts the exact peak between +15% and +51%. The conclusions — the blow-up, the failed hedges, the failed filter — are stable across every ordering tested. The replayer and the M1 source data are in the method notes.
Related: 372 free EAs on the same harness — a third of them never opened a single trade, and 82.6% of the ones that did lost money. Or the arithmetic behind a 5-pip target, which is the one thing that decides whether a scalper like this can work at all.