The duel: BotFactoryFX against StrategyQuant X

The rules

Eight markets, all on daily bars: the Nasdaq 100 (twice, with different cut dates), the S&P 500, the Dow, the DAX, gold, EURUSD and GBPJPY. For each one we picked a cut date, 26 September 2022 for most of them. Both programs could only use the prices before it. Everything after, up to July 2026, stayed hidden: almost four years that neither program saw.

Each one worked its own way. StrategyQuant X ran the founder's own templates: half an hour of building per market, then its chain of robustness tests. BotFactoryFX ran its autopilot: twenty searches per market, about 48,000 candidates, every finalist scored out of 100 by its judge. Both paid the same costs per trade.

How do you know a pick was any good?

“It made money” is not enough. The Nasdaq doubled in those four years, and almost anything that bought made money. So we used a tougher yardstick: chance.

For each strategy we took its real trades in the hidden years and built 500 random twins: the same number of trades, the same direction, each held for the same time, but entered on random days. If the strategy did better than most of its twins, it beat random. A trader flipping a coin beats random about half the time, so 50% is the line to beat.

First result: nil–nil

With each program using its full method, neither approved a single strategy in any of the eight markets. StrategyQuant X built between 16,000 and 38,500 strategies per market, and each step of its chain let fewer through until nothing was left. BotFactoryFX's strictest verdict, 75 or more out of 100 and holding up again on a sister market such as the S&P for the Nasdaq, passed nothing either.

A nil–nil draw tells you both bars are high. It does not tell you which program picks better. So we turned the duel around.

Same exam for both

We gave StrategyQuant X the exam that BotFactoryFX's judge sets, test by test, built with StrategyQuant X's own tools. Does it still make money in years it was not built on? On a similar market? On 4-hour bars? Paying more per trade? With its trades shuffled, with prices nudged, with its settings nudged? And its own walk-forward test. Then, outside the program, the checks it does not have: better than random, enough trades, steady from one period to the next.

Now it approved strategies: 36 across the eight markets. On BotFactoryFX's side we looked beyond its strictest verdict, at every strategy its judge scored 45 or more. Those are not approved. BotFactoryFX calls them close to the bar and sends them to quarantine, to run on new prices before anyone risks money on them.

What StrategyQuant X approved beat random 56% of the time, 20 of 36: barely above the 53% of everything it kept from its builds. What BotFactoryFX scored 45 or more beat random 65% of the time, 91 of 140, with the stops exactly as they were judged. With stops that follow volatility, wider when the market moves a lot and tighter when it is calm, which is how BotFactoryFX sends a strategy to quarantine today, it was 80%.

In one sentence: StrategyQuant X approves more; what BotFactoryFX points to is right more often. But look at where that edge sits. Of BotFactoryFX's 140, 134 scored between 45 and 74, the middle band. Its top scores, 75 or more, were only six. All six beat random with their stops as judged, and six is too few to lean on.

It depends on the market

It was not a clean sweep. On the S&P 500, StrategyQuant X's 13 approved strategies beat random 12 times, the best result of the whole duel, ahead of BotFactoryFX's 18 of 24 with fixed stops. On the Nasdaq, StrategyQuant X managed 7 of 21. BotFactoryFX had its own bad market: the Nasdaq with the 2019 cut, where its picks beat random only 2 times in 14 with fixed stops, and 14 in 14 once the stops followed volatility.

Three things that tripped us up

A fair fight needs both fighters set up properly. Three things went wrong on the way, and we fixed them before counting.

• Our mistake: a spread from the wrong market. StrategyQuant X's high-precision retest carries a fixed spread in its settings. The template was made for the Nasdaq and we copied its number to every market, which in EURUSD meant a spread of 1,503 pips. It rejected everything, of course. We caught it, fixed it and ran the whole round again.
• Weekly bars that were not weekly. One test reruns each strategy on weekly bars. Ten years hold about 520 weekly bars, yet in that test the strategies made between 6,151 and 163,872 trades. Whatever it ran on, it was not weekly bars, so we took that test out.
• A walk-forward that read percentages as days. The walk-forward test kept failing with an error. The template described its windows with a setting StrategyQuant X does not recognise, and it read them as days instead of percentages. Set properly, it worked, and it let through 92% of the strategies that reached it.

None of this shows from the outside. They are settings, and the results depend on them. That is a lesson of its own: a robustness test is only as good as its configuration.

What it cost to run

In the five markets where we measured both, StrategyQuant X used between 43 and 82 times more processor time than BotFactoryFX, and between 25 and 31 times more memory. On a laptop, BotFactoryFX did the same job in under six minutes, still with half a gigabyte.

What this does not prove

• We built BotFactoryFX and we designed the duel. The exam is our judge's. StrategyQuant X ran with one trader's templates, not with the best set-up anyone could make. We tell you the mistakes we caught, and you should still read this as our measurement, not a neutral one.
• It is one stretch of time, 2022 to 2026, and mostly a rising market. Strategies that buy have it easier.
• Strategies on each side resemble each other, so 140 or 36 are fewer independent bets than they sound.
• None of these strategies has proven anything with real money. Beating random in hidden years is a good sign, not a verdict.

The real test has started

On 19 October 2026 we put both sides into quarantine on the S&P 500: StrategyQuant X's 13 approved strategies and BotFactoryFX's 24 strategies scored 45 or more. Each one runs on its own program's engine, on prices that arrive from that day on, and every week both are measured with the same yardstick. Every Saturday morning the score goes up on the Live duel page, whichever way it goes.

Before starting, we tested the set-up on the seven weeks from 1 September, which neither had seen either. BotFactoryFX's strategies won in all 23 that traded; StrategyQuant X's in 2 of the 12 that traded. Read that with care: one or two trades each, and many of them the same trade on the same day. It is a glimpse, not a result.

Then the quarantine itself started, and its first week went the other way: up to 23 October, 1 of BotFactoryFX's 6 strategies that had traded beat random, against 6 of 7 for StrategyQuant X. The second week turned it round again: up to 30 October, 22 of 22 against 4 of 11. Behind BotFactoryFX's 28 trades there are only 5 different ones, because its strategies resemble each other so much, against 11 different ones in StrategyQuant X's 18. Two weeks decide nothing. The Live duel page carries every week as it comes.

Corrected on 1 November 2026. The first version of this article gave StrategyQuant X 2 of 7 in the first week and 6 of 11, with 14 different trades in 21, in the second. Its quarantine prices had been loaded three hours out of step since 19 October, so its daily candles closed at the wrong time; we loaded them again and measured both weeks again. BotFactoryFX's figures did not change.

Changed on 4 November 2026: 13 against 13. From that day the Live duel fields 13 strategies a side: BotFactoryFX plays with the 13 its judge rated highest of those 24, with the scores it gave them before the duel began. Counted that way, BotFactoryFX beats random in 1 of 2 in the first week and in 12 of 12 in the second. StrategyQuant X's figures do not change, and BotFactoryFX's 15 trades are still only 5 different ones.

← All lab notes Try it free