StrategyQuant X · Real funnel
How Many Strategies Do You Need to Generate in StrategyQuant X?
Everyone who starts with StrategyQuant X asks the same thing: how many strategies must the Builder produce before something survives the robustness checks? Here is one honest answer from the logs of a real five-week run.
Quick answer: in our run 130.7 million generated strategies left 669 that passed every check. That is about one in 195,000, or roughly one final strategy per hour of server work. Your ratio depends on your filters, so use the method below to scale it.
The funnel
130.7 million in, 669 out
One dedicated server ran StrategyQuant X almost non-stop from August 4 to September 9, 2026. Fifteen project templates chained in an automated pipeline, genetic evolution, MetaTrader 5 engine, about 7 years of data, timeframes from M30 to H4. All counts come from the SQX project logs.
| Stage | Strategies | Share of previous stage | Share of all generated |
|---|---|---|---|
| Generated and backtested by the Builder | 130,747,057 | 100.00% | |
| Accepted by the Builder filters | 374,397 | 0.29% | 0.2864% |
| Kept by the quality filter and sent to the OOS check | 66,691 | 17.81% | 0.0510% |
| Passed the OOS check | 51,061 | 76.56% | 0.0391% |
| Kept for the heavy robustness checks | 7,270 | 14.24% | 0.0056% |
| Passed every check and reached final storage | 669 | 9.20% | 0.0005% |
What "final" means here: a strategy that passed the Builder filters, the quality filters between stages, OOS, Trade Manipulation, SPP, Randomize Data, Walk Forward and a tick retest. It says nothing about future profit. It only means the strategy survived the tests chosen for this project.
Stage 1
The Builder rejects 99.71% of what it generates
Most of the funnel is lost at the very first step, and that is by design: early filters are cheap, later checks are expensive. In this run the Builder accepted 374,397 strategies out of 130,747,057.
| Why the Builder rejected a strategy | Share of rejections |
|---|---|
| Net profit (OOS) is zero or negative | 20.2% |
| No trades | 18.9% |
| Did not get into the initial population | 16.9% |
| Too few trades | 13.2% |
| Trading months below 85% of the period | 11.7% |
| Open drawdown above 150% of in-sample | 10.4% |
| Many trades close on the same bar | 4.9% |
| Net profit (IS) is zero or negative | 2.4% |
| Yearly return at or below 3% | 1.1% |
| Other | under 0.2% |
If "no trades" and "too few trades" make up a third of your rejections, as they did here, the template is producing many dead candidates. Tightening the building blocks saves more time than faster hardware.
Acceptance by timeframe
Higher timeframes generate faster but accept a smaller share. Accepted strategies per hour is the number that matters for planning.
| Timeframe | Strategies generated | Accepted by Builder | Accepted % | Odds | Accepted per hour |
|---|---|---|---|---|---|
| M30 | 8,328,416 | 54,000 | 0.65% | 1 in 154 | 1,116 |
| H1 | 20,275,251 | 153,600 | 0.76% | 1 in 132 | 2,666 |
| H2 | 18,838,113 | 85,454 | 0.45% | 1 in 220 | 3,067 |
| H4 | 83,305,277 | 81,343 | 0.10% | 1 in 1,024 | 1,433 |
Acceptance by trade direction
| Direction | Strategies generated | Accepted by Builder | Accepted % | Odds | Accepted per hour |
|---|---|---|---|---|---|
| Long and short | 8,238,008 | 117,600 | 1.43% | 1 in 70 | 3,884 |
| Long only | 34,357,650 | 110,401 | 0.32% | 1 in 311 | 2,113 |
| Short only | 88,151,399 | 146,396 | 0.17% | 1 in 602 | 1,354 |
Long-and-short templates were the slowest to generate and the most productive: almost twice as many accepted strategies per hour as long-only, and nearly three times as many as short-only.
Stages 2 to 5
What the robustness checks remove
After the Builder, a quality filter kept 17.8% of accepted strategies for the OOS check, and a second one kept 14.2% of the OOS survivors for the heavy checks. Each individual check then passed about 77% to 85% of what it tested.
| Check | Strategies checked | Passed | Pass rate | Hours | Seconds per strategy |
|---|---|---|---|---|---|
| OOS (out of sample) | 66,691 | 51,061 | 76.6% | 3.2 | 0.2 |
| Trade Manipulation | 7,270 | 6,186 | 85.1% | 0.6 | 0.3 |
| SPP (System Parameter Permutation) | 6,936 | 5,699 | 82.2% | 75.7 | 39 |
| Randomize Data (Monte Carlo) | 6,900 | 5,574 | 80.8% | 260.2 | 136 |
| Walk Forward | 5,506 | 4,594 | 83.4% | 149.6 | 98 |
| Tick Retest | 5,729 | 4,589 | 80.1% | 12.6 | 7.9 |
The checks are not strict one after another in the logs, so read the pass rates per check, not as a chain. Counts are passed plus failed, because the Total field in the SQX retest logs is not filled correctly.
Where the time goes: generation took 190.6 hours, the checks 501.8 hours. Randomize Data and Walk Forward alone took about 410 hours for fewer than 7,000 strategies each. The funnel is narrow at the top and expensive at the bottom. See how long Walk Forward and Monte Carlo take.
Your turn
How to plan your own run
- Measure your acceptance rate. Open
user\projects\<your project>\log\global_log_*.logand read the Builder line:Strategies generated,AcceptedandAccepted strategies per hour. - Measure your survival rate. Count how many strategies reach your final databank after all checks and divide by the number accepted by the Builder. In our run it was 669 of 374,397, or 0.18%.
- Multiply. Strategies to generate = final strategies you want ÷ acceptance rate ÷ survival rate.
- Add the checks. Multiply the number of strategies entering each heavy check by its seconds per strategy. This part usually decides the calendar time.
With the ratios from this run, on the same server, the plan looks like this:
| Final strategies you want | Strategies to generate | Builder time | All work, with checks | Calendar time at 80% load |
|---|---|---|---|---|
| 10 | 1,954,370 | 3 h | 10 h | 0.5 days |
| 50 | 9,771,850 | 14 h | 52 h | 2.7 days |
| 100 | 19,543,700 | 28 h | 103 h | 5.4 days |
| 500 | 97,718,500 | 142 h | 517 h | 26.9 days |
These rows scale our own ratios. They are a reference for the order of magnitude, not a promise for your project: other filters, another template or tick precision in the Builder will change every column.
Good ways to widen the funnel
- Remove building blocks that produce strategies with no trades.
- Run the first pass on a higher timeframe or shorter history.
- Keep cheap filters early and expensive checks late.
- Let heavy checks run on a server that works for days.
Ways that only look faster
- Loosening robustness checks until more strategies pass.
- Comparing strategies per hour between different templates.
- Judging hardware by the Builder only and ignoring Walk Forward time.
- Sharing one machine between SQX research and live terminals.
Hardware
What this means for the server
The run used an Intel Core i9-13900 server with 16 threads in use, 64 GB DDR5 and NVMe storage. It produced about 19 final strategies per day at 80% load. Two things follow. First, a funnel this narrow needs a machine that can work for weeks without being switched off or shared. Second, since checks take 70% of the time, size the server for Walk Forward and Monte Carlo, not only for generation speed.
Speed by timeframe, time per check and a way to compare with your PC are in the StrategyQuant X benchmark. For VPS or dedicated, recommended CPU, RAM and NVMe and current prices, see StrategyQuant X Server: VPS or Dedicated? and StrategyQuant X system requirements and the dedicated servers page.
FAQ
Frequently asked questions
How many strategies do I need to generate in StrategyQuant X?
It depends on your filters. In our five-week run 130.7 million generated strategies gave 669 that passed every robustness check, about one in 195,000. With looser filters you need fewer, with stricter filters more. Use your own acceptance rate from the Builder log to scale the estimate.
What is a normal acceptance rate in the StrategyQuant X Builder?
In this run the Builder accepted 0.29% of generated strategies on average. By template it ranged from 0.06% for H4 short-only to 2.27% for H1 long-and-short. The rate is set by your filters, not by hardware.
How long does it take to get one robust strategy?
On this server, about one hour of work per strategy that passed every check: 692 working hours for 669 strategies. Most of that time went to Randomize Data and Walk Forward checks, not to generation.
Does a strategy that passes all checks make money?
No guarantee. Robustness checks only show that a strategy survived the tests you chose on historical data. They do not predict live results. Winservers.NET rents servers and does not give trading advice.
Which stage removes the most strategies?
The Builder filters. They rejected 99.71% of everything generated. The most common reasons were zero or negative out-of-sample profit, no trades and too few trades.
How can I get more strategies through the funnel in the same time?
Use a higher timeframe or a shorter history for the first pass, keep early filters cheap, and give the heavy checks to a server that can run for days. More CPU threads shorten generation and checks alike.
Planning a long StrategyQuant X run?
Send us your timeframe, data period, filters and the checks you use. We will suggest a dedicated server and tell you what to expect from it.