Prove the edge before you risk the account.
PropQuantX backtests futures strategies on second-resolution data, validates them out-of-sample, and routes the survivors straight to your prop accounts. One terminal, from hypothesis to fill.
Simulated backtest of all seven configurations, 2 Jan 2023 – 19 Jul 2026. Net of costs: $2.50 per side per contract and one tick of adverse slippage on every fill. All seven flat at each session close, and each one sized by its own deployment configuration — its own risk per trade and its own contract cap, not a uniform assumption. No live account traded these figures; every number here is model output. Full hypothetical-performance disclaimer.
This is the stack as it would actually have traded, not seven curves added together. The seven would have taken 3,308 entries between them; the hedge guard refused 293 of those, because an entry that opens against a position already live in the same group is not taken at all. What is plotted is the 3,015 that survived that rule — which is why the total is not the sum of the figures in the table.
The shaded band is the deepest drawdown — peak to trough, . Both the curve and the drawdown are measured on closed trades, totalled by the day each one exits.
Parameters were tuned on windows inside this period, so the total above is in-sample. The honest figure is what each strategy earned on data it had never seen — that one is in the table below.
Instruments in the data set
Strategies
Search the space, not your memory.
Grid and Bayesian search run the same engine your live orders will. Every cell is costed with real tick sizes, session hours, and the flat-by-close rule your prop firm enforces.
- Optuna and exhaustive grid, seeded and reproducible
- 1-second resolution where the fill actually matters
- Conditional filters that scope an edge to its regime
Pick any strategy to see its own equity curve
Return over maximum drawdown, measured walk-forward on data outside the tuning window, and where two search methods disagreed the more conservative number is the one shown. Candidates that beat these ratios in-sample and then failed here were discarded, not re-tuned. Others have been cut for reasons that never show up in a ratio: two turned negative the moment commissions and slippage were charged, two earned too little to justify the drawdown they carried, and one was profitable but ran the same engine on the same market as another here — one bet wearing two names. These seven are what is left, and they are the seven the terminal is configured to trade. The figures beside them are still model output, not a record of those configurations having traded.
Validation
In-sample results are a hypothesis. Nothing more.
Three gates stand between a promising curve and a live order. Most candidates die at the second one, which is the point.
Gate one
Walk-forward folds
Optimize on a rolling window, score on the window after it, never the one before. A strategy that only survives the fit is a strategy you never funded.
Gate two
The held-out bake
A slice of tape the search never touched, opened once. No re-tuning afterward — that would spend the only honest data you had left.
Gate three
Slippage and stack risk
Re-price every survivor a tick worse, then check it against the rest of the stack for hedged pairs and drawdowns that arrive together. Correlation is reported, not silently removed — see the note under the table.
Platforms
From backtest to broker without a rewrite.
Deploy a validated configuration against a live or simulated account and the terminal handles the rest: position reconciliation against broker truth, per-account risk caps, and a flatten-all that means it.
- Reconciles to the broker, never to its own memory
- Refuses to deploy two strategies that would hedge each other
- Exports NinjaScript and Pine when you want the platform's own engine
Bridge or native
NinjaTrader 8
Two ways in. Orders can route through a bridge add-on inside your own NinjaTrader, so fills come back from the platform itself; or take the exported NinjaScript and run the strategy natively, with no bridge in the loop at all.
Webhook
TradingView
Signals leave as webhook alerts and land at whichever broker you have connected. Strategies also export as Pine if you want TradingView's own engine to run them.
Webhook
Tradovate
An outbound adapter posts to your Tradovate-connected endpoint, then reconciles the resulting position against the account instead of trusting local state.
REST API
ProjectX
Direct API access to the platform a large share of funded evaluations run on, which makes it the shortest path from a validated config to a prop account.
Questions worth asking first.
The ones that decide whether this belongs in your workflow.
Where does the market data come from?
Licensed CME data from Databento. The equity index and metals series begin in 2016, Russell in 2017, and the remaining markets in 2020 — all of them running through the present.
Do backtests include commissions and slippage?
Yes. Every figure on this page charges $2.50 per side per contract and one tick of adverse slippage on each fill, so a round turn pays $5.00 in commission and gives up two ticks. Both numbers are configurable per contract.
Can the strategies be used on a prop account?
Yes. Daily loss limits, trailing drawdown, and the flat-by-close rule are enforced in the risk layer, and every account carries its own cap. Check your firm's automation policy first — some require a human at the desk.
What happens if the connection drops mid-position?
On reconnect the terminal reconciles against broker positions before it accepts a new signal. If broker truth and local state disagree, it stops and tells you rather than guessing.
Do I have to write code?
No. The strategy library ships parameterized engines you configure in the UI.
More are still in validation.
What clears the gates gets added here, with its curve and its drawdown attached. What does not clear them does not appear at all.