From an idea
to a strategy you can trust
Build with nodes, code or the AI assistant, then judge it with in-sample / out-of-sample splits, Monte Carlo and honest costs — on real Binance and OKX data. Paper-trade first, go live when the numbers hold.
- Every marketplace listing is a real backtest, refreshed daily
- Fills at next open · maker/taker fees · funding · liquidation
- Your code runs in an isolated sandbox
Robustness report · live from the engine
On-chain activity × OI momentum
Robustness
Out-of-sample
+10.5%
Total return
+5.9%
Max drawdown
-9.7%
Sharpe
0.41
Three ways to build. One engine to judge them.
Drag nodes, write Python or JavaScript, or let the assistant draft it — everything ends up in the same backtest and the same live executor.
Visual canvas
Data → factors → signal → trader. 37 templates cover trend, mean reversion, grids, selection and AI models; every node has an inspector, no code required.
See how it worksCode nodes (Python / JS)
Drop a Python or JavaScript node anywhere in the graph: custom factors, custom order logic, LightGBM / sklearn models trained walk-forward — with syntax checks in the editor.
See how it worksAI assistant that sees your canvas
It reads the current graph and last backtest, proposes targeted patches, writes code strategies and explains why a result looks fragile.
See how it worksNumbers you can argue with, not slogans
A backtest is only useful if you know what it assumed. These are the rules the engine applies to every strategy on the platform — templates, yours, and the marketplace alike.
Real exchange data
Binance and OKX candles, funding rates and open interest, cached locally and pre-warmed. No synthetic prices anywhere.
Next-bar fills
A signal on the closed bar fills at the next bar's open. Limit and stop orders rest on the book and fill only when price trades through them.
Fees and slippage
Maker/taker fees per fill plus configurable slippage. Costs are charged on the way in and on the way out.
Funding and liquidation
Perpetual positions pay real funding each interval; margin is tracked per position and a breach liquidates you, as it would on the exchange.
In-sample / out-of-sample + Monte Carlo
The last 30% of the window is held out. Trades are reshuffled 1,000 times for drawdown and loss-probability bands. The robustness score penalises red flags and losses.
Isolated execution
Python and JavaScript nodes run in a locked-down container with no network and no access to the database or other users.
Data currently available
Only what the engine really ingests today. More feeds are added when they can be backfilled end to end.
Candles
1m – 1d · Binance spot & USD-M · OKX
Funding rates
Binance USD-M · OKX swaps
Open interest
Binance USD-M · OKX swaps
Fear & Greed index
alternative.me · daily
What the templates actually did this year
Ranked by robustness, not by the biggest number. Some lose money — that is the point of showing real results.
#1 · BTCUSDT, ETHUSDT +3 · 1d
On-chain activity × OI momentum
- Out-of-sample
- +10.5%
- Total return
- +5.9%
- Max drawdown
- -9.7%
#2 · BTCUSDT, ETHUSDT +3 · 1d
On-chain fundamentals
- Out-of-sample
- +7.5%
- Total return
- +8.4%
- Max drawdown
- -7.8%
#3 · BTCUSDT, ETHUSDT +3 · 1d
Smart Beta + WQ Alpha
- Out-of-sample
- +9.8%
- Total return
- -5.4%
- Max drawdown
- -17.4%
Let the data speak before you go live
The free account gives you 3 hours of backtesting, 3 strategies and 10 AI chats a day — enough to test your first idea.