Find the edge.
Then prove it.

Statistical proof for your trading strategies, in minutes rather than months. Launching soon.

AI-assisted strategies
L2 backtesting
Forward-walk optimization
BTCUSDT.P · −0.30%ETHUSDT.P · −0.87%SOLUSDT.P · −1.34%XRPUSDT.P · −2.97%DOGEUSDT.P · −2.25%BTCUSDT.P · −0.30%ETHUSDT.P · −0.87%SOLUSDT.P · −1.34%XRPUSDT.P · −2.97%DOGEUSDT.P · −2.25%BTCUSDT.P · −0.30%ETHUSDT.P · −0.87%SOLUSDT.P · −1.34%XRPUSDT.P · −2.97%DOGEUSDT.P · −2.25%BTCUSDT.P · −0.30%ETHUSDT.P · −0.87%SOLUSDT.P · −1.34%XRPUSDT.P · −2.97%DOGEUSDT.P · −2.25%BTCUSDT.P · −0.30%ETHUSDT.P · −0.87%SOLUSDT.P · −1.34%XRPUSDT.P · −2.97%DOGEUSDT.P · −2.25%BTCUSDT.P · −0.30%ETHUSDT.P · −0.87%SOLUSDT.P · −1.34%XRPUSDT.P · −2.97%DOGEUSDT.P · −2.25%

Strategy assist

Describe an edge.
Get production code.

Talk through your idea in plain English. The strategist drafts the Python, validates against the schema, and runs a smoke backtest before you even look at the chart.

  • Streams as you think: interrupt, refine, branch a thread
  • Generates valid, production-ready Python
  • Cites the indicators it chose and why
  • Same prompt, every market, from BTC to small-cap perps
New strategy
Mean-reversion on BTC 15m using RSI and Bollinger bands. Exit when volatility expands.
Ask the strategist…
BTCUSDT.P15m
BTCUSDT.P15m · Bybit
$67,119-0.12%
Equity+$0
Trades
0
Win rate
0%
PnL
+$0
Sharpe
1.84

Long-term backtests

Six years of data.
Every market regime.

Backtest against up to six years of minute-resolution market data: bull runs, drawdowns, and the long chop in between. An edge that only survives one regime isn’t an edge; prove yours across all of them.

Precision

Fills that match the tape.

Most retail backtesters fill at the close. We replay the L2 orderbook level by level, model dynamic slippage by depth, and price funding under Bybit’s margin model, so the curve you see is the curve you’d trade.

  • Level-2 orderbook reconstruction, minute-resolution depth
  • Dynamic slippage from real depth, not a flat assumption
  • Bybit isolated + cross margin, funding, liquidations
  • Deterministic: same inputs, same fills, every run
PriceSize
67257.03.00
67256.55.00
67256.09.00
67255.515.00
67255.020.00
67254.531.00
67254.034.00
67253.553.00
67253.056.00
67252.560.00
67252.062.00
67251.580.00
67,251Mid
67250.574.00
67250.071.00
67249.573.00
67249.050.00
67248.546.00
67248.043.00
67247.532.00
67247.019.00
67246.514.00
67246.011.00
67245.55.00
67245.04.00
BTC mean-reversion · 15m
BTCUSDT.P15mv3
Last run
2h ago
Return
+42.6%
Sharpe
1.84
Win rate
58%
Max DD
-9.4%
TearsheetView detail →

Library & tearsheets

Every backtest, kept.
Every metric, computed.

Pin your strategies. Every run lands in a tearsheet: Sharpe, drawdown, win rate, equity, all of it. Compare versions. Export to CSV. Share what worked.

  • Pin a strategy; every backtest version is kept
  • Tearsheets with Sharpe, drawdown, win rate, equity
  • Compare versions side-by-side
  • CSV export for the runs that matter

Roadmap

Where we are. Where we’re going.

Everything in Now runs today. What follows is where AlphaProve is headed: live trading, new markets, and an AI that does more of the work.

NowLive today
  • AI strategy assist: describe an edge, get Python
  • Candle-by-candle live replay over SSE
  • L2 orderbook fills with depth-based slippage
  • Bybit margin, funding & liquidation modelling
  • Library, version history & full tearsheets
  • Forward-walk validation + Meta-ML
  • Sandboxed strategy execution running fully on our end
NextIn build
  • Full AI strategy creator, end to end
  • ML strategy generator
  • Strategy marketplace
  • Paper-trade a strategy on our servers
  • MCP integration
LaterOn the horizon
  • Stock-market integration
  • Forex pairs
  • Fundamental-analysis signals
  • AI that takes the trades for you
  • Engineering & research blog

Why AlphaProve

Your strategies. Your control.

Build and test trading strategies as fast as you can describe them, with full control over the code, the data, and every assumption, at a price that doesn’t gatekeep.

Build & test in minutes

Describe an edge in plain English or write the Python yourself. Either way you go from idea to a candle-by-candle backtest in minutes, with no boilerplate and no infrastructure to wire up.

Full control, no black box

It's your code, your parameters, your data window. Read every line the AI writes, edit it in a real Monaco editor, and see exactly how every fill was priced. Nothing is hidden.

Priced for real people

Institutional-grade backtesting (L2 fills, six years of history, forward-walk validation) without the institutional invoice. Start free; pay only once you have an edge worth proving.