Supported Kalshi markets.
Turbine can replay candle history where Kalshi data is available for the markets or series your strategy targets. Coverage varies by market and window.
Product · Backtesting
Backtest Kalshi Strategies before deployment.
Backtest a supported Kalshi strategy without Python: describe the rules in Turbine Studio, review the generated YAML, and inspect historical coverage, modeled fills, fees, and trades. Live execution can differ.
Window30 days
Granularity1-min candles
CoverageMarket dependent
Use a supported strategy and review the evidence before starting a bot.
- Open Studio, choose Kalshi, and describe the market, entry, exit, size, and skip conditions.
- Review the generated YAML. The complete Kalshi examples show supported configurations; we ran the rolling BTC example unchanged as the recorded backtest below.
- Request a historical test and compare requested dates with actual coverage. Unsupported inputs or unavailable history do not establish a passed test.
- Inspect modeled fees, drawdown, fills, and the order-log CSV. A zero-trade test does not exercise the entry and exit rules.
- Review a separate Paper Run before deciding whether to deploy live.
For data provenance, read the historical-data guide. For original results and limitations, read the 5,732-backtest research synthesis.
Testing a Polymarket strategy? The Polymarket Backtesting Guide covers the no-code Paper Run workflow and what fill-accurate evidence should include.
We pasted the documented rolling BTC 15-minute example into Studio without changes and ran the default 30-day backtest. The result lost money, and we publish it as recorded.
| Type | Historical Replay: past Kalshi order-book data, simulated orders. Not Paper trading and not live execution. |
|---|---|
| Strategy | BTC 15-minute series, one entry per window (KXBTC15M). Buy 2 YES contracts at or below $0.35 with more than 8 minutes left, at most once per market; sell YES at a $0.75 bid; flatten at 2 minutes to close. |
| Run ID | 915dc99858c87d19a20e1086beaf5d3e3666128bf1c4c31c6f6b1b326502ad90 |
| Dates and coverage | September 19–October 19, 2026 (30 days, UTC). Coverage is marked partial: 2,733 markets were simulated and 3,666 of 6,399 listed markets were skipped; the logged sample is markets past the settlement data frontier. |
| Fills and fees | Simulated against the next executable recorded quote; every fill was a taker fill. Kalshi's fee schedule applied per fill; settlement-aware payouts. |
| Result | 2,456 simulated fills. 406 of 1,228 closing trades were profitable (33.1%). Gross −$64.87, fees $61.51, net −$126.38; max drawdown −$130.00. |
| Files | Public run page · Order-log CSV (first 10,000 of 14,408 order events, through October 8) · Run manifest (JSON) |
| Limits | One 30-day window of one example, not a recommended strategy. Simulated fills have no queue position or market impact. The replay checks max_loss: 5.00 per market, so it kept trading after the run lost $5; a live Bot applies it across the deployment and halts. |
Turbine pulls Kalshi's public candlestick endpoints across the markets your strategy targets and replays them inside the Studio session.
One-minute candles.
Turbine can use OHLC plus best bid and ask snapshots, sourced from Kalshi's public candlestick endpoints.
Default 30-day window.
Studio runs a 30-day backtest by default. Actual replay coverage depends on market history and available archive data.
No notebook setup.
Describe and review a strategy in Studio. Each completed test gets a public page with a downloadable order-log CSV; raw-data downloads are a separate workflow.
Runtime depends on data preparation.
Cached tests can finish quickly. Cold archive retrieval or historical repair can take longer. Inspect actual coverage and assumptions before comparing results from different windows.
One screen. Equity curve, drawdown, trade log. Headline metrics in the strip below.
Headline numbers, fill quality, trade log.
PnL in dollars and as percent. Sharpe, max drawdown, win rate, total trades. Per-fill maker / taker classification, slippage versus signal price, fees applied. The run's public page offers the order log as a CSV.
Turbine simulates next-bar or depth-aware fills, applies Kalshi's fee schedule where modeled, and reports model slippage so you can judge how much the result depends on execution assumptions.
Strategy fills.
Turbine simulates next-bar fills using top-of-book prices or available L2 depth, with maker / taker classification per fill where supported.
Marketable orders.
Turbine models marketable orders against the next bar or available depth. Reported slippage is model-based, not a promise of live execution.
Kalshi fees.
Turbine can apply Kalshi's fee schedule per fill. Reported PnL is net of fees, not gross.
Forced exits at close.
Turbine can flatten open positions at market close so PnL is not left ambiguous.
turbinefi
Notebook + Cron
Generic AI Wrapper
CSV Downloads
Default 30-Day Supported History
Yes
Partial
No
Partial
Same Strategy Spec Live and Replay
Yes
No
No
No
No Infra to Run
Yes
No
Partial
No
Kalshi Fee Model Included
Yes
Partial
No
No
Starter
$29/ month
Starter includes 30 deploys per week. Backtests are included, subject to normal abuse-prevention rate limits.
- Starter · BasicBacktests included, cloud deployments, every supported venue.
- $29 · $99/mo10 · 100 deploys/week
- Pro400 deploys per week with concierge support.
- $199/mo
How can I backtest a Kalshi strategy without Python?
Describe your rules in Turbine Studio, review the generated YAML strategy, and request a backtest for a supported market and data window. Inspect actual coverage, simulated fills, fees, drawdown, and the order-log CSV before considering a Paper Run or live deployment. No Python notebook or self-hosted replay engine is required. Our recorded replay of the documented KXBTC15M example shows every one of those fields, including a net loss.
Where does the historical Kalshi data come from?
Turbine pulls Kalshi public candlestick data and PMXT archive data where configured for the markets and series your strategy targets. The default Studio window is 30 days, but actual coverage varies by venue, market, and available archive history.
How fast is the turbinefi backtester?
Runtime depends on market count, history density, and data preparation. Cached tests can finish quickly; cold archive retrieval or historical repair can take longer. A completed test still needs coverage and trade-log review.
Can I export backtest results?
Yes. Each completed backtest has a public page with a downloadable order-log CSV: every simulated order submission, fill, cancel, and market close, up to 10,000 events per run. Studio shows the equity curve and summary metrics.
What is included in a turbinefi backtest fill model?
Turbine includes one-minute Kalshi candles where available, PMXT L2 archive replay where covered, next-bar/top-of-book or depth-aware fill assumptions, Kalshi fee modeling, and close/settlement handling. Reported Kalshi PnL is net of modeled fees.
Does turbinefi support custom data feeds in a backtest?
Supported Coinbase and NWS fields can be replayed where historical data exists. Support varies by field and venue; a live feed does not automatically have replayable history. Review the edge-data reference for the exact inputs your strategy reads.