Full archives
Recorded order books and outcomes for backtesting and market-microstructure research.
Explore the datasets →Prediction-market data for researchers & builders
Historical Polymarket & Kalshi data.
Explore captured Polymarket and Kalshi order books in Parquet, with free samples, documented schemas and known gaps. Start with the data; choose the archive that fits your research.
No signup to inspect samples. One-time archive purchases are separate from the live API subscription.
Historical sample · Kalshi
NBA · Kalshi · Dec 31, 2025
Phoenix at Cleveland
Total 237 · YES contract
Last sampled midpoint42¢21:14:06 UTC
Midpoint Bid–ask rangePrice / cents
20:43:35 UTC30m 31s sampled span21:14:06 UTC
1 / Q1 · 10:175–6CLE–PHX
2 / Q1 · 8:4412–8CLE–PHX
3 / Q1 · 3:4726–16CLE–PHX
4 / Q2 · 9:2246–28CLE–PHX
Inspect the four source rows
| UTC | Bid | Ask | CLE–PHX |
|---|---|---|---|
| 20:43:35 | $0.48 | $0.49 | 5–6 |
| 20:47:38 | $0.51 | $0.54 | 12–8 |
| 20:56:41 | $0.46 | $0.50 | 26–16 |
| 21:14:06 | $0.40 | $0.44 | 46–28 |
Sport-specific slices
Start with the sport you study. Inspect dates, depth and print coverage before you buy.
Find your sport →Fair-value API
Compare model fair value with available market prices. A separate subscription, with public results.
Open the Edge Finder →Honest measurement
Public trading measurements
Every admitted live fill with a captured close enters the same unconditioned scorecard — currently 46.4% beat that captured close, with mean CLV +0.01¢. We publish the qualifying ledger, the exchange-price correction artifact, and the verifier needed to reproduce it.
And when we get a measurement wrong, we retract it in public. This panel used to headline a 78-point CLV win-rate gap as "proof of signal." We found — and proved with a control — that the statistic was a measurement artifact for in-play trades, and we retracted it. Read the retraction →
- Fills beating our close
- 46.4%n=3,206 measured
- Mean CLV / trade
- +0.01¢admitted live population
- Admitted live fills
- 3,248wins and losses included
- NCAAMB ECE
- 4.39%5,345-game season
Live API and market-price comparison
$49/mo with 7-day free trial: the full real-time API and signed real-time edge webhooks. The multi-venue Edge Finder board stays free to browse.
The comparison, explained
Model fair value and available market quotes
FAIRModel estimate
The model’s probability, expressed as a price.
ASKMarket price
The available quote, with venue and quote age.
EDGEThe difference
Fair minus ask, before fees and execution costs.
A price-comparison reference, not a promise of profit. Open the actual board to inspect available games and quotes.
API features & sport coverage
- Multi-Venue Edge Finder
-
Our calibrated fair value next to the live price on Polymarket, Kalshi, DraftKings, FanDuel and BetMGM — so you can see who has the best number.
- Real-Time Edge Webhooks
-
Register an HTTPS webhook to receive signed matched-price signals on your server. Delivery is subject to signal filters and a 120-second cooldown per endpoint.
- REST API + WebSocket
-
Model win probabilities for covered games, available venue prices, and fair-line endpoints. JSON responses, a Python SDK, and endpoint documentation. Venue and market availability vary.
- 8 API Sports
-
NBA, WNBA, NCAAMB, NCAAWB, CFB, NFL, NHL and MLB from one public API contract.
- Devigged Lines
-
Book odds with the bookmaker margin stripped out, lined up against our fair value — an apples-to-apples reference across every venue.
- Public Evidence
-
Inspect public methodology, benchmark artifacts and a results ledger that includes losses. See the record.
7-day free trial — cancel anytime
$49/mo
Full real-time API access, 30K requests/mo, the WebSocket stream and signed real-time edge webhooks. All included. The Edge Finder board stays free to browse.
Monthly Starter through Stripe: if the 7-day trial ends without a payment method, access ends with no charge. With a saved or added payment method, it renews at $49/month unless you cancel first.
Pro ($149/mo) and Enterprise ($499/mo) plans also available. See pricing →
Just exploring? Create a free Developer account — 500 requests/mo, 5-minute delay on games and edges, no card.
Frequently asked questions
What do I get with the free trial? +
Starter access for 7 days: model probabilities across the eight public API sports, available venue price comparisons, REST endpoints, one WebSocket connection, the dashboard, and one signed webhook endpoint. The Starter monthly request cap is 30,000. Trial renewal terms are shown at checkout. Or try the free Developer tier first (500 req/mo, no card).
What sports are covered? +
NBA, WNBA, NCAAMB, NCAAWB, CFB, NFL, NHL, and MLB are the eight sports in the public API contract. Historical dataset coverage differs by archive; check each product’s schema.
What does "calibrated" mean? +
Calibration compares predicted probabilities with observed outcome frequencies in a defined sample. Our published 2025–26 NCAAMB report measured 4.39% Expected Calibration Error on 5,345 games. ECE is a weighted average of probability-bucket gaps, not a maximum error for every prediction or a guarantee of future performance.
How are predictions verified? +
We publish benchmark artifacts and a filtered public results ledger that includes losses and nonbinary outcomes. The results page reports the measured share of resolved rows carrying an execution identifier or a direct on-chain transaction hash, rather than implying universal coverage.
Do you promise profit or +EV? +
No. ZenHodl gives you calibrated fair value and available market prices as a price-discovery reference. Our public methodology, benchmarks and results let you inspect the evidence, including losses. What you do with it is your call; it is not betting advice or a profit guarantee.
Can I cancel anytime? +
Yes. Cancel from your account page or email [email protected]. The monthly Starter Stripe trial lasts 7 days. If it ends without a payment method, access ends with no charge. If a payment method is saved or added, Starter renews at $49/month after the trial unless you cancel first. The free Developer tier has no trial expiry. Crypto access is prepaid for 30 days, with no trial or automatic renewal.
What can go wrong
Model degradation. Markets adapt. A model that works today may need retraining in 3-6 months as market efficiency improves.
Execution costs. Backtest results assume ~2-3c per trade in fees and slippage. Real execution can be worse in thin markets.
Small samples. Live trading started March 2026. The backtest is strong but live history is still growing.
Regime change. Rule changes, new market participants, or structural shifts can invalidate historical patterns.
Past performance does not guarantee future results. This is a tool for informed decision-making, not a guaranteed profit machine.
Explore the public data and results
Open the Edge Finder to compare model fair value with available market prices. Inspect the public benchmark and results.