Binance Prediction Market Order-Book Data: Full L2, Cadence and Limits
Displayed probability is not an executable price. Full-L2 snapshots preserve the bids, asks and sizes required to model spread and slippage.
Binance prediction-market order-book data records the full observed Predict.fun bid and ask ladders, including price, size, market identity, exchange time and receive time. DepthFeed polls the full book at approximately 1.5-second cadence and retains snapshots from July 17, 2026 forward; it does not claim tick-by-tick matching-engine events.
Why a probability tile is not enough
A displayed 60% chance says nothing about whether a buyer could acquire one contract or ten thousand at 0.60. Execution depends on the best ask, the sizes resting behind it and the levels an order must consume. The spread also determines whether the midpoint was ever directly tradable.
A backtest built from only displayed probability or last trade silently assumes unlimited liquidity at one price. Full depth replaces that assumption with the market state that was actually observed.
What each snapshot retains
- Predict.fun and Binance market identifiers for stable joins
- Bid prices and bid sizes across the observed ladder
- Ask prices and ask sizes across the observed ladder
- Normalized Up/Down outcome identity
- Exchange and collector receive timestamps in epoch milliseconds
- Capture metadata that keeps exact gaps visible
Capture cadence and interpretation
The dedicated collector polls full depth at approximately 1.5-second cadence. A snapshot is the complete observed market state at that poll, not a reconstruction of every add, cancel and fill that occurred between polls.
That distinction matters for queue-position models. The archive supports book walking, spread analysis, depth and state-transition research. It should not be presented as tick-by-tick matching-engine telemetry.
Using full depth in a backtest
At each strategy timestamp, select the newest snapshot available without looking forward. A market buy walks the recorded asks from best to worse until the intended size is filled; a market sell walks the bids. The size-weighted average of those consumed levels is the simulated execution price.
Keep fees, minimum order rules, latency and partial-fill assumptions separate. Full L2 removes the largest liquidity fiction, but a realistic simulator still needs explicit execution rules.
Archive start and historical limits
Recorded full-depth history starts July 17, 2026. Earlier catalogue records do not imply earlier book coverage. Predict.fun's off-chain matching system did not publish a historical snapshot archive, and on-chain settlement cannot reveal orders that never filled.
Key takeaways
- 01The archive stores full observed bid/ask price and size ladders, not only top of book.
- 02Full-depth polling runs at approximately 1.5-second cadence.
- 03Snapshots support spread, depth, slippage and book-walking research.
- 04The data is recorded market state, not tick-by-tick matching-engine telemetry.
- 05Historical full-L2 coverage begins July 17, 2026.
DepthFeed serves Binance prediction-market history and recorded full-depth books over the REST API. Free Explorer tier, no card.
Start freeView pricing