Predict.funIn development

Predict.fun data for bots and cross-market research.

A planned normalized dataset for markets, order books, time series, last sales, and matches—designed to work beside every other KnownAt source.

Join early access Inspect the schema

Free workspace · No card required

DATASETpredict.market_snapshots

PIPELINE PREVIEW

MARKETWill BTC trade above $100k this quarter?Representative normalized record

YES0.624bid 0.621

NO0.376ask 0.379

observed_at2026-08-30 14:32:10Zsourcepredict.fun

DATASETpredict.market_snapshots

SOURCE INTERFACEREST · WebSocket

UPDATE TARGETStreaming target

NETWORKBNB Chain

01 / BUILT FOR RESEARCH

From market question
to testable dataset.

Start with the strategy. KnownAt handles collection, normalization, timestamps, and historical storage.

Compare venues

Study related markets across Predict.fun and other prediction-market sources.

React to live state

Build scanners and bots around normalized order-book and market updates.

Research market structure

Analyze price formation, liquidity, and activity without maintaining another collector.

02 / NORMALIZED SCHEMA

Rows you can
actually work with.

Illustrative normalized schema; final fields may change before release. Types, timestamps, and source identifiers stay explicit so the same dataset works in notebooks, bots, and backtests.

SCHEMApredict.market_snapshots

observed_attimestamp

When KnownAt observed this market state.

available_attimestamp

When the record became usable by a model.

market_idstr

Stable source market identifier.

outcomestr

The represented market outcome.

pricefloat64

Normalized probability-like market price.

source_event_idstr

Traceable upstream event identifier.

SAMPLE ROWSUTC · Predict.fun

observed_at timestamp	market_id str	outcome str	price float64
14:32:10	mkt_8f2…a41	YES	0.5810
14:32:11	mkt_8f2…a41	YES	0.5830
14:32:12	mkt_8f2…a41	YES	0.5820
14:32:13	mkt_8f2…a41	YES	0.5850

03 / SOURCE COVERAGE

Raw source data in.
One query model out.

KnownAt preserves source identifiers and acquisition time while presenting a stable interface over changing upstream APIs.

Markets and statisticsMarket metadata, state, aggregate statistics, and search.

Market microstructureOrder books, last sales, and order match events.

Time seriesHistorical and latest time-series observations.

Live updatesWebSocket topics normalized into the same storage model.

Predict.funsource APIs

KnownAtnormalize · validate · timestamp

Your modellive · history · replay

04 / ONE ACCESS LAYER

Switch time ranges.
Not integrations.

The API shape stays familiar whether you request the latest record, bounded history, or an information set as it was known at a past moment.

Python

knownat

dataset = knownat.dataset(
  "predict.market_snapshots"
)

rows = dataset.history(
  from_="2026-08-01",
  to="2026-08-30",
  as_known_at="2026-08-30T00:00:00Z"
)

REST JSON Parquet

POINT-IN-TIME QUERY

The result does not borrow from the future.

Observed time, availability time, ingestion time, and later revisions remain distinct. Your backtest sees only records it could have known then.

as_known_at=2026-08-30T00:00:00Z

05 / AVAILABILITY

Predict.fun is next in the pipeline.

The page shows the planned normalized interface. Join now to follow coverage, schema, and release progress.

Join early access See founding pricing