What Proofline does not claim.

What a record is not

Not investment advice.

Nothing on Proofline is a recommendation to buy, sell or hold anything. Proofline is not a broker, not an adviser, and not registered as either.

Not a prediction.

A record establishes what happened under the data and assumptions shown. Past behaviour of a strategy says nothing reliable about its future results.

Not a certification that a strategy is good.

Verification is about whether the record is true, not whether the idea is sound. A verified record of a losing strategy is a complete success for Proofline and a bad outcome for its owner.

Not an audit of the person.

Proofline verifies a recorded system. It does not vet the operator, their other claims, or anything they say off Proofline.

What is limited today

Prediction markets only.

Everything operational today is prediction markets. Support for other asset classes exists in the engine as simulated replay and is not offered as a service.

One venue in the live feed.

Evidence currently covers Polymarket markets. Kalshi is not in the live catalog.

Evidence coverage is partial.

Many live markets carry only their official resolution source, and some carry none at all. Where there is nothing, the market says so instead of being filled in.

Verification is pilot-operated.

The end-to-end process is not self-serve. Steps run by hand are stated as such on the verification page rather than implied to be automatic.

The registry is small.

A handful of records, several of them negative. It is honest, not comprehensive.

One recorded dataset to replay against.

You can replay a strategy against exactly one bundled dataset: about eight hours of recorded Polymarket BTC 15-minute up/down markets, 6,312 events, priced from Polymarket and Coinbase. The longer replays shown elsewhere on this site are records of past work, not what you can run here today.

That dataset records one side of the book.

It carries the YES book only, so fill models that price against a NO book are refused rather than estimated. Proofline says which models a dataset cannot support instead of quietly modelling a book the data never recorded.

What the data itself cannot tell you

Paper is not live.

A paper run places no real orders. It can model liquidity and fees faithfully and still miss what happens when your own order is in the book.

Replay is not the future.

A replay is exact about the past it was given and silent about everything else.

Depth evidence has limits.

Where the recorded data carries no proof that size was available, Proofline refuses the fill rather than assuming it. That makes some datasets produce fewer trades, not better ones.

Settlement can differ between venues.

The same real-world question has resolved differently across venues in recorded data. Where that happens it is reported per venue, never merged into one number.

Missing data stays missing.

Gaps are shown as gaps. Proofline does not interpolate a value to make a series look complete.

Platform and operational

Public beta.

Features change, data can be reprocessed, and availability is not guaranteed. There is no uptime commitment.

No email delivery.

Alert and verification emails are not configured. Where the product would send mail it reports the delivery channel as unavailable rather than pretending to have sent something.

Legal review outstanding.

The terms and privacy text are prepared drafts and have not been through professional legal review. They are published as drafts and labelled that way rather than being presented as reviewed.

No third-party audit of Proofline.

No external party has audited Proofline's own systems or claims. The evidence attached to each record is what a reader should judge, not Proofline's word about itself.

If something here is out of date, that is itself an error worth reporting. Tell us. For how each kind of evidence is labelled, read the evidence policy; for how results are computed, the methodology.