Proofline exists because a good result is easy to show and hard to prove.
Proofline started from a trading bot, not from a product idea. The bot ran on prediction markets, and like every bot it had a backtest that said it worked. The backtest assumed that a price you can see is a price you can get.
That assumption turned out to be measurably false. Checked against 199 real arbitrage events and 352 exchange-confirmed fill legs, firing on visible price alone filled 81.7% of the time on one venue and 69.2% on the other. The backtest had been counting the other 18% and 31% as profit. When the same events were re-run with a rule that required recorded proof that the size was actually available, fills went to 100% of 93, and a large part of the edge went with them.
The strategy was not the problem. The record was the problem. It had no way to tell the difference between a trade that happened and a trade that was assumed.
Proofline is the tool that difference required. It records what happened as it happened, computes the result with versioned deterministic code rather than judgement, and keeps the gaps visible instead of filling them in.
Why the losses are the point
The first strategy Proofline verified end to end lost money. A 30-day replay across 2,191 windows produced 504 intents, 504 fills, and a net result of minus $1,500.90 after $158.43 in fees. Every one of those positions was closed by a real recorded settlement, not an assumption.
That record is published, unedited, in the registry. It is the most useful thing on this site: it shows that the execution modelling held up and the strategy still lost, which is a conclusion no screenshot has ever volunteered.
A verification company that only ever published winners would be a marketing company. Negative, limited, refused and unavailable results are valid records here, and they are labelled as such rather than quietly left out.
What Proofline is today
A small, self-funded project in public beta, starting with prediction markets. Verification of a running system is operated by hand as a private pilot. The record registry, the strategy builder, and Evidence are open to anyone and free during the beta.
There is no outside funding to disclose, no customer list to point at, and no third-party audit of Proofline itself. When any of those changes this page will say so. Until then the only thing worth trusting here is the evidence attached to each record, which is exactly the standard Proofline holds everyone else to.
Read further
- Methodology — What is verified, what is not, and how.
- Evidence policy — How evidence is classified and what is never claimed.
- Corrections policy — How errors are reported, reviewed and restated.
- Limitations — What Proofline does not claim.
- Legal — Terms, privacy and disclosures.
Every figure on this page is copied from a sealed engine artifact in the repository. Proofline does not publish a number about itself that it could not be asked to show the working for.