If it's real, prove it here.IF IT'S REAL,PROVE IT HERE.

A profitable screenshot proves almost nothing.

Not because the person is lying. Because the format cannot carry the evidence. Here is what a result has to survive before it means anything:

• When the signal was actually created
• Whether the trade could have filled at that size
• What fees and slippage really cost
• Which orders failed or filled short
• Whether the losing weeks are still in there
• Whether the rules changed afterward

Most backtests answer the hardest questions with assumptions. They assume the fill happened, the slippage was manageable, the fees were small, the rules stayed the same, and the losing stretches still count. Proofline is built to verify what actually happened instead: signals, fills, missed orders, fees, liquidity, strategy changes, and settlement, so the record reflects reality rather than a clean simulation.

In Proofline's first measured strategy, from one live bot on one venue pair over one week, 352 fill legs were matched one-for-one to exchange-confirmed records. Firing on the visible displayed price alone filled 94 of 115 legs on Polymarket (81.7%) and 162 of 234 on Kalshi (69.2%).

What Proofline records

Signals

When the rule fired, and on what data.

Orders and fills

What was sent, what came back, at what price.

Fees and liquidity

What it cost, and whether the size was there.

Failed and partial orders

The ones that did not fill, and the ones that filled short.

Strategy changes

Every version of the rules, and when it took effect.

Settlement and resolution

How it closed, against the venue's official result.

Corrections and restatements

What changed after the fact, and why.

Missing or unverifiable evidence remains visible instead of being filled in.

From a running system to a record someone else can check.

1. 01
   Connect or submit a strategy

   A live system, a paper run, a recorded dataset, or old files and code.
2. 02
   Proofline records activity as it happens

   Raw responses stored before anything is normalized.
3. 03
   Deterministic systems calculate the record

   Versioned code, not judgement. The same input reproduces byte for byte.
4. 04
   Missing data stays visible

   A gap is shown as a gap, never interpolated to look complete.
5. 05
   A sealed record is published

   With its dataset hash, model versions, limitations and full history.

Recording and calculation are automated. Onboarding a system and publishing its record are operated by hand during the pilot, and Proofline will not call this self-serve until it is.

A result is only as credible as the evidence behind it.

Who this is for

Strategy builders

Find out whether the edge survives execution, before capital does.

Bot developers

Prove the fills were real, at the size claimed.

Trading-agent teams

Evaluate an autonomous system on recorded behaviour, not on its own summary.

Creators

Replace a screenshot with something a sceptic can inspect.

Sceptical traders

Check a claim before paying for it.

The limits of a Proofline record are written down rather than implied. Each page below states one of them in full.