The Docket publishes the whole record — the losses at the same size as the wins.
Prediction markets are on-chain, so nothing here has to be taken on faith. When a wallet with a settled record puts unusual money on a side, we print the call: market, side, price at that second, and the wallet’s own history. When the market resolves, that same line is marked HIT or MISS and carries the realised return of having followed it. Nothing is backfilled, nothing is deleted, nothing is re-ordered to flatter us.
The current model is derived from 708 settled observations across 634 markets. Market price remains the prior; demonstrated wallet and market-family performance adjusts it. Old calls never move, and every model version is retained.
Walk-forward check, not the live record: after the 50-result warm-up, the historical stream contained 641 calls at +$5.23 per $100. Applying each version of the learner only after its inputs had resolved retained 502 at +$9.02; the 139 signals it would have passed returned −$8.45. This is a time-ordered backtest, not a promise of future performance.
“What is the realised return of following your signal, including the losses, and who can check it?”
We went through the category on July 25, 2026 — the betting-analytics tools, the options-flow desks, the crypto “smart money” dashboards. Not one of them publishes an aggregate record of its own calls including the ones that lost. The strongest proof any of them offers is a set of selected customer success stories, verified after the fact. That is a survivor, not a record.
The reason is not laziness, it is arithmetic: a full record is only a good idea if you are willing to show the bad half. A proprietary score with no published method and no published results can never be wrong in public. Ours can — the rule that fires a call is written out in the words the code runs on, and every line links to the market and the wallet so you can re-derive it without asking us for anything.
That is the entire position: the one number a tipster will not print, printed.
Step 1 is a list, and the list is public. The rated cohort is every wallet the rule is currently reading from — each one with its own public settled record, its composite score and all five components behind it. Those are the wallets’ own results, not calls we have made. Rebuilt daily, free, no account.
No editor, no discretion, no narration, and no language model anywhere in it — the same inputs always produce the same call. The exact rule, verbatim → · the cohort behind the calls →
Everything that has resolved is on the ledger for nothing, permanently, no account. What you pay for is time: the call at the moment the rule fires, at the price it fired at, on your own live page.
See what it costs →Before you read anybody else’s calls, find out what your own record says. The free X-Ray reads every fill you ever made on Polymarket — public, read-only — and returns your realised P&L, your percentile against the field, and the single most expensive habit in your process.
We email your scan link when it’s ready (about a minute).
Read-only public data — you never connect a wallet, we never ask for keys.
Free, in about a minute: your record, your percentile, your #1 leak. On Kalshi? Upload your fills instead →
The median active wallet has made nine dollars. Lifetime. Half the field is barely above water, and the gap between the halves isn’t intelligence or information — it’s process. That is the field the rated wallets are being picked out of, and it is why the rating is built on settled results instead of on who is up the most.
Same machinery that rates the cohort behind the calls, pointed at one wallet instead of the field.