Verification
Botcappers runs 5 AI handicappers across UFC, MLB, tennis, NFL and college football. Forward selections and historical replay/simulation rows are identified separately and graded against results. Nothing is deleted. This page is the whole ledger reduced to the two numbers that matter, and they do not say the same thing.
Across 5,351 graded picks. This is the number that looks good — and on its own it is close to meaningless, because it says nothing about the price paid.
Across 12 priced selections from the 644 picks actually posted and played forward. This small priced sample is currently negative. Published anyway — that is the point of the page.
A forward-graded pick is distinct from a historical replay or closing-price simulation. They are never blended: historical output is a model diagnostic, not a preposted track record. Football rows have no locked entry price and therefore show ATS diagnostics, not assumed -110 units.
| Bot | Graded | Forward-graded | Historical replay / simulation |
|---|---|---|---|
| BonesUFC / MMA | 1,642 | 130 83-47 +1.35u on 8 priced | 1,512 947-565 no priced sample |
| ChipperMLB | 1,103 | 0 none yet | 1,103 554-549 no priced sample |
| T MoneyTennis | 1,842 | 514 339-175 -1.98u on 4 priced | 1,328 870-458 no priced sample |
| Trip NickNFL | 569 | 0 none yet | 569 378-191 no priced sample |
| HeismannCollege Football | 195 | 0 none yet | 195 136-59 no priced sample |
A -250 favourite needs to win about 71% of the time to break even. A bot that leans on short favourites will post a gaudy straight-up percentage and go backwards at the window. This is the single most common way a public pick record misleads. Monetary figures require an explicit price basis; unpriced football rows therefore carry no synthetic P&L.
Partly, and the honest answer depends on which question you ask. Across 5,351 graded picks the Botcappers bots are 3,307-2,044 (61.8%) at picking straight-up winners. But picking winners is not the same as beating a price. On the picks that were actually posted and played forward — 644 of them — the record at real market odds is -0.63u across 12 priced selections. That live sample is small and is currently negative.
Because a favourite priced at -250 has to win about 71% of the time just to break even. A bot that takes short favourites can post a high straight-up percentage and still lose money. The only number that answers 'did this make money' is units at the price that was available. Unpriced football records therefore report prediction and ATS diagnostics, not assumed monetary returns.
A live pick was published before the event and played forward. A backtest pick is replay validation over history the bot had not seen. They are never mixed here: the live arm is 644 picks (-0.63u), the backtest arm is 4,707 picks (—). A backtest is a model diagnostic, not a track record, and 3 of the 5 bots have no live picks at all.
No. Every graded selection stays on the record permanently, win or lose, and each one carries the risk note written before the result was known. The full ledger is public on each bot's page without an account.
Every number on this page is computed from the same ledgers that power each bot's public record and regenerates whenever the site rebuilds. Nothing here is projected, modelled forward, or filtered to a flattering window. Read the full pick-by-pick history on the record, or on any bot's page — no account required.
