Everything we tested,
including what failed.

Three full seasons, replayed match by match with every price set before the result was known. No selections removed afterwards. Here is all of it.

Season by season

CLV is closing line value — how much better than the closing price the model's picks were, which is the honest measure of whether it saw something real. Simulated ROI is what fixed reference stakes would have returned in the replay — no money was staked, and the number depends entirely on the staking rules chosen. We treat it as reference data from the collector, not gospel.

Model skill against a placebo

The same machinery run on scrambled data, subtracted from the real thing. Anything that survives is signal rather than the testing procedure fooling itself. This is the part that held up every time, including on the season the model had never seen.

Real, replicated, and far too small to bet.

The exam we failed

We found what looked like an edge across two seasons — the model's strongest disagreements, excluding home wins, beating the close by about two points. We froze the rules and tested them on a third season the model had never touched.

It came out flat against the closing price and lost heavily on money. Both pillars broke: conviction stopped concentrating the edge, and the "never back home teams" rule reversed outright. That is what a pattern fitted to the past looks like when it finally meets fresh data.

The full write-up →

Model vs market, league by league

Every settled paper pick is a disagreement with the market — the model claimed an edge, the price said otherwise. This is how those disagreements have gone, league by league. CLV is closing line value: whether the price the model took beat the closing price, the honest measure of whether it saw something real (shown only where a closing price was captured).

The live paper record, 2026/27

The fourth exam runs in public. Every high-conviction call is published before kick-off and settled afterwards, judged on beating the closing price and on profit at the price taken. The bars were written down before the season started and cannot be moved.

22 of 55 fixtures read right · CLV -3.36% on 54 closed price(s) · paper P/L at a £1 reference stake: £-5.77

Leg level: 77 calls published · 66 settled (22 won / 33 lost / 11 push) · 9 pending. Legs on the same fixture rise and fall together — 77 calls across 77 fixtures is closer to 77 separate opinions than 77. Count fixtures, not chips.

Closing line value over time — the running average across settled fixtures. This is the honest version of a tipster's profit graph: it can climb or fall, and it stays up either way. Faint dots are single fixtures; the line is the running mean. Counted per fixture (not per leg), so it can read a touch below the leg-level headline above.

Paper profit and loss over time — the graph every tipster shows, drawn our way: £1 a call, one call per fixture, every loss in, and it stays up when it falls. Three seasons of testing said this line does not climb; watch whether the season agrees.

✓ verified price. The tick beside each odds figure is how long before kick-off that price was captured and written to the ledger by the collector — nothing is priced after the fact, and the closing price it is judged against is the last one seen before kick-off.

12 settled call(s) show no CLV because no closing price was captured for that market — the gap is shown, not filled in.

The early-season window

For a club’s first few games the model still leans on last season’s data, and that is where its ugliest misreads have come from (documented on the wrongs page). The rule was pre-registered on 19 August, before the results it applies to were known: a call sits inside the window if at kick-off either club had completed fewer than 5 league matches this season. Same untouched rows, split two ways — the full record above is always the headline, whichever view looks better.