New to these reports? Start here
- Dotted-underlined words have a plain-English definition — hover or tap them. Every term is also on the glossary page.
- "Null" means we found nothing, not that something broke. Most reports here are negative results, on purpose — knowing an idea doesn't work is the point.
- Two questions get asked separately. First, is the effect real? Second, is it already priced into the betting odds? An effect can be completely real and still useless to bet on.
- A "calibration" row is a self-check. It runs the same method on something already known to be true. If that fails, the whole report is unreliable — so it's reported alongside the findings.
- If a confidence interval includes zero, the real effect might be nothing at all, so no claim gets made.
Middles don't clear the vig — not even with perfect hindsight
2026-08-28. First study enabled by the results join (1936e7d), which
attached final scores to the 10-book line history in unified_odds.db. Middling
is a structural strategy — bet both sides at different numbers, win both if the
result lands between — so it belongs in the category that has actually survived
here (teasers, bonus-bet bands) rather than the predictive category that hasn't.
Break-even: betting both sides at −110 needs the middle to land 4.8% of the time.
Simultaneous cross-book middles, at the closing snapshot
Best available line on each side across all 10 books, realised P&L at actual closing prices:
| segment | middles | hit rate | ROIreturn on investmentProfit as a percentage of the money wagered. +2% means $2 profit per $100 bet. on stake |
|---|---|---|---|
| ncaab/totals | 877 | 2.17% | −2.18% |
| ncaab/spreads | 723 | 2.49% | −2.22% |
| mlb/totals | 165 | 0.00% | −4.62% |
| mlb/spreads | 99 | 25.25% | −0.61% |
| nhl/totals | 88 | 1.14% | −4.87% |
| nba/totals | 54 | 0.00% | −3.93% |
| nba/spreads | 50 | 2.00% | −1.75% |
Every segment loses. Windows exist often (80% of NCAAB games) but average only ~1 point wide, so the result has to land on essentially one number.
mlb/spreads hits 25% — an artefact of the pinned ±1.5 run line, where a
"middle" is just any one-run game — but the juice on both sides still leaves it
at −0.61%, inside noise at n=99.
Perfect-hindsight middles, across the whole pre-game window
The above only tests simultaneous middles. Real middling comes from line MOVEMENT: bet one side early, take the other after the number moves. So this takes the best Over line at any moment against the best Under line at any moment — assuming you bought each side at its single best instant.
| segment | window exists | avg width | hit rate |
|---|---|---|---|
| ncaab/totals | 96.8% | 3.07 | 7.3% |
| nba/totals | 96.9% | 3.56 | 4.8% |
| nhl/totals | 82.7% | 0.72 | 2.6% |
| mlb/totals | 90.9% | 0.70 | 2.0% |
Only NCAAB clears the 4.8% bar, and NBA lands exactly on it. This is an upper bound nobody can reach — it requires knowing in advance when each side peaked. Live, you would capture some fraction of a 7.3% ceiling against a 4.8% cost.
Verdict: dead
Even granting perfect foresight, one segment beats the vigvigThe bookmaker's built-in cut. It is why a coin-flip bet at -110 needs you to win about 52.4% of the time just to break even. by 2.5 points. At the close, achievable middles run −0.6% to −4.9%. There is no version of this that survives contact with execution, and the one segment that looks least bad — NCAAB — is also where limits are lowest and books are quickest to restrict.
Consistent with the earlier finding that arbitrage is "limited before you profit": same family, same reason.
Scope
Feb–Aug 2026 only (the captured window), 1,863 games with both a pre-kickoff snapshot and a linked result. NFL is absent — the capture window contains almost no NFL. Worth one re-check after a full NFL season, since football totals move more than any other market here, but the prior after this is poor.