Favourite-side bias fails: 268 bets, 268-0, and only +1.0% ROI

What we found

We graded the favourite side of a market-bias hypothesis across 268 bets over 52 events. The record: 268-0. Every single bet won.

The verdict is NOT_SUPPORTED.

That is not a contradiction — it is the whole point. The market implied a 99.0% win rate on these positions. We hit 100.0%. The gap between what we were charged and what we delivered is worth +1.0% ROI, and at a raw p-value of 0.231 that gap is indistinguishable from noise at n=268. The market priced these favourites almost exactly right. A perfect record bought at 99 cents on the dollar is a savings account, not a signal.

The useful reading is what this rules out. If heavy favourites were systematically underpriced, 52 events was enough resolution to show it. It didn't. So we stop looking for the edge in the price of near-certainties and go back to the prices where the outcome is actually in doubt.

That is where the rest of today's board lives.

The trends

First, concentrated money is landing pregame. The largest campaign on the board is $216,028 on 215BALTB Over 8, sitting 66% one-sided. Behind it, $101,041 on BOS ML at 99% one-sided and $81,168 on TB ML at 98% one-sided. Total tracked whale money today is $906,280 across 12 campaigns, and the biggest single ticket is $24,703 — meaning this is broad participation, not one account.

Second, the lines are moving with that money, not against it. 335BOSPIT-BOS drifted 49.5c to 52.5c (+3.0) over the last 20 hours. 337NYYTOR-TOR moved the same +3.0, 50.5c to 53.5c. 515STLCHC-CHC went 60.5c to 62.5c (+2.0), and 340SDCLE-SD +1.0 to 51.5c. The lone counter-move: 335AZATL-ATL fading 54.5c to 53.5c.

Third, totals skew hard over. All 8 of 8 flagged total edges point over, led by SEA @ HOU, where the model reads fair at 10.15 against a posted 7.5 — a +2.65 run gap. That is the widest total disagreement on the slate.

Fourth, the arms are pulling in opposite directions. In BAL @ TB, Freddy Peralta is trending the wrong way: an 8.44 ERA over his last 5, 8.18 over his last 3, against a 5.33 season mark — deterioration, not a slump average. Meanwhile in CWS @ DET, Drew Anderson carries a 1.32 ERA over his last 5 with 10.6 K/9.

Fifth, the model leans underdog in 5 spots — BAL ML +130, WSH ML +153, STL ML +167 — which sits directly against where the money is going. The top moneyline read is STL @ CHC: STL ML +167, model 41.6%, EV +11.1%.

What it means

The favourite test tells us where not to spend capital: the 99-cent end of the board is efficiently priced and 268 graded bets could not find daylight there.

The live research book is doing its work elsewhere. VELO_DROP is holding out-of-sample at 25.6% ROI over 101 graded bets, and TRAILING_LATE at 10.4% ROI over 106 graded bets. Those are prices with genuine uncertainty in them, which is exactly where the favourite test says the money is not.

So the next test is the mirror image of this one. We are running the underdog side — the 5 spots the model likes today against one-sided money — and the 8-for-8 over skew, on the same grading standard. Both live where the outcome is in doubt.

All research is probabilistic. 21+ where legal.

What would change our mind: a favourite-side sample that pushes past n=268 and drops p below 0.05 while holding ROI above +1.0% — until then, this idea stays dead.