Do bigger edges win more often? Our 578 calls

No, not on our record so far. We split every settled BAI8 call by the size of its edge, and the wider edges did not win more reliably than the narrow ones. Our 179 positive-edge calls won 77 times: more than the prices implied (about 73) but fewer than our own chances said (about 83.5). The gap between our claim and the result is widest exactly where the edge is widest. That is a weakness in our method, and this piece puts the numbers on it.

The record, band by band

This uses the 568 calls on the Desk record that settled as won or lost with a price, from 4 October to the morning of 8 November 2026. "Our expected wins" adds up our chance on each call. "Price expected wins" adds up the chance the best price implied. Units assume one unit on every call at the quoted price.

Edge band	Calls	Won	Win rate	Our expected wins	Price expected wins	Units
Negative edge	389	246	63.2%	197.5	237.5	+17.4
0 to 3 pts	58	31	53.4%	27.0	26.2	+10.5
3 to 6 pts	50	17	34.0%	23.4	21.3	−10.6
6 to 10 pts	40	16	40.0%	19.1	16.0	−1.8
10 to 15 pts	22	11	50.0%	10.1	7.4	+9.9
15 pts or more	9	2	22.2%	3.9	2.2	−1.5

The total is the +23.9 units on the Desk record this morning. A negative edge means the price was shorter than our fair price; we publish those too, for reasons in why BAI8 publishes negative-edge calls.

What the table says

The narrow band is the steadiest. Calls between 0 and 3 points won 31 of 58, a touch more than we expected and more than the prices implied. These are cases where we roughly agree with the market and found a small gap.

The middle bands are where we overrated ourselves. Between 3 and 10 points we expected about 42.5 wins and got 33. That is the clearest sign on the record that a medium-sized gap often means we missed something.

The 10 to 15 band did well, the 15-plus band did not. Eleven wins from 22 against 10.1 expected, then two from nine against 3.9 expected. Saturday 7 November alone added four losses to the widest band: Gimnasia Mendoza, Pau, Halifax and the Volos draw. Nine calls is far too few to call it a pattern, but it matches what we already suspected and wrote about in why a very wide edge makes us look twice.

Why a wide edge is not a safer bet

Two things are true at once. First, wide edges in football mostly sit on long prices: a 3.50 or a 4.40 loses most of the time even when the chance behind it is right. Win rate alone will always look poor in the widest band. Second, a very wide gap is more often a sign that the research is wrong than that the market is asleep. A bookmaker price is many people's view; ours is one view. When we disagree by 17 points, the honest prior is that we are the ones who missed the injury or the motivation.

The record supports both. Positive-edge calls as a group beat the price, 77 wins against 73.1 implied. But they fell short of our own chances by about 6.5 wins. Our chances on positive-edge calls run high, and they run highest when the gap is widest.

The negative-edge band shows the mirror image. Those calls won 246 times against our 197.5 expected: we underrate favourites, which is what our calibration on 378 calls found too.

What we do with this

We keep publishing wide edges, because hiding them would make the record look better than the method is. We keep flagging them on the card and in previews as the ones to read most carefully. And we will rerun this table as the record grows, because 568 calls split six ways leaves too few in the bands that matter most.

What this does not tell you

It cannot tell luck from error in any band yet; the samples are too small, especially above 10 points. It does not account for closing prices, which are a better test than results. It uses the best price we saw, which not every reader could get.