This one required re-pulling targeted historical snapshots rather than reusing the existing archive, because it needed two prices for the same game: a midweek price and a closing price. Every game was priced from real bookmaker moneylines pulled from the historical odds archive, then de-vigged so the two sides add to 100%. That de-vigged number is the market's honest win probability. A factor only counts as an edge if games matching it beat that probability by more than chance would explain. Weeks 15 through 18, 2021 through 2025 — the part of the season where motivation diverges. 183 games in total: 144 in the elimination arm (a team with nothing to play for) and 39 in the clinch arm (a team already locked in). A control group of week 5 and week 10 games from the same seasons, where every team is still playing for something, to prove the effect was about motivation and not simply about lines improving as kickoff approaches. Both prices de-vigged, then scored on accuracy against the real result, and the difference tested for significance.

edge lab · factor writeup · 8 min read

Late-Week Information Timing: The One Edge That Survived

Q: What the numbers said

Sample: 183 games; Window: 2021–2025, wk 15–18; Signal: z = 3.21 On the elimination arm, closing lines were significantly more accurate than midweek lines against the control — z = 3.21, comfortably past the threshold where chance stops being a reasonable explanation. The clinch arm, at 39 games, was too small to call either way and is reported as inconclusive rather than dressed up. Certified: late-week timing: This is the only factor in the register that earned a certification mark, and it is the reason the product nudges you to lock picks late rather than early in the back half of the season.

Closing lines beat midweek lines at z = 3.21 in late-season games involving teams with nothing left to play for. Waiting for late-week news is a real, usable edge.

Ten of the eleven things we tested failed. This is the one that did not. It is also the least glamorous, which is usually how this goes: the surviving edge is not a secret pattern, it is patience.

What we tested

The hypothesis: in late-season games involving teams that are mathematically eliminated or already locked into their seeding, the closing line is meaningfully sharper than the midweek line — because that is exactly where the important information (who is actually playing) arrives late.

How we tested it

This one required re-pulling targeted historical snapshots rather than reusing the existing archive, because it needed two prices for the same game: a midweek price and a closing price. Every game was priced from real bookmaker moneylines pulled from the historical odds archive, then de-vigged so the two sides add to 100%. That de-vigged number is the market's honest win probability. A factor only counts as an edge if games matching it beat that probability by more than chance would explain.

Weeks 15 through 18, 2021 through 2025 — the part of the season where motivation diverges.
183 games in total: 144 in the elimination arm (a team with nothing to play for) and 39 in the clinch arm (a team already locked in).
A control group of week 5 and week 10 games from the same seasons, where every team is still playing for something, to prove the effect was about motivation and not simply about lines improving as kickoff approaches.
Both prices de-vigged, then scored on accuracy against the real result, and the difference tested for significance.

What the numbers said

Sample: 183 games
Window: 2021–2025, wk 15–18
Signal: z = 3.21

On the elimination arm, closing lines were significantly more accurate than midweek lines against the control — z = 3.21, comfortably past the threshold where chance stops being a reasonable explanation. The clinch arm, at 39 games, was too small to call either way and is reported as inconclusive rather than dressed up.

Certified: late-week timing

This is the only factor in the register that earned a certification mark, and it is the reason the product nudges you to lock picks late rather than early in the back half of the season.

What it means for your picks

In weeks 15 to 18, when your game involves a team with nothing left to play for, do not lock your pick on Tuesday. The information that decides those games — resting starters, a backup quarterback, a coach in his last fortnight — lands Friday and Saturday, and the closing line contains it while the midweek line does not.

For every other week of the season the practical advice is unchanged: take the de-vigged favourite, plan the path, and do not outthink the price. The live board with real book depth is on the odds page, the method is in the methodology, and the honest register of everything that failed is in the Edge Lab.

More: back to the research register · methodology · guides