In a Two-Season Replay, Moving Lines Leaned Toward Newton v3
In a replay of the 2024 and 2025 seasons, when the consensus point spread moved, it moved toward Newton v3's side 54.7% of the time. Here is what that shows, and what it does not.
The short answer
In a replay of the 2024 and 2025 regular seasons, when the consensus point spread moved between the opening and closing line, it moved toward Newton v3's side of the opening number 54.7% of the time. That is 700 of the 1,280 games where the line moved. It is a signal about direction, measured on past games. It is not a live record, and it is not a betting edge.
What we measured
We replayed two past seasons of FBS-vs-FBS regular-season games. For each season, Newton v3's settings were refit using only the seasons before it. These are seasons the team had already studied, so treat this as a backtest, not a blind test. The lines are consensus market lines: the middle value across sportsbooks at the open and at the close.
Of the games whose line moved, the move went toward Newton v3's side 54.7% of the time (95% range: 52.0% to 57.4%). Both seasons leaned the same way, 53.1% in 2024 and 56.3% in 2025. The 2024 season on its own is not conclusive: its range runs from 49.2% to 56.9%, which includes a coin flip. The two seasons together are what make the result worth a forward test.
The comparison is what gives the number weight. Our original rating model, run on the same 1,280 games, showed a much weaker lean: lines moved toward its side 52.3% of the time, a result that chance alone could produce. The new version is the piece that added most of the signal.
The signal grows with disagreement. The bigger the gap between Newton v3 and the opening line, the more often the line moved Newton's way:
| Gap from the opening line | Lines that moved toward Newton v3 |
|---|---|
| Under 2 points | 51.5% |
| 2 to 4 points | 54.1% |
| 4 to 7 points | 59.3% |
| 7 points or more | 62.4% (73 of 117) |
The average move was under two points (1.70), so these are small nudges rather than dramatic swings.
Markets also drift. Over the same games, lines moved toward underdogs 53.9% of the time. When Newton v3 was on the underdog, the line moved its way 57.3% of the time. When it was on the favorite, 50.9%. Once that drift is accounted for, Newton's lean still predicted the direction of the move by about four percentage points (4.2, with a 95% range of 1.4 to 6.9). The same adjustment leaves the original model at 1.7 points, inside the range chance produces.
What it does not mean
It is about direction, not cause. A line moving toward Newton does not mean the market read Newton's numbers or followed them. In the replay, Newton's numbers were not public when the line opened.
It is not a betting edge. Against the spread, Newton v3 went 51.1% against the closing line (749-717, with 32 pushes), below the 52.4% needed to break even at standard -110 odds. Against the opening line it went 52.0%, about break-even. A line that moves your way 55% of the time is not a bet that wins 55% of the time. Not betting advice.
The closing line still picks more winners. Across all 1,498 FBS-vs-FBS games in 2024 and 2025, each with a consensus closing line, the closing favorite won 73.0% of the time. Newton v3 picked the winner in 71.8%, and the original model in 71.7%. Almost all of the gap sits in close games. When the closing spread was under 3 points, which covers about 16% of games, the favorite won 50.6% and Newton v3 picked 48.6%. Those games are close to coin flips for everyone.
It is a replay, not a live record. The forward test of this result is a proposal frozen on October 17, 2026, before any game it covers. It covers FBS-vs-FBS games kicking off on or after October 18 and allows no model changes. It has not been scored yet.
Why it matters
Most fans only see the final number on the screen. This work is about the early read. In the replay, when Newton v3 and the opening line disagreed, the line more often moved toward Newton's side than away from it, and the pattern was strongest when the disagreement was largest. It suggests Newton is picking up some of the same information the market takes in between the open and kickoff. Whether that holds on games played from here on is what the forward test is for.
Where this leaves us
The original model showed only a faint lean here, within what chance produces. Newton v3 shows a clearer one, pointing the same way in both seasons, that gets stronger as its disagreement with the opening line grows. That is the clearest sign so far that the v3 work added something the old model did not have. The closing line is still the more accurate guide to winners, and the forward test will say whether the pattern holds on games nobody has seen yet. How v3 was built is in How Newton v3 Was Built.
Newton’s picks lock at kickoff and are graded in the open, wins and losses both. See Newton’s record and Newton and the betting line for where the model stands today.