Every probability on this site comes from one season simulation. It was frozen on 2026-05-16 and has not been touched since. A reader deserves to know what that means before using any number on the board.
A frozen model is a ruler, not an oracle. It cannot tell you who wins. What it can do is stay still while prices move, and staying still is the only way to see that prices moved at all.
The Garrett trade, as a test case
On 1 June 2026, two weeks after our model went dark, the Rams traded for Myles Garrett, the reigning Defensive Player of the Year (ESPN). The model has never seen that trade.
Our own board watched the market absorb it. On 18 May the Super Bowl market implied 9.5% for the Rams. By 15 August it was 13.4%, with prices moving from around +700 and +800 to +500 at four of the five books that quote this market.
| Market | Model (frozen 16 May) | Market (15 Aug) | Gap |
|---|---|---|---|
| Super Bowl | 12.8% | 13.4% | under a point |
| NFC West | 40.2% | 47.3% | 7.1pp |
The Super Bowl row is unremarkable. The two numbers are close enough that there is nothing to explain. The division row is where the story is: a 7.1 percentage point gap, which on a straight reading looks like the model finding value the market has missed.
It isn't.
The gap is a date, not an edge.
That 7.1pp is the model describing a Rams team without Myles Garrett, because on 16 May there wasn't one. The market has spent eleven weeks pricing him in. A reader who saw 40.2 against 47.3 and read it as an opportunity would be betting against three months of public information.
When the frozen number is worth something
Most teams did not have a Garrett trade. Across the same window, 21 of 32 Super Bowl prices moved by 0.2 percentage points or less, and only three moved a full point either way. For those teams the May model describes roughly the team the market is pricing now, and the comparison carries real information.
What separates the two cases:
- Roster moves the model cannot see. A trade, a signing or a serious injury after the freeze changes a team's true odds without changing its model number. The model is static; the roster is not.
- Absorption is steady, not instant. Books price new information over days and weeks, which is the shape of the Rams moving from 9.5% to 13.4%.
- Size is not signal. A small gap on a quiet team means more than a large gap on a team that has been in the news all summer.
So the method is three questions, and it takes about a minute:
- What does the model say, and what does the market say now?
- What has happened to this team since 16 May? Trades, injuries, coaching changes, anything material.
- If nothing has, the gap is worth a look. If something has, ask whether the gap is about the size of the news.
The Forecaster holds both numbers side by side for all 32 teams. On why gaps open in the first place, see Why the Gap Exists; on checking the market side across books before trusting any single price, Where the Super Bowl board is widest.
The point of a fixed reference
A model honest about its date is more useful than one quietly re-run and never mentioned. Silence about a freeze date invites false confidence. A stated one invites the right question.
The number is a fixed point. Everything you do with it is your own judgement.
Prices are the boards as of 15 August 2026 across bet365, BetMGM, Caesars, DraftKings and FanDuel. Source: ESPN on the Rams' trade for Myles Garrett.