Inside Dimers' new NFL pregame predictions model powering Moneyline, Spread and Total bets

Dimers' NFL betting pregame model has been completely rebuilt for the new season.

Built using machine learning, our model utilizes more than 100 data points for every one of our NFL predictions, from rebuilt proprietary power rankings and an improved approach to factoring in quarterback strength, to team efficiency, pace, form, and game-specific variables.

After six months of in-house development, we've extensively backtested the new model and achieved improved results against our internal performance benchmarks.

Two dedicated models use the same underlying data differently to predict the expected margin of victory and total points, combining to power Dimers' NFL predictions for moneyline, spread, and total best bets, further enhancing the power of a Dimers Pro subscription.

Unlock Dimers Pro at a preseason discount today.

This launch is focused entirely on NFL game lines and the three primary markets you'll find within every NFL prediction on Dimers.

Every NFL best bet and full game prediction you'll see on Dimers will reflect the output of this new model, all designed to bring you the best NFL betting tool out there.

A smarter read on every game; that's the core message behind our new NFL model.

What's new in Dimers' NFL pregame model

More data goes into every NFL prediction

The new model considers more than 100 different data points when predicting an NFL game.

These cover everything from team form, efficiency, and pace to quarterbacks, rest, travel, venue, and the specific circumstances surrounding the matchup.

Machine learning allows the model to interpret those signals together and determine how much they should matter to each prediction and best bet.

Rebuilt proprietary NFL rankings

The team rankings that sit beneath our NFL predictions and best bets have also been rebuilt.

They go beyond wins, losses, and raw past performance by accounting for factors such as opponent quality and the quarterback who was playing at the time.

The ratings are tuned for predictive accuracy before being fed into the wider machine learning model as another measure of each team's underlying strength.

A better understanding of quarterbacks

We've reengineered how the model accounts for quarterbacks, giving it a better understanding of one of the biggest variables in any NFL game, and a position widely considered among the most important in all of sports.

It considers the quarterback expected to play and whether that represents an upgrade or downgrade compared to recent games.

The new system also improves how we rate rookie quarterbacks, giving the model more meaningful information when there is limited NFL history to work with.

Better separation of meaningful football from garbage time

Not every yard or point tells us the same thing about a team.

The new model adjusts offensive and defensive efficiency metrics for garbage-time situations, helping reduce the influence of performance when the competitive outcome of a game has largely been decided.

Built for today's modelling capabilities

Our previous NFL model was built using older technology and modelling techniques that are now considered outdated.

The new model has been completely rebuilt using modern machine learning, which allows us to process more information and model the relationships between different inputs in greater detail. This isn't a tweak to what came before; the model and its key underlying components, including our power rankings, have been rebuilt from the ground up.

How Dimers' new NFL model works

Every upcoming NFL game is assessed using more than 100 data points. Key inputs include:

The same underlying information is then used differently by two dedicated machine learning models:

1. Margin model — Predicts the expected margin between the two teams.
2. Total model — Predicts the expected combined points scored by both teams.

Those two outputs combine to power Dimers' predictions and best bets for our three primary markets: Moneylines, Spreads and Totals.

Why trust the Dimers NFL predictions model?

Six months in the making

This update is not a minor adjustment to our existing NFL model, but developed over the past six months, including rebuilding important underlying components such as our team power rankings and our approach to factoring in quarterbacks.

Extensively tested before launch

Machine learning allowed us to test tens of thousands of different modelling approaches before we selected the final models.

The new model has also been extensively backtested against historical games and has shown improved results against the internal benchmarks and minimum performance standards we set before launch.

More information, used more intelligently

More than 100 data points now contribute to every prediction and best bet.

Rather than treating each piece of information equally, the models learn how those factors relate to the specific outcome being predicted, allowing the same underlying information to be used differently when predicting margin and total points.

Built entirely in-house

Engineered entirely in-house, the result is a proprietary prediction system built specifically for Dimers. You won't find these predictions and best bets anywhere else.

Ready to see it in action? Dimers' new NFL moneyline, spread, and total predictions and best bets are live now for the 2026 season. Tap below to lock in our early bird discount on complete football season access.