A model I built on Rithmm I named My NFL has been a killer this season when it recommends Unders and guess what its says for the Super Bowl.

This NFL season, I’ve been running three different models inside Rithmm. One custom model I built myself, one Rithmm-built model, and one I copied directly from the leaderboard.

All three serve different purposes. But one thing stood out clearly as the season went on.

My primary NFL model finished 26–17 on totals. That’s a 60.5% win rate and a 15.1% ROI. And of those 26 wins, 18 came on unders.

That isn’t random. And it becomes especially important heading into the Super Bowl.

Why Unders Don’t Feel Right but Keep Winning

Most bettors hate unders.

They feel boring. They feel passive. They feel like you’re rooting against excitement. That’s exactly why markets tend to misprice them, especially in high-profile games.

The Super Bowl is the peak version of this problem.

Everyone expects offense. Everyone expects fireworks. Everyone expects stars to deliver on the biggest stage.

Sportsbooks know that. And they price totals accordingly.

Models don’t care.

Why AI Models Gravitate Toward Unders

AI models aren’t trying to predict highlights. They’re trying to predict distributions.

When you strip emotion out of it, big games tend to slow down. Coaches play tighter. Drives get longer. Mistakes are punished. Possessions matter more.

That combination often compresses scoring, even when elite offenses are on the field.

What the model is picking up on isn’t that teams suddenly forget how to score. It’s that expectations tend to run hotter than reality.

That gap is where unders live.

Why This Shows Up Strongest in Totals

Totals are one of the cleanest places for models to express an edge.

You’re not betting on one player. You’re not relying on one moment. You’re betting on how a game unfolds as a whole.

When my NFL model shows a 60.5% hit rate across a full season, that tells me something important. It’s not chasing spots. It’s consistently identifying where the market is leaning too far in one direction.

The fact that 18 of those wins came on unders reinforces that this isn’t about being conservative. It’s about being realistic.

Why This Matters for the Super Bowl

The Super Bowl is the most bet-on game of the year. That also makes it the most narrative-driven.

Public money loves overs in this game. That pressure pushes numbers higher than they should be, especially on totals.

That doesn’t mean every Super Bowl under is good. It means the conditions are ideal for models to surface one when it actually makes sense.

Using Rithmm, I’m not betting unders because I like them. I’m betting them when the model shows the price is off.

That distinction matters more in this game than any other.

The Bigger Takeaway

The most important thing this season taught me is that Rithmm doesn’t force one style of betting.

I can use different models for different purposes. I can trust a copied leaderboard model. I can build my own. And I can still lean on the same underlying logic that keeps showing up across all of them.

When a model quietly goes 26–17 with a 15% ROI, you don’t argue with it. You listen. The under it is and the lin 46.5 now, if you can get it higher of course grab it.

Heading into the Super Bowl, that’s exactly what I’m doing.

‍