← Journal
methodology1 min read

Model vs market: why we blend instead of betting our own number

Our testing showed the market consensus beat the raw model on accuracy — and the blend beat both. Why respecting the market makes predictions better, not weaker.

Momus
Published 4 September 2026

Every prediction site claims its model beats the market. We tested ours honestly and found what almost every honest test finds: on straight accuracy, the sharp market consensus edged the raw model. Then we tested the blend — and it beat both parents. That result now runs through everything Modal publishes.

Why the market is hard to beat

A market price is the aggregated opinion of everyone with money on the line, weighted by conviction. It digests injuries, lineups, weather and news faster than any single system. Treating that as noise to override is how models go broke; treating it as information is how they get sharper.

Where the model earns its keep

Markets are slow exactly where models are fast: quantifiable specifics. The starter's FIP against this lineup, a bullpen run into the ground this week, a rating the crowd hasn't caught up to. Our blended probability starts from the market and lets the model pull it where the evidence is strong — mostly market, corrected by model.

What that means for reading our numbers

On the MLB board you see all three: model, market, blended. When the model and market agree, there's nothing to do. When they diverge by points, that's a flag worth reading — and the written analysis explains which signal is doing the pulling. The engine behind Momus's real-money record bets exactly these blended numbers. We don't publish anything we wouldn't stake.

methodologysports betting marketsexplainers
Share on X →

Frequently asked questions

If the market is more accurate, why use a model at all?+

Because the blend of model and market beats both. The market aggregates enormous information but moves slowly on specific signals — starters, bullpen fatigue, matchup details. The model adds exactly that correction, and the combination is measurably sharper than either alone.

See it on the desk

Every fixture, fully modelled — the correct-score grid, the derived markets, and the written read.

Start free
More from the journal