Betting splits don't reliably measure what the public thinks.
Betting splits track where money landed, not what the crowd believed. MWT Edge measures sentiment independently from betting splits and generates fade signals when consensus reaches predefined thresholds.
Simulated walk-forward results using only pregame data, not actual wagers. Comparative only, using the same game universe, time period, pregame odds window, and sizing rules. Methodology & disclosures.
sentiment signals captured
games analyzed
Why Betting Splits Break Down
Betting splits show where money landed. They do not reliably show what the crowd actually believed. By the time bets appear in the market, public opinion, sharp action, and steam are already mixed together.
Splits are not sentiment
Splits measure ticket and money distribution, not the conviction that shaped the number in the first place. The market often moves before the split tells you anything useful.
The signal gets contaminated
What looks like public money is often a blend of retail action, sharp positioning, syndicate influence, and late steam. That makes splits noisy by default.
The edge is small
Walk-forward backtests compare fading betting splits to fading the MWT Edge sentiment signal over the same games and same market. See backtested results below.
Simulated walk-forward results using only pregame data, not actual wagers. Comparative only, using the same game universe, time period, pregame odds window, and sizing rules. Methodology & disclosures.
How MWT Edge Works
Three steps. No gut calls, no subjective reads. The same process runs every game.
Measure Sentiment
For every game, MWT Edge quantifies sentiment from multiple sources independently from betting splits — capturing what the crowd is saying, not just where the money is moving. Thousands of data points per day, collected and timestamped before tip-off so the same data is available to backtests and live picks alike.
Identify the Crowd
When sentiment strongly favors one side, the model flags the game. It checks that the lean is lopsided enough and backed by a sufficient sentiment sample size to be meaningful.
Fade the Consensus
When the crowd is heavily aligned, the model takes the other side. When betting splits confirm that same lean, it doubles down. The threshold is a uniform two-thirds, chosen as a simple rule rather than tuned to maximize backtest performance — a methodological choice intended to reduce overfitting risk.
What a Signal Looks Like
Every game gets a sentiment profile. When the model flags a divergence, here is the format users see in the dashboard.
Illustrative example. Not a live signal, not a recommendation, not a current matchup.
Walk-Forward Performance
Historical walk-forward results from the current production model. Same season, same pregame odds window, stated staking assumptions. Each pick generated using only data available before that game.
Sentiment is collected and timestamped before tip-off. The walk-forward backtest uses the same recorded pre-tip-off sentiment snapshots, so post-start information does not enter the results.
Simulated walk-forward results using only pregame data, not actual wagers. Comparative only, using the same game universe, time period, pregame odds window, and sizing rules. Methodology & disclosures.
Live Model Record
Live signals from the production model. Every pick generated and locked before tip-off, every outcome recorded. Small sample so far — the record will fill in as it grows.
Live signals generated and locked before tip-off. Small, growing sample. Past performance is not indicative of future results. Methodology & disclosures.
Track the System
Free access to daily model outputs, every signal timestamped and logged before tip-off, and a community of bettors interested in sentiment as a data source for sports betting analysis.