Bet the Process Podcast — Summary

Main themes

• The hosts discuss a difficult stretch in sports betting, the challenge of distinguishing bad luck from flawed models, and the emotional toll of trading.
• They also talk about prediction-market tools, live betting research, personal priorities, election forecasting, and betting-market design.

Betting results and model evaluation

• Rufus says the fall has been mentally demanding, with long workdays and several painful losses. The team’s positions have often lost by narrow margins or landed on especially unfavorable numbers.
• To test whether poor results reflected bad luck or bad pricing, he compared 91 bets across 50 games against Kalshi market prices:
• The bets returned −27% ROI, an unusually poor result.
• Simulations assuming the market price was correct also produced very bad outcomes, suggesting luck may explain much of the slump.
• The team’s overall accuracy measures were similar to, or slightly better than, Kalshi’s.
• Rufus says he remains confident in the process but is using extensive data analysis to keep evaluating it.
• Both hosts discuss how bettors remember losses on key numbers more readily than wins from the many times those numbers do not decide the result.

Live betting research

• Rufus is testing a simulation model that updates at halftime and at possession changes, while recording predictions and market prices in real time.
• Early results look encouraging: the model’s results are about 2.5 standard deviations from random, and market prices tend to move toward its estimates.
• The team is not yet ready to compete with high-frequency trading systems. Its first goal is a manual execution tool that flags favorable prices and lets a trader sweep them.
• Rufus has replaced his college-football model, Massie Peabody, with a new model called Cadet. Cadet has performed poorly so far, though it has generated more market movement than the old model.

Game examples and picks

• In the Ohio–Akron game, the hosts discuss how early-game positions and halftime bets created exposure around several spread numbers. Ohio’s late 91-yard touchdown nearly changed the result, but Rufus says Ohio did not deserve to cover based on its play.
• Rufus’s Bet the Process Heater Pick is Central Michigan, especially on the moneyline, against Eastern Michigan. He sees Central Michigan as potentially undervalued, partly because of quarterback-injury considerations.
• Jeff describes a large Indiana moneyline position against Minnesota. After Minnesota took a late lead, he hedged at what he believed was a small positive-EV price. Indiana then tied the game with a long field goal and won in overtime. He reflects that the decision was difficult to assess because it was made quickly and depended on a late-game model he trusted less than his usual process.

Personal reflections

• Jeff questions whether he still wants to bet second halves. He says losing clusters can make the experience miserable and take up valuable weekends, even though he enjoys other betting activities, such as college basketball and Calcutta contests.
• Rufus says he is excited to trade football at the start of some days, which suggests the work still brings him genuine enjoyment, even when the results are discouraging.
• The hosts discuss drinking socially, personal priorities, and whether Rufus wants children. He says he could imagine being happy either way, but would want children only with a partner he loved and wanted to raise them with.

Other discussion

• Neither host bet on the U.S. election. They would like a prediction-focused post-election analysis, including whether polling or market expectations were wrong.
• On gambling advertising, Jeff favors restricting gambling ads on social media. Rufus suggests requiring gambling companies to fund problem-gambling messages as part of their advertising.
• They prefer prediction markets not to expose traders’ identities by default, while acknowledging that market makers may use identity information to limit abusive behavior, such as quote-spamming or exploiting obvious pricing mistakes.
• Rufus reports that Novig is handling substantial volume and promotes the podcast’s BTP offer.