Two. That is how many of our wins this week came on half-stake picks. Three. That is our total number of wins. The math is simple. The discipline keeps producing the only winners on the board.
Saturday night, Phoenix beat Las Vegas 99-66 outright on PHO +9. Half stake because the prior was suspect. The win was the second-largest cover of the entire WNBA opening week, and it came on a bet we had explicitly downgraded.
Sunday night, Seattle beat Connecticut 89-82 in Uncasville. We had SEA -1.6 at half stake because Magbegor was still out and the road favorite number was hard to trust at full Kelly. The Storm covered without trouble. Our second straight half-stake winner.
Here is what the same three nights produced from the full-stake side. The full-stake STRONG LEAN UNDER on DAL at IND went 211 against a 178.2 line on Saturday. Last night the full-stake UNDER on LVA at LAS went 183 against a 176.8 line. Both losses. The strongest model edges by tier became the largest dollar losses on the slate.
That pattern is not statistically significant after three nights of WNBA games. It is also not nothing. The model is at its most aggressive and least calibrated in the first two weeks of any season because the priors are doing too much work. The picks where we sized down were the picks where we identified a specific reason the prior might be wrong. Two of those have hit. None of the full-stake picks have. We are running an experiment in real time and the experiment is telling us to keep sizing down.
Now the back-to-back rule. We added it Sunday morning after the Vegas postmortem. Any team on a back-to-back gets a tier downgrade in the model. PHO at GSV and LVA at LAS were both back-to-back games. We marked both SKIP.
Phoenix at Golden State went exactly the way the rule predicted. The Mercury, exhausted from the Vegas game, lost by 16. Skip was correct. The rule paid off.
Las Vegas at LA went the opposite direction. The Aces won by 27. Vegas was on a back-to-back after the LVA loss to Phoenix and they still demolished the Sparks because Los Angeles is the worst team in the WNBA right now. The talent gap was so wide that fatigue did not matter. Skip was wrong by a lot.
That is the lesson. A blanket back-to-back tier downgrade is too crude. The model's rolling rating already knows roughly how good each team is. When the absolute talent gap is wider than the back-to-back fatigue effect, the gap wins. The rule needs a talent-gap override.
The refinement, written down here for the record. A back-to-back team's pick gets downgraded one tier UNLESS the model's spread already shows a talent gap of more than the expected fatigue penalty (about 3 points in the WNBA). When the model has Vegas as -2.5 favorites against a team that is otherwise a -8 dog, the back-to-back makes Vegas a closer game, not a flip. The bet stays on. We will apply this starting Wednesday.
Cumulative through three nights. Three wins, six losses on published leans. Down 3.7 units. The bankroll is taking its hits the way an early-season Kelly system always does. The shape of the wins matters. Both half-stake winners covered easily. Both full-stake STRONG LEAN losers were within 7 points of the line. That is a model that is directionally right and still volatile around the line. We do not need to change the model. We need to keep sizing for the volatility.
Tonight there are no games. The first off-day of the WNBA season hits in the slowest part of the week. The model has nothing to bet, the cards have nothing to publish, and we have a chance to recalibrate. Wednesday brings five games on the slate including Toronto's home opener against Seattle at Coca-Cola Coliseum. The Tempo Report ships with that game.
The number that matters at the end of this week is not 3-6. It is whether the discipline keeps separating winners from losers when the priors are at their weakest. So far it has, twice in three nights. We keep sizing down on the picks where the prior smells wrong. We keep going to the line on everything else. The model gets less volatile every game. The discipline does not need to.
[ End Report ]
Share This