Zero. That is our wins on opening night. Two. That is our losses. Negative two units. That is where the live scoreboard sits at 7 AM Saturday morning, less than twelve hours after I published an article telling readers the model had four strong leans on the slate.
I am going to walk through what happened. I am going to be specific about what the model got wrong, because the only thing worse than losing two picks is losing two picks and refusing to say why.
Connecticut at New York. The model said NYL -3.4. The market said NYL -11. We took CON +11.2. New York won by thirty-one. Final score 106-75. Our edge calculation said the line was about eight points too wide. The actual outcome said the line was about twenty points too narrow. The model was on the wrong side of the line entirely.
Here is the specific thing the model missed. The 2025 Liberty had a net rating of +6.1. That is the number our priors fed into last night's model. The 2026 Liberty added Satou Sabally. They also have a healthy Stewart for the first preseason in three years and a Jonquel Jones who took the entire offseason to rest a knee. The model was pricing the 2025 Liberty plus Sabally as a small upgrade. The Liberty playing in their building on opening night with their full new core looked like an entirely different team. They scored 30 points in the first quarter. The Sun never recovered.
The model is built on prior-year ratings. That works for teams whose rosters did not change much. New York's roster changed enormously. The model does not know how to price that until it has watched the 2026 Liberty play five or ten games. We will know more by Memorial Day. Until then, every game involving New York carries elevated uncertainty. We will not be putting STRONG LEANS on Liberty games for at least the next two weeks.
Golden State at Seattle. Model said spread of +1.6 (essentially a pickem). Market said GSV -6. We took SEA +6. Final 91-80 Golden State. Loss by five points off the spread.
This one is closer to a fair beat. The Storm without Magbegor are not as bad as the model expected, but they are not good. The Valkyries got 26 points from Gabby Williams, who the model knew about but did not weight as heavily as it should have. Gabby's playing-time bump from 24 minutes a game in Seattle to 31 minutes a game with Golden State is the kind of usage shift that adds 2-3 points to a team's net rating that the season-prior priors cannot capture. The model will adjust as Gabby's 2026 stats accumulate. For now: Valkyries games where Gabby is the focal point are mispriced.
Toronto at Washington was a PASS. The model said Tempo -0.9, the market had Tempo -1.0. The model agreed with the market within a tenth of a point. There was no edge to bet. Tempo lost by three. The model called it a coin flip and the result was within a possession, which is exactly what a coin flip should look like. No P/L impact, but a useful calibration data point: when the model agrees with the market, the result tends to land where both agree it should.
The actual scoreboard right now: 0-2 on STRONG and MODEST leans, -2 units, 0% win rate. The 2025 backtest hit 57.6% over 151 picks with +10% ROI. Two picks is not a sample. But it is a public sample, and we put it on a scoreboard at /picks because the only way trust gets built is if the wins and the losses both show up the same way.
Tonight's slate. Three model leans on four games. Walking through them.
Phoenix at Las Vegas. STRONG LEAN PHO +9. The market is selling Vegas at -8 because of Wilson and Loyd and Gray. The model thinks the Aces are 1.9 points worse than that line. This is similar architecture to the Liberty pick from last night, with one key difference. Phoenix has Alyssa Thomas, who the model loves. Vegas has the brand. Bet against the brand, take the points.
Chicago at Portland. STRONG LEAN POR +5.5. The expansion home dog. Portland defaults to league-average ratings because they have no prior season data. League-average against the rebuilding Sky in Portland is a reasonable take. The risk: we know absolutely nothing about Portland yet. They could be league-average. They could be terrible. This is more of a directional lean than the model is showing.
Dallas at Indiana. STRONG LEAN UNDER 178.2. The total is the strongest single bet on the slate. The model has 167.4. Eleven points of total edge. Indiana's pace is controlled. Dallas pushes pace but turns the ball over. Both teams play a half-court game when the rotations are still settling. UNDER is the sharp side here.
The two pieces of advice I would give myself if I were reading this. First: shrink position size on opening week. The model is at its weakest when it has the least 2026 data, which is right now. Cut Kelly stake by half until at least May 25th. Second: do not bet against the model when it has a directional view, even when the result feels obvious. The Liberty looked unbeatable last night. They might be. They will also be priced like it by every market in the league within a week. The edge does not live where the consensus is.
Three picks tonight. Tomorrow morning the scoreboard updates and we know more than we knew this morning. That is how this works.
[ End Report ]
Share This