[warmly] Hey, it’s Duke Sizemore. I’ve got my coffee, I’ve got the model open, and today feels like one of those classic NBA contrast days. On one side: a reunion that just makes sense. On the other: a team that looked invincible in June waking up to some very real August math.
We’re going to talk Gary Payton II back to Golden State… and we’re going to talk about the New York Knicks — who closed last season at 1808 Elo — now staring at a preseason reset with Jalen Brunson set for wrist surgery. And in between, we’ve got some sneaky market movement that’s way more interesting than it looks at first glance.
[pause] Alright. Before we dive in, let’s do our regular check-in. How have we been doing?
On the full tracked sample — 1,246 games — the model is hitting 66.8% on moneylines. That’s with favorites winning 68.3% of the time, so we’re basically right in line with how often chalk actually comes through. Brier score sitting at .211, which for a probabilistic model is solid. It means when we say 70%, it behaves like 70% over time.
But zoom in to the last 25 games? We’re at 60% with an average margin error of 14.8 points. And here’s the interesting part — it’s not close games hurting us. It’s blowouts.
[thoughtful] The five worst misses in that stretch? They averaged a 30-point gap from expectation. Detroit hosting Cleveland on May 18 was the ugliest. We gave Detroit a 75.2% win probability… and Cleveland won 125–94. That’s a 37.9-point margin error.
When we dug into it, most of those games had Day-To-Day tags with basically zero impact weight. The model saw “DTD,” but the injury table gave it a 0.0 or 0.1 adjustment, so Elo and home court did almost all the heavy lifting. And when a matchup snowballs — when one team just overwhelms the other — pure Elo doesn’t always widen the spread fast enough.
So this is one of those moments where I’ll say it plainly: the model isn’t failing at picking winners as much as it’s underestimating how ugly a mismatch can get when availability is murky. That’s a tuning problem. And it’s interesting. I love this part — when the numbers tell you where they’re blind.
Alright. Let’s shift to today’s headline.
Gary Payton II is back with the Warriors. One year, $3.9 million, veteran minimum. Cap hit about $2.4 million for 2026–27.
And here’s why this matters beyond nostalgia: veteran minimum deals are low risk, but they’re not low signal. When a team chooses familiarity over a flyer, it’s telling you something about how they view their internal chemistry.
Golden State just finished 5–1 and atop the finalized Las Vegas Summer League standings — 76 total games in the tournament, it’s done. Summer League doesn’t move Elo directly, but it shapes rotation confidence and depth expectations, which absolutely filters into preseason priors.
Payton is entering his 11th NBA season, seventh in the Bay. That continuity has a measurable footprint. When we talk about chemistry in the model, we often reference on/off net rating — that’s simply how the team performs per 100 possessions when a player is on the floor versus when he’s off. Over time, stable on/off signals shrink uncertainty in projections. Bringing back a known quantity tightens variance.
[chuckles] Also… it just feels very Warriors, doesn’t it? Low drama, plug the defensive connector back in, move on.
Now let’s pivot to the Knicks — because this is where the emotional arc of the offseason lives.
New York finished 70–32. They closed at 1808 Elo. That was number one. They surged +41.5 Elo over their final stretch. Beat San Antonio back-to-back, 94–90 and 107–106. Over their last four tracked games? Offensive rating 128.8. Defensive rating 102.6. That’s a +26.2 net rating in a 96.7 pace environment.
Let me translate that. A +26 net over four games is obliteration territory. That’s not “we edged you.” That’s “we controlled everything.”
But here’s the thing about Elo: it’s a rolling measure of team strength, not a trophy. When the season flips, we apply regression toward the mean and roster context adjustments. So New York’s carryover Elo isn’t 1808. It’s 1700.2. Market Elo? 1703.5. Title probability? 7.87%, which is +950 in odds.
And the preseason rank slots them third — behind Oklahoma City at 1750 and San Antonio at 1749.
Why the drop? Two big forces.
One: regression. When you close at a peak, the model asks, “How much of this was sustainable versus a heater?”
Two: Jalen Brunson.
He’s listed Day-To-Day, but he’s set to undergo surgery on his right wrist/forearm area, with at least two months of rehab expected.
[sighs] Two months in preseason math is enormous. Even if he’s back early in the regular season, the uncertainty compresses title probability. The model doesn’t punish optimism — it prices absence windows and re-integration volatility.
So what looks like disrespect — going from 1808 to 1700 — is really just Bayesian gravity. You start with last season, you pull toward league average, then you layer injuries and roster stability.
Actually, hold on — let me pull up Sonar for a second.
[curious] Okay, this is fun. The 76ers were originally the Syracuse Nationals. The “76ers” name came from a contest after they moved to Philadelphia, referencing 1776 and the signing of the Declaration of Independence.
I love that. A franchise literally named by the city’s history.
Why bring that up? Because Philly’s in a weird spot right now. They dipped –4.4 market Elo in a single day, sitting at 1689.5 with a 6.89% title probability, even though there’s headline buzz — Kevin Durant comparing their star power to his Warriors teams, LeBron joining. The market sometimes exhales after hype. It’s not always linear.
Now let’s talk power movers.
Utah posted the biggest single-day jump: +20.3 market Elo. Trend +19.1 since July 29. They’re at 1466 market Elo with 0.82% title odds — long shot territory at 10000.
Here’s the key: this isn’t marked as a reversal. So this is continuation momentum, not noise snapping back. Even though the cited notes are Summer League absences — Trey Alexander (rib), Ace Bailey (back) — the broader market shift is upward.
Actually… one more Sonar check.
[genuinely surprised] The Jazz kept the name “Jazz” after moving from New Orleans to Utah in 1979. One of the NBA’s great name-location mismatches. There is something poetic about “Jazz” in Salt Lake City.
Okay. Back to movement.
Detroit slid –6.5 in a single day. No clear public driver. They’re sitting at 3.18% title probability, 2500 odds. That’s not catastrophic, but it’s notable because it’s not flagged as a reversal either. It’s a fresh downward nudge.
Atlanta ticked up +4.3, but here’s where I want to slow down. Their trendDelta since July 29 is –1.4, and reversal is marked true. That means today’s bump is a wobble inside a slight downward trend — not a directional change. So we don’t suddenly say, “Hawks rising!” We say, “Small give-back inside a drift.”
That reversal flag matters. It keeps us from storytelling off one candle on the chart.
Portland up +2.7, mild continuation. Chicago, Milwaukee, Sacramento all –2.3 with broader –4.2 trend windows. These are quiet board adjustments, not tectonic shifts.
And at the top of the title race? Oklahoma City at 24.31%. San Antonio 21.75%. Knicks 7.87%. Boston 5.9%. Detroit 3.18%.
Holmgren talked about departures being “the business of this game,” and that’s exactly what Elo captures — the business side. Sentiment fades; ratings update.
Now… upcoming games.
There are no preseason games on the slate yet. The next confirmed regular-season tip is October 3: Miami at Toronto.
So instead of a matchup preview, here’s what we’ll be watching when that opener approaches.
One: how New York’s Brunson rehab timeline firms up. If his projected return date tightens, you’ll see injuryAdjustment start nudging probabilities back upward in individual games.
Two: whether Utah’s market bump sustains for more than a week. A +20 single-day jump is loud. Sustained trend is louder.
Three: how Philadelphia’s number stabilizes. If hype converts into lineup certainty, market Elo tends to drift back toward preseason Elo.
[pause] And this is the fun part of August basketball. It’s not about box scores — it’s about probability mass moving around in the dark.
The Knicks story, to me, is the emotional anchor today. A team that closed like a juggernaut — 128.8 offensive rating over their final four tracked games — now opening the cycle with a carryover Elo 100+ points lower and their star guard heading into surgery.
That’s not a fall from grace. That’s the math acknowledging uncertainty.
And if Brunson returns clean, if that +26 net version of New York reappears? The Elo will climb again. Ratings are earned, not inherited.
[laughs softly] Which, honestly, is why I love this sport. We get to watch numbers chase reality in real time.
Alright. That’s the board for today: Payton back in the Bay, Knicks recalibrated, Utah nudging upward, Detroit wobbling, Philly exhaling after hype.
If you’re tracking with me, keep an eye on those reversal flags and injury windows — that’s where the subtle edges live.
[warmly] I’m Duke Sizemore. Thanks for hanging out, thinking through the math with me. We’ll keep the coffee hot and the probabilities honest. See you tomorrow.