🔥🔥

Charlotte-Miami win probability chart:

Is this a clutch assist? He does recognize quickly that the screen has left Gordon open. But how much credit should he get?

www.nba.com/stats/events...

I think most would agree this is a legitimate clutch assist from Jokic (double teamed and finding an open Hardaway on the opposite end of the court): www.nba.com/stats/events...

But....

Here are their respective highlight reels:

Shai: stats.inpredictable.com/nba/jerryWes...
Jokic: stats.inpredictable.com/nba/jerryWes...

To make the case for Jokic, you would need to believe that assists and rebounds count as clutch play. Shai leads by a good margin when it comes to clutch shooting.

stats.inpredictable.com/nba/jerryWes...

Here are the final player clutch rankings for the 2025-26 NBA season. Shai and Jokic finished neck and neck.

the article:
www.espn.com/nba/story/_/...

As pointed out in the article, the relationship flipped this season. Although it flipped in 2022-23 too before reverting back.

Technical note: this chart shows pace for team offensive possessions only. Most traditional measures of team pace combine offense and defense.

Inspired by Zak Kram's article on pace in ESPN, here is an updated chart that shows team correlation by season between offensive pace versus efficiency.

An downward sloping line indicates a positive correlation between fast pace and offensive efficiency.

You can set your watch to the reliability of the rubber band effect in the NBA.

Red = Team overperforms
Blue = Team underperforms

Teams play better from behind. It's pretty much the opposite of momentum.

You can check a box on the inpredictable win probability graphs to display real time, rather than game time, on the horizontal axis.

That can be helpful for a game like LAC-IND where a game's worth of insanity was packed into the final second.

Corralling the Madness in College Basketball Why Don't More Teams Understand Late-Game Situations?

I watched #Kansas nearly blow a 26-point lead last week.

The bigger opponent for CBU wasn't Kansas, it was the clock.

With help from @inpredict.bsky.social, here's what the numbers say about when teams should stop hunting quick 2's & start hunting 3's:

open.substack.com/pub/edwithsp...

Update after last night's games:

New Substack post: March Madness Forecasting Thunderdome

I apply my Kelly-based forecast evaluation method to models predicting the NCAA Men's Basketball national champion.

Link: open.substack.com/pub/inpredic...

Just an absurd win probability chart

A lot of conseqeuntial games in the NHL today.

Here's how playoff chances for each team are impacted based on the outcomes today:

win: 2 points
overtime loss: 1 point
regulation loss: 0 points

Meta-bracketology for the NCAA Women's Tournament.

Of the 3,511 brackets analyzed, there were 2,817 unique brackets (80%).

The most common number of seed upsets picked was 0.

Next most common was 8.

Here is the distribution of pre-tournament perfect bracket probabilities, charted on a log scale.

Pre-tournament, there was about a 3 million to 1 chance that a perfect bracket would emerge from one of these 18,724 distinct submissions.

Some meta-bracketology, using 20,000+ brackets submitted to the NCAA's bracket challenge.

Of the 20,751 brackets analyzed, there were 18,724 unique brackets (90%).

The most common number of upsets picked: 12

Last year, the perfect bracket odds were about 10 billion to 1.

When I looked at this way back in 2010-2014, the odds averaged around 50 billion to 1.

Survival odds by round:

Round 1: 2,800 to 1
Round 2: 2 million to 1
Round 3: 67 million to 1
Round 4: 400 million to 1
Round 5: 1.4 billion to 1
Round 6: 2.6 billion to 1

2.6 billion to 1

That's the odds of a perfect mens tournament bracket this year, if you picked based on @kenpom.com ratings and assume that those ratings represent the true probabilities.

That's better odds than last year, where I estimated a 10 billion to 1 chance.

Beeswarm plot to illustrate the hilarity of Bam's 83 points.

Tonight's Spurs comeback is a good example for why KL-divergence is a better measure of game quality.

Current Excitement Index (total change in win prob):
MIA @ CHA: 8.6 ✅
LAC @ SAS: 8.2

KL-Divergence:
MIA @ CHA: 2.2
LAC @ SAS: 4.4 ✅

A huge thank you to @wikihoops.com for providing game level data on user ratings, allowing me to deomnstrate that using KL-divergence in this way wasn't just a fun theoretical exercise, but an objectively better measure of "excitement".

Using a key concept from information theory, KL-divergence, I create a new version of the Excitement Index that better aligns with how actual humans evaluate the games.

Quantifying Excitement Using information theory to improve the ranking of NBA and WNBA games

New (and first) post up on Substack.

Creating a better way of ranking games with the help of Claude.......Shannon.

inpredictable.substack.com/p/quantifyin...

Here's Desmond Bane's clutch play highlight reel (link in the right column takes you to video of the play):
stats.inpredictable.com/nba/jerryWes...

NBA season leaders in Clutch Win Probability Added.

The market favorite is Ant, but by this metric Desmond Bane deserves to be in the conversation for the Jerry West Clutch Player of the Year Award.