20 years ago today, I recovered enough for the doctors to take me off life support (I used to aim to win at any cost, even my life). I began a bonus life, when I wrote Basketball on Paper, got into sports, met my wife, and lived a lot. Very grateful for 20 yrs of bonus life.
Threshold Win: When a player has more positive net points than the team won by. Also interpreted as: The team wouldn't have won w/o him.
Threshold Loss: When a player has more negative net points than the team lost by. Also interpreted as: The team would have won w/o him.
Thanks for the encouragement since the announcement of my new position. As I've promised the Wizards, I promise that I will work hard, work smart, and aim for the highest levels of success, in hopes that it also opens up opportunities for others...But I won't promise many tweets.
I joined ESPN in 2011 to help build their fledgling Analytics group. We got a lot done before I left in 2014. Since then, several people I worked with carried the group forward. shows off some of what they've done. And now I'm very happy to rejoin them.
When Kyrie plays in a game then goes off the court, his teammates do worse.
When Kyrie _does not play_ in a game, his teammates do better.
That is weird. Can't adjust style in game after playing w/him? It's both O and D, essentially all his teammates. I don't think it's sched.
Sports analytics is so multidisciplinary that it has forced me to learn or re-learn
- network theory,
- evolutionary dynamics
- geometry and trigonometry
- machine learning
- statistics
And that's not counting all I've learned from coaches about their game (hoops and football)
@kirkgoldsberry
Analytics is all about taking the best shot. Oh average, it's layups and threes, but the core is taking the optimal strategy.
And... Late game strategy to maximize win % can be a little different than maximizing points.
A fascinating JJ Redick interview with Joe Mazzulla. It's long, but at around 37:15 or so, the conversation shifted to analytics. The next 20 minutes or so are basketball rocket science. A great listen.
Yet it often takes a player's absence to notice. Last year at this time, people were taking shots at him for his vulnerabilities and saying how KD is better. KD is really good, but Steph is changed-the-game special.
Warriors have an offensive rating of 120.4 with Curry on the court (would comfortably be an NBA record) and 106.2 without him (basically the Charlotte Hornets). Steph is easily their best and most important player.
It has been 33 years or so since my first work on points per possession got published. I was at
@Caltech
, studying hard, playing hard. I had no idea then it would lead to all of this.
Curry may be out of the playoffs, but here is one remarkable stat about him that came up this year. Players with the most career games with 9+ 3pt field goals made:
4. Harden with 9
3. Klay with 14
2. Dame with 16
1. Steph with 41
I reached 25 years of life after death, as of today. I should have died two other times, doing crazy risky things, but the one in 1998 didn't feel risky. Just biking uphill for 78 mi as hard as I could. I celebrate today, the anniversary of when I came off life support. Good day.
It is now 26 years of bonus life after coming off of life support on this day in 1998. I remember looking around the hospital room and being confused, as well as the pain of having a catheter ripped out. But I was alive and ready to take advantage of the time I had left.
It was twenty years ago in November that Basketball on Paper came out. It introduced a lot of analytics to the game. In about a year from now, my second book will come out. Here's a preview:
OKC-GSW: Paul George had the highest correlation of his own net pts to the team. Kevin Durant was negatively correlated to what GS was doing - the better he did, the worse the Ws did - a common thing with KD. I kinda like Huestis as the low usage/good D Roberson replacement.
A Threshold Win is when your individual contribution to a win was as big as the point margin. Exs: If you added 5 net points and won by <=5, that's a Threshold W. A Threshold Loss is when you cost 3 pts and you lost by <=3. Here are the leaders. (Doncic is in the pack at +2.)
In the last ten years, these are the top five players in terms of "threshold wins" in the playoffs, where the player's net points are what carried team to win:
1. Lebron, 32
2. Durant, 15
2. Harden, 15
4. Curry, 12
5. Paul, 9
All the guys still left this year.
Math and science can get you to Mars, to the NBA, to the boardroom, and fundamentally to a state of knowing more than the next person. Don't bash it, don't take it lightly. Study it, learn the process.
(Not paid for by anyone. Just my occasional mention of what's important.)
One of my gripes about a common stat: points off turnovers is really misleading when so many turnovers are dead ball. They should report points off of steals, not off of turnovers.
Just heard about Campbellโs Law: โThe more any quantitativeย social indicatorย is used for social decision-making, the more subject it is to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor." Like sports stats.
Probably the best thing I learned at
@Caltech
was that there are a lot of people smarter than I in this world and that I can find ways to contribute nonetheless.
I submitted the manuscript for Basketball on Paper to the publisher 20 years ago. The basketball analytics industry didn't really exist then (it was just a community), so I guess it's come a long way.
The NBA 3-point revolution is changing defense. This year shows stark decline in defensive performance by bigs, with simultaneous improvement by guards. (See 2019 line relative to the others.) There are things that I need to think through on this, but it's interesting.
In ways, my new book this fall is how you can change players via analytics. This article is about how KP made his body do things that analytics suggested. I'm glad that I was part of the group that helped him, but he put in the mental and physical time.
This is why the corner 3 point shot is overrated, by the way. I did a study 13 years ago, suggesting that the above-the-break 3 has so many secondary benefits to overwhelm the small FG% increase in the corners. And that stretching deeper helps a fair amount.
Preorders are available for my book, Basketball beyond Paper, which is scheduled for release in November. The link has a short overview, as well as advance praise and the list of chapters.
1/ Totally agree with this.
I think there is a strange prejudice in economics that studying sports, movies, game shows etc is not โseriousโ
The truth is that as Ignacio explains, they often provide unusual control and diagnosticity.
One of these days, I will figure out what it is about Q3 that causes Steph Curry to go all supernatural. I have looked into Quarter-related differences before and you can relate to fatigue or adjustments or caffeine. Whatever Curry drinks at halftime, I want some of that.
I saw this line in here "Joel Embiidโs production routinely drops in the second half of games" and looked it up. Embiid has THE BIGGEST 2nd half drop off of any player. Embiid goes from 8.3 net pts/48 to 0.9 in the 2nd half. Next biggest is L Galloway w/7.0 drop.
Trade season is afoot in the
#NBA
.
@KevinOConnorNBA
breaks down all of the teams, players, contracts, and scuttlebutt to keep an eye on as the deal-making starts to heat up:
Clutch performance is somewhat noisy, of course. But offensive usage in clutch time is not. Alpha-players use a lot of possessions. These are the top possession users so far this year, with the rank from last year. Doncic on the list! And actually 12th in O Net Pts.
My eye: I forget about it, but I am, well, cross-eyed and people notice it right away. People who are around me forget about it with time, but first impressions - I look like an idiot because that is the stereotype. It was a crazy low probability event that caused it... 1/14
I don't have numbers for today's game, but historically players going 0/7 from the field or worse and committing 5+ turnovers - they have been bad. Josh Hart did a lot of good things today around the bad shooting/turnovers (generally under the radar guy).
Simulating the NBA Draft can be a useful exercise for teams - forcing them to plan for contingencies of guys get taken higher or lower than expected. ESPN put together a simulation for that, where you can take the role of whichever team you want.
In the playoffs, the entire difference for Boston has been 3pt shooting. In their wins, the net value of 3s is +14 (offense minus defense). In their losses, it's -20. On average, they're winning the non-shooting factors in both wins and losses, like tonight, winning TOV and FTs
Now the individually most inconsistent players from game to game, as defined by an average of percentage of games that are good and bad. Mostly guards, but LaMarcus Aldridge is way up there.
I personally wish there was an NFL Draft show where the talking heads reviewed their prognostications from the last few years. And someone kept track of how many they got right/wrong. That would be entertaining and informative.
22 years ago today, I awoke from life support, the nice nurse ripped the catheter out, and she moved me to a normal room. I had luckily survived pushing my body past its limits in order to win a race. In some cases, giving 110% can kill you.
One of Deandre Ayton's best defensive games of the season against Milwaukee tonight. Made it tough on Giannis around the rim, looked really agile guarding the perimeter, made some fairly instinctual rotations, recovered nicely for blocks. 3 BLK + 12 REB for the
#1
pick.
My work in basketball has been a lot about how to optimally incentivize players to be team players. I don't think I've ever written it that way. At its most basic, it's about making sure that โฆ 1/5
It is possible to have zero minutes and zero seconds played in an NBA game but have a turnover. Justin James did it on 12/26/19, . He threw an inbound pass off the rim and out of bounds. (One of those rare events that threw an algorithm off.)
A quick overview of what I'll be putting into the new book. This is on Royce Webb's new site, Royce having been the longtime ESPN editor who really led a lot of people to basketball analytics via
@johnhollinger
,
@kpelton
,
@TrueHoop
KP was the most receptive player I ever worked with to analytics. It was great working with him. (Pictures with him also made me look like he was my dad because he's so much bigger.)
Let's just keep celebrating
@manuginobili
, the most underappreciated basketball great that ever played (and that I know of). Prepping for him and Tim Duncan was absolutely brutal.
I have data on Cleaning the Glass for the last 15 seasons. 2017-18 was only the 2nd time in those 15 seasons that the Spurs did not have the point differential of a 60-win team or better with Manu on the court.
In six of those seasons they played like a 68+ win team.
Under Ty Lue, the Cavs have won 66% of their games. Under David Blatt, they won 68%. Under Lue, the O got 5 pts/100 poss better than under Blatt, but the D got 5 pts worse.
I lost my sight (and looks) to an infection more rare than winning $500m Powerball. I pushed myself so hard in a bike race to end up on life support. Yeah, life is hard, but all that came before the public ever heard of me... Work harder and smarter.
Hmm, Anthony Roberson has been out since 12/29 and this is how the team has done defensively (in green). I have Roberson as the best floor defender in the NBA this year.
"The hardest thing to coach is decision making."
I would agree. Getting too complicated makes decisions inaccurate or too slow. Getting too simple makes decisions inefficient.
Ben and I haven't explicitly worked together in probably 10 years, but we have a lot of the same guiding thoughts on basketball and analysis. The following are his thoughts on how lower level basketball coaches can use data.
The Warriors with Durant playing with Draymond generates an O efficiency of 123, D efficiency of 108, net of +15. Durant w/o Draymond or Draymond w/o Durant is nowhere near as good. It's ironic, but they need each other more than any pair on the Warriors.
Jacob is projecting not only wins but offensive and defensive performance. Even when wrong about a win total, this kind of projection is good because you can see why/how. (That also can show you got O and D wrong, but total right.)
There are a lot of basketball people behind $Chedda - John Wall, Duncan Robinson, Eric Spoelstra, Buddy Hield, and Alex Len. Crypto linked to sports access. Interesting.
๐
#BitMart
has opened the deposit feature for
@CheddaToken
๐ $CHEDDA/USDT pair will be tradable at 10AM EST, Mar 10
๐ฅ To celebrate the listing, we are giving away 4,816,000 CHEDDA ๐ฅ
๐ Deposit:
๐ Details:
Quinn Cook had 5 threshold losses this year - basically games where the Warriors would have won had he not played - and 0 threshold wins - where Ws would have lost without him. That -5 difference was 3rd worst in NBA behind Kevin Knox and Jabari Parker.
No basketball updates for a week or so. I'm going off grid. There's nothing in my contract that says I can't do crazy risky things. So I'm going rock climbing and ignoring the world for a few days.
Warriors took 32 layups to Toronto's 18, but made just 50%, their second lowest percentage this year. Cousins making 1/6 didn't help.
On the other hand, Toronto had roughly their 10th best shooting game of the year. Danny Green's best O night of the playoffs.
I'm a big believer in the value of metrics, but "metrics fixation" in the way described can be a problem. There are dumb metrics. There are dumb ways to use metrics. Remember that they're a guide to making things better, not a crutch to be lazy with.
Key argument of โฆ
@jerryzmuller
โฉ Tyranny of Metrics book, re: metrics fixation. Essence: โnot everything thatโs important is measurable, and much that is measurable is unimportant.โ Cc โฆ
@FrankPasquale
โฉ โฆ
@nntaleb
โฉ
No sports on TV, so I watched a film on when the NBA last suspended a season due to mysterious illness. That's right, I watched Space Jam. MJ dunked on some evil toons and kept the world from Moron Mountain.
Dear Tankers - Remember 2013 when teams tanked for a top 5 pick? And they all missed out on the best player in the Draft - Giannis at
#15
? (And Gobert at
#27
) On average, top 5 drafting teams pass over 3-4 players who end up better than who they took.
#TankingFails
I'll share this again, in part to thank WIRED for doing a great job on this. It was an interesting set of questions and, seriously, it ain't easy putting my face in a video.
I've been lucky to work with some great people, in and out of sports, on and off of teams, many of whom will never get publicity. It's good to see
@bencfalk
get some from
@SI_ChrisBallard
,
There have been 72 triple-doubles this year, 11 of which have been net negative contributions. Westbrook has 7 of those 11 negative triple doubles (and 12 positive ones). He's had 22 negative triple doubles in the last three seasons, more than the rest of the league combined.
A former student put together a simple classification scheme for players - Superstar, Star, ..., Reserve, Fringe. It comes from a more GM/coach/eye test perspective and is easy to implement. His article
Some rookie progression numbers this year. Doncic has dominated, next three picks have been OK and steady. Trae Young has leveled off after a terrible start.
Rule of thumb for box scores: +/- totals for ATL starters was -60, or -12 per player. +/- totals for the bench was +30, but divide by 5 (not 4) to get +6 per player because there are always 5 on the floor. The rough estimate: starters cost them 12, bench got them 6, losing by 6.
Threshold Wins are when a player has enough total net points to take a team from losing to winning. M Harrell, D Fox, and J Embiid, for instance, all had Threshold Wins last night. Good performances in fairly tight games. Here are leaders in rate of getting them since 2011 seas.
If broadcasters can make a reference to PER in the playoffs (as they just did), we aren't that far away from having them use an even better metric with all the stories it could tell.