Kostya Medvedovsky @kmedved Twitter profile

Pinned Tweet

Kostya Medvedovsky

@kmedved

7 years

@ProfWrightGMU Antitrust hipsterism. Everything old is cool again.

2

0

17

Last Seen Profiles

@MizuhoCommunity

@andy_kingdon

@Chasonpapes

@falconforchrist

@israel_at_war

@AccesDenied4

@Rhenna_Henituse

@RowdyPfeif

@bpritchard152

@art_qm

@Glitch_Smash

@AytunCiray

@Mick2304mick

@KakesksoAp

@7_nanana_777

@FelipeNuru

@louise_bowman68

@rolyatistaylor

@AnikaGillo75442

@katex

@FieryNupes

@stjoes

@heymrshallahan

@ViingoL

@jiyarampate

@MattersonNadia

@GoddessMercyXOX

@JerodMacEvoy

@FCB_PANDA

@IngDAVID1182

@WrenTales

@pekoyukari

@20Shiraish99139

@VmaniakJ

@lisagnv

@paradise_box

Kostya Medvedovsky

@kmedved

11 months

@willystaley @stretch14tw @TheStalwart Carry cash. Never use it. The Zoomers have this right.

4

0

512

Kostya Medvedovsky

@kmedved

4 years

FYI: the growth in positive test results per @COVID19Tracking data maps almost 1:1 with the growth in tests performed : R^2 of 0.97. In other words, reporting new and daily case counts is really just reporting on the ramp up in testing capability. @DanRosenheck @Noahpinion

32

163

424

Kostya Medvedovsky

@kmedved

2 years

@busteddealbot This filing is comedy gold, containing the text exchange below from May 8 (2+ weeks after the Twitter deal was signed), confirming the @matt_levine theory that Musk is just confused about how mergers work.

14

79

351

Kostya Medvedovsky

@kmedved

4 years

@NateSilver538 This is the cost of past reporting on spikes in cases without mentioning increased testing. It makes it much harder to get people to pay attention now after two months of crying wolf. And with the lag, it’s going to be a bit before deaths start to spike proportionately.

33

43

282

Kostya Medvedovsky

@kmedved

5 months

Durant missing the playoffs with one of the preseason title favorites is gonna be the pièce de résistance of my campaign that he's the most overrated player of the modern era.

27

261

Kostya Medvedovsky

@kmedved

3 years

Tatum's defense has fallen off a cliff in DARKO right when he got Covid last year. This sort of dropoff is striking because DARKO is pretty stubborn normally. It takes extraordinary results to get this sort of collapse in such a short time period.

15

47

250

Kostya Medvedovsky

@kmedved

7 years

Everybody get angry. Version 0.1 2018 win projections based on current rosters. Minutes projections done algorithmicly, so lots of errors.

61

111

234

Kostya Medvedovsky

@kmedved

27 days

People often ask what I mean by DPM being an impact, not a talent metric. And I give them some yada yada answer about team context and role, yada yada... Going to send them this chart of Al Horford's experience playing with Joel Embiid instead from now on.

13

18

232

Kostya Medvedovsky

@kmedved

1 year

@NateSilver538 @Redistrict He's gonna rig this thing to make sure "Yes" wins. Clean exit.

13

3

214

Kostya Medvedovsky

@kmedved

3 years

Who would have ever thought coming out of the draft these two would combine for 17 all-star games, 4 MVPs, and 38K+ points.

Burak Can Koc

@burakcankoc

3 years

0

1

7

5

18

208

Kostya Medvedovsky

@kmedved

3 years

@NateSilver538 It's a bearish signal for our ability to catch data fraud like this when the cases we *are* finding are mind bogglingly obvious from spending 20 minutes with the data. Strongly suggests we're only finding the tip of the iceberg.

6

10

198

Kostya Medvedovsky

@kmedved

2 years

@ModeledBehavior We had one. It was called the Manhattan Project.

3

5

187

Kostya Medvedovsky

@kmedved

5 years

2020 NBA win projections based on current rosters. Includes reported trades (e.g., AD to the Lakers), but anyone who isn't signed yet is just not on a roster. Fills in missing minutes with -2 projection. Uses 2019 schedule as a proxy until the actual schedule is released.

23

40

158

Kostya Medvedovsky

@kmedved

4 years

@COVID19Tracking @DanRosenheck @Noahpinion I was briefly optimistic about the slowdown in the growth in cases from ~+35%/day to +25% this week until I ran this correlation. I suspect we're capturing close to *zero* information about actual case-growth in other words and are just tracking the exponential growth in testing.

8

33

144

Kostya Medvedovsky

@kmedved

3 years

DARKO's list of most improved players since the season began. Grant Williams leads, but still a ways to go. Both Ball brothers are top 5!

5

22

134

Kostya Medvedovsky

@kmedved

3 years

I've looked at the predictive power of preseason results before, and there's real value here. I'll do full writeup/tweet thread on the methodology, but a +32.5 preseason net rating suggests we should bump our expectation for the Bulls' regular season ratings by ~1.5 points/100.

Steph Noh

@StephNoh

3 years

I was curious, so I looked it up. NBA Stats has 15 years worth of preseason data. Here's every team that crushed (had > 17.5 net rating). Not exactly a rigorous analysis, but all had between 44-59 regular season wins. Also, Bulls currently the GOAT of preseason NBA history 💪

18

80

517

5

22

133

Kostya Medvedovsky

@kmedved

3 years

First cut of DARKO's 2022 season win projections. Use projected DARKO ratings for next season, combined with minutes projections to generate net ratings. Phoenix!

27

26

131

Kostya Medvedovsky

@kmedved

3 years

@NateSilver538 Somewhat silly, but outdoor mask wearing is less annoying in colder climates, since it has the benefit of keeping your face warm. Especially with cloth masks.

8

0

120

Kostya Medvedovsky

@kmedved

2 years

Happy to say DARKO's win projections were the best performing "player-based" win projection models tracked this year, going 21-8-1 vs. Vegas. They would have also won the ABPR contest. The highest performing model overall was the DARKO preseason-only model as well.

15

13

117

Kostya Medvedovsky

@kmedved

7 months

Here are DARKO win projections for the 2023-24 season. These include both the preseason adjustments I tweeted last night, along with updated adjustments which take into account which players played in preseason (and how many minutes).

37

16

113

Kostya Medvedovsky

@kmedved

2 years

@NateSilver538 To some degree, it wasn't a single policy choice. Rather, it was a rolling series of "keep schools closed for another week" decisions. The result was disastrous, but each incremental decision was rationalized as "well, we were closed last week, and nothing has improved".

7

3

111

Kostya Medvedovsky

@kmedved

4 years

@Noahpinion The universal theme of these types of stories is always '$400K/year isn't that much money once you spend it all.'

5

3

104

Kostya Medvedovsky

@kmedved

4 years

Tatum is leaping.

2

20

108

Kostya Medvedovsky

@kmedved

5 years

Inspired by DeAndre Jordan’s apparent breakout as a free throw shooter last year, I've been working on a project for a bit to project NBA box score stats, and with the season approaching, I figured it was time to tease that. (1/x)

1

13

109

Kostya Medvedovsky

@kmedved

3 years

DARKO projections from the end 2016 for who the best players (age 23.5 and under) would be today. This was Jokic's rookie year - anyone who didn't notice him until later just wasn't paying attention.

7

17

105

Kostya Medvedovsky

@kmedved

4 years

I wrote about the 'padding approach', and NBA box-score stabilization rates here: . This is mostly a recap of work I've done previously, but I wanted to have it somewhere easier to find than some twitter screenshots. ( @knarsu3 ).

10

13

106

Kostya Medvedovsky

@kmedved

3 years

@MelancholyYuga @AlanMCole Reminds me of 'Parachute use to prevent death and major trauma when jumping from aircraft: randomized controlled trial'.

2

17

101

Kostya Medvedovsky

@kmedved

2 years

@busteddealbot lol.

1

11

102

Kostya Medvedovsky

@kmedved

8 months

Presented without comment.

9

17

99

Kostya Medvedovsky

@kmedved

3 years

Current DARKO Daily Plus Minus values. (These will be up on the Shiny App soon).

DARKO - Daily Adjusted and Regressed Kalman Optimized projections

Current DPM Ranks NBA ID,Player Name,Position,Age,DPM,Offensive DPM,Defensive DPM,Box Only O-DPM,Box Only D-DPM,On Off O-DPM,On Off D-DPM 203999,Nikola Jokic,c_pos,29.3,7.01,4.8,2.22,4.36,0.19,4.89...

docs.google.com

15

98

Kostya Medvedovsky

@kmedved

2 years

DARKO. Tracking data integration. Happening soon. Fastest and slowest players in the NBA.

9

92

Kostya Medvedovsky

@kmedved

3 years

@pescami @ADavidJohnson You have confused 'merit' with 'equality of opportunity' here. Very few people could play in the NBA regardless of effort because of the natural gifts required, but NBA players are very clearly the best in the world at their jobs. Sports in general are incredibly meritocratic.

1

3

91

Kostya Medvedovsky

@kmedved

2 years

@busteddealbot @matt_levine This is like, a totally reasonable, good faith question one would ask during the pre-deal signing, due diligence period.

2

1

96

Kostya Medvedovsky

@kmedved

3 years

1

3

89

Kostya Medvedovsky

@kmedved

3 years

Same list for today, projecting for 2026.

24

12

87

Kostya Medvedovsky

@kmedved

11 months

Sure, why not.

Shams Charania

@ShamsCharania

11 months

Sources: Celtics, Clippers, Wizards in serious talks on deal involving Kristaps Porzingis to Boston, Malcolm Brogdon to L.A., Marcus Morris and draft compensation to Washington. More details at @TheAthletic with @JaredWeissNBA :

376

691

7K

2

13

87

Kostya Medvedovsky

@kmedved

2 years

NFL Preseason Results Are Predictive: a Thread I have previously found preseason are fairly predictive of NBA regular season results (). Others have found similar results for baseball (CC @DanRosenheck ). I was wondering the same re: NFL. (1/x)

Kostya Medvedovsky

@kmedved

3 years

I've looked at the predictive power of preseason results before, and there's real value here. I'll do full writeup/tweet thread on the methodology, but a +32.5 preseason net rating suggests we should bump our expectation for the Bulls' regular season ratings by ~1.5 points/100.

5

22

133

2

19

84

Kostya Medvedovsky

@kmedved

3 months

No joke - the only comp I've seen in DARKO is Shaq, and that's a data availability error, where Shaq was already an MVP-level player in the first available DARKO season, but had debuted years before.

NBA University

@NBA_University

3 months

Wemby’s defensive trajectory (blue line 🚀) next to some relevant comps… Something special brewing in San Antonio

30

82

2K

2

4

79

Kostya Medvedovsky

@kmedved

2 years

@NateSilver538 The Boston wastewater provides a good comparison datapoint: As the NYT chart shows, confirmed cases have yet to peak in Boston, but the wastewater data suggests actual cases are *far* higher this time around than ever before.

Joseph Allen

@j_g_allen

2 years

Boston wastewater thru 1/6 (🧵) -the "hint" from 1/3 that it was turning is no longer a hint -levels hit their peak (right on cue based on SA data) -expecting to see a steep drop rather than Dec-Feb plateau like last winter

12

31

157

6

78

Kostya Medvedovsky

@kmedved

4 years

This seems to be an unfortunate side effect of how late we started our testing regime. If we had large-scale testing capability earlier, we would be able to capture *some* information about the growth in cases, but this suggests we're just miles and miles behind this.

2

10

78

Kostya Medvedovsky

@kmedved

3 years

This roughly summarizes NBA twitter's relationship with pbpstats.

Rufus

@RufusPeabody

3 years

I think it's somewhere in Canada but still ( @denniserny )

3

4

42

2

5

75

Kostya Medvedovsky

@kmedved

1 year

This is real. This is happening.

Cory Jez

@JezData

1 year

Damian Lillard is on the verge of breaking the @trailblazers franchise scoring record tonight. But how does his IMPACT stack up among the best scorers in the league? We broke it down using @kmedved and @anpatt7 's DARKO metric

4

21

116

2

78

Kostya Medvedovsky

@kmedved

2 years

538 grades the Celtics as the best team in the NBA (regular season). Only the 2nd best team for the playoffs however. :-(

2

9

72

Kostya Medvedovsky

@kmedved

2 years

Final DARKO win projections below, and updated on the spreadsheet here: A few notes on methodology, focusing on recent updates (which are material). 1/X

6

13

76

Kostya Medvedovsky

@kmedved

4 years

@NateSilver538 It's not just losing seats - you're right nobody owns those. It's the general bloodbath in the House however. Dems lost 27 of 27 tossup seats on Cook's for example. That sort of loss really is worth interrogating, even if it's mostly a failure of expectations.

2

1

71

Kostya Medvedovsky

@kmedved

1 year

If you're in NYC, come see me speak about NBA statistics and DARKO tomorrow!

Advanced Basketball Analytics with DARKO

Learn from one of the world's leading basketball analysts about how he built one of the world's most regarded all-in-one basketball metrics.

www.eventbrite.com

6

13

73

Kostya Medvedovsky

@kmedved

2 years

Alperen Sengun is the best passing big under age 20 (measured by assists per 100) in DARKO's history (1998+). Also, Darko!

6

5

71

Kostya Medvedovsky

@kmedved

2 years

@mattyglesias Very on point targeted advertising with this article. Someone selling their key-words well.

1

2

65

Kostya Medvedovsky

@kmedved

3 years

Excited to be partnering with @TheAthletic to provide DARKO NBA game projections for these playoffs. Thanks to @SethPartnow @nandodifino and @AndrewRDeWitt for making this happen.

Nando Di Fino

@nandodifino

3 years

We have some fun, awesome stuff coming on the betting side over the next few weeks. And not the typical "you have to bet to understand this" bullshit. It's wickedly smart and accessible. Kicking things off -- the DARKO projection system, from @kmedved

0

8

18

10

12

68

Kostya Medvedovsky

@kmedved

20 days

Testing using Gemini 1.5 to check NBA foul calls, e.g., the Brunson lost ball turnover ()). Sorta concludes Brunson flopped? Pretty great!

3

6

67

Kostya Medvedovsky

@kmedved

2 years

I'll do a fuller analysis when I get a chance, but I just want to say that people may be sleeping somewhat on how impressive (and predictive going forward) the Cavs early season start is. A +12.33 SRS in 8 games is no joke. They are very live to come out of the East.

7

6

67

Kostya Medvedovsky

@kmedved

8 months

@NateSilver538 The core issue here seems to be the statement that "I don't care about Republican vs. Democrat", which is a fine position to have, but sort of fundamentally refuses to engage with the topic of "how did red and blue state perform during covid."

3

64

Kostya Medvedovsky

@kmedved

1 year

I believe Miami may be the first team to be outscored in the regular season to make the NBA finals. The DARKO data only goes back to 1997, so I'm eyeballing earlier and could have missed someone? Is Miami the worst NBA finals team ever?

3

7

62

Kostya Medvedovsky

@kmedved

2 years

@NateSilver538 The failure was being unable to acknowledge midway through that the school closures weren't really doing anything material for the public health situation, and that there was no imminent "light at the end of the tunnel" (i.e., "just one more week...").

3

1

61

Kostya Medvedovsky

@kmedved

2 years

Discussing with some NBA people a point raised by @minakimes , that in the NFL, the "nerd" play is more aggressive than the status quo, which does not fit the usual nerd stereotype. It occurred to me, this is somewhat generalizable to other sports. (1/x)

Cameron Soran

@cameronsoran

2 years

We could fix all football analytics in an instant, just like we do in finance, law, medicine, etc.: 'Make the decision, but now you get fired for being wrong.'

44

3

31

10

12

62

Kostya Medvedovsky

@kmedved

4 years

@mattyglesias Or really any other legal issue for that matter. They mostly spend their professional time playing pinochle I think.

0

60

Kostya Medvedovsky

@kmedved

6 years

NBA thread. Did some work on how to project team-level metrics based on year-to-date results. This is related to the work by @knarsu3 on stabilization, (), but slightly different in that it tells you how to *project* these team stats going forward. (1/x)

3

9

57

Kostya Medvedovsky

@kmedved

3 years

Work on DARKO career projections, which includes a survivorship model. Current model result for longest and shortest careers remaining below. (CC @bbstats ).

7

5

58

Kostya Medvedovsky

@kmedved

4 years

Did some work on 3PT% stabilization using the @tangotiger 'padding' method and found a much faster 'stabilization' point for 3PT% than prior estimates. For a season, you need about ~240 attempts before your YTD 3PT% is more predictive than league average of your future 3PT%. 1/n

3

7

57

Kostya Medvedovsky

@kmedved

4 years

@NateSilver538 There's a lot going on here, but I suspect it largely resembles NFL coach conservatism on 4th down. The downside of being overly cautious are disproportionately small relative to the benefits of being viewed as overly optimistic.

3

58

Kostya Medvedovsky

@kmedved

5 months

Appreciate the shoutout to DARKO by @kpelton here on the Lowe Post with @ZachLowe_NBA Around the 24 minute mark.

4

57

Kostya Medvedovsky

@kmedved

1 year

@NateSilver538 @Redistrict Yeah, I was serious. We're clearly at the "saving face" point right now. I was joking yesterday he'd say this was all performance art, but this is cleaner.

3

1

54

Kostya Medvedovsky

@kmedved

4 years

With @anpatt7 's help and inspired by @bbstats and @SethPartnow , I've been working on an "upside" projection model for DARKO. This is *very* early stage, but is too fun not to share. Here's the a first cut at the best prospects age 24-and under.

13

15

56

Kostya Medvedovsky

@kmedved

4 years

Using just the official calls (including AZ) with the 538 "pick your own battleground states" tool, Biden is at 96% to win the election right now, despite the seemingly horrific start to the evening for him. Actually well *ahead* of where he started this evening. @NateSilver538 .

6

4

57

Kostya Medvedovsky

@kmedved

3 years

Good points below. Here's a chart of the best active players in the NBA, ranked by career wins per minute. Will track closely going forward. Very interesting names on this list. New moneyball strategy?

7

5

55

Kostya Medvedovsky

@kmedved

3 years

Working on a survivorship model for the the lengths of various players' NBA careers. Early cuts...

7

5

53

Kostya Medvedovsky

@kmedved

2 years

223,000,001 people who can guard Michael Beasley. DARKO has only identified about 500 of them. Lotta work left to do.

Michael Beasley

@Michael8easley

2 years

7753000000 people on earth…7529999999 of em can’t guard me 😂😂😂😂 #wethemones

433

2K

21K

0

5

54

Kostya Medvedovsky

@kmedved

4 years

@DanRosenheck @gelliottmorris @RileyMcAtee I don't understand the point of debating the 10% vs. 25% chance of a Trump win, etc... *Biden* is going to win, not Trump. The current likely outcomes (~50 to 100 point Dem EV win) is very much in the fat part of both models' curves, not some tail outcome.

3

6

54

Kostya Medvedovsky

@kmedved

1 year

Presented without comment.

5

52

Kostya Medvedovsky

@kmedved

2 years

@benedictevans I hear this is the year of Linux on the PC.

7

3

50

Kostya Medvedovsky

@kmedved

3 years

@juanpablo__u @KDTrey5 When I retire from NBA analytics, the last thing I will do will be to write a @skepticalsports /Dennis Rodman-style magnum opus about how Durant is the most overrated player of our lifetimes.

4

5

46

Kostya Medvedovsky

@kmedved

4 years

@NateSilver538 Sometimes this is the funniest timeline.

1

47

Kostya Medvedovsky

@kmedved

3 years

DARKO's estimate of each team's strength of schedule faced to date. This uses player-level DPM ratings by date, so thus attempts to account for in-season improvements. Will be adding these to soon. CC @NateDuncanNBA @DannyLeroux (mentioned on the pod)

8

10

50

Kostya Medvedovsky

@kmedved

8 years

Here's a very (very) early wins projection for the 2017 season based on the current @Rotoworld_BK depth charts.

24

36

47

Kostya Medvedovsky

@kmedved

1 year

@TAizenberg You’re right. This has been the con the whole time. Spend years building a stat that will hate a player the Lakers don’t have, just in case they later trade for him.

3

5

47

Kostya Medvedovsky

@kmedved

3 years

I'll do a full writeup sometime (maybe on the blog), but for now, here are the biggest winners and losers from preseason based on the SRS-based padding approach I discussed a few nights ago. The Bulls remain the big winner here, with a +28 SRS in 4 games. CC @NBACouchside

4

9

48

Kostya Medvedovsky

@kmedved

3 years

A similar effect is seen in his longevity prediction, with him losing close to a year off his peak. If anyone has found a good list of players who have actually had covid, let me know. I'd be interested in seeing if there's something systemic here.

5

3

47

Kostya Medvedovsky

@kmedved

1 year

Best use case of ChatGPT is so far is to tell stories for my kids: "Make me a story where Elsa meets Cinderella, and they have an adventure together." And then they can make extra requests, like "have them meet Olaf!", etc... It works really well and they love it.

1

2

48

Kostya Medvedovsky

@kmedved

4 years

@NateSilver538 You can always undo it later.

2

1

45

Kostya Medvedovsky

@kmedved

4 years

@ForecasterEnten Man, who are the 16%. Do they all work at Zoom?

3

0

47

Kostya Medvedovsky

@kmedved

3 years

A brief thread on the new DPM 2.0 values on the DARKO site (Shiny App re-launching soon). DPM is a predictive all-in-one metric, on the same scale as RAPM. It is trained on every game in a player's career, weighting more recent games more heavily to maximize prediction. 1/n

1

8

45

Kostya Medvedovsky

@kmedved

3 years

Those who remember the lessons of history are doomed to overfit to it.

2

3

47

Kostya Medvedovsky

@kmedved

7 months

The DARKO web app () may be outgrowing our ability to improve it. If anyone has experience with Python-based data viz frameworks and is interested in helping to upgrade the DARKO site, please DM me. While DARKO is a hobby project, I have some budget

2

9

46

Kostya Medvedovsky

@kmedved

5 years

Looked at turnovers/100 in the NBA since 2005 over time, using the Facebook Prophet package in python. Black dots are observations. The blue line is the forecast. An interesting trend emerges: turnovers start very high, and rapidly decline as the season goes on... (1/x)

5

9

46

Kostya Medvedovsky

@kmedved

3 years

Does PJ Tucker have the largest gap between where NBA/analytics twitter views a guy, and where the various all-in-one metrics are? DARKO, EPM, and LEBRON all view him as somewhere between RL and one of the worst players in the league.

13

4

45

Kostya Medvedovsky

@kmedved

2 years

Since 1998, home teams have won 51.5% of opening tip offs. Why? Not a riddle - genuinely confused why: 1) Teams try to win the opening tip. 2) Why home teams would succeed somewhat materially more often.

14

5

46

Kostya Medvedovsky

@kmedved

6 months

Good thread here. In line with this analysis by @sports_mediocre - DARKO's time decay RAPM metric has Luka 90th.

DARKO - Daily Adjusted and Regressed Kalman Optimized projections

Current DPM Ranks NBA ID,Player Name,Position,Age,DPM,Offensive DPM,Defensive DPM,Box Only O-DPM,Box Only D-DPM,On Off O-DPM,On Off D-DPM 203999,Nikola Jokic,c_pos,29.2,7,4.78,2.22,4.27,0.18,4.89,2...

docs.google.com

NBA Stats

@sports_mediocre

6 months

1/ Luka Doncic is the most OVERRATED player in the NBA This thread will get in the weeds. If you’re a “yay points” person, it’s not for you. I’ll cover: “Luka has no help” Stat padding Why advanced all-in-one metrics continue to misjudge him

240

120

834

7

44

Kostya Medvedovsky

@kmedved

3 years

Just an FYI for those asking - yes DARKO will be back for the 2021 season. I'm reworking all my data sources to use more granular data, which requires refitting the model. Hoping to have the update done this weekend.

4

2

45

Kostya Medvedovsky

@kmedved

2 years

Minor personal (non-NBA) news... I am rather suddenly moving back to NYC to do merger arbitrage (analyst role, non-legal job). Upside: No more recording time in six minute increments or CLE requirements. Downside: nyc housing costs and commute...

2

1

44

Kostya Medvedovsky

@kmedved

5 years

I used the free agency results from the Mock Offseason by @NateDuncanNBA @DannyLeroux @kpelton and @DanFeldmanNBA to generate Mock projected win totals. Uses xPIPM as the player metric. ( @JacobEGoldstein ). I filled out the rotation and MPG myself, so errors abound I'm sure.

Nate Duncan

@NateDuncanNBA

5 years

2019 Mock Offseason Results for subscribers: Subscribe here:

2

0

7

13

7

43

Kostya Medvedovsky

@kmedved

2 years

Good for Ainge getting back in the game. Say what you will, but the man did create the first dynasty to only win one championship.

3

1

42

Kostya Medvedovsky

@kmedved

2 years

@EmilDimanchev @DegenRolf @LenaDevlen Do you mean this?

Thinking Fast and Slow – Replicability-Index

replicationindex.com

2

6

42

Kostya Medvedovsky

@kmedved

1 year

A few DARKO updates: 1) I will be moving the main DARKO projections on the app from showing the next day's projections to showing a "rest of season" projection. The idea is this will align better with what most people are thinking of in terms "X's three point projection is 38%"

1

42

Kostya Medvedovsky

@kmedved

2 years

Daily Sengun propaganda: Highest box-only DPMs in the DARKO era for players under age 20.

2

1

42

Kostya Medvedovsky

@kmedved

3 years

@asymmetricinfo This is a major epistemological question generally. We're taught 'appeals to authority' are a logical fallacy, but for 99% of technical issues, our understanding of an issue relies on some assessment of which experts to trust. That's true even on uncontroversial topics.

4

6

39

Kostya Medvedovsky

@kmedved

3 years

@CrumpledJumper Amazing viz, but I'm gonna flag this for a bit of Y-axis abuse. Size of the nails makes this look like eFG in game 7 is dramatically lower, when we're actually talking about ~2% lower than game 6.

1

0

42

Kostya Medvedovsky

@kmedved

5 years

The rise of the passing big man (from @bball_ref 's Play-Index).

4

9

41

Kostya Medvedovsky

@kmedved

2 years

Question asked in an NBA chat I'm in about how many possessions are needed to trust 5-man lineup data (CC @jessepittsley ). I ran padding values () on 5-man lineups since 1997-98 by game to see how much padding was needed to predict the next game. (1/x)

1

16

40

Kostya Medvedovsky

@kmedved

3 years

Coach Brad Stevens played him sparsely so GM Brad Stevens could sign him to a below market deal in the offseason.