Sign in

Kostas Pelechrinis

@kpelechrinis.bsky.social
305 followers 31 following 690 posts

Information science professor at Pitt | 🏀 analytics at Mavs | Author of Mathletics

PostsRepliesMedia
Kostas Pelechrinis @kpelechrinis.bsky.social · 23/09/2026
Wizard of Boz is the only functioning part of the Steelers offense the last few years. Is he declining? Ofc I did what every fan would: build a Bayesian aging curve (well surface) for kickers. Jury is still out if he’s past his peak but he’s still elite! bayesiana.substack.com/p/growing-ol...
bayesiana.substack.com
Growing old in a young’s man game
A Bayesian look at athletic aging, and whether the Wizard of Boz will decline
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 18/09/2026
And here is a writeup of the methods and other stuff 412sportsanalytics.wordpress.com/2026/09/18/t...
412sportsanalytics.wordpress.com
The Spacing Connectome: What can brain science help us learn about basketball?
This one has been long time in the making in terms of being in my mind for years but at the same time not having the appropriate data (or time) to do something with it. But thanks to SkillCorner th…
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 16/09/2026
Thanks to @skillcorner.bsky.social finally we have some tracking data for European hoops! Only 10 games but enough to look at a fun (imo) idea I always had: thinking of spacing through ideas from functional connectivity of brain regions. So I tried out a few things: sites.pitt.edu/kpele/the_sp...
110
Kostas Pelechrinis @kpelechrinis.bsky.social · 12/09/2026
Known as the "Windrip" divident since .... 1935. Sometimes the difference between fiction and reality is time .... a long time
001
Kostas Pelechrinis @kpelechrinis.bsky.social · 10/09/2026
If you are still not sure about RCV, consider this some (additional) evidence that it can protect us from polarizing candidates. Not necessarily ideologically extreme ones but candidates who generate both intense support and intense rejection bayesiana.substack.com/p/ranked-cho...
bayesiana.substack.com
Ranked-Choice Voting: Against the Tyranny of the Plurality
How RCV can overturn mild pluralities of polarizing candidates in favor of more broadly acceptable candidates
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 06/09/2026
With UCL League Phase starting in 2 days I wrote something about leverage, upside and downside risk for each game - with a focus on teams that are projected to be in the bubble of qualifying for the knockout stage (and of course a lot of AEK...) bayesiana.substack.com/p/the-game-y...
bayesiana.substack.com
Information Matters Relative to What You Expected
A Bayesian interpretation of fixture importance through the lens of the Champions League
000
Reposted by Kostas Pelechrinis
Adam Kinzinger @adamkinzinger.substack.com · 04/09/2026
The VP wants nothing more than a promotion in 2028. Thursday showed again why he doesn’t deserve it. Upon arriving at the briefing room, it took him only 40 minutes to contradict his boss, insult the American military, and ask the country to feel grateful about gas prices going up. (Link below)
1127100533153
Kostas Pelechrinis @kpelechrinis.bsky.social · 02/09/2026
😭😂😅
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 31/08/2026
20 years ago today I left everything I ever knew behind me & came to California to follow my passion for studying further & doing research in CS. There is not a moment I don't miss Greece but if I had the choice I'd do it over again, even if today my "2006 self" is vilified by a vocal minority.
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 30/08/2026
Power laws do not care about your feelings. Which is also why schools should teach actual math instead of fighting over hanging the Ten Commandments on the wall.
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 28/08/2026
Pythia (Πυθία) was Delphi's oracle delivering max ambiguous prophecies. In 🇬🇷 “possible” is πιθανόν. Phonetically they sound as if they share a root, but they don't. Yet, LLMs act like Pythia: when fed a linguistic probability descriptor & asked to assign a number, πιθανόν exhibits max variance [1/2]
100
Kostas Pelechrinis @kpelechrinis.bsky.social · 28/08/2026
Someone is trolling hard in academia. But there is a best case scenario, where we finally move away from using "volume stats" to judge productivity. Sports did it. Time for academia to do it too statmodeling.stat.columbia.edu/2026/08/27/2...
statmodeling.stat.columbia.edu
This University of Chicago business school professor has authored 258 academic papers in 2026 (so far). | Statistical Modeling, Causal Inference, and Social Science
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 25/08/2026
Very few clubs embody better the fact that sports teams in Europe are not "franchises" with the strict meaning of the term but are clubs, embedded in the communities they were created binging with them history beyond sports youtube.com/shorts/msOxw...
youtube.com
A Football Club Built to Remember a Lost Homeland #pitchandculture #football#athens #facts#history
YouTube video by PitchAndCulture
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 22/08/2026
According to some, on Monday I get back to indoctrinating 20-year-olds on partial pooling, redistribution of probability mass, shrinkage towards the mean and the Bayesian mindset.
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 19/08/2026
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 15/08/2026
Broadcasters and sports journalists should stop describing basketball as a “chess match”. Chess is a complete info game and deterministic. There is no bishop capturing the rook 37% of the time. Hoops are poker. Stochastic, dynamic with imperfect-information game of repeated best responses. Iykyk 👇🏻
020
Kostas Pelechrinis @kpelechrinis.bsky.social · 15/08/2026
And here are some more detailed thoughts about lotteries in research funding decisions. The TL;DR version is that they should absolutely be part of the equation and also that we should elicit slightly different information from the reviewers bayesiana.substack.com/p/the-scienc...
bayesiana.substack.com
The Science Draft
Using Uncertainty, Utility, and a Little Luck to Fund Better Science
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 14/08/2026
This isn’t more prevalent bc we dislike admitting how much luck determines outcomes&that even if we've enough info to say a proposal is good, we don’t have enough to distinguish them. We oppose lotteries bc they make randomness that’s already there impossible to ignore www.nature.com/articles/d41...
nature.com
When the best decision is no decision: the rise of randomization in grant funding
If reviewers can’t reliably tell the difference between top grant proposals, why should they pretend they can? A new wave of funders is turning to lotteries to break the tie.
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 13/08/2026
Happy International #lefthandersday! Meanwhile, in 1910s Twitter @ CommonSense1912: “Schools now allow children to WRITE WITH THEIR LEFT HAND bc apparently we must accommodate everyone.” @ SaveOurSchools1910: “Only 3% of adults are left-handed. Now suddenly all these children claim to be one? 🤔”
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 10/08/2026
One thing students often struggle with as we get deeper into Bayesian statistics is that uncertainty is more than just sampling uncertainty. So I wrote something about the different kinds of uncertainty and why the distinction matters bayesiana.substack.com/p/what-3-doe...
bayesiana.substack.com
What ±3% Doesn't Tell You
Knowing something precisely and being able to predict what it will produce are not the same thing.
031
Kostas Pelechrinis @kpelechrinis.bsky.social · 09/08/2026
RIP Don Nelson. We usually throw around the word pioneer too loose today but HE WAS one. Way before “analytics” appeared he was changing basketball; positionless, small ball, point forward, stretch big, space and pace. His impact will endure forever
032
Kostas Pelechrinis @kpelechrinis.bsky.social · 09/08/2026
I read a paper today on using AI in the peer review process. It will eventually happen but I put together some thoughts on the model introduced at the paper and my thoughts in general about peer review medium.com/@kpelechrini...
medium.com
AI Can Help Peer Review But Not Until We Fix What Was Broken Before AI Showed Up
I saw this paper today: “Fighting Fire with Fire: Infusing AI into Peer Review to Sustain Quality Scholarship” at Management Science, and…
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 08/08/2026
I used Enes Freedom's publicity stunt to showcase how three ideas from intro stats quietly explain how a lot of propaganda works. It is not necessarily about lying, but the illusion of frequency: make something rare feel common by controlling the sample medium.com/@kpelechrini...
medium.com
Creating the Problem, Then Winning the Argument: Enes Freedom, Iran, and the Politics of…
When You Have to Create the Problem to Prove It Exists
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 03/08/2026
It is mind boggling seeing people still claiming vax didn’t work, or measures like masks and social distancing infringed their “freedom”. Man we are beyond cooked. We are in the oven 4 hours after cooking was done
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 03/08/2026
Since midterms are coming and we are back at the “we were scammed with the jabs”, I think this points to another example of why probabilistic thinking is a must. Fauci et al. completely botched the communication around the vaccines. And I think this might have been deliberately on their side to
100
Kostas Pelechrinis @kpelechrinis.bsky.social · 31/07/2026
You used to know elections were approaching when cities suddenly became massive construction sites (ofc only for the work to stop once the election was over). Now a few minutes on social media is enough to see the politics of fear in full swing and know elections are coming
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 30/07/2026
This is a pretty interesting experiment. It illustrates why human expertise remains important in causal data science. Without informed guidance an LLM may/will rely on prior/recalled knowledge rather than reasoning from the data. erdogant.substack.com/p/can-llms-d...
erdogant.substack.com
Can LLMs Discover Cause and Effect? A Benchmark Against Bayesian Models and the Problem of Knowledge Leakage.
LLMs may generate convincing answers, but how trustworthy are the results? Hidden knowledge acquired during pre-training can influence conclusions without the user realizing it.
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 23/07/2026
With the usual suspects like Gaines and Travis parading Cunningham's comments, it's worth putting a few things in perspective for those that like to put their brains to thinking rationally: 1. This whole thing has nothing to do with women's sports and supporting women. For god's shake Travis [1/5]
100
Kostas Pelechrinis @kpelechrinis.bsky.social · 18/07/2026
Argentina is known for its approachable yet age-worthy wines. Clos de Rosario Vintage 1987 is their best 🐐 🍷 Enjoy a glass with all publicly available #Messi goals (thanks to StatsBomb) You can also interact with the visualization here: sites.pitt.edu/kpele/messi_...
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 18/07/2026
All (the ones publicly available from Statsbomb) of #Messi 's career goals in an interactive viz. If you have all the goals, you can upload a json file with the data (format described and file is not stored anywhere on the server) and see the new viz. sites.pitt.edu/~kpele/messi...
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 16/07/2026
ok everyone is talking about Messi, but after the 2018 world cup I was invited and gave 3 seminars on sports and football analytics in Argentina. Since then they have won everything and playing again at the final (which I hope they win also). Coincidence? Well, yes 😂 But had the 6-7 meme well ahead😅
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 11/07/2026
This football myth has beeb busted before afaik in many different ways but here is one with Bayesian hierarchical modeling 412sportsanalytics.wordpress.com/2026/07/11/b...
412sportsanalytics.wordpress.com
Big Man, Small Edge: What the Data Actually Say About Tall Goalkeepers and Penalties
Every pundit has said it at some point during a shootout: “you want a big goalkeeper back there”, “height fills the frame”, “height makes the target look smaller&#8221…
020
Kostas Pelechrinis @kpelechrinis.bsky.social · 07/07/2026
1987 Clos de Rosario 🐐🍷
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 06/07/2026
RBFA should do the funniest thing ever: challenge birthright citizenship in the US. It will short circuit the whole WH 😂 /s
010
Kostas Pelechrinis @kpelechrinis.bsky.social · 29/06/2026
Well since I am not travelling to Greek islands again this summer, how else would I feed my crave but with an analysis including them. And it turns out that nature is the worst gerrymanderer medium.com/@kpelechrini...
medium.com
The Gerrymander Archipelago
I applied the same math US courts use to detect gerrymandering to the Aegean and Ionian. The results are alarming. The culprit has been…
100
Kostas Pelechrinis @kpelechrinis.bsky.social · 26/06/2026
It seems that my suggestion for solving the STEM pipeline issue might have to be the standard for ALL students if this trend keeps up. Education is in a really sad state... medium.com/@kpelechrini...
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 25/06/2026
Strange. I thought spreadsheets were supposed to be a coping mechanism for people who can’t play. www.wsj.com/sports/socce...
wsj.com
The Harvard Data Nerd Defending America’s Goal at the World Cup
Matt Freese, who holds an Ivy League economics degree, came out of practically nowhere to secure the job as the U.S. national team’s starting goalkeeper—and his analytical approach has been integral t...
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 17/06/2026
Messi is aging like a 1987 Clos de Rosario 🐐🍷
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 15/06/2026
Galis set some unbreakable records during that (& following) runs. The only way to *try* stopping Galis was “box and 1” BUT the box was for Galis and the 1 for his teammates. This is how great he was. On ball, off ball, transition you name it he did it. He is *basketball* for us Greeks (& Europeans)
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 15/06/2026
39 years ago today Greece won its first European championship in basketball. A championship that boosted Greek hoops to unimaginable heights the following decades. A true David against Goliath moment. Greece beat Yugoslavia twice during the tournament and USSR in the final. youtu.be/rDHWsgOKnPU?...
youtu.be
Hellas - USSR 103-101 Final (Eurobasket) 14/6/87 EΡΤ (16:9 HD)
YouTube video by Aris History Channel
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 14/06/2026
Big part of the NBA finals story are the Spurs double-digit leads blown leads. But this has become more frequent occurrence in modern NBA. And there are reasons for it medium.com/@kpelechrini...
medium.com
No Lead Is Safe Anymore
Despite claims that the modern NBA is repetitive and predictable, first-half leads have become larger, more extreme, and more likely to…
021
Kostas Pelechrinis @kpelechrinis.bsky.social · 12/06/2026
Got to love seeing Mike Brown on the biggest stage explicitly acknowledging the role of luck! #NBA #NBAFinals
1283
Kostas Pelechrinis @kpelechrinis.bsky.social · 08/06/2026
If you want to learn more about the skill paradox and in general the role of luck and the interaction with skill in several environments I cannot recommend enough "The success equation" by Michael Mauboussin
030
Kostas Pelechrinis @kpelechrinis.bsky.social · 08/06/2026
With the 8th different #NBA champion coming up in 8 years (& multiple different teams w deep playoff runs every one of these seasons), this era shapes up to be a great example of the skill paradox: as the aggregate level of skill improves, the difference between he best & the average narrows. [1/3]
131
Kostas Pelechrinis @kpelechrinis.bsky.social · 31/05/2026
We tend to give skill much more weight when attributing success. The uncomfortable reality is that luck plays just as big of a role (and sometimes bigger) 412sportsanalytics.wordpress.com/2026/05/31/t...
412sportsanalytics.wordpress.com
The Spurs, Lottery Luck, and Winning the Right Lotteries
When people talk about NBA lottery luck, they usually focus on one thing: how many spots a team moved up. But that treats all lottery wins as equal. They aren’t. Winning the No. 1 pick in a f…
120
Kostas Pelechrinis @kpelechrinis.bsky.social · 31/05/2026
media.tenor.com
a man stands behind a podium that says draft lottery on it
ALT: a man stands behind a podium that says draft lottery on it
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 27/05/2026
In a few months the Blazers' FO will be: Claude Smith, Gemini Johnson and Grok Jones.
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 26/05/2026
If you are interested in Bayesian analysis and the Bayesian framework in general (which you should imo) I've put together an open course on coursera. Check it out here: www.coursera.org/specializati... #Bayesian #MachineLearning #DataScience
coursera.org
Applied Bayesian Data Analysis
Offered by University of Pittsburgh. Master Bayesian Methods for Data Analysis. Apply Bayesian inference and probabilistic modeling to solve ... Enroll for free.
094
Kostas Pelechrinis @kpelechrinis.bsky.social · 19/05/2026
Something strange is happening. My reviewer 2's reports all of a sudden read like coming from an English major, with its em dashes and several "viable paths here"...very strange #Academia
000
Kostas Pelechrinis @kpelechrinis.bsky.social · 15/05/2026
Two textbook ways to punish defenses showing 2 on ballscreens by #AEKBC First out of short roll punishing weak side help with baseline cuts. If they take away the short roll with a tag, “relay” pass and a baseline cut will also do the job
010