Marek Kwiatkowski

I will be launching a Patreon very soon. I will have to stop compiling TWIC fairly soon unless I can find a way to make it pay. Maybe its time has come and gone. I have some good ideas for a Patreon community but maybe I left that too late too.

Come and join us smartodds.applytojob.com/apply/5kOjpxU3…

I should say also that I believe that this model underestimates Pragg significantly, for reasons that are not yet so clear to me. Personally I have him at least on a par with Sindarov.

This model is significantly stronger than Elo, but like Elo it overestimates Nakamura, who has avoided challenging himself in recent years. In the context of the Candidates, it likely underestimates experienced players with deep opening prep like Caruana and Giri.

Candidates Odds (♚ version): Nakamura 29.0% Caruana 20.3% Sindarov 16.6% Wei Yi 12.9% Giri 9.9% Praggnanandhaa 5.4% Esipenko 4.8% Bluebaum 1.2% The top five are favourites in a match vs Gukesh. See below for explanations.

1 - Dimitri Payet a créé le plus d’occasions dans les 5 grands championnats européens entre 2006 et 2023 (1 350), devant Lionel Messi (1 248). Il est également le seul à avoir atteint le doublé 100 buts/100 passes décisives en L1 sur cette période. Héritage. @dimpayet17

Statistical Rethinking 2026 is done: 20 new lectures emphasizing logical & critical statistical workflow, from basics of probability to causal inference to reliable computation to sensitivity. It's all free, made just for you. Lecture list & links: #calendar--topical-outline" target="_blank" rel="nofollow noopener">github.com/rmcelreath/sta…

It its current form the model overestimates washed-up streamers and underestimates serious professionals who actually play chess. Our team is aware of this issue and we are working to correct it.

It is with a heavy heart that I must announce that I am at it again.

📊 Too much global development data still lives in spreadsheets. But modern analysis requires data acquisition as code. Today I’m pleased to share that the unicefData R package is now available on CRAN, providing direct programmatic access to the UNICEF Data Warehouse. Researchers, policymakers, and analysts can now retrieve 700+ internationally comparable indicators on children and women directly from R — enabling reproducible analytics, transparent workflows, and faster evidence generation. Install install.packages("unicefData") Example library(unicefData) df <- unicefData( indicator = "CME_MRY0T4", countries = c("BRA","USA","IND"), year = "2015:2023" ) Why this matters Reliable evidence depends not only on good data, but also on how easily that data can be accessed, reproduced, and reused. The unicefData package helps bridge the gap between official statistics and modern data science workflows by enabling: • Access to 700+ indicators across countries and years • Integration with reproducible research pipelines • Use of SDMX international statistical standards • Consistent interfaces across R, Python, and Stata Making official statistics more accessible, reproducible, and interoperable is an important step toward strengthening evidence for policies that improve outcomes for children. 📦 CRAN cran.r-project.org/package=unicef… #OpenData #OfficialStatistics #RStats #DataForChildren #ReproducibleResearch

Possibly the best job in the business could be yours:

Our Sports Analytics Team is heading to #SSAC26! Join us for a hands-on workshop on Bayesian Spatial Modeling of #NHL goaltending. From shot difficulty to rink-wide scoring probability, see #PyMC in action! 👇dub.link/e0knYEl #SportsAnalytics #BayesianModeling #Hockey pic.twitter.com/2SddGDo42F

@octonion I love me a Poisson, but the bias identified by Dixon & Coles is real.

@statlurker You and the Poissons don't get along?

For a baseline soccer model in a top club league, which model do you prefer? Poisson Dixon-Coles Bivariate Poisson with covariance Diagonal-inflated Poisson Conway-Maxwell Other?

New open source: cuthbert 🐛 State space models with all the hotness: (temporally) parallelisable, JAX, Kalman, SMC

@octonion Love the curve on these.

My favorite keyboard is still my 1992 buckling spring IBM model M. Bolt-modded and still works great. I've just disassembled it to give it a good cleaning, which it obviously needs. I've also installed Unicomp APL keycaps.