Win Vector’s Dr. Nina Zumel shares how utility methods let you optimally operate an AI/ML classifier model. Please reach out to us at Win Vector LLC for consulting, advisory work, presentations, or training: [email protected] .
Introduction Current computer chess engines include a neat feature: the analysis bar. For example here is the analysis bar from chess.com. White to move, FEN: rnb2rk1/ppp2ppp/3p1q1n/4p3/7P/b1PPP3/PP1BBPP1/R2QK1NR w KQ – 0 1 The analysis bar is on the left, and the white portion is 74 “centipawn units” below the midpoint (matching […]
Introduction In our earlier article “How Many Chess Games are Possible?” we derived that the number of typical short chess games is around 10151. This turned out to be much larger than the “Shannon number” 10120 side-note from Shanon’s remarkable 1950 paper “Programming a Computer for Playing Chess” Philosophical Magazine, […]
Stockfish is a top chess playing engine. We got to wondering how many possible Stockfish chess games are there if Stockfish plays against itself? Could it be that there are only a few stereotypical games? Or are there a lot of games? To answer this we recorded Stockfish playing against […]
How Many Chess Games are Possible? Here is a fun question: how many different games of chess are possible? Counting the number of possible chess games is quite hard, as the numbers are large and chess board positions can be quite complicated. In this note we will try to estimate […]
Mixed populations make correctly measuring price elasticity (the response of sales to price changes) difficult. In fact, standard methods often mis-price. In our latest screencast we show Bayesian methods and Stan can actually “un-mix” such populations and correctly infer price elasticity. This allows for much better price experiment outcomes.
ACM not only hides publicly funded NSF research behind a firewall, even citing it is a premium feature.
Please check out Nina Zumel’s video on how to use Tobit-style methods (and Stan!) to model market demand from censored observations such as capped or limited sales.
I’ve been worrying a bit over the philosophy of creation and dissemination in the era of LLMs (large language models). I feel we are overly impressed with the byproduct or even the bycatch. LLMs do complete some tasks, write code, and produce superficially convincing texts. In my opinion, this is […]
Our group has written many times on how classification training prevalence affects model fitting. Tailored Models are Not The Same as Simple Corrections The Shift and Balance Fallacies Does Balancing Classes Improve Classifier Performance? The Intercept Fallacy Don’t Use Classification Rules for Classification Problems The upshot is: we feel re-balancing […]