Win Vector LLC

A Simplified Mental Model of LLMs

Expected Log-Return as a Risk Control

Choose by Probability, not by Expectation

LLMs Will Benefit from Scratch Workspaces

Maybe Jacobian Conjecture Counterexamples are not Rare

Is it the box or the “yes” that does not fit?

Why You Are Not Getting Full Value From Your Analytics

Who’s the Best Batter? Estimating Probabilities from Unevenly Collected Data

Squeezing the Most Utility From Your Models

What is a Centipawn Advantage?

Exploring Chess Positions and Counts

There are a lot of Stockfish Chess Games

How Many Chess Games are Possible?

How Many Chess Games are Possible?

Uncovering Hidden Price Elasticity

Even citing NSF funded research is an ACM premium feature

Modeling Censored Data with Tobit

I am Tired of Praise of Byproduct

How Logistic Regression Changes with Prevalence

Text as a “Market for Lemons”

The Fermi Paradox Just From Rarity

A Fairly Hostile Subtext

High Dimensional Optimization Remains Hard

Escape the Over Tuning Trap

Reducing Optimization Risk through Portfolio Methods

Modeling Uncertainty For Better Outcomes

Reverse Chess

Toying with the Lambda Calculus

They did use a calculator!

How About Pi ~ 31/32?

Don’t Let a Data Leak Sink Your Project