Ryan's Repository of Random Reflections

This semester I took the machine learning class at UMass, and for my final project I developed models for predicting characteristics of pitches based on the context that they were thrown in.  The context consists of relevant and available information known before the pitch is thrown, such as the identities of the batter and pitcher, the batter height and stance, the pitcher handedness, the current count, and many others.  In particular, I looked into modeling the distribution over pitch types and locations.  This problem is challenging primarily because for a particular context (a specific setting of the different features that make up the context), there is usually very little data available for pitches thrown in that context.  In general, for each context feature we include, we have to split up our data into $k$, groups where $k$ is the number of values that the context feature can take on (for batter stance it is $k=2$, for count it is $k=12$, etc).  Thus, de...

In this blog post, I will introduce an idea I recently came up with to predict the most likely players to get a hit in a given game based on situational factors such as opposing starter, opposing team, ballpark, and so on. I have not written much in this blog on this topic, although I did some work on this topic last fall which you can find here . In that work, I had the chance to explore a bunch of ideas that I had, but ultimately had to back up a few steps and rethink my approach. I think the ideas are still valid, and will continue refine them as time permits. A few weeks ago, I came up with a new approach that is completely different from my other approaches so far, and I will share it in the rest of this post. Defining the Problem Before we dive into the math, let's talk about what exactly we are trying to do. The end goal is to pick the player who is most likely to get a hit on a given day based on factors associated to the games for that day. Some of these facto...

This semester in my mathematics capstone course, students had the chance to develop mathematical models to describe real world research problems.  I took this as an opportunity to research and develop probabilistic models that can be used to predict the outcome distribution of a baseball at bat.  My hope was that I could apply my findings from this research project to my side project of Beat the Streak.  I learned a lot about mathematical modeling in this class, and explored a variety of techniques for making predictions about baseball at bats.  At the end of the class, we wrote a report that introduces the models we came up with, and compares the quality of the predictions they produce.  I have made this report available here .

With the mid way point of the MLB season fast approaching, it is starting to become apparent that I need to finish my program if I want to have any shot at compiling a respectable streak this season. In this blog post, I will talk about one formula that I am relying heavily on in my tool. This formula is based on Bill James' Log5 formula but is modified for non-binary output. In general, the Log5 formula is used to approximate the probability that team A will defeat defeat team B given each of their individual winning percentages. It can be modified to handle batter vs. pitcher matchups as well. There is no mathematically rigorous derivation of the formula, so using it in my tool is a little bit iffy to say the least. However, I think using it will yield a better approximation to batter vs. pitcher matchups than not using it. In it's simplest form, the approximate batting average of batter B against pitcher P is given by: $$ AVG_{B v P} = \frac{\frac{AVG_B \cdot AVG_P}{AVG...

In 2001, MLB.com released the Beat the Streak challenge : A challenge to fans to essentially beat the all time best hitting streak established by Joe DiMaggio in 1941. In that season, Joe DiMaggio had a 56 game hitting streak. The longest hitting streak by any MLB player since that is 45 games. When the challenge was first introduced, fans were asked to pick a (possibly different) player every day who they expect to get a hit. If that player earned a hit, then their streak would increase by 1. Otherwise, it would go back to 0. The first fan to reach a streak of 57 would win the grand prize. Since it's introduction in 2001, the grand prize has grown from $100,000 to $5,600,000, and a number of new features have been added to improve the odds for the fans. Yet, no one has even broken the 50 game streak barrier, let alone win the grand prize. I have been a casual beat the streak player for a few seasons, but I never really took it too seriously. Last season, I decided to use my back...