About

The chat is an ongoing exploration into optimizing the question-and-answer relationship between users and LLMs. The goal is to create a chat experience that understands how to parse user requests, select the ideal data from a finite pool, provide prompt and useful insights, and do it all while minimizing compute and token needs. We hope the chat keeps improving until it is more convenient than clicking through the site itself, but as a work in progress, it may still be several iterations away from that.

The player grades are an attempt to improve on the Savant-style red bar/blue bar evaluations users often make at a glance. They group related stats into individual skills, such as power and contact, then distill those families into grades curved against league performance so it is clear where a player stands among peers. Read the player grades explainer for a deeper look at the grades, their logic, and how we intend to improve them based on our latest research.

Our deepest apologies. This site will always be a work in progress, whether we are improving background logic, pipeline efficiency, or frontend displays. If you see something amiss, please reach out on one of our socials and let us know. We appreciate your attention to detail and will add your concern to our backlog for investigation.

Absolutely. The stories page exists so the community can share interesting ideas in a thoughtful way. If you want to contribute, reach out on social media with a sample or the story itself and we will let you know whether it is a good fit. Two important caveats: since this site is a passion project, there is no economic incentive for contributing stories. We also believe stories should be about community interaction and the sharing of ideas, so we strongly discourage the use of AI in idea creation or writing and may reject stories we suspect used it.

Good catch. Some stats on our site are different from other sites, mostly by design. For example, xFIP is calculated using pitcher fly balls and the league HR/FB rate. Those inputs can vary because there is some subjectivity in what qualifies as a fly ball. Our definition may differ slightly from another site's, causing the final output to change as well. In cases like that, and in similar metrics such as xwOBA, we are confident the same sentiment, or sometimes a slightly better one, is being captured. If a metric looks flat-out wrong, please let us know. We are not immune to mistakes, and something may have slipped through the cracks.

A lot of our metrics are scaled to solve the specific problem in front of them. For example, a zero-baseline scoring system with very small increments worked better for daily matchup scores, while a 50-as-league-average baseline worked better for player trend comparisons. We agree this can be confusing and makes the site worse than it could be. We plan to normalize our scoring so users can move from one metric to another and clearly understand what is being said. For now, please bear with us.

If you have ideas, thoughts, or requests for the site, we would love to hear them. We are certain there are valuable features we have not thought of yet. We are limited by development time, cost, and data availability, but if a feature meets those requirements and fits the spirit of the site, we will add it to our backlog. This site is going to be an eternal work in progress, so hopefully we will all get to watch it grow and flourish over time.

We do not have a roadmap for the site yet. A lot of our current focus is on refining existing features and improving the backend infrastructure of the site, but once we start having more bandwidth for long-term focus on new features, a roadmap to keep the community informed will naturally follow.