Skip to the page

Dugout design system

Dugout answers baseball questions with a small set of components. The kind of answer a question needs picks the component, and the content adapts to the question.

Every component and the rule that picks between them are on these pages. Start with How answers are built.

An example

Asked how one player is hitting, Dugout answers with the season line.

How is Kyle Schwarber hitting this season?

Kyle Schwarber (PHI) was one of the better hitters in MLB in 2026.

Dugout looked up the hitting and defence of Kyle Schwarber (PHI) for the 2026 season.

Kyle Schwarber (PHI)Phillies, designated hitter
StatSchwarberMLB average
Better than average.861.718
Worse than average.235.244
46
1 fewer out than an average first basemanToo few chances to judge−1
Cost 1 run more than an average fielderToo few chances to judge−0.6

Defence is from the 2026 regular season.

The numbers come from the MLB Stats API and Baseball Savant.

Principles

The kind of answer picks the component

The same kind of answer always gets the same component, and its content changes with each question.

Numbers come from MLB

A language model reads the question, and code fetches every number from MLB. The AI never writes a number.

Colour means something

Team colours show whose numbers you’re reading. Dugout green marks the things you can tap.

It works for everyone

Every page meets WCAG 2.1 AA, works at 375px wide and in dark mode, and respects reduced motion.

Where the numbers come from

MLB Stats API

MLB’s official record of every game, including box scores, season totals, standings and postseason series. Dugout uses it for hitting, pitching, team numbers and results.

Baseball Savant

MLB’s public Statcast website. Statcast is the camera and radar tracking system in every MLB ballpark, and it measures things a box score can’t, such as how many outs a fielder saved. Dugout uses it for defence.