About Deep Dugout
Deep Dugout started with a simple question: what would happen if an AI actually managed a baseball team?
Not the abstract "AI in sports" think piece. The real thing. Every pitching change. Every lineup decision. Every bullpen call in a high-leverage October moment. I wanted to build a simulation where an AI manager had to make the same decisions a real skipper faces — and then live with the consequences.
So I built one.
The Setup
Deep Dugout simulates the 2026 World Series between the Los Angeles Dodgers and the Seattle Mariners using real player stats, a full plate-appearance-level simulation engine, and AI managers powered by Claude, Anthropic's AI model.
Why the Dodgers and the Mariners? When the project started in late February 2026, they were the Polymarket odds-on favorites to win the National and American League pennants, respectively. I wanted the most plausible World Series matchup I could get — and the betting markets gave me one.
Each team has a distinct managerial personality:
- The Optimizer (Dodgers) — Data-driven, analytically aggressive. Short starter leashes, leverage-based bullpen usage, platoon-heavy lineup construction. The kind of manager who quotes wRC+ in press conferences.
- The Skipper (Mariners) — Pitching-and-defense first. Trusts starters to find their rhythm, protects the bullpen across a seven-game series, and makes decisions with an eye on the long game. The kind of manager who talks about "feel."
These aren't just prompts — they're fully realized decision-making philosophies that create observable, sometimes dramatic differences in how the two managers handle identical game situations.
The World Series
Both managers are powered by Claude Sonnet 4.6 — same model, different philosophies. The Dodgers won in 7 games, clinching with a Game 7 shutout. Every pitching change, every lineup card, every bullpen call was made by Claude in real time. The game recaps, press conferences, and analysis you read on this site were all generated from those raw decision logs.
That was the project. A complete AI-managed World Series with full media coverage. Done.
Origin Story
Deep Dugout was born from a dog walk conversation about Pete Alonso signing with the Orioles after the Mets lowballed him. "The Mets would be better off if Claude was running the team" led to "What if I built something where Claude actually manages a team?" — and here it is.
Who Built This
Deep Dugout was built by Nicholas De Leon using Claude Code. The simulation engine, AI manager integration, content pipeline, and this website were all developed in collaboration with Claude.
Read about the methodology, see how the content pipeline works, or see the full credits.