# Good Start Labs Trains AI on Railroad Strategy Game to Boost Finance Benchmark Scores

> Good Start Labs is using strategy games to train artificial intelligence models for real-world tasks, Latent Space reported. In a recent experiment, training a 30B model inside the board game 1830 improved its scores on the Finance-Agent benchmark when configured as a multi-turn terminal agent.

- **Topic**: Research
- **Published**: 2026-09-17T14:20:29.574Z
- **Read Time**: 1 min read
- **Canonical URL**: https://highsignal.sh/stories/good-start-labs-trains-ai-on-railroad-strategy-game-to-boost-finance-benchmark-s-c492baf5

## Why It Matters

Reinforcement learning in games provides clear, verifiable outcomes for evaluating model behavior. The startup's results suggest that the structure of an AI training environment directly influences whether strategic reasoning and tool usage can transfer to outside domains like financial analysis.

## Key Findings & Analysis

### Railroad Game Mechanics and Workflow

Good Start Labs tested whether skills learned in 1830: The Game of Railroads and Robber Barons would carry over to financial research tasks. CEO Alex Duffy explained that the game involves bidding on railroad company shares. The lab created tasks where models search a database for game history, write data to Excel spreadsheets, build functions, and calculate answers, simulating typical financial workflows.

### Benchmark Transfer Results

The experiment compared single-turn next-move choices against a multi-turn terminal agent that uses tools to explore and adapt plans in real time. While both setups achieved in-game training goals, only the terminal-agent design produced gains on the Finance-Agent benchmark. Duffy also previously claimed that fine-tuning on the strategy game Diplomacy boosted model performance on customer support and industrial operations benchmarks.

## Primary Sources & Citations

- [Can Skills Learned in Games Transfer to Real-World Work?](https://www.latent.space/p/good-start-labs) — *Latent Space* (Reporting)

---

[← Back to front page](https://highsignal.sh/) | [Daily Brief](https://highsignal.sh/brief) | [All stories](https://highsignal.sh/latest)