Oriel — AI & technology worth noticing, every day.

What the fox say?

Web page · https://entropicthoughts.com/what-the-fox-say

First seen by Oriel on .

AI summary

This article uses OpenSpiel and learned heuristics to analyze Starve the Fox, finding that foxes usually win shallow searches while coordinated sheep require substantially deeper strategy.

  • Starve the Fox has two foxes against 20 sheep, with asymmetric movement, capture, and victory conditions, plus draws after 40 progress-free moves.
  • Plain Monte Carlo tree search running about 1,000 iterations per second produced roughly 85% fox victories in self-play.
  • Sheep must coordinate the flock over many turns, effectively moving gaps away from foxes, while foxes can exploit isolated sheep more directly.
  • Truncated rollouts combined with regression-trained heuristic evaluation produced a stronger opponent at short thinking budgets than random rollouts alone.
  • The learned evaluation values a living fox at slightly more than two sheep and estimates that nine sheep ranks offset losing one sheep.

#game ai #monte carlo tree search #openspiel #board games

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Lobsters: 64, latest observation ; +62 over observed period since

Comments

Comment count reported by the source.

  • Lobsters: 13, latest observation ; +13 over observed period since

Aliases

SourceExternal IDURL
Lobsters7qaac6https://entropicthoughts.com/what-the-fox-say Read discussion