G3.1-FL TrueSkill μ=28.0 ± 2.9, 44.9% win rate

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• G3-F TrueSkill μ=30.1 ± 3.3, 72.9% wins, median score 5,250 on combined-comp1 slice (n=98 canonical games)finding0.896
  Top LLM performance with high win rate and large score.
• G2.5-FL TrueSkill μ=20.5±2.9finding0.878
  Gemini 2.5 Flash Lite, lowest TrueSkill rating
• TrackerAgent TrueSkill μ=28.7 ± 3.6, 53.6% win ratefinding0.855
  Best code agent outperforming six of seven LLMs.
• Sonnet 4.5 TrueSkill μ=26.4 ± 4.9 (n=14, 35.7% win rate)finding0.836
  Mid-field performance with larger uncertainty due to small sample.
• DS-v3.2 TrueSkill μ=23.9±2.7finding0.831
  DeepSeek v3.2 TrueSkill rating
• GPT5.4-N TrueSkill μ=22.6±2.7finding0.807
  GPT-5.4 Nano TrueSkill rating
• Gemini 3 Flash TrueSkill µ = 30.1 ± 3.3 in 98 canonical gamesfinding0.800
  The Bayesian skill posterior mean for G3-F is 30.1 with 3σ confidence interval ±3.3.
• EconomyAgent TrueSkill μ=22.1±3.5finding0.799
  EconomyAgent TrueSkill rating