TrackerAgent and SetRaceAgent have TC tightness τ ≈ 0.2–0.25, looser counters

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

• Two heuristic code agents (TrackerAgent and SetRaceAgent) outperform most tested LLMs.claim0.774
  Calibration that conditional logic can beat cost-efficient LLMs in this setting.
• TrackerAgent TrueSkill μ=28.7 ± 3.6, 53.6% win ratefinding0.772
  Best code agent outperforming six of seven LLMs.
• Claude Haiku 4.5 and GPT-5.4 Nano have TC tightness τ ≈ 0.4, the tightest among allfinding0.760
  These two LLMs bargain with minimal overpayment but low overall efficiency.
• TrackerAgent TrueSkill μ=28.7±3.6 in 98 canonical gamesfinding0.760
  second-highest TrueSkill rating
• TrackerAgent μ=30.8 in 172-game mixed slicefinding0.758
  TrackerAgent uses card counting to achieve high rating, a capability no LLM replicates
• TrackerAgent outperforms six of seven tested LLMsfinding0.757
  In the 98-game slice, TrackerAgent had a higher win rate or TrueSkill than all LLMs except Gemini 3 Flash.
• SetRaceAgent outperforms five of seven tested LLMsfinding0.745
  SetRaceAgent ranked above DS-v3.2, GPT5.4-N, Haiku, G2.5-FL, and EconomyAgent.
• TrackerAgent win rate=53.6%finding0.738
  TrackerAgent won more than half its games