Two heuristic code agents (TrackerAgent and SetRaceAgent) outperform most tested LLMs.

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 outperform most tested LLMsclaim0.868
  author assertion that deterministic heuristics surpass many LLMs
• Two heuristic code agents outperform most tested LLMs, and behavioural traces surface recurring LLM failure modes including overbidding, self-bidding, bankrupt TC initiation, and weak opponent-state adaptation.quote0.851
  Abstract sentence summarising performance and failures.
• SetRaceAgent outperforms five of seven tested LLMsfinding0.838
  SetRaceAgent ranked above DS-v3.2, GPT5.4-N, Haiku, G2.5-FL, and EconomyAgent.
• TrackerAgent outperforms six of seven tested LLMsfinding0.830
  In the 98-game slice, TrackerAgent had a higher win rate or TrueSkill than all LLMs except Gemini 3 Flash.
• The code-agent ordering (TrackerAgent > SetRaceAgent > EconomyAgent) shows information exploitation matters more than greedy quartet-chasing, which in turn outperforms conservative budgetingclaim0.807
  interpretation of what drives success among deterministic strategies
• Code-agent ordering: TrackerAgent > SetRaceAgent > EconomyAgentfinding0.799
  information exploitation outranks greedy quartet-chasing, which outranks conservative budgeting
• Code agents operate on structured data with exact arithmetic, while LLMs must parse natural-language observations and track state across turns; some failures may partly reflect numerical parsing or working-memory limitationsclaim0.791
  discussion of potential confounds
• Card-counting heuristics suffice to outperform most LLMs tested.claim0.786
  TrackerAgent's second-place ranking calibrates the benchmark and highlights LLM shortcomings.