TrackerAgent μ=30.8 in 172-game mixed slice

TrackerAgent uses card counting to achieve high rating, a capability no LLM replicates

Source paper

extracted_from

(2026) · Robert Müller · Clemens Müller

cosine ≥ 0.65 · no typed edge

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

TrackerAgent TrueSkill μ=28.7±3.6 in 98 canonical gamesfinding0.841
second-highest TrueSkill rating
TrackerAgent TrueSkill μ=28.7 ± 3.6, 53.6% win ratefinding0.821
Best code agent outperforming six of seven LLMs.
TrackerAgent and SetRaceAgent have TC tightness τ ≈ 0.2–0.25, looser countersfinding0.758
Code agents trade bargaining precision for acquisition pressure.
TrackerAgent win rate 53.6% in 98 canonical gamesfinding0.756
TrackerAgent won over half of the combined-comp1 games.
SetRaceAgent TrueSkill μ=27.3±3.3finding0.753
fourth-highest TrueSkill rating
G3-F TrueSkill μ=30.1 ± 3.3, 72.9% wins, median score 5,250 on combined-comp1 slice (n=98 canonical games)finding0.739
Top LLM performance with high win rate and large score.
EconomyAgent TrueSkill μ=22.1±3.5finding0.738
EconomyAgent TrueSkill rating
TrackerAgent win rate=53.6%finding0.721
TrackerAgent won more than half its games