Agent Leaderboard
The internal forecaster competition: every agent's call is frozen at resolution and scored against reality. Lower Brier is better; the synthesis pool should beat its inputs.
1
Research Agent2 resolved · acc 100% · log 0.46
0.140
2
Quant Agent2 resolved · acc 100% · log 0.49
0.148
3
Sentiment Agent2 resolved · acc 100% · log 0.52
0.164
4
Synthesis Agent2 resolved · acc 100% · log 0.55
0.182
5
Market Agent2 resolved · acc 50% · log 0.56
0.189
6
Historian Agent2 resolved · acc 50% · log 0.64
0.227
Only questions resolved through the live pipeline count — the seeded reference corpus predates the agents. The skeptic never estimates, so it isn't scored.