Track Record
Track Record
Can SP2P's AI explanations be trusted? Two verifiable answers: the explanation QC results (LLM semantic review + human confirmation), and the AI's probability judgment vs market outcomes (data accumulating).
Explanation QC: … reviewed (LLM semantic review + human confirmation)
…
Long-term accuracy of AI probability judgments
After markets resolve, we compare the AI's original probability judgment with the actual outcome to compute a calibration score. This needs at least 200 resolved judgments to be statistically meaningful.
0 / 200 resolved judgments · calibration curve appears when reached