Connecting to your experimental data…
Recorded decisions
What does the model expect next?
Reveal a decision to request a prediction.
Predictions describe the next sample if the participant continues. Position and quality are predicted separately, so the diamond need not lie on the landscape. The predicted action describes whether the participant will stop or continue; it is not a recommendation. The stopping probability is an uncalibrated model estimate. The cutoff maximizes F1 for this game in the saved validation results. Stopping includes submission and timeout. Previous position error compares the forecast made one step earlier with the latest revealed decision, in percentage points (pp).
Playback advances one decision per second, not at the original speed. The graph shows this participant's own samples; shared information is not yet visualized. The model uses the original shared-information features, although the graph displays own samples only.
See the revealed observations
| Step | Position (%) | Quality (%) | Original x | Observed value | Elapsed (s) |
|---|
Quality is scaled to the estimated maximum of this round's function. This is an analyst's view: participants may not have known that maximum. Elapsed time starts at the first recorded sample, not at the beginning of the round.