Table 1.
Data fields and their sources.
Fig 1.
Finish times and densities according to year and prediction.
Panel A shows a histogram of finish times collapsed across all years. Panel B shows a violin plot of finish time density according to year. Panel C shows a violin plot of finish time density according to the prediction window selected at registration. In Panels B and C, the perimeter of each plot illustrates density, the central point represents the mean, and the vertical line represents +/- one standard deviation.
Fig 2.
Finish times according gender, age and club membership.
Violin plots of finish time density according to club membership (Panel A), gender (Panel B) and age category (Panel C; ‘46+’ is runners aged 46 and older). The perimeter of each plot illustrates density, the central point represents the mean, and the vertical line represents +/- one standard deviation.
Fig 3.
Finish times according to the interaction between demographics and prediction.
Violin plots of finish time density according to prediction and club membership (Panel A), gender (Panel B), and age category (Panel C). The perimeter of each plot illustrates density, the central point represents the mean, and the vertical line represents +/- one standard deviation.
Fig 4.
Scatterplot showing age against finish time.
Males (blue) and females (yellow) are plotted with alongside their lines of best fit.
Fig 5.
Predictions according to year and finish time.
Panel A shows the proportion of predictions collapsed across all years. Panels B and C shows stacked bar plots of predictions according to year and eventual finish time respectively. Prediction proportions less than .05 (Panel A) and .10 (Panels B and C) are not labelled.
Fig 6.
Predictions according to club membership, gender and age.
Stacked bar plots of predictions according to club membership (Panel A), gender (Panel B), and age category (Panel C; ‘46+’ is runners aged 46 and older. Prediction proportions less than .1 are not labelled.
Fig 7.
Prediction discrepancies according gender, age and club membership.
Three panels plotting the proportions of runners whose performance was better (faster) than predicted; matched their predictions; and was worse (slower) than predicted. The demographic variables used to split runners were: club membership (Panel A); gender (Panel B); and age category (Panel C).
Fig 8.
Performance according to the interaction between demographics and prediction.
Panels A, B, and C show stacked bar plots plotting categorical discrepancy—the proportions of runners whose performance were better (faster) than predicted; matched their predictions; and were worse (slower) than predicted. The independent variables used to organise data are club membership (Panel A), gender (Panel B), and age category (Panel C), and how they interact with prediction times.
Fig 9.
Finish times and prediction discrepancies according race repetition.
Panel A shows violin plots of finish time density according to race repetition. Panels B shows stack bar charts plotting categorical discrepancy according to race repetition. Panel C shows violin plots of continuous prediction discrepancy density according to race repetition. These are within subjects data, meaning that the same 483 runners who completed the race at least three times contribute to each violin or bar within the three panels.