System Performance – Possible Cause

An update for those of you still following our progress.. This post explains the latest findings, that we think, might explain why the performance of the system nose-dived, after July 2020.

We’ve noticed our runner “Speed Ratings” have become less effective this year.

The SRs are based on a “race”, which is the combination of “course, direction (left/right), distance, going”. In other words these 4 classifiers are grouped into a single entity, and used as the basis for comparative speed ratings across different courses.

It would appear runner performances, as calculated by our traditional approach to a “race”, has significantly changed this year. This could be due to the lack of crowds (adrenaline), or some other changes to the physical conditions at each course – we’re not sure why. The net effect is that the baseline times for each “race” that we were using are now out of touch which recent performances – which has lead to inaccuracies to the SR scores given to each runner.

In the past, the baseline data for each “race” was re-calculated every 6 months. From the 16th Oct, SRs baseline times are now re-calculated every day (this process now takes around 2 hours).

We’re not sure if this will turn around the performance of the system, but will obviously be monitoring the systems performance following this change to the design.

Update 2020-11-06: This seems to have turned around the systems performance, now showing a profit from 16th Oct – 5th Nov 2020.

Cheers
Nick