09-04-2020, 12:09 PM
|
#19
|
Registered User
Join Date: Jun 2020
Posts: 178
|
Quote:
Originally Posted by JerryBoyle
I'm not sure if it's worth it to be honest, but I've made a note to try it out. I think what I'll do is pick a particular season at any track where there are a decent number of races and a decent number of repeat runners. At the end of the day, the output of the replay will be used to adjust the time the runner ran on a particular race, so evaluating how "good" that adjustment is will be easier if the runners run on the same track next time.
From there, I'll do the first season on my own, trying to capture things that are very simple to encode in rules or steps and are very easy to spot quickly. I think I'll avoid overall analysis that requires domain specific knowledge like "bad trip". With that output, I can hopefully make some adjustments and test how significant those adjustments are. If they're significant, I'll post the job somewhere and see how the crowdsourced version compares.
My initial thought is to try and keep the "jobs" at the race level. That way I can post an arbitrary number of races and let the contractors have at them. At that point, I'm paying for race. I'll keep track of the time I take to do it when I go through the first season and pay accordingly, prorating some reasonable hourly wage by expected time per race.
|
Cool. Sounds like a good plan to test. Good luck. Would love to hear how your experiment goes with this.
|
|
|