Jak Runner Boards Algorithm

Discuss in depth and technical topics.
Post Reply
User avatar
Posts: 1
Joined: 30 May 2019, 20:20

Jak Runner Boards Algorithm

Post by G3catcher » 30 May 2019, 22:50

Jak Runner Boards: https://jakspeedruns.com/

I wanted to create a system that would rank runners based off of how good their times were and not just how many runners they were better/worse than.
For example, in a 2 hour speedrun, one runner has a world record of 1:59:45 and another has 2nd place at 1:59:50. In this case, with the new algorithm, both of these runners will get nearly the same amount of points because both runs are very good and the runner in 2nd place should not be punished because of losing 5 more seconds.
I believe that this new algorithm creates a very close way to rank speedrunners based on the variety of runs that they know and how good their times are for those runs. To move up on these boards, you can make a choice to either focus on getting a really good time in one category and gain points that way, or expand knowledge and just be average at a bunch of categories.

Steps of Algorithm:

1. In the category, take the average of all of the runs.
2. For each run, find the time away from the average.
3. Take the population standard deviation of the times away from the average of each run using the formula: Image.
4. Divide each run's time away from average by the standard deviation to get the number of standard deviations away from the average.
5. If a run is more than 5 standard deviations away from the average, remove the run and recalculate steps 1-4.
6. If the run is faster than the average: Multiply the amount of std. devs. from the median by 5 and add it to the number of runners in the category times 5
7. If the run is slower than the average: Multiply the amount of std. devs. from the average by 5 and subtract it from the number of runners in the category times 5
8. If a runner has run more than one subcategory of a category, multiply their total from that category by 0.75
9. Add together all of a runner's points from each category of every game to get the final score

As of right now, the algorithm is not final and will be changing slightly in the future as more and more ideas will come to mind as to how to improve the algorithm. If you have any ideas to improve the algorithm, feel free to contact @G3catcher#4321 on discord.

Post Reply