Key in your Season 15 statistics!

Season 15 just ended, and I'd been playing daily for the lockboxes. Taskmaster was won on the (5 May, Cinco De Mayo). Next season, we would expect to see some changes in the PVP tournament structure; nonetheless historical precedent in understanding statistics will help us a lot in deciphering what these mean and reforging optimum strategies. 

Statistical primer, is my first blog post. Apologies for the lack of explanations, they will be added as I have time to write them. As you realized, level is a problematic factor. Two ways can correct for it. 

1. Sample restriction

The idea is that if levels 260-299 are causing anomalies in rating, then eliminate them. This regression does exactly that, and restricts the sample to only level 300. The advantage is that it completely eliminates the level variable and so statistical comparisons are more pure. The disadvantage is the sample size becomes a lot smaller, making it harder to draw significant conclusions.


2. More technically, a fixed effect regression.

Fixed effect regression is more versatile and nuanced than restricting the sample. We group agents of equal level into buckets, and compare those buckets against each other. 

See here the R-sq within: 0.68, a very decent number. R-sq between: 0.11, lower, and shows the relationship between all the level buckets.


3. Add more control variables

Adding more control variables could help eliminate the level effect. This is because more variables makes the model more realistic. Since Level is an anomaly, and we expect its effects to be diminished. This helps marginally, the level coefficient decreased from 7.48 to 6.84 (compare to 1st blog post)


Ad blocker interference detected!

Wikia is a free-to-use site that makes money from advertising. We have a modified experience for viewers using ad blockers

Wikia is not accessible if you’ve made further modifications. Remove the custom ad blocker rule(s) and the page will load as expected.