ELO and the VHL part 2

[Enorama 1,980]

If you haven't already checked out my previous article on ELO, introducing the concept and taking a look at last year's VHL, be sure to do so here.

 

For anyone who's been following the VHL standings so far this season (S66), you've probably noticed some trends just on the surface level. Primarily that Helsinki has been relatively strong all year, but really come into their own around game 15, Moscow isn't nearly as bad as last season, and Malmo has a nice arc to their trajectory, probably peaking around the same time Helsinki separated themselves from the pack.

 

Well, I'm here to tell you those same things, but in chart form.

 

First of all, I've charted the cumulative winrate for each VHL team to see kind of a blanket absolute marker of how they've done so far this season. Apologies for the colours, I can't be bothered to fix them here (they'll be fixed on the ELO chart).

 

You can see Moscow started off abysmally, but climbed past Seattle as the latter fell harshly later on. Then there's Helsinki with a degree of separation from an absolutely crowded pack right in the middle of the league. 

 

Now here's the same timeframe represented in ELO form:

You can see that because Moscow has been beating such strong opponents lately, they actually deserve to be right up in the middle pack with the rest of them and they also passed Seattle slightly earlier than on the winrate chart. You can also see a very poignant drop from Malmo starting at the ~25 game mark, it's in the jumble of the pack, but you can pick out the mid-blue line peaking around game 15, briefly in first place, before zig zagging around a bit and eventually falling to just above Moscow.

 

So far, these ELO rankings have been done with a K value of 32, which is a moderate number. Next week, I'll be kicking the K value up to 64, which will bring out some harsher fluctuations and show something more akin to power rankings.

 

343 words.

[Nykonax 1,550]

4 minutes ago, Enorama said:

So far, these ELO rankings have been done with a K value of 32, which is a moderate number. Next week, I'll be kicking the K value up to 64, which will bring out some harsher fluctuations and show something more akin to power rankings.

K 1024