ELO and the VHLM

[Enorama 2,039]

So if you haven't already seen my previous articles on ELO and the VHL, go look em up now, there's four of them. The first explains my basic methodology and some of the terms that I'll be referring to here. I've completely overhauled the formulae and adjusted it to work for both the VHL and VHLM, so I'll dive a bit more into the changes I've made here assuming you've already got a basic understanding of ELO and the VHL.

 

First, a fun graph:

 

This is ELO tracked from S67 Game 1 in the VHLM and tracked across the entire season. Each team starts at 1000 and gains and loses as they win or lose games. The biggest difference you might notice at first glance is that the jumps and dips seem much more nuanced. This is because the formula for ELO gain now takes into account win margin in addition to the wins and losses themselves. Every minute dip and rise is a 1 or 2 goal game and each big jump you see is a 4-5+ goal win. I had to play around with the formula for quite a while before coming up with a graph I was happy with, and this is essentially what I came up with:

 

K = constant (which I've had to lower a ton from the old 32 to the new 6)

C = chance of winning = 10^(Team's ELO/400) / [10^(Team's ELO/400) + 10^(Opponents' ELO/400)]

M = margin of victory

New ELO = Old ELO ± K * (M^2/2) * (1-C)

 

The M^2/2 part is the one I've had trouble with, but after playing around with that and K a bit, I think I'm pretty happy with how this turned out.

 

As for the graph itself right now, you can clearly see the haves and the have-nots in the VHLM with the 1st-6th place teams having a tidy lead over the 7th-11th place teams. If I'm betting on any one underdog to win, it would be the Rush, as they're the ones on the best trajectory upwards leading into the latter part of the season (@Josh will no doubt be happy to hear this).

 

As for the bottom half, I think we should all be a little surprised that Mexico isn't any lower. If you followed my VHL ELO graphs from the season before last, you would have seen how Moscow dropped out of view like a rock, and that is what a truly bad team looks like on a graph like this. Here, however, Mexico maintained a respectable distance from Mississauga and Saskatoon, even passing the Hounds are more than one point this season.

 

Just for fun, because I know @Nykonax would ask, here's what the graph looks like if I plug in K = 32, which is what I was able to use just fine before the shift in formulae to account for win margin. Ottawa and Yukon are way too good, Halifax trashed the end of the season, and Vegas and San Diego managed to turn things around at some point.

 

 

512 words, claiming for Washington 8/25

[McWolf 3,116]

Moscow disappearing below the graph after like 20 games was glorious

[Victor 11,043]

9 hours ago, McWolf said:

Moscow disappearing below the graph after like 20 games was glorious