VHL S67 Pythagorean Expectation

[DMaximus 1,045]

Last season I wrote a post discussing using Pythagorean Expectation to predict the final standings for S66 in the VHL.

If you want the gritty details, check out last season’s post and the footnotes I provided.  Basically Pythagorean Expectation takes Goals For and Goals Against and uses it as the basis for a predictive model.

 

So far this year there has been 1472 total goals in 245 total games which equals 6.008 goals per game. A slight uptick from the 5.88 goals per game mark of S66. That gives us an exponent of 2.273 to use in our equation used to determine each team’s expected winning percentage.

 

Here’s the results:

Team	Points	Expected Win%	Actual Win %	Win Differential
Toronto Legion	80	0.68570	0.77551	4.40072
Moscow Menace	63	0.60215	0.61224	0.49482
Helsinki Titans	59	0.55230	0.57143	0.93734
Calgary Wranglers	55	0.52543	0.51020	-0.74589
Vancouver Wolves	55	0.50000	0.48980	-0.50000
Riga Reign	53	0.54569	0.48980	-2.73881
HC Davos Dynamo	48	0.50818	0.40816	-4.90061
Seattle Bears	47	0.37506	0.40816	1.62207
Malmo Nighthawks	43	0.37585	0.38776	0.58340
New York Americans	39	0.34307	0.34694	0.18937

 

A positive win differential mean that team won that many more games that what was expected. A negative win differential means that team won that many fewer games than expected.

Now we’ll take the expected win percentage, apply it to the number of remaining games to find out the expected point totals for each team. One change I made here from last year is I’m factoring in overtime losses. I took the current percentage of a team’s games that have ended in an overtime loss and applied that percentage to the remaining games in the season.

 

Here’s the expected final standings and point values:

Team	Expected Final Points
Toronto Legion	113
Moscow Menace	92
Helsinki Titans	86
Calgary Wranglers	82
Vancouver Wolves	81
Riga Reign	80
HC Davos Dynamo	75
Seattle Bears	68
Malmo Nighthawks	63
New York Americans	57

 

Last year’s model was under on total point values for almost every team because I looked solely at wins. Hopefully this prediction will be closer to the actual final point values.

 

Last year a commenter asked about factoring in strength of schedule to create a more accurate model. I liked the suggestion so I started taking a look at determining how to include strength of schedule. Little did I know the wormhole this would send me down. It turns out there is not really a consensus on how to best factor in strength of schedule into predictive models.

 

Also with the current layout of the VHL every team plays every other team 8 times, so ultimately at the end of the year everyone will have the exact same SOS. However at this current moment that’s not true.  For example, Moscow has played Calgary and New York 7 times, while only playing Riga twice. That means each team has a slightly different strength of schedule. I emphasize slightly here because, even with using different SOS systems, the actual difference in SOS between each of the teams is slight. Each formula is based around the same concept, take the total winning percentage, give it a weight, take the total winning percentage of your opponent’s opponents and give that a weight. That spits out a Strength of Schedule number.

 

Here’s the results using the NCAA College Hockey RPI/SOS formula[1]:

Team	Opponents’ Win %		Opponent’s Opponents Win %	RPI	SOS
Toronto Legion	0.5233236	0.5584323		0.6053289	0.5486019
Moscow Menace	0.5426905	0.5540421		0.5662090	0.5508637
Helsinki Titans	0.5635152	0.5498092		0.5580923	0.5536469
Calgary Wranglers	0.5316535	0.5586150		0.5408504	0.5510658
Vancouver Wolves	0.5560183	0.5517386		0.5371517	0.5529370
Riga Reign	0.5472720	0.5559206		0.5375732	0.5534990
HC Davos Dynamo	0.5645564	0.5506464		0.5179467	0.5545412
Seattle Bears	0.5708038	0.5499367		0.5188754	0.5557795
Malmo Nighthawks	0.5614327	0.5515389		0.5126707	0.5543092
New York Americans	0.5693461	0.5499324		0.5032609	0.5553683

 

As you can see the numbers are really close together. I’m not sure how much of a difference factoring in the strength of schedule would matter on the predictions made here.

 

---

[1] The Ratings Percentage Index is one of the NCAA's selection criteria for college hockey. It is a sum of .25 times a team's winning percentage, .21 times their opponents' winning percentage and .54 times their opponents' opponents' winning percentage.

The NCAA’s Strength of Schedule formula is 0.28 times a team's opponents' winning percentage plus 0.72 times their opponents' opponents' winning percentage.

[Beaviss 4,949]

 

me reading this