I decided it was time to try a little experiment with in-season power rankings. So the first step is to pick a system to use. I looked at Myhockeyrankings.com, but don't particularly like the additive system they use for SOS (Strength of Schedule) and the results it produces. A different system has the added advantage of being, well, different.
KRACH is a system that is used in college hockey. While it looks fantastic, the system requires some pretty intimidating math for someone with a liberal arts degree. KRACH also needs at least 30 games to get itself sorted out, which makes it less useful for an in-season ranking system for youth hockey. So I settled on a system used by pro and college football analysts that is called Simple Rating System or SRS.
Sports-reference.com defines SRS as "a rating that takes into account average point differential and strength of schedule". It has the advantage of producing results that are easy to interpret, and it is a predictive system rather than a retrodictive system. Essentially that means we can use the rankings to help determine relative strength between teams, or to evaluate the likelihood of one team beating another. Another benefit of this system is that you can fire up Excel and reproduce my results, if you are so inclined.
In the table below, I have calculated the SRS for the Central States Developmental Hockey League (CSDHL) Squirt division. Only league games are used in order to streamline the procedure, and no tournaments have been included. I chose Central States because the NIHL regular season has just begun, and there are not enough games to calculate a regular season SRS. Using the SRS system monikers, MOV stands for margin of victory but hockey people would call it goal differential. SOS is strength of schedule of course, and SRS is the ranking itself. SRS = MOV + SOS.
CSDHL Squirt
Rank | Team | MOV | SOS | SRS |
1 | Bruins | 2.909 | -0.510 | 2.399 |
2 | Falcons | 2.545 | -0.178 | 2.368 |
3 | Hawks | 1.385 | -0.142 | 1.243 |
4 | Blues | 0.900 | -0.228 | 0.672 |
5 | Affton | -0.154 | 0.698 | 0.544 |
6 | Ice Dogs | 0.083 | 0.043 | 0.127 |
7 | Chesterfield | -0.300 | 0.219 | -0.081 |
8 | Sabres | -0.571 | 0.435 | -0.136 |
9 | Glenview | -0.182 | -0.062 | -0.244 |
10 | Huskies | 0.111 | -0.632 | -0.521 |
11 | Winnetka | -3.786 | 0.171 | -3.615 |
12 | Rockford | -3.500 | -0.370 | -3.870 |
I am providing the current standings from CSDHL.org for comparison.
Team | PTS | GP | W | L | T | GF | GA | Goal Diff. | |
1 | Falcons | 22 | 11 | 11 | 0 | 0 | 47 | 19 | 28 |
2 | Bruins | 20 | 11 | 10 | 1 | 0 | 60 | 28 | 32 |
3 | Ice Dogs | 13 | 12 | 6 | 5 | 1 | 34 | 33 | 1 |
4 | Affton | 13 | 13 | 6 | 6 | 1 | 43 | 45 | -2 |
5 | Hawks | 12 | 12 | 6 | 6 | 0 | 43 | 30 | 13 |
6 | Huskies | 10 | 9 | 5 | 4 | 0 | 28 | 27 | 1 |
7 | Chesterfield | 10 | 10 | 4 | 4 | 2 | 24 | 27 | -3 |
8 | Sabres | 10 | 13 | 4 | 7 | 2 | 38 | 41 | -3 |
9 | Blues | 9 | 10 | 4 | 5 | 1 | 33 | 24 | 9 |
10 | Glenview | 9 | 11 | 4 | 6 | 1 | 33 | 35 | -2 |
11 | Winnetka | 3 | 14 | 1 | 12 | 1 | 26 | 79 | -53 |
12 | Rockford | 1 | 6 | 0 | 5 | 1 | 8 | 29 | -21 |
I took the liberty of adding a column for ranking by points on the left and a column calculating goal differential on the right.
There are a couple of things to note. First, of course, is that the GP column for 'Games Played' reveals that not every team has played the same number of games. Second, not every team has played each other yet.
Having said that, here is a bit of analysis worth considering. The Bruins are on top of the SRS rankings despite losing a game to the Falcons. The SRS does not look at wins, losses or ties. However, the Bruins have a lower SOS rating and are on top by virtue of goal differential. The system rewards high scoring teams. I have not taken the step of restricting the differential to say 7 or even 6 goals. Looking at the 67 games played to date, very few have a differential above 6 and I deemed it not worth the effort at this time. As Rockford plays more games, that step may become necessary.
Other differences of note included the Blues and the Huskies. When ranked by points, the Blues are in 9th and the Huskies are in 6th. The SRS has the Blues in 4th thanks to goal differential and the Huskies in 10th due to a weak schedule to date.
Doing a comparative analysis within the SRS, the Bruins and Falcons are on top with a drop off to the Hawks then another drop off to the Blues and Affton. A hypothetical perfectly average team would have an SRS of 0.000, and Chesterfield comes the closest to that team. The Ice Dogs and Sabres are also in that middle tier.
Let me know if you find this useful, and I may keep it going throughout the season or even expand it to include CSDHL PeeWee and Bantam and maybe NIHL Elite and Gold. I don't have a dog in this hunt. If I don't hear from you, then maybe I won't go to the trouble ...
***UPDATE***
If you are interested in the math, you can download the Excel spreadsheet here. The iterative calculations are done on the 'SQ MOV' tab. If you have trouble getting it to work, you might find this video helpful.
You should follow me on twitter here.
Interesting project. Might be worth trying to replicate in another group to see if the statistical premise holds up. When you glance at the scores from this group it appears that they are all over the place - lots of lopsided scores on wins and losses for a lot of the teams, so I'd be curious to see how this plays out in a scenario where things are more even.
ReplyDeleteIt would be great if you could provide your spreadsheet so some of us curious noobs can see the set up/math :)
ReplyDeleteWould be interesting to see how the numbers change over the balance of the season. Great job!.
ReplyDelete