January 9, 2007

2006-2007 Frequency Standings

A little over a month ago Earl Sleek presented a different way of looking at winning. The basic premise is that all that matters is when the "winning team" changes. So you only consider when a team lets in that goal to tie the game or breaks a tie not the one that makes the game go from 5-2 to 6-2. The list below presents most of the relevant results. As stated by Earl Sleek: "Frequency counts are simplistic in that they do not consider how long situations last, but rather just counts how often situations arise." Using the frequency counts one can simulate thousands of games based on those numbers to get expected percentages of wins, losses and ties. Now I'm not giving points for OTL (you can add I 4 points to each team as a reasonable estimate), and I'm giving teams who go to a SO 1.5 points (1 point for L, 2 points for W assumed to be a 50-50 split). The results look reasonable in terms of a season prediction for expected points. In fact I really like this style of logic as it doesn't chase those blow-out games like a Pythagorean prediction would.

You will notice bottom teams struggle to hold onto a lead giving it up over 70% of the time. The best teams often have a great ability to comeback in games with the exception of Detroit the top 10 teams on this list come back at a rate above 60%, Detroit wins as a result of their impressive 58% win percentage when they take the lead.

Personally I think of these as more useful for the Playoff series predictions as I believe that playoff hockey follows slightly different patterns than the regular season, a hypothesis I have for example is that CB (comeback %) dominates other factors in the playoffs so Detroit will do poorly again and Buffalo/Anaheim/Montreal will be the dominant teams and Pittsburgh if they make it could be a dark horse (8th seed goes to Western?), of course that's just a theory (I'll test it when I get more time).

Nlong_nameWLTPTSSWTLOLLWCB
1Anaheim Ducks5514131302.9461%32%48%66%
2Buffalo Sabres5216131252.8153%38%52%67%
3New Jersey Devils4819151192.5249%41%54%61%
4Nashville Predators5121101173.0857%37%47%61%
5Detroit Red Wings522461122.5756%40%58%55%
6Montreal Canadiens4323151103.0544%47%47%65%
7San Jose Sharks552701092.6253%47%70%61%
8Dallas Stars462691072.6645%49%59%61%
9Atlanta Thrashers4229111013.1946%48%45%61%
10Minnesota Wild352720993.0938%51%40%61%
11Vancouver Canucks44308992.7853%42%48%49%
12Calgary Flames45334962.7053%44%52%48%
13Boston Bruins403111963.3953%41%37%53%
14Washington Capitals403210952.6344%49%51%51%
15Pittsburgh Penguins393310943.4346%48%40%59%
16New York Rangers383312933.0047%47%41%51%
17Toronto Maple Leafs383311922.9947%47%41%51%
18Carolina Hurricanes43354912.7147%51%56%52%
19New York Islanders373410902.3845%48%50%41%
20Edmonton Oilers39376872.6048%48%50%44%
21Ottawa Senators42382872.4451%47%56%42%
22Chicago Blackhawks363610862.9446%48%41%48%
23Colorado Avalanche39376863.4450%47%39%53%
24Phoenix Coyotes36388842.6742%53%50%49%
25Tampa Bay Lightning36388833.1347%49%40%50%
26Florida Panthers333811823.1437%56%43%57%
27Los Angeles Kings314210763.1038%57%41%53%
28Columbus Blue Jackets32446732.6640%56%48%45%
29St. Louis Blues244512683.1841%52%29%43%
30Philadelphia Flyers22546532.9841%55%29%37%
The random expected error for this data is around +/-10% (95% confidence intervals - for the Tl, OL, LW, CB), so take the results as you like.
W - wins, L - losses, T - ties (for 82 game season) - estimated using simulation.
PTS - expected points using this method.
SW - average score switching - high numbers likely indicate difficulty maintaining a lead (bad defense/goaltending)
TL - Percentage the team takes lead.
OL - Percentage the opposition takes lead.
LW - Percentage the lead results in win. (otherwise it is tied again).
CB - Percentage of come backs - # of times team re-ties game. (otherwise opposition wins).

3 comments:

David Johnson said...

I'd be interested in the numbers for last season if you could get them. My gut would tell me that LW% would be the more important indicator of playoff success. CB% is probably more indicative of offensive ability where LW% is more indicitave of defensive ability and I think goaldentind and defense are more important in the playoffs. Especially goaltending. Edmonton and Carolina had it last year, Detroit didn't, thus lost. This year Detroit plays great defense (see shots against totals) and have good goaltending so they have the chance to do well in the playoffs.

The truth is, averaging the two is probably the best of them all. A team that can come back from falling behind and can hold a lead if they get it should be a tough foe in any playoff series. This makes San Jose, Dallas and Buffalo look like good picks for playoff success.

Earl Sleek said...

Of course when I used the term "Math Wizards" in the blog post title, I was referring to you.

Great work, next time I'm just going to call it "OK JavaGeek..."

PPP said...

Great work.