Week 8 Playoff Probabilities

Below are the current playoff probabilities as of 3/16/13, with change in PP indicating change from 2 weeks ago. Tables are sortable, click on column heading.

The details of the model used to come up with these numbers has come up a few times, and it’s been awhile since I explained it in detail, so check out the methods section below for a full disclosure.

Western Conference

Final Standings	Team	SAF%	Playoff Probability	Change	Mean Points	Current Points	Current Rank	Home/Road Diff
1	Chicago Blackhawks	54.41	99.9%	1.5%	70.90	49	1	2
2	Anaheim Ducks	47.85	98.9%	6.1%	67.38	44	2	3
3	Minnesota Wild	48.79	64.7%	14.3%	54.85	32	3	-1
4	St. Louis Blues	52.09	69.5%	10.5%	55.74	34	4	0
5	Los Angeles Kings	56.02	65.6%	5.1%	55.04	32	6	1
6	Vancouver Canucks	52.64	64.2%	-5.2%	54.77	32	7	-1
7	Detroit Red Wings	52.00	53.2%	6.7%	53.22	33	5	-3
8	San Jose Sharks	49.97	53.7%	-4.7%	53.03	30	8	1
9	Phoenix Coyotes	51.37	40.0%	-6.5%	50.94	30	9	-4
10	Nashville Predators	46.03	41.2%	-3.2%	50.81	28	10	4
11	Dallas Stars	50.24	36.6%	-9.8%	50.03	27	12	1
12	Edmonton Oilers	46.30	35.6%	-5.9%	49.89	26	14	5
13	Calgary Flames	49.95	33.9%	1.1%	49.34	26	13	-4
14	Columbus Blue Jackets	47.04	22.1%	11.7%	47.96	28	11	-5
15	Colorado Avalanche	48.63	20.7%	-22.1%	46.95	24	15	1

(updated 3.17.2013 at 12:12 AM PST)

Eastern Conference

Final Standings	Team	Score Adj Fenwick%	Playoff Probability	Change in PP	Mean Points	Current Points	Current Rank	Home/Road Diff
1	Boston Bruins	53.81	97.7%	4.7%	64.83	41	4	0
2	Pittsburgh Penguins	52.44	97.3%	18.8%	63.05	42	2	3
3	Winnipeg Jets	51.13	69.2%	23.6%	54.50	32	3	4
4	Montreal Canadiens	52.05	97.7%	14.1%	63.74	42	1	0
5	Ottawa Senators	51.56	75.7%	5.8%	56.04	34	5	2
6	Carolina Hurricanes	51.18	66.9%	3.1%	54.04	31	8	1
7	Toronto Maple Leafs	46.46	60.5%	-9.9%	53.09	32	6	1
8	New Jersey Devils	52.33	55.5%	-11.2%	52.57	32	7	-1
9	New York Islanders	49.83	43.3%	16.0%	50.44	29	9	-2
10	New York Rangers	52.42	43.5%	-3.0%	50.30	28	10	-3
11	Philadelphia Flyers	49.63	29.1%	-13.4%	48.08	27	11	1
12	Tampa Bay Lightning	46.32	21.9%	-11.9%	46.50	25	12	-2
13	Buffalo Sabres	45.62	19.3%	-6.6%	46.32	24	13	2
14	Washington Capitals	48.51	20.1%	-9.6%	45.84	23	14	-1
15	Florida Panthers	48.41	2.3%	-20.5%	40.01	20	15	-5

(updated 3.17.2013 at 12:12 AM PST)

This week I added fenwick close, as well as the difference in rank between these stats. As you can see, SJS are by far an outlier.

Score Adjusted Fenwick%

Team	SAF%	SAF Rank	FenClose%	FenClose Rank	diff
Los Angeles Kings	56.0	1	57.65	1	0
Chicago Blackhawks	54.4	2	55.27	2	0
Boston Bruins	53.8	3	54.81	4	-1
Vancouver Canucks	52.6	4	53.66	6	-2
New York Rangers	52.4	6	53.53	7	-1
New Jersey Devils	52.3	7	53.71	5	2
Pittsburgh Penguins	52.4	5	51.76	10	-5
St. Louis Blues	52.1	8	54.85	3	5
Detroit Red Wings	52.0	10	51.64	11	-1
Montreal Canadiens	52.0	9	53.36	8	1
Ottawa Senators	51.6	11	50.28	19	-8
Carolina Hurricanes	51.2	13	52.73	9	4
Phoenix Coyotes	51.4	12	50.46	17	-5
Winnipeg Jets	51.1	14	50.36	18	-4
Dallas Stars	50.2	15	50.52	16	-1
San Jose Sharks	50.0	16	51.2	12	4
Calgary Flames	50.0	17	50.88	13	4
New York Islanders	49.8	18	49.57	20	-2
Philadelphia Flyers	49.6	19	50.78	14	5
Colorado Avalanche	48.6	21	47.63	21	0
Washington Capitals	48.5	22	46.21	23	-1
Minnesota Wild	48.8	20	47.42	22	-2
Florida Panthers	48.4	23	50.7	15	8
Anaheim Ducks	47.9	24	45.97	24	0
Columbus Blue Jackets	47.0	25	44.04	29	-4
Toronto Maple Leafs	46.5	26	44.8	27	-1
Edmonton Oilers	46.3	28	45.67	26	2
Tampa Bay Lightning	46.3	27	43.89	30	-3
Nashville Predators	46.0	29	45.92	25	4
Buffalo Sabres	45.6	30	44.07	28	2

(updated 3.17.2013 at 12:12 AM PST)

Methods

The model is a monte carlo simulation of the remainder of the season. I simulate somewhere between 10 and 20 thousand seasons, recording the points and standings position of each team every season. Based on these results I calculate the “playoff probability” which is the percent of sims that a team finishes 8th or better in its conference.

There are a lot of ways to simulate the remainder of the season. I choose to use score-adjusted Fenwick, as it it is best predictor of future success. I grab Fenwick by score data from btn, adjust it for score effects, then regress to the % talent based on the average number of games the league has played thus far. I generate Fenwick For and Against from this proportion based on league average Fenwick events per game.

For each simulated season I give each team a shooting and save percentage randomly, based on a normal distribution expected by chance for the remainder games, and integrate this into their current shooting and save percentage as necessary (at even-strength, this isn’t necessary). This gives me a goals per game, and goals against per game for each team for each season.

For the remaining games I generate a end regulation score using a Poisson distribution based on the teams’ goals per game and goals against calculated from above, adjusted by their opponent’s goals and goals against per game. For tie games, I randomly choose a winner. For games within 1 goal, I adjust for empty net by turning 1 goal games into tied games 12% of the time.

I’ve simulated enough seasons to ensure that the average points per game for remaining games comes out to about 1.118 over a sufficient number of seasons, and the distribution of points follows a similar distribution based on actual data.

The variables that really end up mattering in the model are games remaining, whether or not those are home vs. away, opponent’s remaining, and of course SAF. In total home/away games remaining ends up mattering the most because generally teams are huddled around the mean for SAF, and the strength of the opponents remaining usually ends up being pretty similar.

An important point about this method, is that I have not formally evaluated the sensitivity and specificity of the model. After doing this for awhile, this approach seems the most realistic (with the exception of excluding PP and PK situations, OT scoring), and gives results that most closely align with the point distributions expected..

Western Conference

Eastern Conference

Score Adjusted Fenwick%

Talking Points