Tuesday, April 12, 2011

Predicting the Playoffs is Simple

With the NHL Stanley Cup Playoffs kicking off tomorrow, many people are sitting down with a pencil and a bracket today, myself included. And as I make my best Nostradamus impression, I keep telling myself those pieces of advice we tend to repeat in these times. That defense wins championships. That the Sharks will choke. That it’s about the match-up not the seed. That the seed is meaningless.

Well, that isn’t entire true. The Sharks part is, but the seeds aren’t exactly meaningless. They just aren’t all that great either. Over the last 4 post seasons the higher seed has only won 36 of the 60 matchups, or 60%. Hardly impressive.

But what if there was a statistic just sitting out there that was more accurate in predicting series winners? Even more accurate than a particular crab-eating macaque monkey?

It turns out – there is. Considerably more accurate.

Like how about going 50-10 over the same last 4 postseasons?

83.3%.

That’s only missing 16.7% of the time. You cannot even get dentists to agree on a toothpaste with that kind of consistency.

How about correctly predicting 15 of the last 24 upsets?

How about going a near perfect 14-1 in 2008?

Maybe you know about Simple Ranking System, even if you aren’t quite sure how it works. I doubt you were aware of how spot-on it has been predicting postseason series winners.


SIMPLE RANKING SYSTEM

Who exactly concocted SRS is unknown, but Doug from Pro-Football-Reference introduced it to a wider audience almost five years ago in a blog post.

Despite its name, even in a 16 game NFL season it isn’t that simple to figure out. An 82 game NHL season is even more complex. Luckily since its growth in popularity in the stat-heads world it has spread to other sports and a sister site at the Sports Reference Network, Hockey-Reference.

Doug does a better job explaining it, but it works like this:

If the Penguins are playing the Lightning, who normally allow 2.85 goals per game, and the Penguins score 3 times, they are +.15 (3-2.85).

If the Penguins allow 2 goals in the same game, when the Lightning are averaging 2.94 goals per game, the Penguins are another +.94 (2.94-2).

So on the game they are +1.09.

Over the course of the season you do this for every game and average them for a team’s overall SRS rating.

Essentially it judges how a team does factoring in how they are expected to do. Something like grading each game performance on a curve based on the competition.

What caught my eye, as I looked for a leg up in the yearly email playoff pick ‘em I do with my buddies, was the Flyers SRS last regular season. For a 7th seed, it was high. Second highest in the East.

It got me thinking just how well SRS did at predicted playoff series. And as it turns out, pretty damn well.

All SRS ratings come from the wonderful people at Hockey-Reference. They are awesome.

2010

SRS went 12-3 during the post-season, outperforming seeding which was an unimpressive 8-7. It predicted 4 of the 7 upsets, including Philadelphia’s first round upset of the 2nd seed Devils and their 2nd round victory over the Bruins.

The highest SRS rating in the NHL belonged to your eventual Cup champions, the Chicago Blackhawks.


2009

The SRS ratings once again out-did seeding, 11-4 to 9-6.

SRS correctly predicted each time a higher seed beat a lower seed, while also predicting 2 of the 6 upsets, including the Pittsburgh Penguins 7-game series winner over the Capitals.

The Penguins, who had the second highest SRS in the East, won the conference.

The Detroit Red Wings lead the West in SRS and won the conference.


2008

This was the year SRS ratings really nailed the bracket, going 14-1, predicting 5 of the 6 upsets. It also correctly predicted both Conference Champions, and the Stanley Cup Champion.

The higher seed was only 9-6.


2007

Once again it did a wonderful job predicting upsets, hitting 4 out of 5, and going 13-2, compared to seeding which was 10-5.

And once again it did magnificent job suggesting who would be competing for the cup in June.

The Ottawa Senators, who lead the East in SRS, won the conference. The Ducks, who were number 2 in the West, took the conference and the Cup.


2006

This by far was the worst showing for SRS. Which actually isn't all that surprising. I’ve written before about the once-in-a-generation bizarre 2005-06 season.

Left was right, up was down, all 4 top seeds in the West lost in the opening round and Jonathan Cheechoo, now a member of the Worcester Sharks of the AHL, won the Rocket Richard Trophy with 56 goals.

SRS correctly predicted only 2 of the 7 upsets, and merely match seedings, both going 8-7.


WHAT ABOUT 2011?

So what does this mean for this postseason? What does SRS project?

Well, nothing too drastic, save for the Rangers knocking off the top seed in the East, the Capitals, in the opening round.

Do I agree? Probably not (although it should be noted that obviously SRS has done a ridiculously better job than I have over the past couple years).

I understand it though. Despite the divide in points and seeds, the Capitals have only won a few more games than the Rangers. Their Goals For and Goals Allowed are actually quite similar. The biggest difference is that the Caps have lost in OT, picking up the extra points.

What SRS can’t see is the fact that Ryan Callahan won’t be suiting up for the New York.

Other than that, like many experts, SRS predicts the Vancouver Canucks to be your 2011 Stanley Cup Champs, with the Boston Bruins playing runner-up.


EAST

1st Round
Rangers (8) over Capitals (1)
Flyers (2) over Sabres (7)
Bruins (3) over Canadiens (6)
Penguins (4) over Lighting (5)

2nd Round
Flyers (2) over Rangers (8)
Bruins (3) over Penguins (4)

Conference Championship
Bruins (3) over Flyers (2)


WEST

1st Round
Canucks (1) over Blackhawks (8)
Sharks (2) over Kings (7)
Red Wings (3) over Coyotes (6)
Predators (5) over Ducks (4)

2nd Round
Canucks (1) over Predators (5)
Sharks (2) over Red Wings (3)

Conference Championship
Canucks (1) over Sharks (2)


STANLEY CUP FINAL

Canucks (W1) over Bruins (E3)


WHAT ABOUT THOSE SHARKS?

Have the Sharks choked these past few years? According to SRS, except their 2009 first round loss to the Ducks, not really. It predicted the Hawks beating them in the conference champions last season, their loss to the Stars in 2008, and their loss to the Red Wings* in 2007.

* Originally read "loss to the Predators". Thanks to M4ck for the correction.

11 comments:

Anonymous said...

Is there a way you can tweak this to take into account injuries or trades or late-season trends? Because it's clearly not the same team over 82 games.

If you could add in some of those variables, you might be able to increase your accuracy further.

Anonymous said...

Could this be more accurate if only the second half of the season was taken into account? I would guess that Buffalo's rating is far higher in the second half of the sesason than it was in the first half.

Anonymous said...

First of all this is frigging awesome and one of the coolest new stats I've read about in a long time.

I do have a question about how this is calculated. In your example where Pittsburgh plays the Lightning, and the Lightning are allowing 2.85 goals a game, is the 2.85 used the figure for the Lightning at that point in the season, or is it based on their final goals allowed at the end of the season?

Justin said...

Good question. It is based on the end of the season averages.

Justin said...

Absolutely, I just stumbled on this early this morning, but I'm definitely looking on a system that puts more weight on recent games.

Anonymous said...

"If you could add in some of those variables, you might be able to increase your accuracy further."

Statistical models don't work like that. More variables makes it easier to fit past data, but not necessarily predict future outcomes.

Anonymous said...

In terms of predicting playoff outcomes, it might be interesting to recalculate the SRS to only include the teams that made the playoffs. That would eliminate some of the statistical bias that comes from having an easier schedule.

Justin said...

That's the best thing about SRS. Because Strength of Schedule is such a strong factor in the statistic, there is no bias towards team with easier schedules (i.e. the Eastern Conference)

kosagm63 said...

What about mining the data from later in the season. It is often said that the team that plays the best 2nd half and plays well in April and May win's the cup. Apply the same SRS to the second half of the season or say, since the trade deadline and see how SRS performs as a predictor.

CoachZ said...

Can we get a new prediction for Round 2 in the east?

Bookkeepers Sunshine Coast said...

I watched every game because it was very entertaining game. It's really hard to predict what team to go in the playoffs but there are teams which dominate the game.