HFBoards (http://hfboards.hockeysfuture.com/index.php)
-   By The Numbers (http://hfboards.hockeysfuture.com/forumdisplay.php?f=241)
-   -   What Variables Predict Playoff Success? (http://hfboards.hockeysfuture.com/showthread.php?t=1261987)

 Chalupa Batman 09-19-2012 10:23 AM

What Variables Predict Playoff Success?

Given a playoff series between Team A and Team B, what variables can be reliably used to estimate Team A's chance of winning the series (prior to the start of the series)?

You'll hear a lot of things bandied about by pundits (usually in early April). Some off-the-cuff thoughts:

Relative seeds?
Points earned in the regular season? (In the season prior? In the season prior to that?)
Goals for / goals against in the regular season? (In the season prior? In the season prior to that?)
Playoff success in the most recent N seasons?
Power play success / penalty-kill success?
Average age of the team?
Amount of prior postseason experience on the team?
Team A's record against Team B during the season?

Ideally, we'd be able to come up with a logistic regression of sorts that takes in all sorts of available information to predict how Team A and Team B would fare if they squared off in the playoffs. We could come up with an R^2 value (and other descriptive metrics) and improve upon those values.

Parallel to that, we could also look at what types of teams regularly outperform the predictions. Are these teams special in some way that would improve the model? Or are they just special?

This could also lead into insights about how to construct a team that's playoff-optimal.

 Chalupa Batman 09-19-2012 10:23 AM

This may or may not be a starting point.

From HOH, the weakest team to win the Cup in recent memory: http://hfboards.hockeysfuture.com/sh....php?t=1195909

Quote:

Originally Posted by TheDevilMadeMe (Post 50138971)
Regular season rankings of Cup winners since 2000

 team overall seed GF GA PP PK 2000 Devils 4th 4th 2nd 7th 3rd 4th 2001 Avalanche 1st 1st 4th 3rd 3rd 19th 2002 Red Wings 1st 1st 2nd 3rd 2nd 8th 2003 Devils 4th 2nd 14th 1st 30th 1st 2004 Lightning 2nd 1st 2nd 10th 16th 10th 2006 Hurricanes 3rd 2nd 3rd 18th 17th 19th 2007 Ducks 3rd 2nd 9th 7th 3rd 5th 2008 Red Wings 1st 1st 3rd 1st 3rd 8th 2009 Penguins 8th 4th 5th 18th 20th 8th 2010 Blackhawks 3rd 2nd 3rd 5th 16th 4th 2011 Bruins 7th 3rd 8th 3rd 20th 16th

Interesting to see two Cup winners with below average defenses (2006 Canes, 2009 Penguins), but no Cup winners with below average offenses. And only the 2003 Devils have an offense ranked below 9th.

Penguins and Bruins are the only Cup winners to not be top 4 teams in the league during the regular season.

2003 Devils' PP is the single biggest weakness of any Cup winner, though they made up for it with the #1 PK.

Obviously, such an analysis fails to take into account changes that a team made during the regular season.

The recent Kings (terrible offense in the regular season) broke some major trends

Quote:

Originally Posted by Hawkey Town 18 (Post 50141179)
Didn't have time to look up Shot Differential, but here's Goal Differential added to this.

 team overall seed GF GA PP PK G DIFF 2000 Devils 4th 4th 2nd 7th 3rd 4th 4th 2001 Avalanche 1st 1st 4th 3rd 3rd 19th 2nd 2002 Red Wings 1st 1st 2nd 3rd 2nd 8th 1st 2003 Devils 4th 2nd 14th 1st 30th 1st 6th 2004 Lightning 2nd 1st 2nd 10th 16th 10th 3rd 2006 Hurricanes 3rd 2nd 3rd 18th 17th 19th 6th 2007 Ducks 3rd 2nd 9th 7th 3rd 5th 6th 2008 Red Wings 1st 1st 3rd 1st 3rd 8th 1st 2009 Penguins 8th 4th 5th 18th 20th 8th 9th 2010 Blackhawks 3rd 2nd 3rd 5th 16th 4th 2nd 2011 Bruins 7th 3rd 8th 3rd 20th 16th 2nd

Goal Differential appears to be one of the better indicators as to who is a contender, with only one team finishing lower than 6th.

Quote:

Originally Posted by imayagainknowanson (Post 50177103)
Let's see performance in the playoffs.

 team G/ G G/ A PP PK S/G SA/ G 5-5 F/A 2006 Hurricanes 2.92 2.40 24.0 85.4 28.8 27.0 1.06 2007 Ducks 2.76 2.14 15.2 86.8 30.5 27.6 1.42 2008 Red Wings 3.27 1.86 18.9 85.7 36.5 23.6 1.62 2009 Penguins 3.29 2.67 20.6 83.3 32.5 29.0 1.09 2010 Blackhawks 3.54 2.82 22.5 83.3 31.8 29.6 1.22 2011 Bruins 3.24 2.12 11.4 84.4 31.9 31.4 1.82
No one can even catch Canes killer powerplay game. Also, they have better PK than Penguins, Bruins and Hawks. And only that powerhouse Red Wings team gave less shots per game than Canes.
Most of the teams are better on paper mainly because Carolina defense has no sexy names but on ice everything was exceptional for team from Raleigh.

Red Wings
Ducks/ Hurricanes/ Blackhawks
Bruins/ Penguins

 Chalupa Batman 09-19-2012 11:08 AM

Very interesting - thanks! I missed this originally, and it's definitely a starting point.

The rankings given make me wonder what will be more predictive - raw metrics (such as power-play percentage), or relative metrics (such as a team's rank in terms of power-play percentage). My layman's guess is that the relative metrics will fare better (as they normalize for the league) although we'd want to note outliers as being particularly exceptional (positively or negatively).

(And I'm still below my quota of five parentheticals per post, so here's a fifth).

 Hockey Outsider 09-19-2012 12:22 PM

I'll speculate that regular season points have become less predictive since the advent of the shootout. The shootout determines the outcome of many regular season games, but they don't determine the outcome of playoff games.

I would guess that a model that excludes shootout wins/losses (perhaps treating them all as ties) would be a better predictor of playoff successthan a model that uses total points.

 wgknestrick 09-19-2012 12:41 PM

I remember reading analysis that stated that GF had a higher correlation than GA. I would also think team Corsi post ASG would be another high correlation metric (ie LA Kings this year). I think the key is to limit the sample period to be both large enough to provide accurate data, but also small enough that it captures a "real time" snap shot of how the team is currently playing. In general, I'd think post ASG is a great period to sample.

True factors:
anything with goal differential (5v5)
anything with shot differential (5v5)

Noise (minor helpers):
special teams
team SV%
Points
Net Penalties

 Beef Invictus 09-19-2012 01:27 PM

It would be interesting to look at the last month or two of the season to see who was best down the stretch and into the playoffs.

 reckoning 09-19-2012 10:19 PM

This has long been on my to-do list of stats I want to research, but know I'll never find the time for.

But one factor I've noticed over the years that I suspect matches up well with playoff success, is to remove one-goal games from a teams record, then just look at their record in the other games. The logic being that close games are often decided by breaks, and those will even out over time. The other games are a better indicator of how good a team is. It's similar to Pythagorean, but without putting too much weight on blowout games.

For example, last year the Kings had a terrible record in one-goal games (17-29, 27th overall), but had a very good record in the other games (23-13, 4th overall) . The other two "surprise" Cup winners , the '86 Habs and '95 Devils, had similar disparities in their records in those two categories; though who knows if any trends could be found in a long-term analysis.

Other factors that I suspect (but can't prove) may be better than just looking at total points:

- a teams record in the second half of the season.
- a teams GF/GA ratio at even-strength.
- in addition to a teams PK%, look at how often they are shorthanded. The logic being that the best way to stop a strong powerplay is to limit their powerplay opportunities.

 overpass 09-19-2012 11:09 PM

Quote:
 Originally Posted by Hockey Outsider (Post 54445531) I'll speculate that regular season points have become less predictive since the advent of the shootout. The shootout determines the outcome of many regular season games, but they don't determine the outcome of playoff games. I would guess that a model that excludes shootout wins/losses (perhaps treating them all as ties) would be a better predictor of playoff success than a model that uses total points.
The LA Kings were 9-15 in the regular season in games that went to overtime, all of which were decided in a format that doesn't apply to the playoffs. One reason among others why they were better than their regular season record.

I think any 4-on-4 and shootout results should be ignored for the purposes of playoff analysis. The Kings won 31 games, lost 27, and were tied in 24 games while playing under playoff-style rules. They won 9 and lost 15 under regular season rules.

I wonder if record versus playoff teams is more predictive of playoff success. Haven't done the research, but I know LA performed pretty well against playoff teams last regular season. They didn't beat up on the weaker teams like Vancouver and St. Louis did.

Trades at the deadline make playoff vs regular season analysis difficult. Going back to the Kings again, it appears that trading Johnson for Carter was a real upgrade that showed up in the playoffs. See also Edmonton picking up Dwayne Roloson in 2006.

 Czech Your Math 09-19-2012 11:24 PM

I agree that SO and probably OT points shouldn't be counted, since it differs from playoff hockey.

Seems like this would be a prime candidate for regression, with many possible variables (including lagged variables for playoffs).

There's some things that just aren't going to be predictable. Sure, teams improve at the deadline, but they also win the Cup or make the finals and then miss the playoffs or lose in the first round the next year... or don't come close to repeating that success in future years, despite having a similar core of players. Sometimes the stars just align and everything clicks.

Still, it would be an interesting and worthwhile study.

 Czech Your Math 09-19-2012 11:30 PM

Quote:
 Originally Posted by wgknestrick (Post 54446017) I remember reading analysis that stated that GF had a higher correlation than GA. I would also think team Corsi post ASG would be another high correlation metric (ie LA Kings this year). I think the key is to limit the sample period to be both large enough to provide accurate data, but also small enough that it captures a "real time" snap shot of how the team is currently playing. In general, I'd think post ASG is a great period to sample. True factors: anything with goal differential (5v5) anything with shot differential (5v5) Noise (minor helpers): special teams team SV% Points Net Penalties
You mentioned most of the factors. Points and/or wins (excluding SO/OT games) need to be included. Some sort of lag factors for playoff series (playoffs series in season before, maybe even further back) should also be used.

I think you're right. Despite the old adage that "defense wins championships", offense seems to be at least as good of an indicator as defense for playoff success (as far as winning the Cup).

 seventieslord 09-20-2012 12:22 PM

Quote:
 Originally Posted by reckoning (Post 54459993) This has long been on my to-do list of stats I want to research, but know I'll never find the time for. But one factor I've noticed over the years that I suspect matches up well with playoff success, is to remove one-goal games from a teams record, then just look at their record in the other games. The logic being that close games are often decided by breaks, and those will even out over time. The other games are a better indicator of how good a team is. It's similar to Pythagorean, but without putting too much weight on blowout games. For example, last year the Kings had a terrible record in one-goal games (17-29, 27th overall), but had a very good record in the other games (23-13, 4th overall) . The other two "surprise" Cup winners , the '86 Habs and '95 Devils, had similar disparities in their records in those two categories; though who knows if any trends could be found in a long-term analysis. Other factors that I suspect (but can't prove) may be better than just looking at total points: - a teams record in the second half of the season. - a teams GF/GA ratio at even-strength. - in addition to a teams PK%, look at how often they are shorthanded. The logic being that the best way to stop a strong powerplay is to limit their powerplay opportunities.
This sounds like just using goal differential. Which I think is as strong an indicator as anything. Not sure if it's better or worse than goal differential, but I think the results by either method would have to be pretty similar.

 Hammer Time 09-21-2012 02:24 PM

Coaching changes may also be a major variable that impacts the quality of a team.

2009 Pens -
Therrien: 27-25-7, 46.7% Fenwick
Bylsma: 18-3-4, 52.8% Fenwick

There was a significant improvement in the team after Bylsma's hiring. Using full season stats would probably have underestimated the Penguins' chances.

On the other hand, if you just use the stats for the 25 games after the coaching change, you would be using a small sample size with what is probably an unsustainable winning percentage.

Perhaps there's a middle ground, such as valuing games played after a mid-season coaching change more than those before it?

 typicryptic 09-21-2012 03:37 PM

What I suspect but, like the other poster, haven't proven, is that the GF/DA differential is important; however you also need to consider the standard deviation of the differential. That is, a team with a lower standard deviation from the average goal differential is more likely to have playoff success than one which has a higher standard deviation (implying that the average is more volatile in that there are more blowouts to cover heavier losses rather than consistency).

 Analyzer 09-21-2012 05:27 PM

Quote:
 Originally Posted by Taco MacArthur (Post 54443363) Team A's record against Team B during the season?
This means nothing.

5 on 5 scoring is key imo. PPs are rare for the most part, so you have to be able to score 5 on 5, especially if the game heads into ot.

The next thing I'd look at are impact players. Players who have a tremendously good shot and some skill to get them room. Players with terrible shots can win it too, but the players with amazing releases are always a threat.

Goaltending. If the goalie has been hot lately, or is just naturally skilled.

A team just making it into the playoffs could be more of a threat to anyone in the 1st round than the top team, as this bottom seed team has been playing playoff hockey for 2-4 weeks now, while their opponent will just be starting.

A player who catches fire. If you get a goalie, like say, Halak and a player, say Cammalleri who catch fire at the same time then teams are in trouble. You have a goal scorer to fill the opponents' net and a goalie to keep his net empty.

There's which role player can step up and be that guy who score timely goals.

There's a coach who can see a problem and address it before it causes a problem.

There's other options and then there's just pure and simple luck.

 DL44 09-21-2012 06:09 PM

this info comes in handy for weighing of power rankings in-season.. will have to incorporate.

Quote:
 Originally Posted by Fish Invictus (Post 54447151) It would be interesting to look at the last month or two of the season to see who was best down the stretch and into the playoffs.
I was also going to suggest this. Getting hot down the stretch is a common expression; I wonder if it has any basis in reality as an indicator of playoff success.

 Kershaw 09-21-2012 08:30 PM

Quote:
 Originally Posted by Fish Invictus (Post 54447151) It would be interesting to look at the last month or two of the season to see who was best down the stretch and into the playoffs.
Kings and Devils both were two of the top 2 teams down the stretch.

But then again, Pittsburgh was THE top team down the stretch.

 green lantern 10-02-2012 12:45 AM

Quote:
 Originally Posted by Tortorella (Post 54501265) Kings and Devils both were two of the top 2 teams down the stretch. But then again, Pittsburgh was THE top team down the stretch.
Since the lockout, the presidents trophy has lost more first round series than they won (3-4) In fact the only one to survive the last round in the last 4 was the canucks which needed OT in the 7th game in 2011. Where am I going with this?
I dont have exact stats on it but generally the presidents winner clinches 1st overall or at least 1st in the conference with a few games to spare. And of course we know generally that means they get their black aces into games and rest up their big guns and starting goalie. Even if they dont, the players know the games are meaningless.
Now 3 of the four 8th seeds to knock out the presidents team made the third round with 2 of them making the finals. So a possible explanation is this; you have an 8th team which has been in playoff mode for weeks since it literally has been do or die recently.
So you have one team playing playoff games for weeks playing a team which has had exhibition games recently.

 Bonesy99 10-02-2012 12:53 AM

I am actually surprised that no one has mentioned health. To me there is maybe 4-6 legit contenders each year entering the playoffs. The team who stays the healthiest is likely to win. So to me i see variables around durability being very important.

Man Games Lost
GP over the past couple of seasons
Stuff like that which could really wear a team down.

The NHL is very well balanced now. The penguins could lose any given night, even though they may be the best team. Health is an important factor that comes in to play.

 Talks to Goalposts 10-02-2012 12:22 PM

Quote:
 Originally Posted by CanadianHockey (Post 54500485) I was also going to suggest this. Getting hot down the stretch is a common expression; I wonder if it has any basis in reality as an indicator of playoff success.
I believe that's been looked at and found to be a relatively weak predictor of success. Team's that are "hot" going into the playoffs get no real advantage from it.

Its best to use the whole season in most cases, in predicitive hockey stat terms even a single season is a relatively small sample size.

I think the same studies showed that post trade-deadline games had more meaning than earlier season games though.

If you are going to consider a major mid-season change in roster or coaching I'd say you should consider:

1. Only shot based data, goals and wins will be too "streaky" over half a season's worth of games.
2. Only massive changes in team strength. Something that demands attention rather than that which can be easily explained by normal variation.

 Feed Me A Stray Cat 10-02-2012 12:24 PM

Quote:
 Originally Posted by Talks to Goalposts (Post 54716809) I think the same studies showed that post trade-deadline games had more meaning than earlier season games though.

 Beville 10-02-2012 02:06 PM

 Talks to Goalposts 10-02-2012 02:58 PM

Quote:
 Originally Posted by Feed Me A Stray Cat (Post 54716845) This could potentially be explained by trade deadline acquisitions, I think.
That's the idea. Although injuries, coaching changes or breakout players could also be factors for a team.