Friday, March 07, 2008

Friday Roundup

Discussion time, peeps.

Rink Reader Kyle shot me an email the other day that I'd like to throw open to the group:
[Effusive praise omitted]

I have a question re: goalie stats. In many of your game previews, you refer to a goalie's winning percentage, goals against average, or save percentage against a particular team. Sometimes you even use career numbers against one club.

When you get the chance, can you explain to me the value in these numbers? With more experience as a baseball fan, I can see how this applies to a one-on-one matchup such as pitcher vs. batter. But it seems to me a goalie's performance on a given night is better predicted by his most recent outings, not his history against one team. Putting aside Huet's more accomplished set of defensemen in Montreal, don't other factors such as team penalties, winning streaks and injuries (and chance?) play more important roles in determining a goalie's success on a given night? And is it likely that a goalie would have a problems with the style of a particular opposing coach or a particular opposing forward? I guess my concept is that while maybe Zdeno Chara does well in his individual matchups with Ovechkin (save Monday night), it is that type of individual skater matchup that better accounts for how many goals are scored in a game (regardless of goaltender, who sits in his crease and tries his best to see the puck). You also have clubs like the Caps who have changed personnel and philosophies during the course of the season, or just off nights from goalies (difficult to predict a goalie's future success from one or two showings).

[More over-the-top praise omitted]
[Still more embarrassing kudos omitted]
Below is my response to Kyle, but I'd like to know what everyone thinks on the topic:
I pretty much agree with everything you're saying - using career records against an opponent as some sort of predictory device is dubious, at best. The sample size is usually tiny, and even where it's not, the personnel changes and small differences between "good" games and "bad" ones can be misleading (and if a goalie gets absolutely shelled once - he'll never get his numbers back down).

But these numbers can sometimes creep into a team's psyche and they start thinking that the only way they can beat a guy is to be perfect - which leads to overpassing, shots wide, etc. Roberto Luongo a couple of years ago is a perfect example. Also, players and even teams develop tendencies, and if an opponent picks up on them, it can make a goalie's job that much easier. And, of course, with teams playing each intra-Divisional opponent eight times a year, these stats can take on some significance.

At the end of the day, I just love stats and all the various splits and permutations. They're fun. But they don't tell you what's going to happen in any given game. After all, games are played on ice, not on paper.
So what do you all think?

On a mostly unrelated note, if you were Bruce Boudreau, how would you handle your goalies this weekend, with afternoon games in Boston on Saturday and at home against the Pens on Sunday?

Elsewhere 'Round the Rinks:

Make sure to take a look at the stuff I posted "after hours" yesterday - just scroll down past this post.... Here's another Mirtle poll for you guys not to vote in.... Sergei Fedorov knew he was a Cap before he was a Cap. Huh?... I've pinned the "Milestones" post under the Quick Links sidebar so it's readily accessible, and please feel free to add any suggestions you might have in the comments. Oh, and next up? More goals than Glenn Anderson or Pat LaFontaine ever scored in a single season.... Just sayin'.... Happy 45th Birthday to former Cap Mike E-A-G-L-E-S Eagles.... One year ago today we recapped another Caps loss (that included a juicy Alex Ovechkin postgame quote), went free agent defenseman shopping, clubbed Matt Barnaby and looked at a pretty picture, and two years ago today we actually got to recap a Caps win.

Daily Awards
  • Hart: Erik Ersberg (40-save shutout win)
  • Ross: Bryan McCabe (4 points)
  • Norris: Bryan McCabe (Game-winning G, 3A, +2, 3 SOG, 3 BkS)
  • Vezina: Erik Ersberg (40-save shutout win)
  • Richard: Sean Avery, Nik Antropov, Jeff Carter (2G each)
  • Calder: Erik Ersberg (40-save shutout win)
  • Aiken: Dan Ellis (L, 3 goals allowed on 11 shots against in just 6:09 of work)

12 comments:

Charlie G. said...

Stats are fun...and often confusing. The stat I like most about Olie this season is his won-lost record over the past two months...not too shaby.

Rage said...

Basically you both are right. Splits are largely silly, but fun to look at.

However, I think the role of psychology is downplayed a bit in analyzing sports statistics. Streakiness doesn't exist statistically ("it's just regression to the mean!"), but it sure makes sense logically. Hasn't everyone had moments where they're just locked in (in terms of execution, regardless of results).

But yeah, the fact remains that the limit of P(Colton Orr being an ass) approaches 1 as ice time approaches 20 minutes. That's indisputable.

Rage said...

And holy crap, why are the Senators so terrible?

Paul Nichols said...

I'd start Olie on Saturday and Huet on Sunday - if he's available. I thought Johnson was going along at a good clip, but they haven't played him recently.

Or maybe Olie should start against the Pens, since he's the #1 and it's the NBC game of the week.

hotdog88gt said...

Goals and assists, wins and losses. That's about all I need to know. The rest of it is just BS. Goalies W-L records are always subject to whether his team scores enough goals to win.

Stats can spun to prove or disprove just about anything.

Here's more BS: over the weekend I met a guy named Randy and his son, Cole. It occurred to me the next day that my roommate during my freshman year in college was named Randy Cole!!!

NS said...

Olie vs Bruins
Huet vs Pens

JP said...

Here's a hypothetical:

All else equal (i.e. the team in front of the goalie), would you rather face a guy who is 10-0 with an .850 save percentage in his last 10 games or a guy who is 0-10 with a .950 save percentage in his last 10 games?

Does the first guy "just win"? Is the second guy "due"?

~Mark said...

This weekend:
Olie less, Huet more. (you do the math)

As for numbers versus a team... I think it's a grain of salt kind of thing, but usually gets tossed out once the puck drops anyway. Take our last 2 wins for instance.

What I really find useless is the Caps career record against [insert team name here]. What possible value is it that the Caps organization has won or lost against another organization when the current players weren't even alive during most of the years the stats were accrued? I just don't see the point.

Jan said...

I agree with JP when he says "But these numbers can sometimes creep into a team's psyche and they start thinking that the only way they can beat a guy is to be perfect - which leads to overpassing, shots wide, etc."

But the other side of that coin is the slight psychological edge that a goalie may have when he is, like 12-1 career vs. a particular team. You know: the age old saying that so-and-so "owns" Team X. And if the goalie feels confident, upon whatever irrational basis, still it's got to trickle down to his mates. Human nature and the nature of team play being what it is.

I am sure that they all know intellectually that that career-against-one-team stat doesn't amount to a hill o'beans, but no matter. An edge is an edge, regardless. If you think you're going to win, often you just DO win. The mystery and wonder of sport.

Eric said...

With my shooter(s), I'd rather face a guy who is 10-0 with an .850 save percentage in his last 10 games. My odds are better. He can "just lose" as easily as he can "just win".

Brian said...

Quote: "What I really find useless is the Caps career record against [insert team name here]. What possible value is it that the Caps organization has won or lost against another organization when the current players weren't even alive during most of the years the stats were accrued?"

Ok, well, it may be a different sport (as well as college), but ask Clemson basketball about playing in Chapel Hill, or the Terps playing football against Penn State, and the importance and meaning of those numbers.

Drew said...

You could isolate a cohort of current goal tenders with a certain amount of experience to drive up the sample size and then look at the predictive power of past stats to any given match up.

Would be an interesting exercise if the data were easy to get. But that approach would introduce bias problems (more tenured goalies = more lineup changes).

Just as everyone is pretty much agreeing here - you are likely to find these stats are not that good a predictor. The model would have a lot of outside factors that you cannot/would be really hard isolate, control or account for.

Recent performance - say over last 3 to 10 games against ANY team is likely better. BUT the issue is, is it marginally better or a lot better? Hard to tell.

The only folks I know of with the time, data savvy and motivation to review this live in the Vegas area. Call up a large sports book - I bet they (or their consultants) have looked at this as 1 possible factor in setting odds on hockey games.

Would be pretty fun to know what stats they DO use overall. Who has a hook up?