Monday, September 8, 2008

I'm a stat man



Man, I never thought I'd get a chance to link to that video. This is already a great day.

Anyway, I'm a touch behind on this (story of my life), but Staples has a fairly interesting post up about the battle tension between "traditionalists" and "stat guys," or those who, broadly speaking, prefer to rely on their own observations versus those who like to rely on numbers and shit. Lowetide has, as per usual, a very smart take up on it already, but there's a few things I thought were worth adding (and I wanted to encourage more people to go read it, because if nothing else, I think it's a useful primer to a debate that's entirely worth having).

Before I start though, I should point out, in case it isn't readily obvious, that I really like the new stats work that's being done around the internets. I'm pretty much a fan of anything that encourages people to think more, and a lot of the new stats shows, if nothing else, a real willingness for people to engage with hockey beyond the usual beer-and-bitch way. I certainly don't have the patience or inclination to come up with this stuff on my own—which makes me appreciate those who do all the more—but most of the stats I've come across have changed the way I look at hockey, like to the point that I'm trying to figure out my Corsi number in beer league games (incidentally, the Corsi numbers still say I suck).

Anyway, to get to my first point, as a stat guy, one thing that frequently bothers me is the above characterization, which kind of boils down to, in my mind, people who watch versus people who read and do math. Staples isn't the worst offender in this particular post, but he is drawing on a tradition that I think is just entirely unfair. There is a certain thread that basically casts "traditionalists" or "gut thinkers" as these kind of grizzled, sports-watching veterans, people who spend their lives tucked into a fedorah and a trench coat at the back of an arena, coffee spiked with JD, soaking up the kind of knowledge that can only come from freezing your ass off on bench seats; on the flip side, stat guys are latte-drinking metrosexuals, crunching numbers on their MacBooks (yeah, I have one: want to fight about it?) and barely even bothering to turn on the radio to catch a game, because what could watching hockey possibly tell you about it, when you have all these beautiful numbers?

Obviously, I think that's crap. I would actually argue—even if I'm biased—that the difference between traditionalists and stat guys is that one group is satisfied in the knowledge they have, and the other is inquisitive and wants to understand the game as best as possible. I wouldn't go so far as to say one approach is necessarily better than the other—it's a game, interact with it however you like—but that doesn't mean their isn't a profound, almost prideful ignorance in a statement like the one Staples attributes to Robin Brownlee: "I'm not the least bit interested in these numbers. I know what I see and I know what I think." Forget even if the new statistics are of any use: that's a man who loves hockey enough to make it his career openly saying he doesn't give a whit about anything that might disagree with his established way of thinking.

Not that I think the so-called "new stats" would. As Matt Fenwick likes to frequently point out, generally speaking if a stat disagrees greatly with what you see on the ice, it's probably misguided. I think us stat guys—or at least certainly myself, and I assume from my interactions with other people—use stats to help us confirm or refute our observations, so we can better understand what we're seeing.

Just as a personal example, I never liked Raffi Torres: I generally thought he was a marginally useful player, made far more popular then he deserved due to a penchant to take the body and a bull-headed attitude in the offensive zone. I guess there's no real harm in my thinking that, but after delving into some of the new numbers, I got a better understanding of him. Simply put, I'd say I was half-right: offensively, Raffi has some issues, but it turns out he was a fairly useful player, because he could manage to at least play the other team's quality players to about a draw. Anyway, though, the point is that I used the stats to further my understanding of what I was seeing, not as any kind of substitute.

(Which isn't necessarily to say that statistics aren't a substitute for watching the game: obviously a large reason why we have statistics at all is so we can tell what happened/what kind of players these are without having to watch every single NHL game in a season. Still, that holds true for old statistics or new: I would simply argue that the new ones are, again, trying to give us a more in-depth understanding than the old.)

Now, again, Staples' post is hardly one of the worst offenders out there for this, but there's still a definite undercurrent of this kind of experience-versus-computers mindset. Far from being guys who only analyze numbers, though, I think it's better to characterize them as people who both watch games and analyze numbers, versus people who just watch games. To put it another way, Staples says, "... scouts might rely on statistics to help them identify which players they should go see and focus upon, but the scouts always test the numbers with their own observations": conversely, then, I'd say stat guys always test their own observations with the numbers. I'll leave it to the individual to decide which method might be more effective, but I think it's fairly obvious where I stand.

The second thing I wanted to point out was Staples' characterization of Desjardins work. Though hee does praise it—as well he should, Behind the Net should be essential reading for anyone who wants a broader perspective of the game—Staples adds the following caveat: "Desjardins' work on quality of competition lead[s] the way [in new statistical analysis], even if it is based on the old botched concept of plus/minus."

This is mostly for the benefit of anyone who comes across Desjardins' numbers (and if you read the Oilogosphere, it should be fairly regular occurence), but to characterize the numbers he uses as based on "the old botched concept of plus/minus" is to misunderstand what Desjardins is doing. He does use a plus/minus system, but it's far more complex, and I would argue far more effective, system. Far from just taking the standard tally, Desjardins comes up with a player's plus/minus per 60 minutes based on a sum of the differences between the goals for and against when a player is on-ice and off-ice. Standard plus/minus only takes into account on-ice statistics, with no thought to either how the team performs without said player, nor how often said player is on the ice. Desjardins aptly named advanced plus/minus isn't entirely perfect, but it does factor in both per-minute contributions and a player's standing versus the rest of his team (the lack of the latter is one of the chief problems with plus/minus), which puts it far ahead of the old plus/minus. That's something that's important to consider, especially when we're debating the usefulness of certain stats.

Anyway, back to looking up '90s one-hit wonders on YouTube.

38 comments:

Addicted-to-oil said...

I too am a stats guy with a macbook, and I too am a guy that listens to/watches every game I am able to. I take into account what I see/hear on the ice, and the things that I don't catch watching/listening to the games. Stats show so much more about the game than just watching the game. You can't watch everything that happens during a game, so taking a look at some stats is a great way to enhance an already great knowledge of the game.

When it comes to some of the more advanced stats it's totally hit or miss. Some are great, some are crap, but that doesn't mean we should ignore them.
I'm constantly trying to come up with new stats for my blog.
One of the big ones I'm working on is finding an injury quotient so we can finally see how important a loss it is when someone gets injured on a team.
It's not going to be foolproof, but it could offer a broader view on something like injuries.

I think Staples is far too extreme in some of his views. I believe there's a lot of us out there who fall in a happy median where we watch the game and take a look at the stats.
Solely watching the game doesn't tell you the whole picture. Solely looking at stats doesn't give you the whole picture either.

LittleFury said...

What we have here is your basic tension between the Apollonian and Dionysian impulses that reside within us all and shit.

LittleFury said...

Also that White Town track stands up pretty well. Scatman less so.

DMFB said...

Yeah, that White Town stuff is the shit.

As for Scatman John, I have absolutely no idea how that ever became popular. I'd blame the mid-'90s swing craze, but even that doesn't seem close enough.

boopronger said...

I cant believe that the guy who sings scatman looks like that. Never would have guessed.

All this talk about stats guys and traditionalists, why dont some of you put your money where your mouth is and have a big competition. If each side thinks that they are right then just see who can predict more stats/standings for the upcoming season?

Dana J said...

I did a double take today when Jason Greggor said on the radio that he didn't believe in soft minutes. He fits into the traditionalist point of view, I guess, but it still blew me away.

Vic Ferrari said...

dana j:

Wow. I don't know if 'soft minutes' is really something Gregor can choose to 'believe in' or not. He can choose to ignore it, though.

Jason Gregor seems like a really nice guy, but I think this is more about being oblivious to a lot of things that happen in a hockey game than it is about being a traditionalist.

Vic Ferrari said...

dmfb

Nice post. I think it's easier for some people to embrace most of the stuff that MC, myself, Matt, Lowetide and a couple of others do.

Partly because that's their predisposition, and mostly because this stuff largely reaffirms the things they see in a hockey game anyways.

I mean if, when MC had come into the Oilogosphere and started splashing numbers around, if he had shown that Lupul and Comrie were real difference makers and that Dvorak and Pisani were even worse than they seemed, you probably would have shook your head and disregarded him.

And if you've ever listened to sports talk radio, you know that lots of folks like to talk about momentum changing goals, big saves at big times, etc., and they see some things in hockey games that I'm pretty sure never happened. Those folks aren't going to embrace this stuff, because inevitably it will fly in the face of what they know about the game.

Tyler said...

I did a double take today when Jason Greggor said on the radio that he didn't believe in soft minutes. He fits into the traditionalist point of view, I guess, but it still blew me away.

Far be it for me to steal an opportunity for Dennis to come here and tell this story but he's got a story about Gregor telling him that Matt Greene is better than Jan Hejda. Guy's a bit of an idiot, that's all.

Oilman said...

Vic,

I've never understood your stance on the "big goals", "momentum changing" stuff. Surely you can't think that "clutchness" in hockey doesn't exist? Not everything that doesn't fit into models (many of which I've come to appreciate over the past year or so BTW) can be tossed aside as luck can it?

Dennis said...

Even though he's no longer ours it's still nice to see Raffi getting some love.

I was as frustrated as the next guy about his inability to score and his inexplicable passing but the guy could keep his head above water playing the toughs and that certainly matters. BTW, not sure what Hitch has planned but I'll be defintely watching some early Colu games to see if he puts Raffi/Peca back together again. That was really the first time I realized just how valuable Raffi could be and since then I watched to see who he was playing against just to make sure he could keep it up.

One thing I'm reading more and more that holds true is that stats have to jive with what you see; of course, some guys really can watch games and not see anything and some folks aren't watching enough to really see anything in the first place.

Back when Marchant was an Oiler you had a tonne of guys talking more about the big scoring opps he blew rather than appreciate just who the guy was out there playing against. The same held true for Grier as well; albeit on a lesser scale. Later on, Horc became the new Marchant as the next supremely underrated player. There were all kinds of debates about Horc not being a real #1 pivot when folks didn't even have a clue what to look for.

Anyway, the whole thing of Hedja vs Greene was actually an argument I had with Gregor's producer, Will Fraser. I still can't get over the fact they let Hejda walk so one night I was bored and I called into Gregor's show and Fraser was at the helm and he basically told me that Foote was making Jan look so good in Colu and that Greene was by far the better dman and that if my idea was correct that in the summer of '07 the Oilers chose to bet on Matt over Jan, the Oil choose correctly.

The thing with Staples is that he's so dismissive and/or hesitant to embrace anything else other than his error stat because he has a healthy ego and he wants to be the man. There's nothing wrong with that other than the fact he's betting on something that's soaked in objectivity and as someone else said the other day on some other blog, there are lots of "errors" made in a game and sometimes the goalie bails you out and sometimes they don't.

You need scoring chances and GFON and GAON bathed in the light of matchups and that's something I might be able to attempt this winter if something pans out for me that allows me such leisure time.

I tried it out last year for a home game vs the Stars where I counted scoring chances and listed the forwards on the ice for and against and if I can do that this year while nailing all players on the ice on both sides, then I think that's a better gague than just who makes an error on a play that winds up resulting in a goal.

Vic Ferrari said...

That would be terrific Dennis.

If you just record the time on the clock, which team got the chance, and the quality of the chance (by some simple metric, 10 cents and 25 cents or something else) then it would take nothing for me to put a script online that takes those numbers and churns out the scoring chance +'s and -'s.

I may very well lean in the other direction too far, but a guy needs to count the 10 cent scoring chances as well as the five bell types imo. Otherwise Maurice/Babcock teams are always going to be devalued a touch, and Quenneville/Crawford teams are going to be flattered methinks.

Just as well that you didn't do it last season though. The Oilers won a lot of games that they were badly outchanced in, and managed to find ways to lose most of the games that they had the better of the chances in (often due to that crappy PP mid-season). Just the way the cookie crumbled. So you would have had 90% of message board kids pointing at your numbers and concluding that getting outchanced is a good thing. Yeesh.

The dice have no memory though, this season should be good that way for 80% of the teams, and the Oilers have an 80% chance of being one of those squads.

BTW: I'm sure I said this before, but Jeff Ward was talking about the minor league Oilers, and he said that the Oilers keep track of scoring chance +/- at evens. They weighed that strongly.

Kent W. said...
This comment has been removed by the author.
Kent W. said...

conversely, then, I'd say stat guys always test their own observations with the numbers.

That's about where I land. I also find stats useful to help evaluate guys I can't watch a lot. I love hockey and obsess over it, but even I can't be watching every single game every night.

In addition, the eyes can lie...no matter how thoroughly "experienced" a veiwer is. By way of illustration:

I had a thing against Anders Eriksson last year - because I thought the guy was awful almost every time I watched him play. A senior member on a Flames messageboard disagreed with me - he seemed to think that Eriksson was being unfairly maligned by the various and sundry (because almost everyone with two eyes last year saw that Bubba was drowning most nights).

Anyways, this guy basically made it his years mission to shield Eriksson from criticism. He accused anyone who spoke out against Bubba of being either a victim of group think or just generally uneducated. His conviction was so great that I went to the numbers - both traditional and new - to see if I indeed was out to lunch.

Turns out, Eriksson DID suck last season by just about every measure. Which, of course, made sense to me. The adversary in question, however, would hand-wave away any sort of numerical evidence that was presented to him. That was all just mumbo-jumbo to him because he *saw* it otherwise. He'd grown so convinced of his position that nothing could shake him of his conviction.

Bruce said...

You need scoring chances and GFON and GAON bathed in the light of matchups and that's something I might be able to attempt this winter if something pans out for me that allows me such leisure time.

Sounds promising Dennis. Of course you can already get GF/GA ON from the summaries, though they are bathed in the exact same light as Errors; a goal has to be scored for those match-ups to register. Those many other plays where the goalie makes the stop to thwart the opportunity and/or cancel the gaffe -- or where the first baseman pulls down the errant throw and makes the tag -- will require a lot more detailed film study. More power to you if you want to evaluate all scoring chances, but it will be a HUGE undertaking.

a guy needs to count the 10 cent scoring chances as well as the five bell types imo.

I agree with this, Vic, with the obvious caveat that the ranking of the scoring chances will be itself a very subjective measure which (again like Errors) can be defined only imperfectly and will rely on the observational skills, hockey knowledge and experience of the "scorer".

Vic Ferrari said...

Bruce:

Scoring chances is a means to an end, not the end in itself. If we do this we'll see what it meshes with.

Errors (whatever the hell they even are, I've never been a Staples reader) and scoring chances are both subjective, leading some people to think that they somehow equate.

In fact if you honestly track scoring chances for a game, and most people who have coached probably do so automatically anyways, just subconsciously it will take you down a different path altogether. And it will be surprising how often scoring chances and shots don't happen at the same time at all. Yet, as if by magic, at the end of the game the shot clock is usually really representative of these scoring chances.

I think that's because possession is driving both.

Now every team is different, just nowhere near to the extent that most hockey fans seem to think.

And unless a person is so lost in the magic that they can't be saved, in short order they'll realize that there is no such thing as a rope-a-dope coaching strategy. Just that sometimes you roll a whack of sevens even when you get your assed kicked in trivia.

Or so I think, Bruce. It's worth a go, I hope we draw you into it for a few games early next season, and I hope you do it honestly. It's a bit tedious, and you feel like the biggest geek in the world when you do it, so make sure your wife and kids aren't around.

The path to reasonableness is paved with scoring chances, Bruce.

Dennis said...

Vic: Bruce has hitched his wagon to Staples.

And I can defintely do the clock thing as long as it's visible when I'm fastforwarding and rewinding.

Bruce said...

Scoring chances is a means to an end, not the end in itself.

Yes of course, Vic. They're just another (imperfect) measurement system. The only measurement system that matters is the "official" one which is measured along the following hierarchy:

Goals
Individual game wins
Playoff qualification
Playoff series wins
Stanley Cup wins

... and all else is secondary bragging rights.

I do however believe that any reasonable measurement of scoring chances would provide extremely important insights into the unit currency, goals, and it pisses me off no end how for years the commentators will occasionally throw a number at you from that period or that game but never ever does one ever get a chance to see any sort of comprehensive result. They would of course be subject to the same sorts of limitations as the RTSS numbers that are pretty widely criticized in the 'sphere. (I like them, accepting and accounting for their limitations, primarily everything is counted by zeros or ones.)

In that vein I do like your idea of five-and-dime weighting of scoring opps; it would be important to be consistent about such judgements whether a goal is actually scored or not. One could count both a player's raw conversion rate (points over scoring chances) and his "slugging percentage" (how many $$$ he banks on the scoreboard compared to the sum of all those 10 and 25 cent chances). Same thing in reverse for goalies and team defences; it might be a useful measuring stick for bad goals. Interesting possibilities here all around.

Errors (whatever the hell they even are, I've never been a Staples reader) and scoring chances are both subjective, leading some people to think that they somehow equate.

I don't know how you're in a position to suggest they do or don't "equate", given your professed ignorance on one of those measurements. Fact is there's a ton of stats that are subjective. It's a subjective business: goals themselves can wind up being the opinion of Colin Campbell, and lesser stats are the opinions of lesser officials in the league hierarchy. Ultimately a decision is made on hundreds of plays a game, whether something is a shot or a hit or a giveaway; and a hit is a hit for Jarret Stoll or Scott Stevens, even though one is far likelier to affect the actual game than the other. So maybe we should have 10 and 25 cent and whole-loonie hits while we're at it.

No reason why observers of the game who are willing to put in the time and effort shouldn't contribute their own measurements, be they who made the errors that led to the goals or who created the scoring opportunities that led to the goals. Either way it's bound to be subjective.

In fact if you honestly track scoring chances for a game, and most people who have coached probably do so automatically anyways, just subconsciously it will take you down a different path altogether. And it will be surprising how often scoring chances and shots don't happen at the same time at all.

I have watched the game like that since long before I ever coached, and I wouldn't be surprised at all should a weak relationship between scoring chances and shots be demonstrated. A scoring chance can be a hit post, a missed shot, a missed pass, broken up by a great defensive play, or a save. Or two or three saves, for the goalies that leave short rebounds(is that two or three scoring chances, or just one?).

Yet, as if by magic, at the end of the game the shot clock is usually really representative of these scoring chances.

I don't believe in magic, and I don't believe that conclusion either. Show me a season's worth of honestly compiled scoring chances before you make such a claim. Or better yet, ten seasons' worth. Fact is shots on goal are what we have, and they're better than nothing, but they don't "equate" to scoring or winning all that well, and I bet they wouldn't "equate" to scoring chances either.

I think that's because possession is driving both.

And I think that on balance the best scoring chances come against the flow of play. To oversimplify, if one team gets 20 * 10-cent chances and the other team gets 12 * 25-cent chances, who do you like to win?

Now every team is different, just nowhere near to the extent that most hockey fans seem to think.And unless a person is so lost in the magic that they can't be saved, in short order they'll realize that there is no such thing as a rope-a-dope coaching strategy. Just that sometimes you roll a whack of sevens even when you get your assed kicked in trivia.

I dunno, Vic, throughout the Dead Puck Era it seems to me that counterattack hockey has been pretty successful. I'm not sure that I'd call Jacques Lemaire's infamous trap or Scotty Bowman's celebrated left-wing lock rope-a-dope exactly, but they are both designed to counter possession and create opportunities going the other way. Don't even get me started on Dallas. On the "good" side of the ledger, the zone defence worked well for MacT against the President's Trophy winners in '06. And on the bad side, think of all those games between two trapping teams that were unfit to watch that we watched anyways.

For that matter the Lively Puck era was in large part a counterattack "transition" game. The Oilers certainly played it that way, I observed it and I got the shots on goal data to prove it. Can't speak to numbers of scoring chances, but if Wayne Gretzky was involved a scoring chance was usually worth 50 or 75 cents, let's put it that way. It was all about quality over quantity.

Or so I think, Bruce. It's worth a go, I hope we draw you into it for a few games early next season, and I hope you do it honestly. It's a bit tedious, and you feel like the biggest geek in the world when you do it, so make sure your wife and kids aren't around.

Sure, I'll be happy to give it a go for as long as it proves sustainable. I think it's a really important measurement. To be really useful it would also answer questions like these:

-- was the scoring opportunity created off a turnover or through possession?
-- was it created on the rush or generated entirely from the offensive zone?
-- did a goal result, or was it thwarted by a save, a defensive play, or an offensive miss?
-- did the team generating the opportunity gain the zone, recover a dump-in, or cause an O-zone turnover in the play leading up to the chance?
-- did they possess the puck for less than 5 seconds? 5-15 seconds? longer?
-- was the puck in the offensive zone for less than 5 seconds? 5-15 seconds? longer?
-- was the goal scored off a faceoff? 5-15 seconds after a faceoff? longer?

... and on and on. There's no end to information one could compile and which would be potentially useful. For one thing such might establish a more meaningful relationship between puck possession and puck position. The former may drive play, but the latter has its own effect on the scoreboard.

The path to reasonableness is paved with scoring chances, Bruce.

The path to wins is paved with goals.

Bruce said...

Vic: Bruce has hitched his wagon to Staples.

Dennis: I have hitched my wagon to considering new ways to measure, consider and interpret the game. Like it or not, that includes both you and Staples and a great many others who contribute to a greater or lesser degree. Rather than "hitch my wagon" to any one individual, I have freefallen within the Oilogosphere as a whole.

While I greatly value those opinions of others, surely I have demonstrated a capability of thinking for myself by this point.

When it comes to considering new ideas I call it being Open Minded. I recommend it.

Dennis said...

Bruce: if it's alright with you, I'll just keep on believing what the Edm media tells me!

DMFB said...

The path to wins is paved with goals.

I think what we're after is figuring out what the path to goals is.

Bruce said...

I think what we're after is figuring out what the path to goals is.

No duh. It's interesting, and "important" as any of this stuff goes. Any statistical light that can be shone on that process will provide insights into the game itself.

I'm just saying that no matter how sophisticated we become in finding relationships between flow of play and goals, errors and goals, Corsi numbers and goals, shooting percentage and goals, scoring opportunities and goals, etc., the "true" measurement system will remain the goals themselves. Everything else is a means to that end. Let's hope we can find one path which is more smoothly paved with reasonableness than another, but I'm guessing there is sufficient complexity here to collectively test our keenest intellects. Of which there are a few in this particular loop; I look forward to this discussion moving ahead as we (finally!) get into the new season.

Doogie2K said...

stat guys are latte-drinking metrosexuals, crunching numbers on their MacBooks (yeah, I have one: want to fight about it?)

So do I. Regular or Pro?

And if you've ever listened to sports talk radio, you know that lots of folks like to talk about momentum changing goals, big saves at big times, etc., and they see some things in hockey games that I'm pretty sure never happened.

The funny thing with those things is, I swear they happen more often than the stats guys give it credit for (almost never), but less often than sportscasters/radio callers like to think (pretty much every fucking goal, considering how few of the things there are). But as I noted on LT's last night, I'm the only one around here who'll give sports psych the time of day, so that's not surprising.

@Dennis: It still strikes me that scoring chances no less subjective than errors, at least in terms of sorting out the quality of the chance, and that the concept of tracking who fucks up more often on plays has pretty obvious merit (I'd be baffled if coaches didn't mentally track that as sure as they did chances). Given what you said this morning -- "there are lots of 'errors' made in a game and sometimes the goalie bails you out and sometimes they don't" -- if it were practical, would you be more amenable to the concept of errors resulting in scoring chances against, rather than just goals? It could be as simple as "Smid missed his check but Garon got across in time" (none of this primary/secondary stuff; too much work, and the most important mistake is usually pretty obvious anyway), and can be figured at the same time as who was on the ice for the chance. Again, it's not like we don't do that anyway just in the course of watching and commenting on the game.

Oilman said...

-- was the scoring opportunity created off a turnover or through possession?

As a nut-job goalie, I'd be willing to bet that at even strength, the scoring ops created off turnovers have a higher completion percentage and vice versa when on the PP...which of course would fly in the face of possession driving results at EV. I'd think that offensive zone turnovers have as high a correlation to winning as any other stat does..I know they kill me.

Oilman said...

that of course should be defensive zone turnovers...

Ender said...

I'm going to start out by agreeing with bruce. I understand that Dennis and Vic might just skip my post because I said that, but hey, we were talking about not ignoring things that might enrich your view of the game, right?

Let's be anecdotal for a moment. I remember one game the Oilers ('86 or so) were playing the Bruins. The Bruins got out of the first period up 10-3 (or so) in shots, but the Oilers were up 1-0 or 2-0 (Sorry, can't be bothered to look it up). I also remember watching the '06 playoffs cringing in terror every time Spacek (Same as Grebeshkov, early this season) touched the puck and screaming at Roloson to get into position. As far as I know, I'm the only one who was doing these things. Yes, Spacek was ok in the attacking end, but anywhere else to my eye he was a liability. Roloson was almost cheating to one side of another, most of the goals against were short-side goals, but he had a great SV%. Why? Because he got a barrage of shitty shots because the defense held it together.

My point? Shots on goal do not equate to winning games. They equate to puck possession. Puck possession does not win games. Goals do.

So, from the get-go, I throw SOG and SV% out the window. But Goals against average depends on the quality of the shots. Well, to check that, as Dennis said, scoring chances would come into play. But as Vic said, that would make Detroit look worse than they are (due to qualifying what counts as a scoring chance. So, unless Detroit is worse than they are, scoring chances are out the window.

The bottom line is that in talking about any specific stat, we find that certain styles of play skew the stat to the point that you don't have a strong enough correlation to argue causality.

Now the error stat. This at least follows the "what makes the puck actually go into the net" idea that I've already pointed out, you know, wins games. Sure, it's biased. Sure, it's not mathematically sound. That said, it's not hard to see that it's the only stat that I've seen so far that directly follows the path of the puck into the net. It's more subjective than +/-, but it's more focused, and focused is what we're looking for here. Does it need work? Probably. Would it be aided by a bunch of people all evaluating errors on plays? Definitely.

Hell, anybody can use any stat to try to argue that their POV is correct. Unfortunately, without doing actual statistical tests (like Chi-squared tests) you're not going to get an answer that's meaningful to anybody that either doesn't already agree with you or already respects your opinion. It's all rhetoric, and passing off your mathematical prowess (which, given the math I've done in university doesn't really strike me as any particular prowess at all) as fact or at least justification for your eye is ridiculous.

Seriously, I don't care if you want to use stats to try to reinforce ideas in your head. Fine. Just understand that you're reinforcing and rationalizing. This, folks, ain't math.

Ender said...

I should also add that "almost cheating" should be "always cheating" and that I started this argument, not Staples, right here: http://www.stillnoname.com/posts/show/574

NotLeeFogolin said...

I skim a few 'sphere blogs, watch as many games as I can and consider myself an inquisitive sort of person. But whenever I load up Behind The Net, thinking today might be the day I delve in, I'm reminded of how ignorant I am in western history (or some such pail of knowledge), become distracted and agitated, and immediately switch activities. Same when I read blog comments that are longer than five lines.

Dennis said...

___Given what you said this morning -- "there are lots of 'errors' made in a game and sometimes the goalie bails you out and sometimes they don't" -- if it were practical, would you be more amenable to the concept of errors resulting in scoring chances against, rather than just goals?____

D2K: This is what I'm trying to get at. Goals are the ultimate of course but I want to see who's on the ice when SC's are created and who's on the ice when they're allowed.

Doogie2K said...

I see that, Dennis. I'm just asking if you think pointing the finger at one particular player for bollocksing the play, in the course of chance-tracking, would be of further analytical value. It seems to go part and parcel with the creation of the scoring chance of the first place, and it's really the sort of thing you tend to do anyway when describing the plays in question, from what I remember. (Besides, "hockey is a game of mistakes" is one of the oldest cliches in the book, right up there with "we need to work harder, go to the net, create more chances.")

DMFB said...

the "true" measurement system will remain the goals themselves. Everything else is a means to that end.

I'm not trying to be obtuse here, I just think a statement like that misses the point of further analysis, which is what I was trying to point out. It seems uncomfortably close to me to saying something like, 'You guys have fun playing in the sandbox, but at the end of the day, it's the scoreboard that decides the win.' I mean, no shit, but the whole point of doing this is to find out why the scoreboard says what it does: saying "goals will always be the ultimate judge" is devaluing the analysis before you even bother to undertake it. Understanding the game is the end, not winning it.

Regular or Pro?

Regular, but it's the black one, for style points.

LittleFury said...

Understanding the game is the end, not winning it.


Is it still okay to watch hockey because you enjoy it on its aesthetic merits or because there's nothing else to do when its winter in Edmonton? Stats, shmats.

mike w said...

Man, do we need the season to start soon...

James Mirtle said...

I'll second that.

And request some OMC.

Doogie2K said...

Regular, but it's the black one, for style points.

When I bought mine, the black one was basically $80 for the upgrade package and $120 for the paint job. I got the white one with the upgrades.

DMFB said...

I got the white one with the upgrades.

Well and good, but I bet yours doesn't impress the ladies.

Bruce said...

//the "true" measurement system will remain the goals themselves. Everything else is a means to that end.//

I'm not trying to be obtuse here, I just think a statement like that misses the point of further analysis, which is what I was trying to point out.


DMFB: Not at all. I have made several comments above which are strongly in favour of further and deeper analysis. I'm merely pointing out that the scoreboard is the ultimate currency measuring success and failure, and any analysis should be anchored to that fact.

It seems uncomfortably close to me to saying something like, 'You guys have fun playing in the sandbox, but at the end of the day, it's the scoreboard that decides the win.'

It is both that simple, and not. Of course we should dig around in that sandbox to see what contributes to that end result. We must consider a hodgepodge of disparate factors like possession, turnovers, scoring opportunities, execution, goaltending, coaching, intangibles, and dumb luck. Some of those will be greater factors than others, and some of those will be harder to measure than others. But by all means we should try.

At this point my default position is fairly close to Ender's:

The bottom line is that in talking about any specific stat, we find that certain styles of play skew the stat to the point that you don't have a strong enough correlation to argue causality.

That's the challenge, all right.

Doogie2K said...

Well and good, but I bet yours doesn't impress the ladies.

How did you know? :(