Why?

Jonathan Willis
January 24 2009 10:35AM

Statistics and mathematical modeling have changed the way society functions.

Meteorologists use new methods of gathering information, especially from the atmosphere, to create a far grander database than any that human society has used for that purpose before. They feed these data points into supercomputers, and create complex models that involve a multitude of variables -- often variables that are in a constant state of flux or that are difficult to get firm information on.

Edward Lorenz, who served as an army weather forecaster during World War II, and then studied and taught meteorology, coined the “butterfly effect” as shorthand for how tiny variations could affect weather models: “Does the flap of a butterfly’s wings in Brazil set off a tornado in Texas?” He pioneered the field of chaos theory largely in an effort to better predict meteorological changes. Scientists continue to use long-reaching models to predict damage caused to the earth by global warming and human impact, while similar (albeit less complex) models are used all over the Earth for long-range forecasting.

The world economy, tracking the financial dealings of seven-billion plus through a dizzying array of legal jurisdictions, tax laws and the like, is tracked through a series of macroeconomic statistics; the ones produced at Berkeley use GDP, CPI, unemployment rate, corporate profits, change in business inventories, housing starts, interest rates, exports, and personal savings rate, among others. It’s on the basis of similar economic models that men like the Governor of the Bank of Canada or the Chairman of the Federal Reserve issue projections and make their decisions -- decisions that affect millions.

Even the advertising industry has made extensive use of mathematical modeling. Taking statistics compiled by government and the private sector, they break the population into demographics, and target their ads to different sectors of the population. By doing this they hope to better market their products. In 2001, The Coca-Cola Company alone spent nearly $2 billion on advertising -- money that was largely distributed on the basis of demographic information compiled by market research companies.

Even sports teams have turned more to mathematical models in recent years. The book Moneyball, based on the success of Oakland Athletics’ GM Billy Beane, is generally considered to be the work that brought statistical work into the mainstream in baseball. Theo Epstein, originally a PR man with the San Diego Padres, worked his way up the ladder and in 2002 was hired by the Boston Red Sox as General Manager. He was 28; the youngest GM in the history of Major League Baseball, and a man who’d never played the sport at even the high school level. Using sabremetric principles (math pushed by the Society for American Baseball Research), his team has won two World Series championships in his six years at the helm.

These are just a few examples; in virtually every scientific field, whether “hard” science or the social sciences, mathematical modeling is a primary tool for extrapolating all kinds of factors. It’s used in other sports, it’s used in scenarios that are for less predictable than what occurs in the carefully regulated confines of the arena, and often it’s used with data far less complete than we have available for free from the stats page and play-by-play reports at NHL.com.

In most of these other fields, suggesting that your powers of observation and gut feeling were a better method for prediction than mathematical models using a vast database would get you laughed out of a job.

In hockey, it’s the only commonly accepted practice -- and I have only one question. Why?

74b7cedc5d8bfbe88cf071309e98d2c3
Jonathan Willis is Managing Editor of the Nation Network. He also currently writes for the Edmonton Journal's Cult of Hockey, Grantland, and Hockey Prospectus. His work has appeared at theScore, ESPN and Puck Daddy. He was previously founder and managing editor of Copper & Blue. Contact him at jonathan (dot) willis (at) live (dot) ca.
Avatar
#1 Rick
January 24 2009, 11:02AM
Trash it!
0
trashes
+1
0
props

I don't think observation has ever been argued as the only acceptable practice. Where the debate gets heated is in determining the weight of advanced statistics VS observation and experience.

Good article all in all.

One critisism would be that using data as a predictor of weather misses as a pro argument. Les Nessman's Eye Witness Weather was way more reliable than the modern weatherman and that harkens all the way back to the '70's.

Avatar
#2 Librarian Mike
January 24 2009, 11:11AM
Trash it!
0
trashes
+1
0
props

I think a big part of the problem is that so many of the people who are running teams/the league are not actually qualified. Their big selling point is that they played for awhile and maybe won a few cups. Sure, that experience can be valuable but there does seem to be a mentality in the NHL that having played there automatically means you know what you're doing. Can you imagine if the Jets named Brett Favre as their head coach: They would immediately become the laughingstock of the NFL.

Avatar
#3 Jonathan Willis
January 24 2009, 11:16AM
Trash it!
0
trashes
+1
0
props

Rick wrote:

Where the debate gets heated is in determining the weight of advanced statistics VS observation and experience.

Among reasonable folks (like Brownlee, for instance) that's where the debate rests. Among less reasonable folk, the value of statistics at all is pretty debateable.

Les Nessman’s Eye Witness Weather was way more reliable than the modern weatherman and that harkens all the way back to the ’70’s.

I watched the odd WKRP in Cincinatti show on Prime, I think it was (because obviously I didn't see it when it first aired). Les was always pretty funny, but there aren't too many videos of him hanging around the internet.

Avatar
#4 Jonathan Willis
January 24 2009, 11:43AM
Trash it!
0
trashes
+1
0
props

Rick wrote:

One critisism would be that using data as a predictor of weather misses as a pro argument. Les Nessman’s Eye Witness Weather was way more reliable than the modern weatherman and that harkens all the way back to the ’70’s.

I was out to dinner with some friends last night, one of whom works at the air-traffic control tower at the local airport. Part of her job is to update weather observations.

Anyways, she was saying last night that the weather report will often switch from rain to sunshine based on the fact that she wrote down "clear" when she reported the conditions.

I don't know how much modelling local weather forecasts use; my hunch would be not much. I was thinking more long the lines of the folks who track hurricanes, tropical storms and the like.

Avatar
#5 Jonathan Willis
January 24 2009, 11:46AM
Trash it!
0
trashes
+1
0
props

@ Librarian Mike:

Neil Smith agrees with you.

Avatar
#6 The Towel Boy
January 24 2009, 11:55AM
Trash it!
0
trashes
+1
0
props

This site is becoming Jonathan Willis Nation...seriously dude...do you sleep at night? You're a blogging machine.

Avatar
#7 speeds
January 24 2009, 12:03PM
Trash it!
0
trashes
+1
0
props

Jonathan Willis wrote:

Among reasonable folks (like Brownlee, for instance) that’s where the debate rests. Among less reasonable folk, the value of statistics at all is pretty debateable.

I don't think that is true. "New" statistics that they haven't been brought up, sure, those may be met with resistance. But those that prefer to rely only on their eyes aren't generally afraid to use goals or assists to bolster their argument.

Avatar
#8 Jonathan Willis
January 24 2009, 12:03PM
Trash it!
0
trashes
+1
0
props

The Towel Boy wrote:

This site is becoming Jonathan Willis Nation…seriously dude…do you sleep at night? You’re a blogging machine.

I write as a way to relieve stress and such, so I've got hundereds of text files on my computer on all sorts of topics that I can pick at random and use for ideas. Plus, it usually only takes between 15 minutes and half an hour to get a rough draft, and the whole thing's wrapped up inside of an hour (usually).

Still, my wife's a patient woman.

Avatar
#9 speeds
January 24 2009, 12:05PM
Trash it!
0
trashes
+1
0
props
Jonathan Willis wrote: Among reasonable folks (like Brownlee, for instance) that’s where the debate rests. Among less reasonable folk, the value of statistics at all is pretty debateable.

I don’t think that is true. “New” statistics that they haven’t been brought up, sure, those may be met with resistance. But those that prefer to rely only on their eyes aren’t generally afraid to use goals or assists to bolster their argument.

Avatar
#10 Jonathan Willis
January 24 2009, 12:06PM
Trash it!
0
trashes
+1
0
props

speeds wrote:

I don’t think that is true. “New” statistics that they haven’t been brought up, sure, those may be met with resistance. But those that prefer to rely only on their eyes aren’t generally afraid to use goals or assists to bolster their argument.

You're right, of course, but goals and assists are pretty straight-forward; you just need to count. It's when the mathematics get more complex (for example, counting events against players and valuing those players based on ice-time, points-per-game, etc.) that they don't make the jump.

Avatar
#11 Dropping Deuces
January 24 2009, 12:24PM
Trash it!
0
trashes
+1
0
props

It sounds to me that Willis is trying out for the part of John Nash in A Beutiful Mind. Russel Crowe be damned!

Avatar
#12 mikemc8
January 24 2009, 12:29PM
Trash it!
0
trashes
+1
0
props

As a huge fan of the Oakland A's and of the methods of SABRmetrics (although I haven't made a good enough effort to understand them like the back of my hand), I love reading your stat breakdowns and I hope that this becomes a bigger part of the hockey world.

I'm sure the time will come, baseball too was once a very old-school system, but it is in the process of changing and hockey is perhaps just 5-10 years behind in its thinking.

Avatar
#13 Cam
January 24 2009, 12:32PM
Trash it!
0
trashes
+1
0
props

@ Jonathan Willis: I enjoy the intricate nature of stats, which is why I suppose that is why I got 100% in University Statistics. I love reading well thought out analysis' of players backed up with stats.

More and more I have found that the people that seem to be talking out of their butt crack don't seem look at a single stat. Or if they do they pick the one they like the most and use that to make universal declarations that are asinine.

One thing I appreciate about a statistical machine like yourself is that you are always willing to hear the intangible arguments surrounding some stats. People are more random than weather at times and there are often things that affect the game that can't be told by statistics (like team support and the mood in the room and leadership etc). Once you couple statistics with insider information (thank you Robin and Gregor) it really starts to make a lot more sense.

Avatar
#14 speeds
January 24 2009, 12:34PM
Trash it!
0
trashes
+1
0
props

Jonathan Willis wrote:

You’re right, of course, but goals and assists are pretty straight-forward; you just need to count. It’s when the mathematics get more complex (for example, counting events against players and valuing those players based on ice-time, points-per-game, etc.) that they don’t make the jump.

Tom Benjamin wrote about the danger of being able to count stats awhile ago, with regard to faceoffs. I can't find the article (I think it may be lost forever from when his site remodelled), here is another blog talking about it (maybe 1/4 to 1/3 of the way down the webpage).

I don't want to misremember or misstate his argument (so if anyone remembers better, or has a working link, it would be appreciated), but what I took from his article as his opinion was that faceoffs were overrated by conventional wisdom because they were now counted.

Avatar
#15 Fiveandagame
January 24 2009, 12:57PM
Trash it!
0
trashes
+1
0
props

@ Jonathan Willis: My statistical analysis has revealed that Sheldon Souray is statistically the most manly Oiler in existence and the Jonathan Willis really likes math.

Jonathan, I suggest a book for you called "Blink" by Malcom Gladwell. It's subject matter will interest you as it confirms what you're saying, to a point, but it also points out that the split second intuitive decision you first make, is most often the correct one. It points out that your brain is making thousands of unconscious decisions every second. It further points out that the more you have to explain your immediate intuitive decision the more likely you are to change your mind.

It also points out the failures of relying on only statistical data and number crunching. While those numbers will tell you what you need to know, it cannot predict with absolute certainty what the future will be.

If your statistical analysis leaves out one variable, it could be the difference in drafting Marc Pouliot and Zach Parise.

Avatar
#16 Jonathan Willis
January 24 2009, 12:58PM
Trash it!
0
trashes
+1
0
props

Dropping Deuces wrote:

It sounds to me that Willis is trying out for the part of John Nash in A Beutiful Mind. Russel Crowe be damned!

The funny thing is, I'm a good math guy, but I don't have a ton of experience in statistics. For example, in the Lecavalier article a little ways down, it didn't even occur to me to adjust his Corsi for the team until I started taking heavy flack, and it's something I should have done in the first place.

I'm learning, in other words.

Avatar
#17 Jonathan Willis
January 24 2009, 01:00PM
Trash it!
0
trashes
+1
0
props

@ speeds: Tyler has some good stuff on that at his site, and Staples picked up on it too. I really wonder how important faceoffs are in the grand scheme of things; I think we do tend to place too much importance on it in all likelihood, but that's a guess not a hard statement.

Avatar
#18 Dropping Deuces
January 24 2009, 02:08PM
Trash it!
0
trashes
+1
0
props

Fiveandagame wrote:

My statistical analysis has revealed that Sheldon Souray is statistically the most manly Oiler in existence and the Jonathan Willis really likes math.

I was just on the Oilers website and the picture of Souray makes him look like a hypnotist. I believe his is using his mind powers to receive man crushes from all of Oiler Nation.

Avatar
#19 Doogie2K
January 24 2009, 02:09PM
Trash it!
0
trashes
+1
0
props

@Fiveandagame: Well of course you can't rely on stats 100% of the time, because even if this is a deterministic universe, we simply don't have the tools to unlock our destiny: the model is far and away too complex, and may always be. But as a tool for checking what your eyes are seeing, they're perfectly valid, and you can make reasonable guesses from them just as sure (and sometimes more so) as you could from your eyes. And yet, because they don't understand them, people have an allergic reaction to them.

Avatar
#20 Jonathan Willis
January 24 2009, 03:28PM
Trash it!
0
trashes
+1
0
props

Cam wrote:

One thing I appreciate about a statistical machine like yourself is that you are always willing to hear the intangible arguments surrounding some stats. People are more random than weather at times and there are often things that affect the game that can’t be told by statistics (like team support and the mood in the room and leadership etc). Once you couple statistics with insider information (thank you Robin and Gregor) it really starts to make a lot more sense.

Take Sean Avery, for example. His underlying numbers (QualComp, Corsi, etc.) show a heck of a hockey player, and if those existed in vacuum, he'd be a guy to pursue before.

Of course, knowing what we know about him now, I wouldn't want him within 1000 miles of my hockey team. There's definitely personality/human factors that need to get taken into consideration, and when it comes to coaching you absolutely need to look at the process (i.e. physical scouting) as opposed to results.

But I've yet to see an explanation for why the eyes of one biased observer is a better tool for determining results than indepth statistical analysis.

Avatar
#21 Robert Cleave
January 24 2009, 03:29PM
Trash it!
0
trashes
+1
0
props

I don’t know how much modelling local weather forecasts use; my hunch would be not much. I was thinking more long the lines of the folks who track hurricanes, tropical storms and the like.

Numerical models drive every single weather forecast, from the local forecast at your friend's airport to broad scale climatology. Human input happens a lot less frequently than people might suspect, likely to the overall detriment of the product.

That aside, the best thing that advanced stats offer us is the opportunity to assess players beyond the "saw him good" approach. Context, in other words. Why aren't more people talking about the year Mike Smith is having in T. Bay? Look at his numbers versus, say, Kiprusoff's. He has a lousy team in front of him, so the win total is lower, but he's had a significantly better year by any other worthwhile measure. He's on a poor team in the South, so he doesn't get much media attention, but looking at the numbers can help us go beyond that bias. The math behind this stuff isn't that complex. Why be scared of it?

Avatar
#22 Hockey ain’t Baseball, stats-lovers - OilersNation
January 24 2009, 03:30PM
Trash it!
0
trashes
+1
0
props

[...] me the sports are just too different to allow statistical data to overtake live scouting. Willis mentioned Sabremetrics as a key to the success of the Red Sox, and while I agree it had something to do with [...]

Avatar
#23 Jonathan Willis
January 24 2009, 03:31PM
Trash it!
0
trashes
+1
0
props

@ Robert Cleave:

Mike Smith's season has been amazing. Probably lost in the mess that is Tampa Bay.

Avatar
#24 Deans
January 24 2009, 03:48PM
Trash it!
0
trashes
+1
0
props

Contrary to public belief and practice, some things in life just aren't quantifiable. How do you put a statistic on an important facet like desire and heart. Everyone knows that desire and sacrifice are key components to any successful team. Probably just as important as scoring. Did the 06 Oilers have a 4.6 desire/sacrifice rating? Was that rating 1.3 higher than the rest of the Western Conference? Statistics are a vital assest for evaluation and prediction, however, we would be sticking our heads in the sand if we were to believe they told the whole story.

Avatar
#25 Jonathan Willis
January 24 2009, 03:53PM
Trash it!
0
trashes
+1
0
props

Deans wrote:

Contrary to public belief and practice, some things in life just aren’t quantifiable. How do you put a statistic on an important facet like desire and heart. Everyone knows that desire and sacrifice are key components to any successful team.

You can't, and I wouldn't argue that that you could. I would argue that on-ice results (offensive and defensive ability) can be measured better with statistics and math than with scouting, because of the limitations of an individual scout (no matter how capable).

A player's character, off-ice behaviour, willingness and desire; these are all things that need to be answered through a combination of on-ice observation, and off-ice discussion. I'd argue that those attributes can't be effectively guaged solely by on-ice observation, either.

Avatar
#26 speeds
January 24 2009, 04:23PM
Trash it!
0
trashes
+1
0
props

Deans wrote:

Contrary to public belief and practice, some things in life just aren’t quantifiable. How do you put a statistic on an important facet like desire and heart. Everyone knows that desire and sacrifice are key components to any successful team. Probably just as important as scoring.

As important as scoring? Two things matter in hockey, goals scored and goals allowed. Anything else is only important insofar as they lead to goals scored and goals allowed.

Avatar
#27 Ender
January 24 2009, 05:08PM
Trash it!
0
trashes
+1
0
props

Why?

For the same reason Vic (one of the staunchest supporters of hockey stats) chirped off about how goalies shouldn't get as tired as other players because they don't really do anything outside of spurts.

Nobody wants to let go of their own biases. The old stats are hugely inadequate for looking at much of anything. The new stats have not been held up to review (in the peer review sense, since you're talking about science. That's the thing. The hockey stats are not actual stats. There have been no actual studies. They are data sets that people take, put together with other observational things, and try to draw conclusions from them. There is no rigor. There is no repeatable experiment. There's a series of data sets that people attribute to meaning "something" when realistically we don't have a good way of mathematically modeling anything this fast or fluid.

Let me put it this way. You can model a game as simple and algorithmic as baseball. You can model the effects of X given a control group. However, physicists *still* cannot accurately measure n body gravity problems where n > 2. It's an issue with static vs dynamic.

That's not to say the whole field is lost. It's not. But instead of trying to say "this data set implies X" how about everyone steps back for a minute and says "We need to try to use these methods to build a model which should lend itself to prediction."

Comparing hockey stats to anything in the hard or soft sciences is apples and oranges.

Avatar
#28 Ender
January 24 2009, 05:17PM
Trash it!
0
trashes
+1
0
props

Oh, and business and hockey stats are also apples and oranges. In the business world, you make a product, send it out into the world, and see how it's received. You can replace the word "product" in that sentence with "ad" or "hype machine" if you like, but the point stands. The only comparable things between business stats and hockey are a)How many tickets did the team sell, b)How many more tickets will the team sell if it's doing well/making the playoffs and c)How much money should we charge for those tickets due to demand.

Of course then your hockey stats *are* business stats, so they can both be apples.

Avatar
#29 Ender
January 24 2009, 05:20PM
Trash it!
0
trashes
+1
0
props

@Doogie:

If you're using stats to confirm or deny your eyes, aren't you running into the old correlation fallacy that if two things are correlated it "means" something? I mean, realistically there's enough data out there to confirm just about anything to your eyes, depending on how rigorous (predictive) you want to be.

Avatar
#30 Jonathan Willis
January 24 2009, 05:31PM
Trash it!
0
trashes
+1
0
props

@ Ender:

Such work hasn't neccessarily been done by the folks in the Oilersphere, since it's a developing field, and people wroking on it for teams are obviously going not going to just share the information, but what about the work done by Alan Ryder and others at his site? That would seem to fit into your point.

Besides the point was less about direct comparison and more about the fact that systems far more complex than hockey are studied via math and statistics, so why should hockey be immune to that sort of study, rather than a direct comparison between the stuff being done right now and a paper submtted for peer review.

Avatar
#31 Jonathan Willis
January 24 2009, 05:50PM
Trash it!
0
trashes
+1
0
props

Ender wrote:

Comparing hockey stats to anything in the hard or soft sciences is apples and oranges.

I'm going to disagree with you a bit here, because even though the fields are vastly different, much of the math will still cross over.

Take chaos theory - the same mathematical process is used to study epilepsy, the stock market, and weather patterns. I mean, the fields couldn't be more different, but the math crosses over.

Now, we aren't talking about anything as complex as chaos theory, but it would seem to me that the same processes that allows us to account for multiple variables in a complex (but very observable) environment would translate across the fields, no?

Avatar
#32 Ender
January 24 2009, 05:56PM
Trash it!
0
trashes
+1
0
props

@Jonathan

I know I muddled a few points together, but the thing is that hockey is too fluid to deal with the kind of statistical analysis that you're talking about. Every one of your examples are "Let's roll the dice and see how they land." The issue is that in hockey they often don't land. I mean, we're talking about Quantum Mechanical level of complexity here. At best you're going to be able to work with probabilities, and even at that, the only meaningful things you're going to be able to talk about is how players play as a team due to scope.

And that's assuming players are intelligent robots. What I mean by that is that quanta only act as quanta. They don't have "off-days" or get "psyched out" by a goalie. They are perfect players. Hockey players are not. Without interviewing/knowing the players or the room dynamic you're never going to know the validity of your assumptions. If Torres seems to be a streaky player, can we reasonably predict how long the streak will be? How many games? How many games on average? If we boil it to a macroscopic level (as to not look at the causes *for* the streak) we might be able to find those averages. But what exactly do they tell us? Presumably he doesn't have bipolar disorder on a strict schedule, so it would likely be things going right and wrong in his life. Can we assume his life to be reasonably static with barrels of money and a revolving door of teammates and cities? Probably not.

I get what's going on in the oilogosphere, I really do. Just when I think about it at any sort of analytical level, I just wonder what the purpose is. People have said they want to "understand the game" better. The issue is that there are so many extenuating circumstances that we don't know about. I mean, take the "Moreau waved Pouliot out of the circle" thing or the "Garon hates Peeters" thing. At the level of study that's being done, you need reasons to make it make tangible sense, and more often than not, huge stretches are being made.

Avatar
#33 Ender
January 24 2009, 05:57PM
Trash it!
0
trashes
+1
0
props

Jonathan Willis wrote:

Ender wrote: Comparing hockey stats to anything in the hard or soft sciences is apples and oranges. I’m going to disagree with you a bit here, because even though the fields are vastly different, much of the math will still cross over. Take chaos theory - the same mathematical process is used to study epilepsy, the stock market, and weather patterns. I mean, the fields couldn’t be more different, but the math crosses over. Now, we aren’t talking about anything as complex as chaos theory, but it would seem to me that the same processes that allows us to account for multiple variables in a complex (but very observable) environment would translate across the fields, no?

I think this is where we're getting held up. I'm saying that the dynamics of hockey are every bit as complicated as chaos theory and Quantum Mechanics.

Avatar
#34 Ender
January 24 2009, 05:58PM
Trash it!
0
trashes
+1
0
props

Ender wrote:

Jonathan Willis wrote: Ender wrote: Comparing hockey stats to anything in the hard or soft sciences is apples and oranges. I’m going to disagree with you a bit here, because even though the fields are vastly different, much of the math will still cross over. Take chaos theory - the same mathematical process is used to study epilepsy, the stock market, and weather patterns. I mean, the fields couldn’t be more different, but the math crosses over. Now, we aren’t talking about anything as complex as chaos theory, but it would seem to me that the same processes that allows us to account for multiple variables in a complex (but very observable) environment would translate across the fields, no? I think this is where we’re getting held up. I’m saying that the dynamics of hockey are every bit as complicated as chaos theory and Quantum Mechanics.

I should add that while it's as complicated as your examples, it also has a psychological component which makes it that much more mathematically complex.

Avatar
#35 David S
January 24 2009, 07:41PM
Trash it!
0
trashes
+1
0
props

Ender wrote:

Ender wrote: I should add that while it’s as complicated as your examples, it also has a psychological component which makes it that much more mathematically complex.

And a physiological component. Both of which are non-linear variables far beyond outcomes that can be replicated.

Avatar
#36 Ender
January 24 2009, 08:23PM
Trash it!
0
trashes
+1
0
props

David S wrote:

Ender wrote: Ender wrote: I should add that while it’s as complicated as your examples, it also has a psychological component which makes it that much more mathematically complex. And a physiological component. Both of which are non-linear variables far beyond outcomes that can be replicated.

Exactly.

Avatar
#37 RLH
January 24 2009, 09:04PM
Trash it!
0
trashes
+1
0
props

JW, I'm going to take a wild stab at the title of your Masters thesis. Let me know if I'm right:

A Subjective Analysis of the dietary Evolution of Polar Bears from B.C 7500 to Present

:D Am I even close?

Avatar
#38 RLH
January 24 2009, 09:09PM
Trash it!
0
trashes
+1
0
props

@ Ender: Despite the earlier Avery comment, I'll step out here and say that a player with heart and spirit has a significantly better chance of looking favorable from a statistical perspective than one that does not. In other words, I can drive my score higher by loving what I do, even if I'm not particularly good at it. Sure, someone with God-given talent may do better than me. The reality is, however, that both of us are going to show up with good numbers because I simply want it more.

I wish I had some statistics to back that up, but you're right that the quantification of spirit is very difficult.

Avatar
#39 Ender
January 24 2009, 09:22PM
Trash it!
0
trashes
+1
0
props

@RLH

I'm not really talking about "spirit" in the Rudy sense. I'm more talking about how much focus one has. It's one thing to love your job, but it's another to bring your A game when your marriage is falling apart (for example). Some people are better at dealing with stress than others, so I see the overlap between that and what you're calling spirit, but I don't think they're the same thing.

And you're right. There's something to be said for "who wants it more."

Avatar
#40 Jonathan Willis
January 24 2009, 10:01PM
Trash it!
0
trashes
+1
0
props

RLH wrote:

JW, I’m going to take a wild stab at the title of your Masters thesis. Let me know if I’m right: A Subjective Analysis of the dietary Evolution of Polar Bears from B.C 7500 to Present Am I even close?

And I've been struggling with the title. Thanks!

Avatar
#41 Jonathan Willis
January 24 2009, 10:16PM
Trash it!
0
trashes
+1
0
props

@ Ender:

Certainly if you want to predict it with game-in, game-out regularity we're talking an incredibly complex system, despite the fact that it's simple on the surface. But is that level of precision necessary?

Basically, I view being a general manager (or a business owner or whatever) as making a series of smart bets. If, for example, we knew that a European defenseman playing a certain level of opposition and producing certain numbers relative to his team would translate well to the NHL 70% of the time.

Then there are other things - look at the contract Philly signed Daniel Briere to. Guys generally get paid on offense and reputation; obviously as a GM you want to target guys who are undervalued, so paying for somebody who brings much more to the table than offense would seem to be the way to go, and some of the advanced statistics can help us there.

You want to know which bets are good ones to make - and I think advanced stats can tell you that.

Now, I apologize if that didn't make sense; I just got home from hockey and I fear that I'm not coherent right now.

Avatar
#42 Tyguy
January 25 2009, 02:52AM
Trash it!
0
trashes
+1
0
props

Deans wrote:

Contrary to public belief and practice, some things in life just aren’t quantifiable. How do you put a statistic on an important facet like desire and heart. Statistics are a vital assest for evaluation and prediction, however, we would be sticking our heads in the sand if we were to believe they told the whole story.

I agree with these statement 9 times out of ten , with =/- of 2%, 99 times out of a 100.

Avatar
#43 Doogie2K
January 25 2009, 10:28AM
Trash it!
0
trashes
+1
0
props

If you’re using stats to confirm or deny your eyes, aren’t you running into the old correlation fallacy that if two things are correlated it “means” something? I mean, realistically there’s enough data out there to confirm just about anything to your eyes, depending on how rigorous (predictive) you want to be.

Not necessarily. You know from talking to me personally that one of the things that bugs me the most about some of the stats stuff done in the Oilersphere is treating the Pearson correlation as some sort of all-purposes Truth Detector. The most you're ever going to get out any of these numbers, I suspect, is correlative in quality, and so the most you can imply is a relationship, rather than direct cause and effect, but I still think that they can at least tell you something about how a player's performed in the past, if not necessarily how they'll perform in the future (though you can make intelligent guesses, no doubt). I don't think there's too big of a logical leap from looking at a guy's GD and SD on the ice and his team's GD and SD without him and concluding that he's driving things in one direction or the other, especially since it's something you can go back and see for yourself. ("Hey look, that Horcoff guy is pretty good defensively.") If anything, it's QualComp-type stats and indices of zone time that are a bloody mess, though I respect the idea they're trying for, at least.

Statistics are a vital assest for evaluation and prediction, however, we would be sticking our heads in the sand if we were to believe they told the whole story.

No one method of analysis will tell you everything in a complex system with a large number of confounding variables. For example, I don't think I've seen any definitive statement on whether Staios or Smid is "the problem" on the third pairing, or if it's the forwards that are causing trouble for both of them. In any case, you're never going to be right 100% of the time when making predictions, because as Ender points out, the system you're talking about is at quantum-level complexity. I never understood how anyone could say hockey was less complex than baseball, because hockey is a fluid system of ten frequently-changing moving parts and two semi-stationary ones, whereas baseball is a discrete system with mostly-consistent starting conditions, but I digress.

And a physiological component. Both of which are non-linear variables far beyond outcomes that can be replicated.

Well, the physiological variables can probably be modelled a lot more readily than the psychological ones. The psychological model is another hugely complex system involving a good chunk of the brain, which we still don't have the technology to map and model at the single-neuron level (though we're getting there), and even if we could, it's a system in flux, i.e. it continues to change with time. The physiological model, at least, is largely mechanical, and there's an entire field (which I'm currently studying) devoted to studying biological systems from the perspective of a mechanical engineer. There's still a ton of variables there, but I suspect they're a little more manageable, because we understand a lot more of the underlying interactions of exercise physiology than we do psychology. While there are still some aspects of physiology that don't quite match our models neatly, like Newtonian physics, it's usually good enough for everyday purposes. Hell, we still use the Hill equation to model the force-velocity relationship of muscle more than 80 years after his paper was first published in 1938.

I've talked about both components in the past at some length, and I think both have to be considered in the discussion. That being said, I do wonder if certain things show up over time as trends. For example, one of the things I'd like to see at some point is some more data on the leading-trailing-tied splits, and who scores the "true" GWG: the one that breaks the tie for good, not simply the "Opponent + 1" goal. I think that would be a better measure of "clutchness" (a concept many Statzis are allergic to, despite the very clear scientific evidence that such a thing exists from a psychological point of view, probably because it's too complex to realistically model, particularly with very little access to the most vital data), and tell us something new about some players. I suspect Hemsky would have a pretty decent "clutch" rating. I suspect "heart" and "desire," as well as better fitness, would naturally manifest themselves in those that are more consistent and more "clutch."

Avatar
#44 Ender
January 25 2009, 10:37AM
Trash it!
0
trashes
+1
0
props

Jonathan Willis wrote:

@ Ender: Certainly if you want to predict it with game-in, game-out regularity we’re talking an incredibly complex system, despite the fact that it’s simple on the surface. But is that level of precision necessary? Basically, I view being a general manager (or a business owner or whatever) as making a series of smart bets. If, for example, we knew that a European defenseman playing a certain level of opposition and producing certain numbers relative to his team would translate well to the NHL 70% of the time. Then there are other things - look at the contract Philly signed Daniel Briere to. Guys generally get paid on offense and reputation; obviously as a GM you want to target guys who are undervalued, so paying for somebody who brings much more to the table than offense would seem to be the way to go, and some of the advanced statistics can help us there. You want to know which bets are good ones to make - and I think advanced stats can tell you that. Now, I apologize if that didn’t make sense; I just got home from hockey and I fear that I’m not coherent right now.

I'm not really saying that it's impossible to get the answers that you want on any scale. Faceoff numbers, for example, are reasonably repeatable. It's also entirely possible that some players *will* be consistent enough to hold up to this kind of analysis. It just won't be all of them.

And here's me bringing my biases in, but I'd love to see some sort of statement of intent once in a while. People present information, and sure, it's interesting, but there is generally no conclusion. There is no indication of what scale or scope people are working on or what they're trying to get at with their writing.

Take the LeCav piece for example. It's all well and good to put up stats that indicate that divisional play makes a big difference with regards to output. LeCav + St Louis will likely be better than Hemsky + Horcoff most days, but their stats will always be higher due to the divide between the skill levels of the divs. At that point it's easy to call it obvious to anyone who's watching the game, and move on. That said, it's important to test the obvious, if you have any numbers to back it up.

The issue is that at that point, what are you really trying to say? I mean, there have been all sorts of rumors of dissent in the Bolts' locker room. Coaching and management have always been a problem. Plus, these players have played against soft opp for so very, very long. It's entirely possible that LeCav would flourish in the NW because it would be a challenge again (and some people like challenges) and entirely possible Hemsky would flounder in Tampa due to the lack of challenge. The only way you're going to be able to know that is by knowing the character of the players, which bloggers won't.

Now, your question was "why?" so here's the answer. Anyone with eyes can see that both LeCav and Hemsky are top-level talent, regardless of who is better on paper. That said, GMs can actually talk to players and ex-coaches, etc to get to know the player. The "seen him good" crowd has a point in that if you're paying attention you can at least divide players up into tiers. From there, sure, you could crunch the numbers, but since we're talking about betting, I bet you that talking to the personel would be more informative than the numbers 99 times out of 100.

Avatar
#45 Jonathan Willis
January 25 2009, 11:25AM
Trash it!
0
trashes
+1
0
props

Ender wrote:

From there, sure, you could crunch the numbers, but since we’re talking about betting, I bet you that talking to the personel would be more informative than the numbers 99 times out of 100.

Really? 99 times out of 100? Then why do NHL franchises consistently make stupide decisions?

Look at off-season free agency. Daniel Briere's widely hyped, signed to a contract that never ends, and hailed as a difference maker. Than he doesn't live up to expectations - from QualComp, we know he was playing sheltered minutes in Buffalo. That matters.

Turns out, there are a whole bunch of players who are hailed as "difference makers" but then fail to live up to expectations - and using only QualComp, we should know better.

Avatar
#46 Ender
January 25 2009, 01:13PM
Trash it!
0
trashes
+1
0
props

Jonathan Willis wrote:

Ender wrote: From there, sure, you could crunch the numbers, but since we’re talking about betting, I bet you that talking to the personel would be more informative than the numbers 99 times out of 100. Really? 99 times out of 100? Then why do NHL franchises consistently make stupide decisions? Look at off-season free agency. Daniel Briere’s widely hyped, signed to a contract that never ends, and hailed as a difference maker. Than he doesn’t live up to expectations - from QualComp, we know he was playing sheltered minutes in Buffalo. That matters. Turns out, there are a whole bunch of players who are hailed as “difference makers” but then fail to live up to expectations - and using only QualComp, we should know better.

Because people aren't doing their research. Are you really telling me that if they talked to the personnel in a responsible way and actually watched when Briere was on the ice they wouldn't notice the level of competition?

I mean, you're assuming people aren't lazy here.

Avatar
#47 Sean
January 25 2009, 04:01PM
Trash it!
0
trashes
+1
0
props

Once each game is taped and information for each player on the ice (and the puck) is tracked serious progress will be made - not that there hasnt been huge progress already.

As more and more information becomes available mathematical models are going to become a huge competitive advantage.

Jonathan lets go apply for jobs at the Oilers ;)

Avatar
#48 Jonathan Willis
January 25 2009, 05:11PM
Trash it!
0
trashes
+1
0
props

Ender wrote:

Because people aren’t doing their research. Are you really telling me that if they talked to the personnel in a responsible way and actually watched when Briere was on the ice they wouldn’t notice the level of competition?

If it was just Briere, but every summer there are a ton of guys who get picked up this way - and honestly, if you're gonna bet 65 million or whatever the number on Briere's contract is, you do your research.

They aren't picking up a guy for their fantasy roster - it's their job and it's a truckload of real money that's on the line.

I don't mess around at work with 1000$ on the line; forget about with 65 million. I don't buy the explanation that they're all lazy - I'd suggest that they just don't know who he's playing against.

Avatar
#49 Ender
January 25 2009, 06:22PM
Trash it!
0
trashes
+1
0
props

Jonathan Willis wrote:

I don’t mess around at work with 1000$ on the line; forget about with 65 million. I don’t buy the explanation that they’re all lazy - I’d suggest that they just don’t know who he’s playing against.

And I'd suggest that if they don't know who he's playing against, they're not doing their research.

Look, I'll give you Briere. But for every Briere there are a dozen Hejdas who come out of nowhere (I say this, ccertain that you'll pull numbers from his euro league days, but I still think it's beside the point). I'm not saying stats can't help that. I'm saying that stats *cannot* tell the whole story. However, between "seen him good" and access to people around the player, not using stats *can* tell the whole story. Not saying it definitely will or *should* but I still think it *can*.

Whether or not people use the tools at their disposal for the best result is another story altogether.

Avatar
#50 Mike
January 26 2009, 11:42AM
Trash it!
0
trashes
+1
0
props

Ender wrote:

But for every Briere there are a dozen Hejdas who come out of nowhere (I say this, ccertain that you’ll pull numbers from his euro league days, but I still think it’s beside the point).

Is that really beside the point? I would say a gem toiling in a Euro league is EXACTLY where the "seen him good" approach will fail, simply for lack of resources.

It's both easier and cheaper to pay an intern to put all the numbers in a spreadsheet and flag the guys who float to the top.

Hejda only "came out of nowhere" if you didn't know where to look.

Comments are closed for this article.