Analytics be damned!!

[n1ck13]

Here's my answer to you n1ck13:

Prove that momentum exists.

That's a cop out... My point is that because something has yet to be proven doesn't mean it doesn't exist. I try to weigh what we know vs what we don't. If hockey statistical analytics continue to evolve and grow and in ten, fifteen, twenty years can't prove momentum exists, I would certainly say it is a probability it does not exist. I am not a statistician, so I would not attempt to be one, (a side note, this is the problem, people playing statisticians) rather, I have a master's degree (working on a doctorate in sociology) that has required me to study portions of statistics. I feel comfortable interpreting data, but am not nearly skilled enough to go about attempting to prove/disprove anything. However, hockey statistical analysis is still too infintile in its very nature to be making those extreme conclusions. I am always open to new ideas and a preponderance of evidence (this isn't a court of law), however, that preponderance has yet to be met for momentum.

 

[Sproo]

Statistical analysis in hockey will always be exponentially less informative than baseball.

In baseball getting discrete data points is pretty easy. Pitch 1, pitch 2, ... final pitch. Each pitch has a clear outcome for the pitcher and the batter.

In hockey there are no obvious data points. You can't slice the whole game into pitches like baseball, so you have to deal with a time variable (time on ice), but not only that there are usually about 12 other variables that you need to account for at any given time t-- the interactions and positioning of all other players on the ice.

I'm not going to pretend I know much about corsi, but unless it's a complicated algorithm which analyzes games on a computer resulting in each player's relative contribution to preventing/scoring goals, then I'm presuming all these variables are just being collapsed into some vastly inferior statistical categories.

Just because it is mathy it does not mean it is an accurate model of how games are won and lost. It is an accurate model of the invented statistical categories.

The onus is on the model-makers to refine it to capture what actually happens in real life (e.g., the Leafs winning games), and NOT on the leafs to refine their team to fit the model.

 

[The Legend]

I really don't care what people choose to do with their time, but the advanced statistics crowd drive me nuts because they DO take the numbers as facts. Many try not to or at least make it seem like they don't, but they do because every argument is supported with the numbers. The ones that buy into the advanced statistics NEVER have an argument that goes against the numbers which simply doesn't make sense as they are not fool proof.

The increased use of stats (they really aren't so advanced) in the mainstream makes me happy because it means that the days of the former player-turned commentator who tosses out opinions based on emotion and bias are beginning to come to an end.

 

[Gallagbi]

Here's my answer to you n1ck13:

Prove that momentum exists.

Leafs Boston Game 7?
Boston Chicago Game 6?
Pretty clear momentum shifts.

I'd bet that there are more scoring chances within 2 minutes of a team scoring a goal than your average 2 minutes of game play.

 

[TieClark]

The increased use of stats (they really aren't so advanced) in the mainstream makes me happy because it means that the days of the former player-turned commentator who tosses out opinions based on emotion and bias are beginning to come to an end.

It doesn't help anything... James Mirtle is a terrible reporter not because he's a poor writer but because every single article he makes is somehow linked to advanced statistics.

 

[Ricky Bobby]

Having Bolland and Clarkson as players who can play along the boards, cycle and hold onto the puck in place of transition players in Grabo and CMac should change our analytics quite a bit.

Having Bernier able to play the puck should also change those analytics.

Having a healthy Gardiner who barely played in the regular for us and Ranger who should be better than Kostka should also help those analytics.

Different season. Different numbers.

 

[Sproo]

1) Momentum doesn't exist

Just because you can't adequately define it mathematically it does not mean it does not 'exist'.

A player's psychological state clearly as an effect on their performance. E.g., someone who is confident in their abilities tends to perform better than someone who is second guessing their abilities.

If a whole team is in a similar heightened mental state, and if their opposition is beginning to doubt their abilities, then you have a situation which I think most people would agree defines a momentum shift in the game.

If we had the ability to do brain scans on the fly and have a way of mathematically representing people's brain states, then wouldn't you want to include this in your statistical analysis?

 

[hd1344]

I think the big issue that turns off many people with these "advanced stats" is that they don't adequately take into account the quality of the Corsi events during the game.

All these Corsi events don't take place in a vacuum. Dynamic game situations affect every event, including teams' systems and tendencies.

It's sometimes tough to argue in favour of Corsi when you watch teams effectively keep everything to the outside, resulting in many low-quality attempts, and have your own team score once or twice on higher-quality attempts.

Would the Leafs want to possess the puck for longer periods and direct more shots at net? For sure. But not at the expense of affecting their systems and structure.

 

[The Winter Soldier]

No it doesn't, possession in advanced stats are measured using shots and there wasn't a shot attempt there. Also, if you make the right play like that over time it will push the puck into the other teams end more often, result in more shots for your team and improve your advanced stats.

Puck possession = % of passes = Fenwick = Corsi = Scoring Chances

http://www.thescore.com/buzz/articles/220380-talking-with-advanced-hockey-stats

Question what does this prove?

Absolutely Nothing.

 

[jakapono24]

Just because you can't adequately define it mathematically it does not mean it does not 'exist'.

A player's psychological state clearly as an effect on their performance. E.g., someone who is confident in their abilities tends to perform better than someone who is second guessing their abilities.

If a whole team is in a similar heightened mental state, and if their opposition is beginning to doubt their abilities, then you have a situation which I think most people would agree defines a momentum shift in the game.

If we had the ability to do brain scans on the fly and have a way of mathematically representing people's brain states, then wouldn't you want to include this in your statistical analysis?

But you don't have any evidence of the claim that confidence has an effect on performance. It may appear to be 'common sense', but there is no evidence to support that claim. I'm not denying that confidence would appear to be beneficial, however we can't arbitrarily assign a value to confidence (and thus momentum) if we cannot show any tangible and relevant effect on a team or individual player.

 

[TieClark]

But you don't have any evidence of the claim that confidence has an effect on performance. It may appear to be 'common sense', but there is no evidence to support that claim. I'm not denying that confidence would appear to be beneficial, however we can't arbitrarily assign a value to confidence (and thus momentum) if we cannot show any tangible and relevant effect on a team or individual player.

Of course there is evidence.. every player who has played even semi-competitive hockey knows momentum is a huge factor in the game. There's a reason players go out there some nights and look amazing and other times look like **** and it's not "luck" or a freak occurrence.

We don't need to assign a "value" to anything... again you're thinking like a statistics formulator.

 

[YearsintheWilderness]

I find advanced stats the epitome of tedium. Talk about taking the soul out of hockey.

You can take anything in life and dissect it to the nth-degree, and for something like cancer research, okay, but that degree of analytics isn't a fit for a free-flowing, emotional, game like hockey. It robs it of its passion and mystery and magic -- the very reasons we watch it.

There are some things we don't need to know. In fact, not knowing them allows us to enjoy life more. If I'm enjoying watching two distant birds soaring amongst the clouds at sunset I don't need to know their exact speed, their relative resistance to gravity, or the unfulfilled potential of their wingspan. I will just enjoy the moment, live in the moment, and move on, fulfilled. And it's the same when it comes to watching hockey.

One final example: Beethoven.

You lose something, the soul of something, when you eviscerate and analyze it too much. Imagine going to a live Beethoven concert in his time, or to Furtwangler conducting a Beethoven symphony. One person sits back, closes their eyes, and allows the music to transport them into the ineffable realms of the composer's mind when he wrote it. Pure magic!

Then there's the person beside him -- probably an art critic -- who is frantically and furiously jotting down all the C's, and C-sharps, and key changes, and tempo changes, and attempts to formalize it.

No thank you.

Why would you want to deprive yourself of the magical experience of enjoying your favorite team in a life and death battle of athleticism, passion, tradition and emotion with thoughts like, "Oh, that will probably drop his relative-Corsi by half a percentage point"?

I am going to remain happily content with my goals, assists, points, penalty minutes, plus-minus, and face-off percentage... and, as always, focus on the only statistic that truly matters to me: did the team with the maple leaf on their sweater win or lose? Simple.

If someone else wants to devour the soul of hockey, chew it up and spit it out, fine. But I want no part of such folly.

But for many people, the stats add to the passion and mystery. It's simply the way a person sees things and what aspects of the game provide entertainment.

Also, a fairly absurd caricature of what someone interested in stats is like.

 

[The Winter Soldier]

ADVANCED STATISTICS
Mirtle: Why the Leafs luck may run out this season

“Even if the season had continued to 82 games, the Toronto Maple Leafs may still have finished among the top five luckiest teams [in the last five years],†Vollman writes in Hockey Abstract, a statistical tome released this year that is meant to mimic what Bill James’s various Baseball Abstracts were to that sport in the 1980s.

“They finished in the top five in the Eastern Conference standings despite being outshot 1543 to 1264. There was not a single Leaf [player] with whom Toronto outshot their opponents throughout the course of the season. They started the lowest percentage of shifts in the offensive zone in the entire league, the second most in the defensive zone, and their puck possession rate was also second lowest in the league.â€

My question is who compiles these stats? Like Scoring Chances. Some 5th grader living in their parents basement? Pretty subjective what a scoring chance is. Shots on goal is pretty subjective even in the old fashion sense in some buildings. The analytics put a lot of faith in puck possession stats.

Let's hope the Leafs continue to be one of the luckiest teams in the NHL, and damn the advanced stats groupies.


http://www.theglobeandmail.com/spor...luck-may-run-out-this-season/article14623017/

 

[jakapono24]

Of course there is evidence.. every player who has played even semi-competitive hockey knows momentum is a huge factor in the game. There's a reason players go out there some nights and look amazing and other times look like **** and it's not "luck" or a freak occurrence.

We don't need to assign a "value" to anything... again you're thinking like a statistics formulator.

Except you haven't provided any real evidence; it's all anecdotal evidence. As for the "reason players go out there some nights and look amazing and other times look like ****", it could be preparation, fatigue-level, random variation, etc. We can't really know for sure.

 

[jakapono24]

But for many people, the stats add to the passion and mystery. It's simply the way a person sees things and what aspects of the game provide entertainment.

Also, a fairly absurd caricature of what someone interested in stats is like.

Bingo! We can choose to enjoy hockey either way or (as I prefer) both ways!

 

[The Legend]

Just because you can't adequately define it mathematically it does not mean it does not 'exist'.

A player's psychological state clearly as an effect on their performance. E.g., someone who is confident in their abilities tends to perform better than someone who is second guessing their abilities.

If a whole team is in a similar heightened mental state, and if their opposition is beginning to doubt their abilities, then you have a situation which I think most people would agree defines a momentum shift in the game.

If we had the ability to do brain scans on the fly and have a way of mathematically representing people's brain states, then wouldn't you want to include this in your statistical analysis?

Logically - I agree with you. It makes sense that human beings might push themselves a bit harder if they are feeling "it". But the results simply don't back this up. It's possible that - as human beings - feeling "it" leads to overconfidence, which leads to poor decision making. It's also possible that high level athletes don't actually deal with those emotions.

 

[James Mirtle]

It doesn't help anything... James Mirtle is a terrible reporter not because he's a poor writer but because every single article he makes is somehow linked to advanced statistics.

Ah, yes, this old bit.

It's funny I get accused of writing about "nothing but Corsi" all the time, but the reality is that it's actually rarely mentioned. You can search for yourself - on The Globe's website: 1,200+ articles, eight mentions of Corsi, ever.

All of my work is easy to sort through - as this link shows, very few of the stories recently are linked to or about advanced statistics.

I've been a lurker and reader on hfboards for a long, long time, but this criticism lately is really off base and frankly absurd. I understand that this type of analysis isn't for everyone, but I'm not doing it out of bias or Leafs hatred or for any other reason than it's another way to understand and write about the game.

The regression analysis I wrote about this week (wrote about, not invented, by the way) was used by analysts to predict the 2011-12 Wild team that was first in the NHL in mid-December would struggle to close the season, and that the eighth-seeded Kings that year were a contender despite their low position in the standings. Among many, many other things that have nothing to do with the Leafs.

No one in the traditional media made those predictions that season.

This stuff isn't going away - in fact, the work I see some people doing online is getting more impressive (and accurate) than ever. It's fine if some people aren't interested and find this a bit arcane - that I understand. It's a niche. But these constant potshots aren't fair and have little basis in reality.

Thanks.

 

[James Mirtle]

My question is who compiles these stats? Like Scoring Chances. Some 5th grader living in their parents basement?

Well, a site like behindthenet.ca scrapes all of its data from NHL.com's game information and presents various computations of that. It's explained on the site.

The site is developed by Gabriel Desjardins, an engineer who graduated from Berkeley and has worked for several NHL teams. His work has been vetted by basically every other hockey stats analyst as it's very publicly available and widely used.

 

[MrLegend28*]

It should be obvious to most that analytics play a role but if you try and hang your hat on them as an absolute truth then you're only fooling yourselves.

The art of properly analysing stats takes years of University and then years of experience before you can even hope to be correct more than half of the time.

The truth about numbers is that they can be presented in such a way that proves a point but also in another way that disproves the same point. They are complimentary at the very best.

In terms of possession stats:

Possession is a key stat in all sports, but all sports also have a strategy, often called the counter-attack, that is built on NOT having possession and taking advantage of odd man situations created when the possession team is caught overloading (to keep the puck its necessary). This will always be effective against possession strategies. When these two strategies come head-to-head it will always be decided by execution.

Speedy teams are built to counter attack otherwise you aren't really taking advantage of your strengths. Of course this doesn't mean the goal is to concede possession at every chance like we tend to do. High shooting % come from the odd-man rushes that the counter-attack creates. The Leafs problem is that we may not have the defence to sustain this strategy over 82.

Possession is overrated. its efficiency that matters.

 

[Penalty Kill Icing*]

Since James Mirtle is a proponent of advanced stats and views this thread, I propose following questions:

1. In 2013 season, top 5 teams in PDO were Pittsburgh, Toronto, Chicago, Columbus, and Anaheim. Since PDO is shooting % + save %, I believe the luck factor kicks in (as claimed by supporter of these stats), and that these teams should regress. Does that mean that team like New Jersey that finished 28th in the list didn't have luck?

2. Let us change to Fenwick now (which is corsi minus the blocked shots). This one had New Jersey, Los Angeles, Chicago, Boston and Montreal in top 5. While Toronto was 30th. How did it happen, that NJ did better than every team in the league in this stat, yet it was ranked 9 points behind Toronto which was the last placed team in this statistic. How can a team, ranked 1st by Fenwick end up being the 9th worst? And the worst ranked team end up being the 9th best?

3. Corsi (since it is another version of Fenwick) tells the idential story. And hence, I pose the same questions.

4. One common thing I noted in all these stats was that Chicago was a common, thereby one could say that Chicago would be a good pick to win it all. But then I ask, do I really need all these advanced stats to predict that? Their point total and goal differential are enough for one to conclude that they are a really good team, and thus a good bet for cup champions.

5. In 2011-12, LA was ranked top 5 in all stats. But no-one then had picked them to win the cup, and they surprised everyone indeed. Goaltender (for the most part) doesn't affect Fenwick/Corsi, and we all know how Quick played then. My point is, LA had advanced stats in their favour all this time, but only in playoffs when Quick put up Conn Smythe performance; did that work by skaters amount to anything. Also, LA was top 5 in these stats, yet made the playoffs as an 8th seed. Makes you question if these stats are reliable.

5. Let us go further back. In 2010-11, Vancouver, Washington, Philly, Pittsburgh and SanJose ranked top 5 in NHL. Playoffs had Bruins, Canucks, Sharks and Lightning as top 4 teams. In comparison, top 5 teams according to Fenwick were SJ, Chicago, Tampa bay, Detroit and CALGARY. Corsi had San Jose, Detroit, Pittsburgh, Chicago, Vancouver. PDO (or the luckiest team) had Boston, Vancouver, Nashville, Phoenix and Philadelphia. I don't see much correlation between the rankings by these "advanced stats" and the final standings (be it regular season or playoffs). Yes, PDO had the 2 cup finalists ranked at the top, but when the Leafs are at the top of that list, why do we have articles that say "Leafs will regress" rather than "Leafs will win the cup" (which I don't think they were going to win last year anyways).

My point by raising these questions is simple: I see more noise than signal with the data that is presented to me by these stats. I see more anomaly than trends in these data. At the end of the day, I find goal differrential to be a far more reliable stat than whatever these advanced stats are trying to show me. Every year in past 6-7 years ( I went only that far back), the team with best goal differential won the president's trophy. Do advanced stat guarantee that? Further, at least one of the top 4 teams in goal differential makes the cup final.


Bottom line: These so called advanced stats don't show any trend to me. I see more flaws with this than trends. Hence, I don't want to rely on these bunch of numbers (is what I will call them) for even a tiny moment.

 

[The Legend]

Since James Mirtle is a proponent of advanced stats and views this thread, I propose following questions:

1. In 2013 season, top 5 teams in PDO were Pittsburgh, Toronto, Chicago, Columbus, and Anaheim. Since PDO is shooting % + save %, I believe the luck factor kicks in (as claimed by supporter of these stats), and that these teams should regress. Does that mean that team like New Jersey that finished 28th in the list didn't have luck?

2. Let us change to Fenwick now (which is corsi minus the blocked shots). This one had New Jersey, Los Angeles, Chicago, Boston and Montreal in top 5. While Toronto was 30th. How did it happen, that NJ did better than every team in the league in this stat, yet it was ranked 9 points behind Toronto which was the last placed team in this statistic. How can a team, ranked 1st by Fenwick end up being the 9th worst? And the worst ranked team end up being the 9th best?

3. Corsi (since it is another version of Fenwick) tells the idential story. And hence, I pose the same questions.

4. One common thing I noted in all these stats was that Chicago was a common, thereby one could say that Chicago would be a good pick to win it all. But then I ask, do I really need all these advanced stats to predict that? Their point total and goal differential are enough for one to conclude that they are a really good team, and thus a good bet for cup champions.

5. In 2011-12, LA was ranked top 5 in all stats. But no-one then had picked them to win the cup, and they surprised everyone indeed. Goaltender (for the most part) doesn't affect Fenwick/Corsi, and we all know how Quick played then. My point is, LA had advanced stats in their favour all this time, but only in playoffs when Quick put up Conn Smythe performance; did that work by skaters amount to anything. Also, LA was top 5 in these stats, yet made the playoffs as an 8th seed. Makes you question if these stats are reliable.

5. Let us go further back. In 2010-11, Vancouver, Washington, Philly, Pittsburgh and SanJose ranked top 5 in NHL. Playoffs had Bruins, Canucks, Sharks and Lightning as top 4 teams. In comparison, top 5 teams according to Fenwick were SJ, Chicago, Tampa bay, Detroit and CALGARY. Corsi had San Jose, Detroit, Pittsburgh, Chicago, Vancouver. PDO (or the luckiest team) had Boston, Vancouver, Nashville, Phoenix and Philadelphia. I don't see much correlation between the rankings by these "advanced stats" and the final standings (be it regular season or playoffs). Yes, PDO had the 2 cup finalists ranked at the top, but when the Leafs are at the top of that list, why do we have articles that say "Leafs will regress" rather than "Leafs will win the cup" (which I don't think they were going to win last year anyways).

My point by raising these questions is simple: I see more noise than signal with the data that is presented to me by these stats. I see more anomaly than trends in these data. At the end of the day, I find goal differrential to be a far more reliable stat than whatever these advanced stats are trying to show me. Every year in past 6-7 years ( I went only that far back), the team with best goal differential won the president's trophy. Do advanced stat guarantee that? Further, at least one of the top 4 teams in goal differential makes the cup final.


Bottom line: These so called advanced stats don't show any trend to me. I see more flaws with this than trends. Hence, I don't want to rely on these bunch of numbers (is what I will call them) for even a tiny moment.

You seem to quote quite a large number of statistics that correlated to success....so I'm not sure what you're talking about with not seeing a trend.

 

[Penalty Kill Icing*]

You seem to quote quite a large number of statistics that correlated to success....so I'm not sure what you're talking about with not seeing a trend.

My argument is that advanced stats aren't telling me anything that I don't know of. And I see more anomaly in these advanced stats than I see in regular stats like goal differential.

I didn't set out to say what happens is complete opposite of what advanced stats indicate. Since I already wrote out a long post, I won't repeat all of it again. I'd like to highlight how advanced stats were wrong in predicting NJ v/s Toronto's performance.

I am a math geek myself, but I raise strong objection to these advanced stat proponent who view them as 'end all be all', while I don't see any empirical proof of them being any better than a stat like, say, goal differential.

Maybe it does need more research. But definitely not in the avenues contemporary statisticians are heading towards.

 

[Sproo]

Logically - I agree with you. It makes sense that human beings might push themselves a bit harder if they are feeling "it". But the results simply don't back this up. It's possible that - as human beings - feeling "it" leads to overconfidence, which leads to poor decision making. It's also possible that high level athletes don't actually deal with those emotions.

What do you mean "the results simply don't back this up"? Not only does it seem intuitively true that people perform better when they are confident, but there is also a lot of research that backs this up.

Also, people almost universally overrate their own abilities. My guess is we do this in an attempt to capitalize on this phenomenon- If I boost my own confidence by convincing myself that I am better than I am at something, then I will likely perform a little bit better.

And of course high level athletes have to deal with emotions. Everyone has to deal with emotions (unless you have some severe cognitive impairments).

 

[Jimmy Firecracker]

Logically - I agree with you. It makes sense that human beings might push themselves a bit harder if they are feeling "it". But the results simply don't back this up. It's possible that - as human beings - feeling "it" leads to overconfidence, which leads to poor decision making. It's also possible that high level athletes don't actually deal with those emotions.

How can one even quantify human emotion and momentum? It's not something that can be measured or counted. High level athletes don't deal with emotion? Are you serious? We're talking about one of if not the most emotionally charged game on the planet!

No evidence or results to back up momentum? I'll give you a few.

Boston's comeback against us in Game 7. They scored to make it 4-2, and the Leafs were on their heels the rest of regulation, with Boston not letting up on the gas pedal.

Toronto vs. Montreal opening night this season. After Parros was injured, neither team was the same afterwards, and the game was much slower paced until the final minute when the Habs were trying to tie it up.

Game with Toronto vs NYI last year. After two fights that had happened, the momentum switched to the Islanders favour, and they ended up coming back and winning that game.

Sitting there, and telling people emotion and momentum don't exist is just plain wrong, and I question whetheer or not you've ever played sports in your life.

 

[Sproo]

Ah, yes, this old bit.

This stuff isn't going away - in fact, the work I see some people doing online is getting more impressive (and accurate) than ever. It's fine if some people aren't interested and find this a bit arcane - that I understand. It's a niche. But these constant potshots aren't fair and have little basis in reality.

I think you have it a bit backwards here. It's the corsi enthusiasts that are telling people that the traditional idea of assessing teams and players based on, you know, actually observing a hockey game and checking the standings is arcane. It is pretty arrogant.

Hockey is a very dynamic of a game, and has many variables which change from shift to shift, and second to second. A good model should capture all these variables. It is literally exponentially more complex to accurately model a hockey game than it is to model something like a baseball game which has discrete outcomes for every pitch. The traditional NHL stat categories does a terrible job capturing all these things, and now, the advanced stats do a slightly less terrible job. Neither will soon replace the need for serious hockey people to make player/team assessments based on watching hockey.