Speculation: Oilers hire Tyler Dellow

[Ogopogo*]

The purpose of corsi, from my understanding, is that it is an approximation for possession. Within that there is value. Obviously it is not a 100% accurate in any circumstance, however it tends to give us a larger sample size than strictly shots on goal while also seeming to be fairly accurate with regards to both good and bad players. It accurately assesses the lack of value a guy like Luke Gazdic has, while demonstrating the value of having an Anze Kopitar.

If Team A shoots 40 times but only 24 hit the net, while Team B hits the net with all 4 of their shots they'll be given the same shot total but who would you assume had possession of the puck more often? At a team level, that's all corsi really is. At an individual level I can see the flaws that it may have, but it still works in the same way. Yes, there are 10 players on the ice but certain players are obviously gonna be better at helping move the puck in a positive direction.

Also i'm not sure what you meant by "nobody takes the time to measure individual contributions".

Corsi doesn't measure possession, it measures shot attempts - those are two different things. I see it as arrogance from the Corsi community to even refer to their flawed numbers as possession numbers. Possession can be measured in several ways, but shot attempts is not one of them. Possession can be time of possession or even number of touches.

If Kopitar played with Gazdic on his wing all season long - every time one of them is on the ice the other is as well - what do you say then? Kopitar and Gazdic both have the same Corsi - who is good? Who is not? That is the problem with Corsi - put a bum out with good players and his Corsi is good. Put a good player out with bums and his Corsi suffers - it is NOT an individual measure no matter how much the cult wants to believe it. Corsi is a guess at best. It would be like saying that the murder rate in Edmonton has dropped since I moved here so I have a good Corsi for murder rates. Really? Is that all on me? I am a small part of the whole - it makes no sense to pin it on me. The concept is the same with 12 players on the ice as it is with 1.2 million people in the metro Edmonton area. You cannot pin a group stat on an individual - it is completely inaccurate and little more than a guess.

When I say "nobody takes the time to measure individual contributions" I mean nobody measures what individuals do on the ice. We can all see when players make mistakes - whether they lead to shots, goals or lead to nothing at all. We can also see when players make a particularly good play - something above average so to speak. Those may lead to goals, shot attempts or nothing as well. If we were to accumulate all of the + marks for a good play and subtract all of the - marks for a poor play for each individual then we would have some real individual measurement. It would be very easy to see individual contributions or lack thereof for each player in the NHL. But, instead, people would rather use the lazy man's Corsi and have numbers that really don't mean anything because they are team stats.

The entire point of "advanced stats" is to provide insight. The idea of "Moneyball" is what people are trying to replicate for hockey and they are doing it completely wrong. Would it make any sense at all to calculate the number of hits against a team and blame that on the 3rd baseman - like Corsi does? Absolutely not. What about the hits for a team and credit that to the same 3rd baseman? Is it really all on him? Nope.

Baseball is wonderful in that, it is easy to measure accomplishments of the hitters because hitting is an individual act within a team concept. If I hit, you can measure it. If I strike out, you can measure it. All of that is completely on me. In hockey, aside from goals and assists, there really aren't any other numbers that you can credit an individual for. Corsi is a very weak attempt to pin an all-around performance measure on an individual - it is like giving a 3rd baseman a score for the teams hits for and against. To measure true individual performance you have to watch the game, hockey is a sport where you cannot quantify an all around performer with a ridiculous team stat like Corsi. A true +/- of actual individual play where a player gets credit for good plays and debited for gaffes is the way to do it. Corsi will never be the answer - the number is influenced far too greatly by the other 11 players on the ice. My idea isolates an individual and measures them on their performance alone - that is what the entire point of advanced analytics is. Find the undervalued players with better ways to measure - not confuse the world with crap like Corsi.

 

[SPIRIT]

The idea of "Moneyball" is what people are trying to replicate for hockey and they are doing it completely wrong. Would it make any sense at all to calculate the number of hits against a team and blame that on the 3rd baseman - like Corsi does? Absolutely not. What about the hits for a team and credit that to the same 3rd baseman? Is it really all on him? Nope.

Baseball is wonderful in that, it is easy to measure accomplishments of the hitters because hitting is an individual act within a team concept. If I hit, you can measure it. If I strike out, you can measure it. All of that is completely on me. In hockey, aside from goals and assists, there really aren't any other numbers that you can credit an individual for. Corsi is a very weak attempt to pin an all-around performance measure on an individual - it is like giving a 3rd baseman a score for the teams hits for and against. To measure true individual performance you have to watch the game, hockey is a sport where you cannot quantify an all around performer with a ridiculous team stat like Corsi. A true +/- of actual individual play where a player gets credit for good plays and debited for gaffes is the way to do it. Corsi will never be the answer - the number is influenced far too greatly by the other 11 players on the ice. My idea isolates an individual and measures them on their performance alone - that is what the entire point of advanced analytics is. Find the undervalued players with better ways to measure - not confuse the world with crap like Corsi.

Excellent post.

The idea of comparing Moneyball to Moneypuck misses a massive fundamental part of what the Moneyball concept was.

They determined that a previously unrecorded and cared about statistic, on-base - when isolated - was a massive contributing factor to success.

This is not possible in the hockey world. There is no magic algorithm, other than "if you score more, you win".

You are unable to isolate variables in hockey like you can in baseball. It's a wash.

The second you get a new set of circumstances, you change any relevance previous isolation of the variables have revealed to you.

A new line-mate, a bump on your knee, that changes e-v-e-r-y-t-h-i-n-g.

You say, "but sure, it is about trends".

That's hog-wash.

You take the Babcock model of "what have you done for me lately".

Advanced stats can take a hike.

Most people here attack Dellows because they're envious of the fabled 'blogger-turned-Oiler-executive' path he took.

A smart man with a keen hockey eye, who knows his stats.

This is no better than a smart man with a keen hockey eye.

 

[azashi]

Excellent post.

The idea of comparing Moneyball to Moneypuck misses a massive fundamental part of what the Moneyball concept was.

They determined that a previously unrecorded and cared about statistic, on-base - when isolated - was a massive contributing factor to success.

This is not possible in the hockey world. There is no magic algorithm, other than "if you score more, you win".

You are unable to isolate variables in hockey like you can in hockey. It's a wash.

The second you get a new set of circumstances, you change any relevance previous isolation of the variables have revealed to you.

A new line-mate, a bump on your knee, that changes e-v-e-r-y-t-h-i-n-g.

You say, "but sure, it is about trends".

That's hog-wash.

You take the Babcock model of "what have you done for me lately".

Advanced stats can take a hike.

Most people here attack Dellows because they're envious of the fabled 'blogger-turned-Oiler-executive' path he took.

A smart man with a keen hockey eye, who knows his stats.

This is no better than a smart man with a keen hockey eye.

Ugh, man, both these posts are so one-sided.

First of all, let me be clear on this - Corsi is not the be-all, end-all of stats. There are serious weaknesses involved with it, particularly if teams start paying players for Corsi, or playing for Corsi. Other advanced stats, like PDO, are borderline useless the way they're being thrown around by some people.

Corsi is a great stat, with **all things being equal**. The moment that teams start pumping shots from crappy locations in order to bump their Corsi, or that individual players start shooting muffins from the point to get a better Corsi, it loses its value.

But let's look at your post in detail.

>They determined that a previously unrecorded and cared about statistic, on-base - when isolated - was a massive contributing factor to success.

Not exactly correct. Advanced stats in baseball were about showing how players were under/over valued based on traditional stats, permitting teams to sign quality players (with flaws, according to traditional stats and thinking) for much less money.

Baseball is the ideal scenario for advanced statistics, but given the success of advanced stats in football and particularly basketball, it is obvious that advanced stats apply to fluid games.

>This is not possible in the hockey world. There is no magic algorithm, other than "if you score more, you win".

With all due respect, your first sentence is completely wrong. Chicago and LA are huge proponents of advanced stats and have won 4 of the last 5 Stanley Cups. Your second sentence applies to every single team sport on the planet, but is incomplete. What your second sentence fails to address is *how* to score more.

>You are unable to isolate variables in hockey like you can in hockey. It's a wash.

I assume you meant "You are unable to isolate variables in hockey like you can in baseball". Which is false. It's more difficult, but it's been done, it's been applied, and it's resulted in success (remember when everyone thought that losing Byfuglien, Campbell, and Versteeg meant the end of Chicago as a contender?) If you don't want to accept hockey as an example, look at basketball. Basketball went from "big man rebounds, isolation scorer with a second option" style of play in the 90s and early 2000s to a true team game in the 2010s, almost completely due to advanced stats. In basketball, it used to be all about the "next Jordan", now it's about "the next Spurs".

>The second you get a new set of circumstances, you change any relevance previous isolation of the variables have revealed to you.
>A new line-mate, a bump on your knee, that changes e-v-e-r-y-t-h-i-n-g.

First of all, that's a significant exaggeration. Putting McSorley on Gretzky's left wing doesn't turn Gretzky into Horcoff or Belanger. It just means that McSorley isn't going to bury nearly as many pucks as Messier, so Gretzky isn't going to get as many assists, but conversely McSorley is going to put up a whole lot more points than he would if he was playing with Billy Carroll.

Second, you're failing to account for the law of averages. Crosby isn't healthy all the time, neither are Malkin, Toews, or Kopitar. The value of advanced stats in hockey, today, isn't about what the players did last week or last month, but what they did the last season, two, or three.

I want to finish this post by stating that I absolutely I agree with some - not all - of the sentiments you and Norm are exhibiting. Willis, Lowetide, and to a lesser extent Dellow have become fascinated by advanced stats as the next iteration of baseball's OPS or basketball's PER. Hockey is a more complicated game than either baseball or basketball, but you can still measure advanced statistics in it. Where advanced stats fall apart in hockey is if you start playing for the advanced stats without context. PDO and raw Corsi are prime examples of stats that are deeply misunderstood by both the deniers and the proponents. The deniers refuse to give the stats any value, while the proponents fail to acknowledge the inherent weaknesses in a system that gives no consideration to shot quality or make assumptions that "everything will average out" - like they do with PDO.

 

[Replacement*]

^

I would respond in more detail to some of the above except that its nearly impossible to parse when somebody doesn't properly use quote tags. Really this isn't hard to learn. If you learned that much about advanced stats you can learn how to wrap quotes instead of the mess above. thanks

Anyway heres the LA Kings argument coming again and now citing that they are "huge proponents" of advanced stats.

No wait, they use advanced stats. Saying more than that is an attempt to make a more convincing argument except that it isn't. I'll go out on a line here saying you are either in your 20's or 30's. I state that because you seem to have no familiarity with what the main cause and effect is in LA. It may not even occur to you as you watch the Kings play that they play Sutter hockey. The same sensible gameplan most Sutter reknowned teams played but especially the NYI who were the Al Arbour blueprint for subsequent Sutter coaching of the game.

In fact even Daryl Sutters previous coaching gig, he had enormous success coaching a fairly limited team all the way to the SC final. A performance so impressive that the LA KINGS convinced Sutter to come out of retirement and coach the LA KINGS. Because they knew his model (the same one he was using in Calgary) was going to be what the Kings required.

If one watches the Islanders of old, or that flames team, or the Kings the strategic lineages are clear. Get puck deep, limit turnovers, play north south. When you don't have the puck fall back, eliminate time and space, pressure over the ice. When you do have the puck utilize solid puck support and quick passing game. This isn't due to advanced stats its an Islanders blueprint that Sutters brother went over each summer and discussed in detail. Its the same damn system play. Oiler fans should recognize it.

If we want to look for causality in the Hawks and Kings success theres more significant variables at work than "Advanced stats". Namely solid D first of all. Strength at Center and solid coaching. Both clubs also look hard for players that will fit well into a team concept. Both orgs have an eye for this kind of player.

To state as an argument that the "Hawks and Kings are huge proponents of advanced stats and have won 4 of the last 4 SC's: makes it seem as if you are invoking that as a chief reason for the respective successes. That in turn seems to be a disingenuous argument, that by invocation, makes it seem you are lacking in more substantiated arguments. In other words its a weak argument.

 

[harpoon]

Still, if you're Spector and Tychkowski your comments are public, and engaged in with respect to employ. Maybe I'm old fashioned and feel that people in such employ should act somewhat responsibly in respect of their credibility (unlike myself as a random messageboard poster) but I'm raising there is a cost to Tych and Spector doing this if they are proved wrong. Its one thing to have a dislike and to further rumor, its another if it has connection with what you do for a living. Why engage in that cost.

This warrants repeating. Certain media types (and to be fair, certain posters at this site) have a personal dislike for Dellow that extends well into the unreasonable.

In order to break down game film and provide raw stats Dellow does not need to be in Edmonton.

In order to break down game film and provide raw stats Dellow does not need to be a nice person and a cordial fellow poster.

Hilarious that Spector and Tychkowski outed themselves as having no clue about the workings of the Oilers ... and as being jealous little weasels to boot.

 

[tempest2i]

This warrants repeating. Certain media types (and to be fair, certain posters at this site) have a personal dislike for Dellow that extends well into the unreasonable.

In order to break down game film and provide raw stats Dellow does not need to be in Edmonton.

In order to break down game film and provide raw stats Dellow does not need to be a nice person and a cordial fellow poster.

Hilarious that Spector and Tychkowski outed themselves as having no clue about the workings of the Oilers ... and as being jealous little weasels to boot.

 

[Cloned]

Edited.

 

[Replacement*]

This warrants repeating. Certain media types (and to be fair, certain posters at this site) have a personal dislike for Dellow that extends well into the unreasonable.

In order to break down game film and provide raw stats Dellow does not need to be in Edmonton.

In order to break down game film and provide raw stats Dellow does not need to be a nice person and a cordial fellow poster.

Hilarious that Spector and Tychkowski outed themselves as having no clue about the workings of the Oilers ... and as being jealous little weasels to boot.

Yep. Still don't know why paid journalists would besmirch their reputation by engaging in such falsehood. (if indeed there is no smoke or fire)

I don't know if its because of the blurring of lines today between posting, blogging, tweeting, professionally writing but its a slippery slope for *professional* media to engage in.

The other factor is journalists that just post crap for more hits. I would think there would be inevitable cost to that but in terms of hits its not the case. Be purposely misleading in media and incite and state falsehoods and the hits just keeps coming. Although the credibility also takes a hit. If it wasn't lost before..

 

[syz]

Corsi doesn't measure possession, it measures shot attempts - those are two different things. I see it as arrogance from the Corsi community to even refer to their flawed numbers as possession numbers. Possession can be measured in several ways, but shot attempts is not one of them. Possession can be time of possession or even number of touches.

If Kopitar played with Gazdic on his wing all season long - every time one of them is on the ice the other is as well - what do you say then? Kopitar and Gazdic both have the same Corsi - who is good? Who is not? That is the problem with Corsi - put a bum out with good players and his Corsi is good. Put a good player out with bums and his Corsi suffers - it is NOT an individual measure no matter how much the cult wants to believe it. Corsi is a guess at best. It would be like saying that the murder rate in Edmonton has dropped since I moved here so I have a good Corsi for murder rates. Really? Is that all on me? I am a small part of the whole - it makes no sense to pin it on me. The concept is the same with 12 players on the ice as it is with 1.2 million people in the metro Edmonton area. You cannot pin a group stat on an individual - it is completely inaccurate and little more than a guess.

When I say "nobody takes the time to measure individual contributions" I mean nobody measures what individuals do on the ice. We can all see when players make mistakes - whether they lead to shots, goals or lead to nothing at all. We can also see when players make a particularly good play - something above average so to speak. Those may lead to goals, shot attempts or nothing as well. If we were to accumulate all of the + marks for a good play and subtract all of the - marks for a poor play for each individual then we would have some real individual measurement. It would be very easy to see individual contributions or lack thereof for each player in the NHL. But, instead, people would rather use the lazy man's Corsi and have numbers that really don't mean anything because they are team stats.

The entire point of "advanced stats" is to provide insight. The idea of "Moneyball" is what people are trying to replicate for hockey and they are doing it completely wrong. Would it make any sense at all to calculate the number of hits against a team and blame that on the 3rd baseman - like Corsi does? Absolutely not. What about the hits for a team and credit that to the same 3rd baseman? Is it really all on him? Nope.

Baseball is wonderful in that, it is easy to measure accomplishments of the hitters because hitting is an individual act within a team concept. If I hit, you can measure it. If I strike out, you can measure it. All of that is completely on me. In hockey, aside from goals and assists, there really aren't any other numbers that you can credit an individual for. Corsi is a very weak attempt to pin an all-around performance measure on an individual - it is like giving a 3rd baseman a score for the teams hits for and against. To measure true individual performance you have to watch the game, hockey is a sport where you cannot quantify an all around performer with a ridiculous team stat like Corsi. A true +/- of actual individual play where a player gets credit for good plays and debited for gaffes is the way to do it. Corsi will never be the answer - the number is influenced far too greatly by the other 11 players on the ice. My idea isolates an individual and measures them on their performance alone - that is what the entire point of advanced analytics is. Find the undervalued players with better ways to measure - not confuse the world with crap like Corsi.

Nothing like a far-flung hypothetical being used to claim that something else is a far-flung hypothetical.

I wish logic classes were mandatory.

 

[Ogopogo*]

Nothing like a far-flung hypothetical being used to claim that something else is a far-flung hypothetical.

I wish logic classes were mandatory.

Yes, you could use them.

 

[Bryanbryoil]

Even if the Oilers did fire mudcrutch that doesn't mean they stopped using analytics. The Oilers currently have (to public knowledge) at least three different groups of analytics. One is Dark Horse analytics, another is a group of bloggers and hockey writers and the other is Dellow. I believe Dellow was hired as Eakins' main guy.

Teams will always look for an edge. Perhaps Dellow is an inferior analytics guy to Dark Horse if his voice has less clout than it initially did? We have no idea really. It's about thinking out of the box and also figuring out which players bring the most bang for the buck. For me I look at our team and feel that we need guys that can win battles consistently, crash the net and break the puck out of our end. Every team should know who's on the trade block and who potential UFA's are. Scout the crap out of them, see which guys fit what we need and then do your best to go out and get them. When you target a guy like Nikitin initially that is a pretty telling sign that someone dropped the ball, likely Howson in this case.

This. I'm still waiting for someone to explain to me why (shot attempts for - shot attempts against) is such a bogus stat? Is it slightly disingenuous for analyzing individual play, although Corsi Rel tries to alleviate this, yeah sure; but for team play, corsi is pretty damn intelligent.

MessierII sums it up nicely. This is why if anyone wants to do this right they will hire a bunch of individuals to keep track of significant stats league wide. Ie. passes that get to guys in prime scoring areas that aren't converted, goals/scoring chances, puck battles won/lost, etc.
The problem is that most bloggers obviously don't have these resources at their disposals so that is why I take what most of them have to say and corsi as a raw stat with a big grain of salt. Then you have the eye test, because I'm sorry but you can't tell me that EVERY analytics guy is = the best analytics guys, there's always a drop off so the data that one guy has based on his own eyes maybe completely different than the data from another guy. There are many things that teams keep track of beyond goals-assists-points and +/- but corsi is a flawed stat with the exception for what it was intended for, goalie workload evaluation.

The big hole in advanced stats is the scoring chance. The point of hockey is to maximize your own scoring chances and minimize your opponents. All shot attempts aren't equal that's why it's a misleading stat,

 

[Bryanbryoil]

A prime example , we could outshoot a team 40-2 and lose 1-0 if they get 2 shots in the slot and all of ours are from the blueline with no screens or deflections. It's the quality of the shot and the circumstances surrounding it that matter the most. I'm not sure that analytics would call for a big body presence in front of the net to help our PP for example when it is much needed and hopefully Pouliot fills that role now.

 

[Replacement*]

Anybody else find it odd that such terms as *ADVANCED* stats, Hockey Analytics etc are used as the name of these things?

That's as if to say no other forms of hockey stats are advanced. It speaks to self aggrandizing turf claim. jmo
Next, Hockey Analytics somewhat infers that its the sole way to analyze hockey.

Even the names of these things are offputting to me.

Moneyball was never laying claim to already in use vernacular in baseball or inferring anything overtly assumptive about itself in title. Other than perhaps fiscal efficacy through stats which is not exactly what we're getting in Hockey with 4.5M advanced Pouliot analytic contracts...

 

[Gunnersaurus Rex]

Here's your advance stat... W v L

The rest is a bunch of crap

 

[McGoMcD]

Here's your advance stat... W v L

The rest is a bunch of crap

No offense, but this is nice to posture and say. However what is a GM suppose to do exaclty to put a team that wins more games than it loses together? Obviously it is a good idea to look at stats or should you just get all the 4th liners from the best teams in hockey and expect to win?

I think people are just too black and white. Stats are all about gray areas, you are better of using them than not, you just have to realize when the phenomenon you seek to understand is highly correlated with other things and can't be isolated.

 

[Gunnersaurus Rex]

No offense, but this is nice to posture and say. However what is a GM suppose to do exaclty to put a team that wins more games than it loses together? Obviously it is a good idea to look at stats or should you just get all the 4th liners from the best teams in hockey and expect to win?

I think people are just too black and white. Stats are all about gray areas, you are better of using them than not, you just have to realize when the phenomenon you seek to understand is highly correlated with other things and can't be isolated.

This is the problem in my opinion. Paralysis by analysis. Too much looking at the grey areas. If, as a GM you cannot tell if a player is good or bad by watching him, your should not be a GM. We all know the good players from the bad players. We don't need advance stats to tell us that.

If you watch a player and he sucks, but then after the game you hear the advance stats that say he's a potential Conn Smythe winner, how is that relevant?

 

[McGoMcD]

This is the problem in my opinion. Paralysis by analysis. Too much looking at the grey areas. If, as a GM you cannot tell if a player is good or bad by watching him, your should not be a GM. We all know the good players from the bad players. We don't need advance stats to tell us that.

If you watch a player and he sucks, but then after the game you hear the advance stats that say he's a potential Conn Smythe winner, how is that relevant?

Sure, I agree. I don't think the advanced stats are saying a guy who sucks is great. Any way, point is you have to look at stats, no getting around that. We all admit the only thing that matters in the end is winning vs losing. The bottom line is every other GM is also looking for the best players, so you always have to try and get a leg up. I agree though, it is a mistake to over rely on the advanced stats, or any stat for that matter.

 

[TheRebuild]

All boils down to some nerd, doing geek stuff on a computer with spreadsheets. That ain't sports.

 

[McGoMcD]

All boils down to some nerd, doing geek stuff on a computer with spreadsheets. That ain't sports.

Haven't you watched the movies, Geeks always win in the end. Sure the jocks look unbeatable, but the nerds always win out.

 

[McQuixote]

>They determined that a previously unrecorded and cared about statistic, on-base - when isolated - was a massive contributing factor to success.

Not exactly correct. Advanced stats in baseball were about showing how players were under/over valued based on traditional stats, permitting teams to sign quality players (with flaws, according to traditional stats and thinking) for much less money.

Baseball is the ideal scenario for advanced statistics, but given the success of advanced stats in football and particularly basketball, it is obvious that advanced stats apply to fluid games.

Excellent point. I think this is where the debate gets lost in the trees. MONEYBALL isn't really about stats. It's about being forward thinking, exploiting market inefficiencies, investigating new sources of information and re-evaluating old ones, and letting go of your biases and paradigms.

But you can't stop doing this. You can't just have Corsi replace +/- or PDO replace sv% and leave it at that. You have to keep checking for these biases and paradigms (for the record, I think most advanced stats advocates are better at this than the detractors, but not always).

I think one of the problem with both camps is that many people, especially the laypeople, conflate the stats with the analysis, and forget to actually do the latter. And it is much harder in a game like hockey than baseball, though I don't think it's impossible.

 

[Master Lok]

All boils down to some nerd, doing geek stuff on a computer with spreadsheets. That ain't sports.

Don't worry, keep waiting goons are coming back into the picture I'm sure.