The Advanced Stats Thread Episode VI: RIP To Our Databases

[chosen]

Hockey analytics is way past it's infancy stage.

I'd bet that in 5 years, some of the "obvious" conclusions being reached now are no longer looked upon in the same way. Sure, I could be wrong, but that's the way I'd bet.

 

[ManUtdTobbe]

Sure, but that's the thing with any research based subject, not just hockey... Hockey analytics infancy was when Corsi was "invented", we have legit xG models now, that's not infancy.

 

[chosen]

The stats are good and 100% objective, it's how people use them that might not me good.
The best way to do it is to read up on analytics as a whole and use lots of stats to help form your opinion, don't look for a one column stat, they'll never paint the full picture.

+/- is a useless stat, don't ever use it for anything.

Edit: Also, you seem to be doing it wrong. If you look for stats to support your preconcieved notions (+/- agreeing with your opinion that Duguay was overrated in this case) then you're doing it wrong.

So, if I thought Girardi sucked and then saw that analytics supported it, I would be doing it wrong? Everyone has opinions before looking into the analytics. It's the way humans think. That will never go away. Otherwise there would be no point in watching. Scouts could just look at stats and make their selections based solely on that. All of us have preconceived notions. They might turn out to be wrong or they might not.

 

[chosen]

Sure, but that's the thing with any research based subject, not just hockey... Hockey analytics infancy was when Corsi was "invented", we have legit xG models now, that's not infancy.

Maybe infancy is the wrong word, then. I doubt it has reached its potential. How about calling it an adolescent?

 

[silverfish]

So, if I thought Girardi sucked and then saw that analytics supported it, I would be doing it wrong? Everyone has opinions before looking into the analytics. It's the way humans think. That will never go away. Otherwise there would be no point in watching. Scouts could just look at stats and make their selections based solely on that. All of us have preconceived notions. They might turn out to be wrong or they might not.

Everyone is going to use these metrics differently, right? We're human.

Basically, if you thought Girardi was good, and then the metrics said he was terrible, and you said to yourself: "these metrics are obviously flawed because Girardi is clearly good", then you'd be using it wrong.

Similarly, if you thought Girardi was bad, and then the metrics said he was bad, and you said to yourself: "ha! I'm obviously right because the metrics agree with me", then you'd also be using it wrong.

'Advanced' stats, to me, for the fans, should be to increase their knowledge of the game. What makes a player good, what makes a player bad, how can I challenge what I already know. For front offices, 'advanced' stats should be used as a way to complement decision making.

EDIT - (This was after Mac n Gs, Chosen, and ManUtdTobbe liked this post, for full disclosure, they may not agree with the below )

That being said(!) if for some reason, we lived in this weird vacuum where NHL front offices could ONLY use advanced stats or ONLY use their eye-test, I'd take the 100% stat FO over the eye FO any day of the week.

 

[silverfish]

I do not understand these commonplace 8 year massive deals that we're seeing. The Sabres effectively just bought four years of Eichel UFA. He'll be 30 when the deal expires.

Then there's the discussion of whether or not Eichel is even worth $10m AAV . Has he proven that yet? His 5v5 numbers leave a lot to be desired. Across his two years in the league he has 57 5v5 points with a 1.6 points per 60. That's 2nd liner production. I have no doubts that Eichel will be one of the league's best, but, I don't know if I'd have thrown him a $10m AAV right away. How much more could he possibly demand if say, you bridge him after his ELC? $12m AAV? At least you buy more UFA years and if his 5v5 production doesn't increase, and his PP production decreases, then maybe you save yourself some money.

 

[Machinehead]

You need a wealth of +/- evidence for me to buy into it, but if you're a minus player every year for literally several years in a row -*cough* GIRARDI IN THE PLAYOFFS! *cough*- then there might be -*ahem* Cam Fowler *hack, sputter*- something to it.

Sorry, got something in my throat there.

But then again, very few players who are bad at corsi have a good +/- and vice versa, because corsi leads to real results in the long-run.

Unless you're Kris Russell. Still not sure what to make of him.

 

[Off Sides]

They are Buffalo, if they somehow screwed up and Eichel decided he wanted out as soon as he could hit UFA or by selling off only 1 UFA year, that would be worse for them than overpaying to get 8 years.

____________________


I don't post often in this thread but if the individual player stats are objective, why not use other objective math in order for a team to decide if they should rent, sell, sign a player, make a trade, etc. Basically use objective math to roster build towards the goal of winning a cup in some future year instead of just going with subjective stuff like we think we are pretty good and adding this or that will make us better?

If the goal is to win a cup and the math says the probability your team is going to have to beat the odds two or three times in different playoff series in order to win a cup that year, why would a team rent? If the math still says even after you make a trade the probability did not improve above the other teams who already had a better probability, why make that trade?

GM, coaches, some fans use the phrase anything can happen, while that is true, the probability of anything happening... a team who is an underdog in two or more playoff series out of the four they have to win, actually winning them all, the probability must be extremely low.

If the probability of a team selling and using those picks in some fashion raise a future probability of winning 4 series some other year, wouldn't the smart move be to go that direction instead of just subjectively believing anything can happen?

I guess that rambling post is more meant to question, if one of the reasons to use stats is to take away subjectivity, why only use it for player evaluation instead of using something similar for a teams probability to win a cup and then using that objectivity to make moves? Why is subjectively saying anything can happen, or he's a good player to rent, or sign, etc, be okay if it's not okay to use subjectivity in player evaluations?

 

[silverfish]

31 different definitions of scoring chances...

https://www.nhl.com/bluejackets/news/scoring-chances-october-3/c-291513024

I don't post often in this thread but if the individual player stats are objective, why not use other objective math in order for a team to decide if they should rent, sell, sign a player, make a trade, etc. Basically use objective math to roster build towards the goal of winning a cup in some future year instead of just going with subjective stuff like we think we are pretty good and adding this or that will make us better?

If the goal is to win a cup and the math says the probability your team is going to have to beat the odds two or three times in different playoff series in order to win a cup that year, why would a team rent? If the math still says even after you make a trade the probability did not improve above the other teams who already had a better probability, why make that trade?

GM, coaches, some fans use the phrase anything can happen, while that is true, the probability of anything happening... a team who is an underdog in two or more playoff series out of the four they have to win, actually winning them all, the probability must be extremely low.

If the probability of a team selling and using those picks in some fashion raise a future probability of winning 4 series some other year, wouldn't the smart move be to go that direction instead of just subjectively believing anything can happen?

I guess that rambling post is more meant to question, if one of the reasons to use stats is to take away subjectivity, why only use it for player evaluation instead of using something similar for a teams probability to win a cup and then using that objectivity to make moves? Why is subjectively saying anything can happen, or he's a good player to rent, or sign, etc, be okay if it's not okay to use subjectivity in player evaluations?

They are. Statistical modeling is occurring everywhere and readily available. Odds that one team will win a certain game, make the playoffs, make the finals, win the cup. It's out there, and being done.

I think Manny is launching his on Corsica today for the start of the season.

 

[Greg02]

They are Buffalo, if they somehow screwed up and Eichel decided he wanted out as soon as he could hit UFA or by selling off only 1 UFA year, that would be worse for them than overpaying to get 8 years.

____________________


I don't post often in this thread but if the individual player stats are objective, why not use other objective math in order for a team to decide if they should rent, sell, sign a player, make a trade, etc. Basically use objective math to roster build towards the goal of winning a cup in some future year instead of just going with subjective stuff like we think we are pretty good and adding this or that will make us better?

If the goal is to win a cup and the math says the probability your team is going to have to beat the odds two or three times in different playoff series in order to win a cup that year, why would a team rent? If the math still says even after you make a trade the probability did not improve above the other teams who already had a better probability, why make that trade?

GM, coaches, some fans use the phrase anything can happen, while that is true, the probability of anything happening... a team who is an underdog in two or more playoff series out of the four they have to win, actually winning them all, the probability must be extremely low.

If the probability of a team selling and using those picks in some fashion raise a future probability of winning 4 series some other year, wouldn't the smart move be to go that direction instead of just subjectively believing anything can happen?

I guess that rambling post is more meant to question, if one of the reasons to use stats is to take away subjectivity, why only use it for player evaluation instead of using something similar for a teams probability to win a cup and then using that objectivity to make moves? Why is subjectively saying anything can happen, or he's a good player to rent, or sign, etc, be okay if it's not okay to use subjectivity in player evaluations?

There's a (small) group of people the use Bayesian probability to make their life decisions. It's fascinating but a bit reductive. The big problem with this is making sure that your priors are correct.

 

[Off Sides]

31 different definitions of scoring chances...

https://www.nhl.com/bluejackets/news/scoring-chances-october-3/c-291513024



They are. Statistical modeling is occurring everywhere and readily available. Odds that one team will win a certain game, make the playoffs, make the finals, win the cup. It's out there, and being done.

I think Manny is launching his on Corsica today for the start of the season.

I'd think that sort of model if it works would be pretty valuable, the old school thinking that just making a team better actually does a whole lot to improve the math towards them winning the ultimate goal has always been a pet peeve of mine.

Even today, many advanced stat proponents would like the team to fill up on players who do well within them, but my next question would be, is even that enough to really change a teams chances at winning if another team already has been doing that?

Seems to me there has to be timing to moves more so than just loading up. Sure it makes them them marginally better but is there a long term point to just improving a little given players age, move on, cap hits grow, etc

To use the Rangers as an example, subjectively I think they will be a decent to good team this year, but I don't think any amount of realistic moves puts them into top 3 in the east in terms of winning the cup, yet if they are even near to playoff position at the deadline they likely rent or at least will not sell. It just does not make sense to my subjective opinion to make, for example, a ~4th in the east team slightly better to still be ~4th in the east.

__________________________

Also think this will be an interesting year for the Rangers in terms of Shattenkirk and Smith in particular. Two players who played well within their roles according to the advanced stats but will now be asked to play well in a larger role, see if the theory that players like that were just being under utilized or incorrectly utilized or if they only put up those stats due to the role they were in. I'm not sure the quality of competition stats add up. If true QOC is near useless, imo there seems to be something wrong with how it is being weighed rather than it not having a relatively big impact.

 

[silverfish]

I'd think that sort of model if it works would be pretty valuable, the old school thinking that just making a team better actually does a whole lot to improve the math towards them winning the ultimate goal has always been a pet peeve of mine.

Even today, many advanced stat proponents would like the team to fill up on players who do well within them, but my next question would be, is even that enough to really change a teams chances at winning if another team already has been doing that?

Seems to me there has to be timing to moves more so than just loading up. Sure it makes them them marginally better but is there a long term point to just improving a little given players age, move on, cap hits grow, etc

To use the Rangers as an example, subjectively I think they will be a decent to good team this year, but I don't think any amount of realistic moves puts them into top 3 in the east in terms of winning the cup, yet if they are even near to playoff position at the deadline they likely rent or at least will not sell. It just does not make sense to my subjective opinion to make, for example, a ~4th in the east team slightly better to still be ~4th in the east.

__________________________

Also think this will be an interesting year for the Rangers in terms of Shattenkirk and Smith in particular. Two players who played well within their roles according to the advanced stats but will now be asked to play well in a larger role, see if the theory that players like that were just being under utilized or incorrectly utilized or if they only put up those stats due to the role they were in. I'm not sure the quality of competition stats add up. If true QOC is near useless, imo there seems to be something wrong with how it is being weighed rather than it not having a relatively big impact.

This just leads us into the normal discussion about any trade, right? You have to weigh the cost versus how much better it makes your team. Is a +2% increase at the Cup worth trading this asset? If yes, make the trade. If no, don't make the trade.

 

[Off Sides]

This just leads us into the normal discussion about any trade, right? You have to weigh the cost versus how much better it makes your team. Is a +2% increase at the Cup worth trading this asset? If yes, make the trade. If no, don't make the trade.

Basically yes, but more so does that 2% improvement by trade also lower the future chances by 2% or more? And does that 2% improvement put the team any % above the team who was already 2% above them pre-trade?

 

[silverfish]

Basically yes, but more so does that 2% improvement by trade also lower the future chances by 2% or more? And does that 2% improvement put the team any % above the team who was already 2% above them pre-trade?

Right. This still seems to me like your average, run-of-the-mill, trade thought. I'd expect any GM worth his weight to go about this process regarding transactions.

 

[Off Sides]

Right. This still seems to me like your average, run-of-the-mill, trade thought. I'd expect any GM worth his weight to go about this process regarding transactions.

I'm not so sure, seems to me even Smith for example last year, it was more like well we have an anything can happen chance perhaps, so let's just trade for him and hope for the best. If he works we will re-sign him to a pure UFA contract, if he does not work out, Eric Staal for example, well we gave it a shot.

 

[silverfish]

I'm not so sure, seems to me even Smith for example last year, it was more like well we have an anything can happen chance perhaps, so let's just trade for him and hope for the best. If he works we will re-sign him to a pure UFA contract, if he does not work out, Eric Staal for example, well we gave it a shot.

I think I'd disagree. But, if you really feel this way, then that's a brief glimpse into what you think of JG and the Rangers FO. And I don't mean to put you on the spot, I believe HFNYR overrates the absolute shit out of JG. He's on this weird, can do no wrong pedestal, and it's been that way since he was placed into the organization. When Sather was GM, every time Sather ****ed up, it was Glen's fault. Every time Sather made a good move, it was "wow, JG is really putting his mark on the organization!". People still blame Sather for the 2015 draft debacle, but do you really believe Gorton wasn't working the moves that day when they knew he was going to takeover the GM duties on 7/1? They let the outgoing Sather run that draft? I very much doubt this.

Hindsight is 20/20, but that '15 draft was horrendous. Talbot trade was horrible. Hagelin trade was horrible. I suspect no player selected by the Rangers in that draft will play more than 100 career NHL games. Gropp is trash. Kovacs already busted. Zboro is in the ECHL, Morrison hasn't developed at all and is playing an overage year right now, don't know enough about Bernhardt, too early to tell on Huska. Best pick of this draft likely ends up being Saarela.

Anyway, back on topic

I think the Rangers were keeping these things in mind when conducting both the Staal and the Smith trade. But there are hundreds, if not more, factors at play here. A 2% increase at the cup odds in a year where Henrik Lundqvist is still on the time might be worth the 2% decrease in a year down the line as a consequence of not having those draft picks.

 

[Off Sides]

I actually agree with most of your first paragraph. I think Gordon is better than Sather, but I don't see a real omg they fully committed to a direction that makes sense approach either.

The 2nd, I subjectively believe, and believed at the time, that none of that was worth it.

 

[silverfish]

Cody Franson: 1 yr @ $1m or Brendan Smith: 4 yrs @ 4.35m?

 

[Filthy Dangles]

good news about Eichel....this is a very repeatable skill that is contingent on talent. Shot and scoring chance metrics can vary greatly year by year based on linemates and team system/structure. Ability to carry the puck is as repeatable as it get.




also really like how Kelly melds the stats with the eye test.

Hockey analytics is way past it's infancy stage.

Debatable, depends on your definition. At their heart, they're still basically recorded shots logged by human beings.

Cody Franson: 1 yr @ $1m or Brendan Smith: 4 yrs @ 4.35m?

Neither please.

 

[Machinehead]

31 different definitions of scoring chances...

https://www.nhl.com/bluejackets/news/scoring-chances-october-3/c-291513024

And isn't it funny how everyone is good at their own definition?

 

[silverfish]

Thought I finished the trended shots plot. Scraped the first period of TML/WPG. Found a bug. Nice.

As I'm working on mine, though, I just want every to take a look at the ones on Corsica this year in the live game trackers. That thing is beautiful.

Fixed!

Not gonna have the heat map by tomorrow though. Oh well.

 

[Machinehead]

Thought I finished the trended shots plot. Scraped the first period of TML/WPG. Found a bug. Nice.

As I'm working on mine, though, I just want every to take a look at the ones on Corsica this year in the live game trackers. That thing is beautiful.

Now I can watch Hayes be horrible in full 4k!

 

[ManUtdTobbe]

Connor McDavid sported a 1.8 ixG last night, i'm not sure but that sounds insanely high.

 

[Mac n Gs]

Surprised it hasn't been said in here yet: Josh Manson signed a 4 year deal with a $4.1M AAV

 

[silverfish]

Surprised it hasn't been said in here yet: Josh Manson signed a 4 year deal with a $4.1M AAV

Worth every penny, and then some.