Dbrownss
Registered User
- Jan 5, 2014
- 31,359
- 8,734
Blues Trade Proposals Part XXXIII
- Thread starter EastonBlues22
- Start date
- Status
I wish people would learn that advanced stats are a tool for solidifying the eye test and they don't tell The Whole story, and can almost always be made to say anything.
Advanced stats are basically for confirming hypotheses. It's so frustrating seeing posters say a player is good or bad because of corsi or zone starts and ignore everything else.
In that thread one guy just lists zone starts and p/60 and argues parayko isn't that good, ignores that parayko had poor puck luck (low shooting %) and things that parayko does really well in the efront fensive zone such as breaking up rushs and recovering for others mistakes.
Advanced stats are basically for confirming hypotheses. It's so frustrating seeing posters say a player is good or bad because of corsi or zone starts and ignore everything else.
In that thread one guy just lists zone starts and p/60 and argues parayko isn't that good, ignores that parayko had poor puck luck (low shooting %) and things that parayko does really well in the efront fensive zone such as breaking up rushs and recovering for others mistakes.
Dbrownss
Registered User
- Jan 5, 2014
- 31,359
- 8,734
None of my dreams make sense but yeah, it happened.
I can only imagine waking up from that and saying....wtf.
i will admit that I dont know enough about AS to really make a counter argument using them but it just seems that they are entirely subjective.
stl76
No. 5 in your programs, No. 1 in your hearts
- Jul 2, 2015
- 9,068
- 8,352
I believe he looks at stats, makes up his mind, then ignores stats that don't confirm his opinion. Context be damned! Who can be bothered to actually watch players before forming immutable opinions about them, anyway?
This is really a terrible way to go about player evaluation. Like you say, advanced stats should be for confirming (or possibly negating...at the very least calling into question) hypotheses.
The best was that guy saying Gardiner and Parayko were both sheltered #3's when HE HIMSELF posted Gardiner had 53.3ozs% & 0:20pkTOI while Parayko had 49.6ozs% & 1:33pkTOI. One of these things is not like the other...
Last edited by a moderator:
It's easy to look at "advanced stats" (and I still take issue with that term... what's so "advanced" about counting?) and keep a narrow view, because for the entire history of hockey that's what fans have been taught to do: goals are goals, assists are assists, that's it, goodbye and goodnight. It takes waaay more time to look at and understand all of these other things, especially considering that there's no real, centralized, agreed-upon place to find this stuff. The NHL website is absolute garbage, the other "hobby sites" like hockeydb, natural stat trick, hockey-reference have a little of this, a little of that, and a lot of times they're counting things differently and it's not always readily evident how. Zone starts are a good example: some places split it up into offense/defense, some places add the neutral zone in it... so what happens to the NZ starts in the places that don't count that...?
For me, it doesn't completely ruin it. It's a good start, a good way to understand loosely and with implicit margins of error what is happening in games. People that act like they're the end all be all answer to everything don't understand what's going on, and imo ruin the conversation for people who actually have good things to say and use the info well. Most hockey fans aren't going to write a thesis on corsi every time they want to say something about it, and it's nearly impossible to qualify everything and account for every aspect when you're talking about it. But corsi (just as an example) isn't supposed to do that: it's a snapshot. Snapshots are valuable, they just aren't the answer to everything, and nobody who knows what they're talking about will ever pretend like they're supposed to be.
But dumb people will, and so now every time someone brings it up the reaction is "oh boy this guy is misusing the stat." And sometimes they are, But that's no reason to just write off the whole endeavor. Having more shots for than against is a good thing. Having more goals for than against is a good thing. It's not the only way to be good(!), but more often than not it's a quality and reliable indicator. If someone is good and their advanced stats are sub par, it just requires more information, it doesn't mean the whole thing is bunk. The same way Ratties run of fluke goals a few years back didn't all of a sudden make him Wayne Gretzky all of a sudden (that's a "non-advanced" stat, just a lot of goals going in off his butt over a few games), Schenn's low 5v5 p/60 last season didn't make him garbage all of a sudden. It's just an indication to take a deeper dive.
BUT, if you *do* take that deeper dive, and the circumstances match what the stat sheet is showing you, or if you adjust the stats to compensate for the circumstances, then we ought to be listening to what those numbers are telling us instead of dismissing them for voodoo. Idk, just my opinion I guess.
Majorityof1
Registered User
I don't have a problem with advanced stats, but I have a problem with the way many people use them.
First, they represent only the sliver of the whole picture. Take GF%. Player A has a >50% GF%, and player B has <50%. Well looking at that one stat, player A is the superior player. Its obvious, when he is on the ice, goals get scored for his team more than against, and that is the point of hockey. However, what if I told you that player A was playing on a really good team, with 2 great linemates, playing against lesser competition, and getting exclusively offensive opportunities while player B was on a horrible team playing with horrible players, getting buried in the d-zone in the toughest situations against the best the NHL has to offer. Now, who is the better player? Its not so black and white when you add that extra data. Player A could be better, or maybe Player B would be better if he had the same opportunities as A. When testing a hypothesis in a lab, you hold everything else constant and vary one thing. It is impossible to normalize all other factors when comparing players because players play games to prove which team is the best on that given day, not which player is the best overall.
The second problem I have with how people use advanced stats is that they do not say what people think they say. We have no way to measure what we want to measure individually because of all the variables discussed above. So we try to find proxies for those. But the proxies are imperfect. Corsi is used to determine possession, but it is just a proxy. A team could come down with the puck, be effectively funneled to the edges, rip off a poor shot immediately, get a rebound, pass it back and be forced to shoot wide as all lanes were blocked . That exchange took 10 seconds. The other team than takes it down the ice, cycles the puck for 20 seconds, gets a great shot and scores. The CF% for the first team is 67%. The CF% for the second team is 33%. Based solely on Corsi from that exchange, the first team is better on possession and/or defense. But in reality, the second team had the puck twice as long and was more effective in stopping the other team, forcing them into bad shots, while they could take their time and get off one high quality shot.
The third problem is sample size. Stats are aggregates of a lot of data. The more of it you have, the more accurate your data. A player plays 1,000 games, including hundreds of clutch situations, and he's money. Then he plays 1 playoff game, does poorly, and he is a crappy player who chokes when it matters. The larger sample is more important than the smaller. Outliers can ruin a small sample, but get smoothed over with enough data. OR Sobotka was a Goal per game player last year. Is he really a goal per game player, or did he just have a goal in the one game he played.
Those are all extreme examples, yes, but hopefully they show my point.
First, they represent only the sliver of the whole picture. Take GF%. Player A has a >50% GF%, and player B has <50%. Well looking at that one stat, player A is the superior player. Its obvious, when he is on the ice, goals get scored for his team more than against, and that is the point of hockey. However, what if I told you that player A was playing on a really good team, with 2 great linemates, playing against lesser competition, and getting exclusively offensive opportunities while player B was on a horrible team playing with horrible players, getting buried in the d-zone in the toughest situations against the best the NHL has to offer. Now, who is the better player? Its not so black and white when you add that extra data. Player A could be better, or maybe Player B would be better if he had the same opportunities as A. When testing a hypothesis in a lab, you hold everything else constant and vary one thing. It is impossible to normalize all other factors when comparing players because players play games to prove which team is the best on that given day, not which player is the best overall.
The second problem I have with how people use advanced stats is that they do not say what people think they say. We have no way to measure what we want to measure individually because of all the variables discussed above. So we try to find proxies for those. But the proxies are imperfect. Corsi is used to determine possession, but it is just a proxy. A team could come down with the puck, be effectively funneled to the edges, rip off a poor shot immediately, get a rebound, pass it back and be forced to shoot wide as all lanes were blocked . That exchange took 10 seconds. The other team than takes it down the ice, cycles the puck for 20 seconds, gets a great shot and scores. The CF% for the first team is 67%. The CF% for the second team is 33%. Based solely on Corsi from that exchange, the first team is better on possession and/or defense. But in reality, the second team had the puck twice as long and was more effective in stopping the other team, forcing them into bad shots, while they could take their time and get off one high quality shot.
The third problem is sample size. Stats are aggregates of a lot of data. The more of it you have, the more accurate your data. A player plays 1,000 games, including hundreds of clutch situations, and he's money. Then he plays 1 playoff game, does poorly, and he is a crappy player who chokes when it matters. The larger sample is more important than the smaller. Outliers can ruin a small sample, but get smoothed over with enough data. OR Sobotka was a Goal per game player last year. Is he really a goal per game player, or did he just have a goal in the one game he played.
Those are all extreme examples, yes, but hopefully they show my point.
Majority, I think we're basically saying the same thing, you just think that AS are trying to be general in scope, and I think they're not. The fact is, we use metrics that don't exactly measure what they purport to all the time.
Home runs in baseball: a home run off of Kershaw is more impressive than a home run off of, idk, Jaime Garcia. It still counts as a home run. Hitting in a deep lineup is easier than hitting in a thin one, a home run still goes down on the sheet as a home run. It doesn't really matter if you hit all of your home runs off the guy in the bullpen with an 8.00 ERA, if you hit 40 of them it still counts. Gross Domestic Product: GDP doesn't really measure the economy per se, but it's a pretty good indicator. It's close, not perfect. Those that use GDP as if it was perfect don't understand it, but it is useful for many things. And the economy is waaaaay more complicated than a hockey game. GDP is still a widely-used, widely-accepted metric.
Your corsi argument is a good one, and it's a concern to be aware of for sure. But the problem is that it's not what we've seen most of the time. Obviously it's not all the time, just look at what happened between us and Minnesota last playoffs. Teams that have a better corsi percentage have a higher likelihood of winning. They aren't guaranteed anything because of things you call out like shot quality (which we absolutely can and do quantify - see HDSC) and QoC (which isn't nearly as significant as the players you're playing with, as opposed to against), but you have a better likelihood. So, if you want to get more shots, you need to acquire players that are going to come out positive on the shot ledger. You can control for things like linemate effects by seeing how they do with and without their regular partners (WOWY), fenwick vs corsi, primary vs secondary assists, game state situations, roll aaaalllll of those things together, and at least get a good idea of who you might want to dedicate some extra scouting resources towards. Or who you might want to start playing more/less. Or whether or not that long-term contract you're considering is really worth it. Not because it's guaranteed and fool-proof, but because it increases the likelihood of being right, and decreases the likelihood that your opinion is based purely on ideology or bias. And I'm not using bias there in a derogatory way: bias is a way of life, impartiality is impossible. AS are guilty of this too. Too often it measures only one "brand" of success. There are other ways of being successful (or not) than just generating shots - and I am using corsi here just as a stand in - but those show up in other places if you're paying attention and keeping a wide view.
Of course sample size matters. Of course none of this stuff is black and white. Nobody is saying that it is - at least nobody that actually knows what they're talking about. It's ok to disagree about what good/bad results mean over small sample sizes, that's part of the fun of watching guys develop. It's ok to see different things in the proverbial tea leaves of a spreadsheet. But just because you can reach different conclusions from something doesn't mean that it's pointless: if everyone agreed then fantasy sports wouldn't be a multi-billion dollar industry. It just doesn't make any sense to me to disregard advanced stats simply because some knucklehead on an internet message board thinks Parayko is worth Kadri and a 2nd and uses corsi to "prove" he's right. There are dozens of other metrics proving (legitimately) that he's wrong.
Home runs in baseball: a home run off of Kershaw is more impressive than a home run off of, idk, Jaime Garcia. It still counts as a home run. Hitting in a deep lineup is easier than hitting in a thin one, a home run still goes down on the sheet as a home run. It doesn't really matter if you hit all of your home runs off the guy in the bullpen with an 8.00 ERA, if you hit 40 of them it still counts. Gross Domestic Product: GDP doesn't really measure the economy per se, but it's a pretty good indicator. It's close, not perfect. Those that use GDP as if it was perfect don't understand it, but it is useful for many things. And the economy is waaaaay more complicated than a hockey game. GDP is still a widely-used, widely-accepted metric.
Your corsi argument is a good one, and it's a concern to be aware of for sure. But the problem is that it's not what we've seen most of the time. Obviously it's not all the time, just look at what happened between us and Minnesota last playoffs. Teams that have a better corsi percentage have a higher likelihood of winning. They aren't guaranteed anything because of things you call out like shot quality (which we absolutely can and do quantify - see HDSC) and QoC (which isn't nearly as significant as the players you're playing with, as opposed to against), but you have a better likelihood. So, if you want to get more shots, you need to acquire players that are going to come out positive on the shot ledger. You can control for things like linemate effects by seeing how they do with and without their regular partners (WOWY), fenwick vs corsi, primary vs secondary assists, game state situations, roll aaaalllll of those things together, and at least get a good idea of who you might want to dedicate some extra scouting resources towards. Or who you might want to start playing more/less. Or whether or not that long-term contract you're considering is really worth it. Not because it's guaranteed and fool-proof, but because it increases the likelihood of being right, and decreases the likelihood that your opinion is based purely on ideology or bias. And I'm not using bias there in a derogatory way: bias is a way of life, impartiality is impossible. AS are guilty of this too. Too often it measures only one "brand" of success. There are other ways of being successful (or not) than just generating shots - and I am using corsi here just as a stand in - but those show up in other places if you're paying attention and keeping a wide view.
Of course sample size matters. Of course none of this stuff is black and white. Nobody is saying that it is - at least nobody that actually knows what they're talking about. It's ok to disagree about what good/bad results mean over small sample sizes, that's part of the fun of watching guys develop. It's ok to see different things in the proverbial tea leaves of a spreadsheet. But just because you can reach different conclusions from something doesn't mean that it's pointless: if everyone agreed then fantasy sports wouldn't be a multi-billion dollar industry. It just doesn't make any sense to me to disregard advanced stats simply because some knucklehead on an internet message board thinks Parayko is worth Kadri and a 2nd and uses corsi to "prove" he's right. There are dozens of other metrics proving (legitimately) that he's wrong.
Celtic Note
Living the dream
- Dec 22, 2006
- 16,940
- 5,732
None of my dreams make sense but yeah, it happened.
The only dream I ever had about the Blues was basically a memory recall of when they tore down the Arena.
That place was awesome. It wasn't some over engineered atmosphere. That place was raw and rowdy.
Yeah I usually don't have any hockey-related dreams either, at least none that are realistic and as detailed as that one was. I remember seeing a poll on this board and everything, lol.
I never experienced the old Arena but I don't think I've ever met someone who didn't love it. Sounded fun.
Majorityof1
Registered User
We are kind of saying the same thing. My arguments are for statistics in general, not advanced hockey stats. "Lies, damn lies and statistics". I am saying stats are not inherently bad. They are a useful tool in the hands of someone who knows what they mean and maybe more importantly knows their limitations. But like all tools they can be worthless at best or destructive at worst in the hands of someone who doesn't know how to use them (and very destructive in the hands of someone who does know how to use them and manipulates them for their own purposes).
The stats themselves are not trying to be or do anything. They tell a small part of a very large story. You cannot hope to get the whole story from any one stat. You will probably never get the whole story from stats. But coupled with what you see with your eyes, and what other stats say, you can fill in a good bit of the picture. However, a lot of people ignore the context and other stats. Yes, we can measure quality of chances, but how often are those used in conjunction with Corsi as a percentage of times someone mentions Corsi? How often are zone starts mentioned? QoC? etc? Most people want to find the one magic stat that tells it all, or they ignore the stats that don't support their side, focusing on the ones that do. It doesn't work like that.
taylord22
Registered User
- Mar 30, 2009
- 1,529
- 323
This is a great post. (And great discussion). My take...
Hockey statistics are largely measures for experimentation vs. reporting. They're actionable, not truly reportable (in the truest sense). They were created to provide avenues to experiment and succeed, or experiment and fail quickly. We're not going to get anything close to actual data science until we can track directly from the puck. And even then, that doesn't tell the full story. The Minnesota series is a great example. As good as Jake Allen was, Minnesota's offense was incredibly dull and linear. He barely had to cross his crease outside of jam plays. There are really no models that take the actual threat of an attack into account.
I think it's fair to call current models a decent start -- and are especially valuable in giving insights into results significantly above or below the norm -- but I say that more from the perspective of them being actionable. They're less meaningful to the media and fans.
Last edited:
You're not wrong! I just think you need to cut people a little slack haha. Like I said earlier, I just don't expect a doctoral thesis from somebody every time they want to talk about somebody's impact on the game beyond just the raw points they score.
If we were really drilling down deep, we would prefer to know about things like ice condition or whether or not the goalie had a bum knee. Let's hold all things equal and not try to equate things that aren't equal... that makes enough sense. Our spreadsheet would have 1000's of very precise categories that would be next to impossible to digest and understand in any meaningful way. Only goals scored against perfectly healthy goalies with a .923 save percentage in their home barns go in the "goals scored against perfectly healthy goalies with a .923 save percentage in their home barns" column. It's just cumbersome to be completely comprehensive, especially when we're dealing with large amounts of data, and long ongoing conversations spanning season after season. So we use shorthand, understanding (hopefully) that there is error involved. We smush things together, try to control for what we can, and get pretty close to what is actually meaningful on the ice. Remarkably close! And for those willing to put in the effort, you can really drill down pretty far, and answer so many of those burning questions for yourself (until the guy running the website where the data is hosted gets hired by the Panthers). But that isn't always possible, so for shorthand purposes, corsi will do just fine most of the time as a back-of-napkin indicator.
As taylord said, we try things out, we see if they work, and as better things come along, we adopt them instead. That's science in the field for ya! The good news is the stakes are about as low as they can be: none of us are making actual decisions here (until we get hired by the Panthers). No real mortal danger of malicious intent with someone reading something differently in a CFrelTM% stat than you do.
But I guarantee that for every instance of a stat being "wrong," there's 100 guys out there whose eye test was wronger. Lou Korac didn't even notice the goalie change in tonight's game. The stat sheet did though. Both have bias, but the stat sheet is biased in predictable and controllable ways, and I'll take that every day of the week.
Aaaaaaaaaand, the Parayko/Kadri thread is closed. That didn't take very long ...
I was shocked it lasted that long
There have been a few that closed within the hour
Somebody suggested in a post a while ago that Blues fans stop responding with "Matthews" to trade proposals involving Parayko. My answer is NOPE. That is the response they will get, again and again and again. NOBODY, IMO, in the Leafs organization other than Matthews gets it done for Parayko. And yes, that includes Nylander and Marner. Maybe then these ridiculous threads will stop.
Falco Lombardi
Registered User
Yeah until I'd be willing to accept something other than Matthews, I'm not going to respond with anything else.
Eventually they'll get the point.
Eventually they'll get the point.
Majorityof1
Registered User
There is a term in Economics called Abstraction. It means drilling down to the essence of an economic interaction, ignoring the unimportant variables, and creating a workable model that can explain behavior. You can't possibly model every single thing that could potentially affect an economy, so you hold certain things constant and go from there. However, if you abstract too much, you run the risk of ignoring pertinent factors. So in your example, you have to ignore ice conditions and hope they normalize out over extended use. Maybe if a player's home has significantly worse ice, you keep it in the back of your mind. But things like QoC, zone starts, etc, are not too much. There is a middle ground between using 1 stat and 1000.
You keep treating stats like they are a living breathing thing. The stat sheet didn't notice a goalie change. Someone keeping the stats did, and they are just as susceptible to human error as Korac.
Finally, I am not denouncing stats. Nor am I expecting a doctoral thesis. If a poster has an underlying understanding of the limitations of stats,, then they don't have to explain everything. If someone who understands stats says, "his Corsi is really high as well as his other defensive stats, granted he's a bit sheltered but not so much as to have a major impact, ergo I think he is a good defensive player". That sentence didn't lay out a doctorial thesis on what every stat was. But it took them into consideration, didn't ignore them. Now if I disagree with the conclusion based on the eye test, I can dig deeper into that argument. But people, like the poster on the main boards who started this discussion, who just lay out stats, misinterpret them, expect them to be the end-all-be-all and don't even acknowledge mitigating factors, they give advanced stat users a bad name. Especially because when challenged they automatically cite their use of stats as a trump card. "I used stats, you didn't, so my argument is automatically superior.: Never mind that the stats were faulty and misused, they used them so its better.
Last edited:
Majorityof1
Registered User
Exactly. I was the one who said don't mention Matthews. You are throwing fuel on the fire to encite the slightly more rational Leafs fans. The thread is dying and its always someone who bumps it from the 2nd page to say "Matthews" and then the thread goes for 4 more pages before its shut down.
"Eventually they'll get the point"? Do you know the definition of insanity? Repeating the same behavior and expecting a different result. They post a bad Parayko offer. One of us says Matthews hoping to kill the thread and stop future Parayko asks. Neither happens. Rinse and repeat ad nauseum. Either you all are crazy for expecting it to stop, or you all secretly like antagonizing each other.
bleedblue1223
Registered User
- Jan 21, 2011
- 51,940
- 14,925
Sometimes I just want to watch to Parayko threads burn down, so throwing more fuel on the fire is quite enjoyable.
Dbrownss
Registered User
- Jan 5, 2014
- 31,359
- 8,734
My only suggestion is not to even respond to the thread. Who cares how bad the offer is..... If we use don't respond it would die in one page.
It is one thing to open up a discussion, but most of the offers are so bad, it doesn't merit responding. And to take it a step fruther, their own fans will correct them and tell them how bad the original offer is.
It is one thing to open up a discussion, but most of the offers are so bad, it doesn't merit responding. And to take it a step fruther, their own fans will correct them and tell them how bad the original offer is.
Celtic Note
Living the dream
- Dec 22, 2006
- 16,940
- 5,732
It certainly wasn't a very family friendly place. It was loud as hell. The concourses were crammed. Seats and steps were steep.
But it was certainly intimate. It felt like a community of fanatics and it made games that more intense.
It was very working class.
- Status
Ad
Upcoming events
-
-
Game 3 Belleville Senators @ Cleveland Monsters - Series tied 1-1Wagers: 2Staked: $5,025.00Event closes- Updated:
-
Game 3 Providence Bruins @ Hartford Wolf Pack - Series tied 1-1Wagers: 2Staked: $50.00Event closes
- Updated:
-
Game 3 Hershey Bears @ Lehigh Valley Phantoms - Hershey Bears leads series 2-0Wagers: 3Staked: $2,590.00Event closes
- Updated:
-
