Jump to content

The value of advanced stats


Recommended Posts

Advanced Stats track very well over long sample sizes.

 

Over 10 games, you can get wacky results

 

Anyone who has taken the most basic of college or university level stats courses would understand the problems with making conclusions based on a 10-game sample. 

  • Like 1
Link to comment
Share on other sites

22 minutes ago, Commandant said:

Advanced Stats track very well over long sample sizes.

 

Over 10 games, you can get wacky results

 

Anyone who has taken the most basic of college or university level stats courses would understand the problems with making conclusions based on a 10-game sample. 

When can simply use one stat over ten games that is a darn good one to give you a pretty good idea how team is doing...it is wins-losses. And they say teams are generally expanding their scouting staffs, so the eye-test must also be seen as pretty valuable tool. Advanced stats and colourful graphs are fine for the geeks and bean counters to entertain themselves with, but beyond that is just mostly white noise and overrated mumbo-jumbo.  IMHO.

Link to comment
Share on other sites

32 minutes ago, DON said:

When can simply use one stat over ten games that is a darn good one to give you a pretty good idea how team is doing...it is wins-losses. And they say teams are generally expanding their scouting staffs, so the eye-test must also be seen as pretty valuable tool. Advanced stats and colourful graphs are fine for the geeks and bean counters to entertain themselves with, but beyond that is just mostly white noise and overrated mumbo-jumbo.  IMHO.

 

Wins and losses are important over ten games.  So is Vegas going to win the cup?

 

 

The fact is that while no stat is perfect... possession stats are a better predictor of future wins than win loss records. This has been examined over large samples and the correlations are significantly higher. 

 

You can call them bunk, but there is a reason why all 31 teams have analytics depts now... why Vegas is looking into them when setting betting lines, why they are being widely used for a number of other things as well. 

Link to comment
Share on other sites

44 minutes ago, Commandant said:

 

Wins and losses are important over ten games.  So is Vegas going to win the cup?

 

 

The fact is that while no stat is perfect... possession stats are a better predictor of future wins than win loss records. This has been examined over large samples and the correlations are significantly higher. 

 

You can call them bunk, but there is a reason why all 31 teams have analytics depts now... why Vegas is looking into them when setting betting lines, why they are being widely used for a number of other things as well. 

The only thing I will say is that I've never heard of any research stating what exactly is a "large enough sample size" when it comes to the NHL. For some statistics it would seem to be that 10 games may be enough of an indicator of correlation, and for other statistics it might take 40, others 80, others 200, etc.

 

How do we know that everything will even out as they should by the end of a regular season? Statistics are extremely valuable but how can we know that 80 games will be enough of a sample size in and of itself. I'm thinking there's a chance that some seasons, certain teams just defy the norm because 80 games may not have been a large enough sample size for "things to even out" as well.

 

 

Link to comment
Share on other sites

32 minutes ago, xXx..CK..xXx said:

The only thing I will say is that I've never heard of any research stating what exactly is a "large enough sample size" when it comes to the NHL. For some statistics it would seem to be that 10 games may be enough of an indicator of correlation, and for other statistics it might take 40, others 80, others 200, etc.

 

How do we know that everything will even out as they should by the end of a regular season? Statistics are extremely valuable but how can we know that 80 games will be enough of a sample size in and of itself. I'm thinking there's a chance that some seasons, certain teams just defy the norm because 80 games may not have been a large enough sample size for "things to even out" as well.

 

 

 

As with any statistic.... the bigger a sample size you have, the better it is. 
 

A large enough sample size to have a perfect statistic, in any statistic like this is infinite.  

 

However the larger the sample size you have, the more you reduce the margin for error.... as you increase the sample it gets better and better, with a smaller percentage of errors. 

 

 

Its not about having a perfect stat though, that is impossible. 

 

Its about having the best statistics you can have. And possession is the best indicator of future wins that we have found up to this point in time. 

Link to comment
Share on other sites

as with any stat in sports, a smooth combination of knowledge of the game, and understanding of stats gives you the best results for evaluation.

 

I used to not quite understand the impact of corsi and fenwick when it first was introduced, because it revolved around a certain way to play the game, and there were varying ways to play the game. Upon further study, I realized it was a stat that helped immensely in evaluating how well a team was playing the game in a way that yields the highest % of success. Of course there are exceptions, possession hockey seems to be the trend the last several years, it has the best chance to yield successful teams. Are there teams who can be successful playing dump and chase? sure, are there teams that can be successful playing the trap? sure, are they likely to be as successful on a percentage based study as teams who play possession hockey? Not likely.

 

But if you were to combine the eye test with these stats, you could see that a team or player is still very good, who deviates from these stats, you can also see there are many teams and players who are successful when highly ranked by these stats. Weber is a prime example, He always ranks average or below average in corsi and fenwick, yet if you watch him play with a knowledgeable eye, you quickly determine he is a tremendously good Defenceman. The key factor here is he simply doesn't  play the game in a way that suites the advanced stats themselves, but it is still a style of play that he performs extremely well. Because of how exceptional he is at playing defence the way he does, if you compare him to any other player who is simply good or average, but plays the same way, your eye test would determine, these Defencemen are not very good, your typical 5 - 7 guys.

 

The same can be said the other way around, if you look at the top of the spectrum of the corsi/fenwick stats, you will find guys like Torey Krug, Colin Miller, Shattenkirk, and Justin Schultz all parked in the top 20. Are they top 20 D-men? No of course not, in fact they are just good top 4 D-men who can get the job done on a relatively good basis night after night, nothing tremendous about them at all. However they are all swimming along this advanced stat pool beside guys like Burns, Doughty, Ekblad, Faulk, Werenski, Hamilton, and Hedman, all top Defensemen for their teams, truly exceptional players. So you see, the point is there are exceptions to every rule, that is especially so in statistics. of course if you look at our team for instance, it looks like we should be the best team in the league according to the advanced stats, however if you combine it with the eye test, you quickly see we are highly inconsistent throughout a game and generally shoot ourselves in the foot because of it lately. We play the game the right way, generate everything we should be generating to create success, but because of certain pockets of time throughout a game where we are especially bad, we are blowing games left and right. We can be a corsi/fenwick dream team all we want for 48 min out of 60, but if in those 12 min we self destruct, it will not be grasped by an average/percentage system like advanced stats properly, hence, we are the exception presently.

 

When you find a situation that falls out of the norms of stats and advanced stats, they are simply exceptions to the rule, it does not mean the statistic is grossly wrong and should not be used. It simply means there are exceptions, and it is easier to properly evaluate teams and players a like, by combining both, eye and paper to determine what you are truly looking at. And that my friends, always gets you the closest to reality than not. If all you do is look at the paper and the stats, you will never get it right, if all you do is watch the game, you will never find all the details that determine why a player is being successful or not.

  • Upvote 3
Link to comment
Share on other sites

Extremely well said. 

 

That said there is one advanced stat where Weber excels.  HDSC%

 

High Danger Scoring Chance %

 

What does that teach us?  Weber's Corsi sucks cause he gives up lots of perimeter shots, but he does not give up many shots from the front of the net. 

 

This shows two things....first, no one stat tells the whole story.  Always look deeper. Corsi without usage, without zone starts, without opponents, is not a great stat.  And corsi with HDSC included is even better.  Secondly, it shows that the eye test is still important.  Stats aren't the whole story, but they are a very good tool that enhances knowledge when combined with the eye test. 

Link to comment
Share on other sites

Stats are important. The more stats the better at understanding the game. 

 

Stats however can be entered incorrectly. We've seen that with giveaways and takeaways around the league. They can also be still in beta and not necessarily say what they should say. You can also have stat anomalies or evolving views on stat analysis. 

 

There are things I've always had beef with. The disregard for quality of shot is one. I understand the value of stats is to simplify and understand trends but any stat that tries to treat every shot as the same is broken to me. I'm also recently having a real issue with who tracks "high danger scoring chances" because watching the game I see guys like Pacioretty taking these really bad shots near a goalie and it treated like it was a high danger scoring chance simply based on where he was on the ice. Again, I know it's more about an average and looking all players on that average, but it's getting frustrating to hear someone try to defend Pacioretty when 2/3 of his shots could be stopped by Jonas Gustavsson but because he gets close to the Monster I'm supposed to be nice about it and just hope he eats more Lucky Charms. Spare me. 

 

But what the high advanced stats right now are saying isn't that the Habs are doing great but losing. It's saying the Habs show good habits to win. That's a hard concept to wrap your head around at times but think of it like losing weight. If you work out regularly, eat well, drink water, and stay away from drugs, there's no guarantee you'll lose weight. But those good habits on average will lead to weight loss. That's what the stats are really about. With better luck, and better defence, the Habs should have a better shot to win games. But lots of shots isn't a guarantee to win, it's just better than not taking shots. 

 

TL:DR, advanced stats are vital but that doesn't mean they are gospel. It's less about guarantee and more about good habits. 

Link to comment
Share on other sites

Since I posted snark and then went away for a few days, I thought I'd add comment now.

 

In 11 games, Montreal has absolutely destroyed the competition in control of the puck, but been thrashed on the scoreboard. I think 660+ minutes of play is more than enough to demonstrate how play is going on the ice. This has been consistent, game after game.

 

Yes, Carey Price is looking a little bit like André Racicot out there. But the simple fact is we've taken more shots on opposing nets than any other team in the league, and we're not scoring. I keep seeing hockey analysts saying, "abnormally high or low shot percentages will end up evening out over the long haul." As MoLG said above, the quality of the shot cannot be quantified in statistics, no matter how you try. And this applies at both ends of the ice.

 

You can outshoot your opponent 60-15, but if you had 3-4 tough shots out of your 60, and they had 7-8 out of their 15, they're going to win, every single time.

 

Possession stats cannot easily indicate the quality of the scoring chances allowed or taken. 

 

I get why they exist. It's impossible to watch all 1271 games in a regular season of the NHL. Hell, I probably only watch about 15-30 games in any given NHL season. Most of those will be the team I cheer for. The "Eyeball" test requires attention most of us do not have time to perform. If you can quantify play into numbers that can be crunched, you can analyze play to a degree... it's true. But those numbers never tell the complete story. What gets me are NHL analysts and commentators crunching these things and treating them like they can discover some deep hidden truth if they twist them inside out in enough detail, and these are guys that are actually paid to watch all 1271 games in a season. I don't get it. 

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...