Does anyone miss hockey analysis before analytics?

[SnowblindNYR]

Seems like whenever a team wins a lot of games by large margins now instead of people being impressed they start yelling PDO!!!

I remember during the 11-12 season the Rangers and Bruins were neck and neck but the Bruins most of the season but at one point in the season Boston ended up destroying teams. I had an argument with someone on HF and he said that Boston winning games by larger margins meant they're less fluky than the Rangers because each close game is closer to a loss than a blowout. So the Rangers were more likely to benefit from a bounce here and there than the Bruins who would have to have a lot more go against them to lose a game they won 6-0. As much as I argued with him I still think that's a sound argument. Now the Rangers ended up with more points and went deeper in the playoffs. But I kind of miss the days teams weren't penalized by public opinion for winning in a dominant way. The new age way of thinking of this seems backwards. Anyone else feel this way?

Edit: Just to make it clear, the example of 11-12 was BEFORE analytics became as big of a thing today.

 

[MCR74]

Yep. All this XY%FU garbage means nothing to me. l watch hockey with my eyes, not a stat sheet.

 

[MadLuke]

But I kind of miss the days teams weren't penalized by public opinion for winning in a dominant way

This must by a typo ? How does team penalized by public opinion for winning in dominant fashion versus in close games ? (And goes directly against the example you just gave, maybe I am misreading the negative)

Lot of hockey analysis we hear are from ex player-coach and not that analytics based.

 

[Empoleon8771]

The problem with analytics are the people who use them, not the analytics themselves. Just like any sort of statistical data set.

"Analytics" aren't this magical thing, they're basically just shot differentials. That's it, any sort of "advanced stat" is just a measurement of shots. The people who overuse them think they're gospel and the people who hate them think they're nonsense, but both sides are wrong. Analytics are a supporting point to an argument, not the entire argument. It's the same thing as the eye test or basic stats.

I think the largest failure of "analytics" was calling them "advanced stats", because they are not "advanced stats" whatsoever. The reality is that "advanced stats" are just some measurement of shot differentials, that's it. They're effectively +/- for shots.

CF is just a shot attempt and CF% just shows what percentage of shot attempts your team is taking. FF is just unblocked shot attempts and FF% just shows what percentage of unblocked shot attempts your team is taking. xGF is just shot attempts multiplied by a location probability factor that converts that shot attempt in that specific location into an "expected goal".

None of these stats are these magical stats, both from the arguments of "they're the truth" or "they're made up".

 

[Jumptheshark]

The problem with hockey analytics is about 80% of the people do not actually know how to PROPERLY explain them and how to APPLY them to the game.

On nearly every board at least once a day, you get someone posting some analytic stuff and just go "there, that proves it" that is not how analyze and explain analytics.

I have worked with business analytics as a hotel manager for 30+ years and it amuses me how some people fail to understand "the explanation" part of analytics and the comparison part.

The top Analytics guy who wrote a great book on the subject and now works for the kings even says hockey analytics is still in its infancy due to the fact how much hockey is a team sport compared to baseball

 

[wintersej]

Guess what. Everyone has been using analytics since the beginning of hockey. Goals, assists, plus minus, wins, save %…it’s all stats and analyzing of stats. Shocking.

Much like those stats, the new stats don’t tell the whole story.

Things fall apart on this site because too many people who use the new stats don’t really know how to use them to analyze and how to use them in concert with scouting and too many people dismiss them as being pointless and “this ain’t baseball” or whatever and it just starts all sounding like me trying to talk politics with drunk uncle at Thanksgiving.

 

[SnowblindNYR]

This must by a typo ? How does team penalized by public opinion for winning in dominant fashion versus in close games ? (And goes directly against the example you just gave, maybe I am misreading the negative)

Lot of hockey analysis we hear are from ex player-coach and not that analytics based.

My example was from 2011-12 before analytics was as big of a deal as now. Nowadays if a team wins in a dominant way people complain about PDO being too high.

 

[Trash Panda]

Modern analytics empowers too many people to ignore the subtle nuances that make hockey the greatest game on earth.

It’s a game full of emotion, momentum, physicality, and flat out luck sometimes. Gotta closely watch the games to appreciate what is truly going on.

 

[MadLuke]

My example was from 2011-12 before analytics was as big of a deal as now. Nowadays if a team wins in a dominant way people complain about PDO being too high.

They will complain much more if they have an high PDO and wins in non-dominant fashion despite it.

If they actually dominate (and people take score effect into the outshooting out possession domination) it should show up in the advanced stats as well or at least the more they dominate the better.

 

[Empoleon8771]

My example was from 2011-12 before analytics was as big of a deal as now. Nowadays if a team wins in a dominant way people complain about PDO being too high.

No they don't. If a team is getting outshot 45-15 every game but is winning 3-2, that is not a sustainable way to win going forward and likely won't last.

People say things like that because small sample sizes can make really skewed results and outlier wins likely won't be sustainable over a full season. PDO is a measurement of the sustainability of their performance, with the expectation that the average team will trend back towards a PDO of 1.000 over an infinite sample size.

 

[SnowblindNYR]

No they don't. If a team is getting outshot 45-15 every game but is winning 3-2, that is not a sustainable way to win going forward and likely won't last.

People say things like that because small sample sizes can make really skewed results and outlier wins likely won't be sustainable over a full season. PDO is a measurement of the sustainability of their performance, with the expectation that the average team will trend back towards a PDO of 1.000 over an infinite sample size.

I mean a perfect example is the Canucks. It's not just the blowout wins it's also I guess shot differential. But their blowout wins are a big reason for the unsustainable PDO.

 

[MikeyMike01]

There are people who unironically think reality is wrong, when it doesn’t match the models.

 

[wintersej]

And if you are getting annoyed with analytics now, just wait until all this player positioning data starts getting fed into machine learning models. It’s gonna be like people complaining about calculators in the classroom right before it became smart phones and laptop.

Being able to understand what all the players are doing on the ice together and at what speed and exactly where they are and exactly where the puck is and exactly how quickly it moved around or if it was tipped or…everything… it’s gonna be super interesting to see how things advance when you understand the systems at work and blaa blaa…

 

[Petey O]

 

[Empoleon8771]

I mean a perfect example is the Canucks. It's not just the blowout wins it's also I guess shot differential. But their blowout wins are a big reason for the unsustainable PDO.

People mention PDO regarding the Canucks because they're getting outshot but somehow has a massively positive goal differential. The point of mentioning PDO is that their winning ways are likely not sustainable going forward, and they'll likely begin losing more games if they continue to get outshot.

So far on the year, the Canucks are on average getting outshot 28-30 in a game but are winning games an average of 4.3-2.2. Does that seem like something that is sustainable going forward? No, which is why their PDO is so high.

You start getting into dicey territory when you try to convert shots into expected goals, but in its purest form, analytics aren't complicated.

 

[wintersej]

My example was from 2011-12 before analytics was as big of a deal as now. Nowadays if a team wins in a dominant way people complain about PDO being too high.

Ok and here is an example of not analyzing stats well.

Ok, yes, those blow out wins are totally exploding their PDO and that is a valid thing to look at.

But then people find the result they want to see and stop digging.

You know which team is the luckiest in the NHL when the game is within one goal?

Those same Vancouver Canucks. So that kinda makes the previous argument kind of garbage, right?

This site just ends up being a major case study in how to lie with statistics. If you are at least a little bit clever you can cherry pick things to tell a narrative you want. And I don’t think it’s something folks do maliciously, it’s just people like to find the facts that assure what they already believe. Just like drunk uncle at Thanksgiving.

 

[WarriorOfGandhi]

Don’t give me these newfangled “assists” and “penalties in minutes”! I know everything I need to know about a player by looking at their daguerreotypes in the newspaper!

 

[Breakers]

It’s broad samples that is the problem

If you say so and so isn’t a good 5v5 team
But you look at 10 games sample for instance individually and they carried the play 5v5 in 6 of those games

But in the losses they got slaughtered from a play driving standpoint

Should the bad play in those 4 losses outweigh the good from 6 games?

 

[Empoleon8771]

It’s broad samples that is the problem

If you say so and so isn’t a good 5v5 team
But you look at 10 games sample for instance individually and they carried the play 5v5 in 6 of those games

But in the losses they got slaughtered from a play driving standpoint

Should the bad play in those 4 losses outweigh the good from 6 games?

The problem here is trying to make declarations of whether a team is "a good 5v5 team" based on a 10 game sample size.

Over a larger sample size, those odd peaks and valleys should even out and give a good idea for what a team is.

 

[Romang67]

How many people do you know who actually understand statistics to the point where they 1) understand what the numbers they're looking at signify and 2) understand at a higher level what they're trying to show using those numbers?

I don't mean just in hockey, but in every facet of your life where you come across some sort of analytics.

Turns out the proportion of hockey fans who 'get it' isn't much greater.

Even when people who don't actually get it use it, I don't think I find it any more irritating than I do with people whose argument comes down to the eye test, and if your eye test disagrees with them, that must mean you don't even watch hockey.

 

[666]

"Advanced stats" have a small role to play but most people misuse them without understanding context. For example, we constantly see team based stats like xGA used to compare players across teams. Isn't there some guy at The Atlantic who has some "formula" that assigns a player a single number now? It's just absurd and a clear sign that anyone who does this sort of thing can not critically think.

 

[JKG33]

As annoying as the analytics crap is, it's sure made it easy to tell who's worth taking seriously vs those who just watch stat sheets

 

[Breakers]

The problem here is trying to make declarations of whether a team is "a good 5v5 team" based on a 10 game sample size.

Over a larger sample size, those odd peaks and valleys should even out and give a good idea for what a team is.

That was just an example for this situation

20 games is the minimum at least for underlying numbers

 

[ClydeLee]

My example was from 2011-12 before analytics was as big of a deal as now. Nowadays if a team wins in a dominant way people complain about PDO being too high.

People still just said that team was being more lucky or just not, oh this goalie is hot who isn't expected to be thay hot or that line is great.

 

[Chips]

Yep. All this XY%FU garbage means nothing to me. l watch hockey with my eyes, not a stat sheet.

People who use analytics do so in a flawed way more so than the analytics themselves are “useless”


Likewise the vast majority of fans “eye tests” are unequal and near useless unless you’re taping games, rewatching, staring at a player when he doesn’t have the puck rather than strictly seeing what he does sometimes with it


That’s one good thing about analytics is they help account for a lot of the little things like positioning and the million seemingly meaningless passes and what not that add up. That’s when you have to combine the eye test to determine which factors are actually coming j to play