Adjusted stats- stop using them

[MXD]

Why did scoring peak in the 80s? A lot of reasons. Shabby goaltending is certainly one of them. But the talented wasn't as diluted as it is now. A lot of teams can't ice two credible scoring lines anymore. Lots of good players, but not many good NHL scorers.

... There were lots of teams that weren't able to ice two scoring in the 80ies, but i see your point.

 

[BostonAJ]

Adjusted stats don't work. They're an abject failure from the moment they're conceived.

For one thing, greatness in hockey, with few exceptions (save percentage, goals against average), is not to be measured with a calculator. Maybe it's just me: I want to know what a player did, I want to know how he actually played the game, what he meant to the game, what he brought to the table, and what he didn't bring to the table. I don't want to know what a calculator tells me he might have done.

They also fail to take into account changes in the game. No adjusted formula can accurately take into account the impact that Bobby Orr had on the game.

Why did scoring peak in the 80s? A lot of reasons. Shabby goaltending is certainly one of them. But the talented wasn't as diluted as it is now. A lot of teams can't ice two credible scoring lines anymore. Lots of good players, but not many good NHL scorers. It's what hockey people were looking for. They were putting a premium on offensive skill. The one-dimensional scorers had more of a place in the late 70s and the 80s than any other time in the game's history. And the offensive defenceman suddenly became a factor, too. Defencemen played an offensive game throughout their developmental careers, and they could do it once they reached the NHL.

That's not to say that you can blindly look at the numbers. You can't do that. In any circumstance. 100 points now is different than 100 points 20 years ago. I won't deny that. But part of the reason that 100 points was attainable was that teams put a premium on offence that they don't have now, the offensive defenceman was becoming a big factor (they aren't as much of a factor now), and you had the two best centres ever ripping the league apart.

But you can't say that "Yzerman would have had x points in 1989, and Sakic would have had x points in 2001." Yzerman had more than 150 points in 1989, won the Pearson Award for best player in the league, as voted by the players. Maybe if Yzerman has top-notch talent to play with, he hits 175 points. And maybe if he didn't play against shabby opposition (I would argue five of the six worst teams in the league in 88-89 were in the Norris) for 32 games, Yzerman winds up around 120 points.

I cannot express enough how much I agree with this post. HOH section has gotten insane with the statistical comparisons while rarely bothering to describe a player. Hockey is not math.

 

[Ohashi_Jouzu*]

...They also fail to take into account changes in the game...

...They were putting a premium on offensive skill...

Good post, sorry to edit so much of it out. First of all, I don't think adjusted stats are complete failure. As a tool, they can be useful.

But back to the intended topic. Yes, changes in the game. Definitely agree. Every level of hockey, right down to on-ice coaching, has evolved like a business. Evolution towards focusing on control, limiting uncertainties, and prolonged return taking minimal risks. Perhaps a main reason why a lot of the observable changes in the game seemingly revolve around keeping the puck out of the net (obvious emphasis on responsibilities of forwards in all 3 zones, "the trap", defensemen must be more mobile and aware than tough, goalie gear has been allowed to balloon out of control, likely under some reasoning of safety, etc), rather than putting pucks in.

In fact, despite the average player being much better trained, developed, and reputedly skilled than generations before them, a recent barrage of rule changes and adjustments to tolerances in defensive play have been necessary to increase goal scoring in today's game.

Ultimately, if any discussions or comparisons across eras are going to be attempted, some kind of tool is going to be necessary to give context to player accomplishments. Since there is no way of separating elements such as increase in goals due to elimination of the two-line pass, or effect due to increase in roster size, or number of teams, effect of the trapezoid, effect of "the trap" etc, we're left to "adjusted" stats to help us. For example, if you're discussing scoring, average league scoring in one year versus another is certainly relevant. Sure, lots of elements are tied into it, but that's why it's a "tool", and not "proof".

Are adjusted stats perfect? Not even close. But even having watched a LOT of hockey over the past 20+ years, I need a tool to reconcile an doubts/uncertainties and the fact that I have had access to MUCH more hockey (and hockey knowledge) over the past 10+ years compared to the previous 10+. These kinds of stats certainly help. It doesn't facilitate the level of comparison actually watching provides (how sure are their passes? how quickly do they change direction? how strong are they in the corners? how often do they seem to find themselves in the right place at the right time? do they go down before the shot more often than they "should"? etc.), but there is still a place for them in discussion imo.

 

[seventieslord]

Adjusted +/- is a stellar statistic but lets not go there right now.

Adjusted scoring totals are a handy little tool for comparing single seasons between players of similar stature. But nothing more. There are too many issues with adding up career adjusted totals and they are wrought with the same problems you'd run into when adding up actual career totals; that is, that they are not in any way an indicator of who was better or more dominant. They are one little piece of the picture, nothing more.

The methodology used to create these adjusted stats is flawless, IMO, but the point about margin of error is very important, too.

by the way, about the league getting "watered down" so that teams can't ice two scoring lines anymore, doesn't it also follow that the goalies should be watered down at the same rate? Why does more teams have to mean less scoring thanks to less good scorers? Couldn't it just as easily mean more scoring thanks to less good goalies? Logically, the dropoff at both positions would be about the same, and it shows based on what scoring totals look like in Europe and the north american minor leagues. There is more at play than the size of the league. The game evolves all the time. Goals today are worth much more than they were 25 years ago, and the fact that there are 30 teams instead of 21 is probably not even in the top-10 reasons why.

 

[golfortennis]

Goals today are worth much more than they were 25 years ago, and the fact that there are 30 teams instead of 21 is probably not even in the top-10 reasons why.

At the risk of a threadjack, I'd be curious to know what the top 10 reasons are.

 

[Fredrik_71]

Ajdusted stats are usefull in a context. The same can be said about the +/- stats.

/Cheers

 

[XploD]

I have some major issues with your thoughts. I mean, if you think we shouldn't use adjusted stats, why should we use stats at all? I mean the unadjusted stats mean absolutely nothing when trying to compare differnt players from different eras.

So then we're left only with observation to compare these players, and to be honest, that's quite arbitrary. So basically what you're saying is that we shouldn't compare players from different eras at all?

The best tool for comparing players from different eras is still the adjusted stats. At least then there's a starting point which no one can argue. That's not the case with unadjusted stats and observation.

 

[seventieslord]

At the risk of a threadjack, I'd be curious to know what the top 10 reasons are.

I never said I had the answers, and if I did, that would be a threadjack. However, I did explain why I don't think the size of the league had anything to do with it. More evidence: One of the most prolonged low-scoring eras of all-time was the O6 era... when there were, of course, six teams. More teams led to more scoring. Even more teams after that, led to even more scoring. If we want to relate size of the NHL to scoring levels there is ample evidence that it makes it go both up and down, so if we wash those two things out we can start looking for real reasons and not phantom ones.

 

[rallymaster19]

I have some major issues with your thoughts. I mean, if you think we shouldn't use adjusted stats, why should we use stats at all? I mean the unadjusted stats mean absolutely nothing when trying to compare differnt players from different eras.

So then we're left only with observation to compare these players, and to be honest, that's quite arbitrary. So basically what you're saying is that we shouldn't compare players from different eras at all?

I never said we shouldn't compare players from different eras. In fact, the first thing I said was I enjoy comparing them.

The best tool for comparing players from different eras is still the adjusted stats. At least then there's a starting point which no one can argue. That's not the case with unadjusted stats and observation.

I agree that of course the game changes over time and unadjusted stats don't change with the game, they stay the same. But where I disagree with you is that unadjusted stats are actually the stats no one can argue. A point by Yzerman in '89 was the same as a point by Messier in '89. A point by Orr in '75 was the same as a point by Lafleur in '75 and so on. Of course the game changes, and everyone who watched the game will take that change into consideration when comparing players; and this is where observation is key to go along with those raw stats.

The biggest problem I have with unadjusted stats are they produce an absolute figure that some people assume is 100% accurate, when in reality there is typically a margin of error +/- some 15-20%. That means they provide absolutely no benefit or insight when examining players that are even somewhat close to one another. With the Yzerman/Sakic example, showing Sakic had the marginally better adjusted numbers adds absolutely nothing to the argument because the margin or error is high enough to deem any conclusion from the numbers irrelevant. That means Yzerman vs. Sakic offense-only is best argued with raw stats (goals, points, PP time, PK time, PPG, faceoff %, strength of schedule etc.) and observation (quality of linemates, defensive habits, leadership abilities etc.).

Now, like what Dark Shadows said, if someone comes up to you claiming Ovechkin is better than Howe because of his raw stats, you could counter with adjusted stats because in that case the comparison isn't close at all and the margin of error will be slim enough that it is almost mathematically impossible to prove Howe was not the better player.

Really, my only point is that in any close comparison between two players, pointing out that one player has the other beat with adjusted stats adds nothing to the argument.

 

[Ohashi_Jouzu*]

Really, my only point is that in any close comparison between two players, pointing out that one player has the other beat with adjusted stats adds nothing to the argument.

This seems to be what some people are getting panty hitches from, and I still can't find where anyone did this.

 

[overpass]

I agree that of course the game changes over time and unadjusted stats don't change with the game, they stay the same. But where I disagree with you is that unadjusted stats are actually the stats no one can argue. A point by Yzerman in '89 was the same as a point by Messier in '89. A point by Orr in '75 was the same as a point by Lafleur in '75 and so on. Of course the game changes, and everyone who watched the game will take that change into consideration when comparing players; and this is where observation is key to go along with those raw stats.

Adjusted stats are aimed at the fact that a point by Lafleur in '75 is not the same as a point by Yzerman in '89. They are scored in different contexts and have different values to winning.

Also, if we really want to be accurate, a point by Yzerman in 89 is not the same in value as a point by Messier in 89 if one of them plays on a team with a wide-open style and the other plays on a team with a slow-paced, defensive style.

Finally, I'm really not sure why you think that adjusted stats have a higher error bar than unadjusted stats. Based on what? If you are trying to model contribution to winning, I'm pretty sure adjusted stats have a smaller error bar than unadjusted stats.