Oh, Right, The Error

Rather small sample size, Jonathan ;)

But each to his own.

One other thought. . . You say you are uncomfortable with Dryden's theory and can't buy into the Error stat as a result.

On one level, this is an entirely fair criticism.

But on another level, I believe you are contracting yourself.

I say this because you make arguments all the time based on points per 60s minutes, and the "point" is based on the notion that not all players on the ice when a goal is scored are equally responsible for the goal going in the net. Some are more responsible, so they get awarded a "point."

Dryden's notion -- as it relates to both the point and the error -- is thi: that suddenly the game breaks down, suddenly one team becomes vulnerable, and a goal is scored.

With the point stat, the idea is you want to reward those players with a point who caused that breakdown and carried through the attack.

You don't want to reward every player on the ice with an assist, because you're not interested in what happened before the crucial breakdown. You want to focus in on that breakdown and what came after it, and describe it with a statistical code.

So how do you go about doing that?

You could do this subjectively, and that way guys who effectively screened the goalie or made a big hit, but didn't touch the puck, might also get assists.

But, in their wisdom, the folks who formulated the "assist" stat chose a different route. They chose a mechanical way to award assists.

Of couse, they didn't want to give an assist to everyone on the ice, we can assume that. They clearly didn't think that would make sense. They wanted to reward only those players who were most closely tied to the goal being scored.

So they figured if they tracked back from the goal scorer to the two guys who passed to him, that would generally finger the right players, the guys who had all worked together to make the other team vulnerable to a goal being scored and then putting the puck in the net.

In a similar way, I'm tracking back from the goal to identify who is most responsible for the goal against. There are advantages to doing this mechanically, by counting passes for the assist stat, but I'm doing it subjectively, by attempting to identify the moment when the breakdown happens, and counting from there.

Now, when plus/minus was invented in the 1950s, it was done because some folks felt that all ten players on the ice should be recognized for the goal being scored. Some saw this as an advance.

Still, the goal and the assist remained as extremely popular stats, and the continue to be used by fans today. I would argue that when fans use the point statistic to make their own point, they are buying in to the notion that some players are more responsible for goals than others.

Anyway, what I'm doing here is addressing your point that plus/minus is the way to go. I don't think it is, and based on the times almost everyone uses the "point" statistic in argument, I think they are also voting with their feet, so to speak, and saying they're more comfortable with a stat that identifies those who are MOST responsible for a goal.

Now, with the Error, there is no tracking back passes, because defensive players don't have the puck. But the concept is the same, and in practice it amounts to essentially tracking back two passes. You go back those two passes and almost always you will find you have found the moment when play breaks down and one team becomes vulnerable.

If you don't think it's possible to identify that moment where that vulnerability begins, that is a reasonable enough position.

But I think this can be done with a reasonable degree of accuracy. It won't be perfect. It won't get it right all the time. But my argument is that it will get it right more often than just giving a minus mark to everyone on the ice. It will do a better job than mechanical plus/minus. This is the basis of our disagreement, I know, so we'll have to agree to disagree on this.

It's worth noting that in a similar way, I believe Dennis King will be able to identify scoring chances with a reasonable degree of accuracy and he will be doing valid work. Yes, he will be making subjective calls, but this is hockey we're talking about, not brain surgery, not rocket science. It's not so hard to identify a scoring chance, not so hard to identify the moment that a team's defence breaks down and the team is vulnerable to being scored upon. That is my basic proposition.

What's an MSM? Is it like, the media conglomerate in Edmonton?

It seems that Mr. Staples would make postmodernists of us all, no doubt.
If +/- isn't perfect, and GAON/60 isn't perfect, and the error stat isn't perfect, does the latter get us any closer to pure statistical equilibrium? If all are flawed, then give me the mathematical over the subjective, or give me death.

That said, I cannot in good conscience deride another's attempt to understand the lord's game.

A bunch of people subjectively pulling X number of names out of a 6 name hat, as it pertains to an observable phenomenon creates a mathematical (statistical) model. You just need more than one person making the subjective choices based on the same phenomenon.

It's math, and it's the basis of all experimental science.

Not arguing for or against the Error, but rather pointing out that saying "it's subjective" doesn't nullify its results. Having only one person analyzing it does. Much like the "scoring chances" stat.

Thought I'd jump in on this discussion. My point in enjoying Staple's attempt to track the goals more closely comes from my looking at the pre-established stats last season. Look at the Oilers PP last season - especially in terms of Defence.

Grebeshkov scored 1 powerplay point for every 49.5 minutes he played on the PP last seaon.

Gilbert scored 1 powerplay point for every 20.58 minutes he playd on the PP last season.

For every 9.28 minutes Grebby was on the ice on the PP last season a goal was scored

For every 11.58 minutes Gilbert was on the ice on the PP last season a goal was scored.

Who was better on the PP? Gilbert because he simply scored more points? What about the fact that more goals were scored in relation to ice time for Grebby? One would assume that the play started with Grebby more, and ended with Gilbert more.

No stat accounts for this. I think the error assignment would help identify anomolies like this - especially on the PK.

Take a look at some of the PK stats I delved into this season:
http://addicted-to-oil.blogspot.com/2008/09/hot-water-is-back-and-look-at.html

The error stat would definitely help determine who on the PK was more effective.
At the very least it'll create a larger distinction between those who are truly skilled, and those who get lucky a lot.

I have never looked at the scoring chances stat as a realistic thing because of it's dependence on point of view.

I am still undecided on the error stat value, but I like the project because it allows me to consider a differing point of view on a game. I may think differently than Staples that say...Garon wasn't to blame on the Huskins goal (I agree he was)and that would provide discussions and insight.

I would be curious to see the project results for the year, compare them to plus minus and tough minutes and see if there is a correlation.

I'm not worried about subjectivity in either scoring chances/errors; I think it can be largely compensated for with multiple viewpoints.

@David: That's a spirited defense, and well-argued to boot, bt I still don't buy it, and here's why: I don't think that choosing the goal as the time to track from works from a defensive perspctive.

For each goal, the offensive forwards are taking full advantage of the defensive breakdown; as a result, players are awarded points. It's a direct indicator of offensive prowess, because for every offensively successful moment, we have a goal.

Now, there are many defensive breakdowns during a game, but not all forwards capitalize equally. Take two forwards in a gam against Calgary. Forward A makes three defensive errors against Jarome Iginla. Forward B makes identical errors against Wayne Primeau. Despite the fact that both players behaved exactly the same, Forward A may end up -3 while Forward B remains even. The primary driving force behind errors is opposition, while the primary driving force behind points is teammates. The error attempts to tribute things primarily driven by the offensive player to the defensive player.

As a result, I don't really believe in Dryden's theory - the mistakes that he feels throw the game out of equilibrium actually occur all the time; it's the offensive player that makes them critical.

Mostly, this is what I was thinking last year too, but I was forcefully reminded of it when I decided to shadow the Error process this year.

David: I think that the fundamental problem is trying to argue the value of the error stat versus others like assists, points, or plus/minus.

The reason is simple - for the latter stats, we argue what value the information has. Simply put, is it useful?

For the error stat, there's a much greater fundamental problem. We're not arguing whether the stat is useful, we have to start by asking if it's even true.

And that's the bottom line on any subjectively assigned statistic. We can question the value of second assists, but you generally accept that they can be assigned correctly by any reasonbly knowledgable viewer. (Unless you have a conspiracy theory about Schremp intentionally having his counting numbers reduced, but I digress...).

Similary with plus/minus, it's of limited value, especially between teams. That doesn't change the fact that a monkey with a VCR can figure out who was on the ice at the time of the goal. It's a true statistic.

You yourself have changed the way you're assigning errors this year. It means that last year's numbers are not directly comparable to this year's numbers. If it was used league-wide, it would mean that the errors coming out of 30 different rinks would be judged on 30 different standards.

We already see variations in how hits and shots on goal are recorded in various rinks, and that's an order of magnitude easier to assign than the error stat.
Until the NHL puts in place a unified system of error-counting robots, I don't see how you'll be able to come out from under the black cloud of "is this even true?".

Jonathan, what right do you have to criticize the defensive stat for being inaccurate because it only tracks defensive problems on goals when you yourself cite GAON/60 in your post? We should use goals, because they are the defensive errors that are the most critical. Besides, let's not forget the error stat isn't meant to be a definitive read on the player's defensive ability; it's meant to be a more accurate replacement of the +/- stat, which is meant to measure whether the player creates or harms his team more by being on the ice. That stat should be tied to goals.