Wednesday, 15 July 2015

My Errors, Being Human, and My Central Position

Recently there was an individual who was running rampantly throughout Twitter yelling at how wrong I am.

For those that do not know, I frequent the online community HFBoards quite often. Like Twitter, I've found it to be a good place to both advertise about my work and discuss it.

I do this because discussing my work with individuals is the best way for me to learn. It's how I have improved. I know so much more about hockey and hockey analytics than I did months ago, let alone years ago.

This unique individual was calling me out for being wrong all the time (interestingly enough, right after an HFBoards banning). Which was ironic, since he was wrong about what my actual position was nearly every time.

So let's talk about this being wrong all the time...

Now don't misunderstand me here, I'm not saying I'm never wrong. That would be extremely foolish. I get stuff wrong as often as any other human being.

I can usually suss out the source of my errors in analysis from being overly optimistic and trying to reason out when the numbers point out a negative. It's almost never an issue with the numbers; I just can sometimes argue myself away from what some of them say.

I've made an entry error in my data before. This was the only time I've caught myself doing it, but I do not dismiss the possibility that there are others... I just have not found them or have had anyone report them yet.

If you follow me on Twitter, I'm sure you see the plethora of typos that go out. I tend to think faster than I write, and so I throw out garbled Tweets that have mixed thoughts in them. It's not often that I have to edit some typos and grammar in blog posts that are pointed out to me by those with a far more keen eye than my own.

So I'm not perfect, and I know it, but what this guy is trying to say of me is different.

There's a difference between being human and making mistakes, and having a fundamentally flawed central position.

So what is my central position?

Well it's simple and can be boiled down to this:

1) I believe hockey to be a goal scoring contest, where the team with the most goals wins.
2) I believe that in the long run the better, more-deserving team will win more often not, but in the short term that is not always the case.
3) I believe certain teams, players, and actions tend to beget more wins than others.
4) I believe that certain events have a relationship with winning within the current state environment of hockey.
5) You can combine the above factors, track the particular events, and then predict players who in the long run will improve a team's chance at winning.

Statistics do not tell you everything, but they do give you more information.
Statistics do not make decisions for you, but help you make more informed, and better, decisions.

Simple. It seems logical, at least to me.

There are many different places where this central position may be applied: player evaluation, prospect predictions, optimizing rosters, etc.

So to help this individual, I have tried to get him to screen shot where "I was wrong" so I can RT to everyone where he's right and discuss where he may misunderstand what my position was, as he is prone to do often.

He didn't comply though.

He only felt it necessary to reply what the subjects were on, without actually quoting what I did or say that was wrong. He would then reply that there's no point to him taking a screenshot, as I would just ignore them.

Not sure why a screen shot is considered ineffective but a general tweet is not, but alas, here we are.

So, without further ado, here is where he has vocalized an issue with me, and me using it as a lesson on many key areas of analytics and what I, Garret Hohl, believe to be true.

Subject One: Alexander Burmistrov

Scoring versus driving shot differentials

Where ever there is controversy or conflicting opinions, I investigate, discuss, and report on it. It's my job to do so. And some players are more prone to these discussions than others, which erroneously leads people to believe that I am subjectively biased to these individuals because I talk about them more.

It's simply false. I talk about those being talked about. In analytical blogging, you want to teach new ideas, refute false ones, and also gain readership. These three reasons incentivize one to talk about the controversial players more often.

My avatar at HFBoards is a picture of Burmistrov in the KHL simply as satirization of those who believed such things of me.

Now, Burmistrov is one of these players. Why? Because he's an effective player but not an effective scorer. It's the truth.

The eye-test, especially for the layman hockey fan, is prone to certain players and events. We tend to remember certain events more than others, and we tend to look for certain events as well. We also do not notice or look for any non-events, as in noticing when things do not happen.

Anyways, Burmistrov, who is a great Corsi driver, but a poor scorer. These players tend to help their teams more than people give them credit for.

Even GMs undervalue shot differentials relative to scoring. Eric Tulsky, now of the Carolina Hurricanes, showed that Corsi driving provides more wins/dollar than scoring does.

It doesn't mean that Burmistrov is a perfect player, far from it, but these players can be useful too. If used right.

Using player types correctly

Speaking of which. There was a time during the 2014 summer when Fear the Fin asked me if Alexander Burmistrov could be better than Mark Scheifele.

I answered truthfully, that it is possible. I even mentioned that at that time it was possible that Burmistrov could be a more important addition for the Jets.

That irk someone the wrong way, mostly because they did not understand my point.

At the time, Scheifele had posted slightly better scoring pace in his NHL rookie season, but worse shot differentials than Burmistrov at the same age. Scheifele was also a pretty solid scoring forward in the OHL, and out scored Burmistrov in their draft season, as well as post superior NHLEs for the following seasons.

The Jets in 2013-14 were a middling shot differential team, with poor goaltending, but decent finishing.

My point with Burmistrov being potentially more valuable to the Jets was that he filled more of need.

The Jets needed to improve their shot differentials outside of their top line of Andrew Ladd, Bryan Little, and Blake Wheeler. Burmistrov presented a possible remedy there.

This was the opposite situation as the Los Angeles Kings, when they added Marian Gaborik. Gaborik was never that great of a performer in shot differentials, but could score effectively. The Kings as a whole were the opposite. Gaborik was more valuable to the Kings than he would have been to other teams.

Luckily for the Jets, they were able to fix some of this from other sources.

The Jets added Mathieu Perreault in the summer. Perreault has long been a fancy-stat All Star. He both scores at an extremely efficient point per minute pace while also drives shot differentials extremely well.

Adam Lowry also joined the Jets and turned out to be very much like a Burmistrov-lite. While a year older, Lowry performed similarly to Burmistrov's past performance in shot differentials, although scored at a lower rate.

There were also a few other things that improved the Jets shot differentials. Dustin Byfuglien shifted back to defense where he performs much better. Zach Bogosian was swapped for a minor upgrade in Tyler Myers. Deadline additions of Jiri Tlusty and Lee Stempniak provided depth to the Jets bottom six.

The most interesting change though was in Mark Scheifele. I had assumed, wrongfully, that Scheifele would likely improve more-so in scoring than in shot differentials. The opposite happened though.

If there is one area one could actually point out me being wrong in this situation, it would be in Scheifele's development arc.

I'm fine with that.

I don't think many would have expected Scheifele to develop that way and the team I cheer for having a young player exceeding reasonable expectations is a good thing.

Aside: There is however some evidence that Scheifele's development may be over exaggerated, since most of his best minutes were when between Michael Frolik and Mathieu Perreault, two players that tend to out shoot opponents with or without Scheifele. However, I will still hope that it is a positive sign.

On off-ice issues

At the end of the 2012-13 season, there were a lot of rumours going around about Alexander Burmistrov and Claude Noel.
People began to question Burmistrov's character.

I tend to be quite agnostic when it comes to these discussions.

People tend to discuss a player's character as a static variable. It's a mistake to do so (although if I am to be frank and fair, analytical types fall for the same mistake too).

People change. Not just in maturation too.

How often do two individuals become friends, schism, and then come back together again? How bout couples?

Relationships and such are quite dynamic and complex. So is the psychology of an individual. To typecast players as immature or mature is oversimplification.

I never made a point that Burmistrov did not have off-ice issues or conflicts going on. What I did do however is discuss that he could not or that the issues could be more complex than how many painted them out to be.

Subject Two: Jiri Tlusty and Drew Stafford

Over at HFBoards there were those that mentioned they preferred Jiri Tlusty over Drew Stafford. There reasoning was that the on-ice impact value probably wasn't too different and Tlusty likely cost less in Cap Hit.

I made one little mention that went like this:

Now to be completely fair, the sentence "Tlusty tends to outscore Stafford on both 5v5 and PP" is extremely vague. That's my fault. It was also on a Friday that I posted this, and I tend not to visit HFBoards on the weekend, so I never got the chance to clear up any questions.

Also, in being honest, I'm not 100% sure exactly what I was being specific to, although I can show what it was possibly about.

Whenever you increase sample size for hockey, you have trade offs. You improve the number of events, thus decreasing the impact of an outlier; however, you lose recency in how a player may be developing or aging.

For those reasons, I normally use three years, although I have sometimes used two in my past work at times as well.

Over the past three seasons, Tlusty had scored 1.71 points per sixty minutes at 5v5, while Stafford comes in at 1.70. The 2012-2015 sample indeed has Tlusty out pacing Stafford at a per minute rate, although the differences are so small it's too marginal to be significant to say one is exceptionally better than the other.

The contrast is much more stark on the power play, where Tlusty has paced at 3.02 points per sixty minutes, and Stafford has paced at 2.49.

However, it is also true that Stafford has been historically better in 2010-2012 range, both 4 and 5 seasons ago, while Tlusty was entering as a full time player into the league.

There is the argument about Stafford being negatively impacted by linemates with the Buffalo Sabres, but the Carolina Hurricanes have been a fairly week team as well. In addition, since 2011 Tlusty's linemates experience a +0.60 goals for per sixty minute increase with Tlusty, while Stafford's linemates only experience a +0.29 increase.
So, yes, it would be technically true --if I were discussing the past three seasons between the players in terms of point production per minute-- however the difference is so marginal that I'm skeptical that was my point when I first looked at it.

What was I probably talking about in the vague one sentence post now trying to be forced out of context: goal scoring.

Tlusty scored at a 1.51 goal per sixty minute pace on the power play and a 0.87 pace at 5v5. Stafford meanwhile has posted 0.83 goals per sixty on the power and 0.75 at 5v5.

It's unfortunate part of the nature we have with the vocabulary in hockey. Scoring points or scoring goals is both still scoring.

Which ever way is true, it's still my fault for placing a vague one sentence statement and not explaining when others were confused.

In the context of the original argument though, I'd still take Jiri Tlusty at the price he will likely cost over what Stafford did end up costing. It's definitely not something to be going out all nutbar like over Twitter. C'est la vie.

Subject Two: Andrew Copp and Jim Slater

Possible does not always mean most probable

It's funny that the person re-mentioned this discussion, as the last time I wrote on this blog was because they didn't get the point... and they still don't.

There were a great many getting excited about Andrew Copp, many calling him a bluechip prospect or A prospect. Some placed him in the same conversation as Nikolaj Ehlers and Nic Petan.

Don't get me wrong. I like Copp. He's a smart player, fun to watch, and I can see him possibly becoming something similar to Adam Lowry, and be a plus bottom-six piece. But I don't put Lowry or Copp in the same tier as Petan or Ehlers.

Anyways, being the person that I am, when many talk about best case scenarios, I use an example of how there are other possibilities. In this case I simply pointed out how Jim Slater was once a highly thought of prospect who even outscored Copp in the NCAA.

This person took it as though I was saying Copp wouldn't be better than Slater. An odd position to have for someone like me, who is always talking about probability not equating to destiny.

The linked post above contains the context that is being ignored by the individual; however, I have something to add since now he's just saying I was wrong in saying Slater outscored Copp.

With the PCS project, Josh Weissbock has era-adjusted all scoring numbers for players for every league, provided we have the data for it.

Slater and Copp's era-adjusted points per game comes out as:

Again, I'm not saying Copp will not be better than Jim Slater. In fact, I'm hoping he will be better. Slater was once a prospect that was both highly thought-of and also posted relatively better scoring numbers in the NCAA does add context in creating reasonable expectations of what is possible.

I mean, even if Andrew Copp --a fourth-round selection-- becomes only a Jim Slater --a first-round draft pick that played over 500 games and scored over 130 points--, he's still providing amazing value for a mid-round selection.

Final Thoughts

Honestly, I'm surprised this is all the guy could find on me.

It boils down to:

1) Him not understanding my actual position or being ignorant of the context of the discussion. (AKA as him being wrong haha)
2) Me not being able to predict Scheifele developing into a better shot differential driver for the 2014-15 season than scorer.
3) Me being far to vague in a single sentence and not providing context of time frame.

If this is really the extent to me being "often incorrect" or whatever... I'm doing better than I even thought or expected.

I can live with this.

No comments:

Post a Comment