Faceoff Elo?

Steerpike · Aug 16, 2014

So I was watching a replay of an Avs-Kings game and noticed that predictably the kings used their last change to send out Kopitar to shut down Duchene. This led to Duchene having to take a bunch of faceoffs against Kopitar and I'd reckon that the average players Duchene/Crosby/Seguin/Giroux have to face off against are way better than average at faceoffs. This should lead to their faceoff percentage being much worse than it would be if they faced average competition.

Is there a magnificent record of faceoff match ups somewhere that I could use to run some sort of Elo rating system to determine how players rank? Has this sort of analysis already been done somewhere?

Thanks

EDIT:

Also, how exactly do they determine who wins a faceoff? Is it entirely dependent on the faceoff taker or does his team have to gain possession of the puck? When the puck is sent laterally often it becomes a battle between wingers. If one center has far stronger wingers than average it could largely inflate his numbers eh?

Caeldan · Aug 17, 2014

bedsheetrubber92 said:
So I was watching a replay of an Avs-Kings game and noticed that predictably the kings used their last change to send out Kopitar to shut down Duchene. This led to Duchene having to take a bunch of faceoffs against Kopitar and I'd reckon that the average players Duchene/Crosby/Seguin/Giroux have to face off against are way better than average at faceoffs. This should lead to their faceoff percentage being much worse than it would be if they faced average competition.

Is there a magnificent record of faceoff match ups somewhere that I could use to run some sort of Elo rating system to determine how players rank? Has this sort of analysis already been done somewhere?

Thanks

EDIT:

Also, how exactly do they determine who wins a faceoff? Is it entirely dependent on the faceoff taker or does his team have to gain possession of the puck? When the puck is sent laterally often it becomes a battle between wingers. If one center has far stronger wingers than average it could largely inflate his numbers eh?

I think it's the team that gets possession off the draw. So yes, wingers influence to some extent (I'm not sure what the split of drawing towards a winger vs drawing towards a defenseman would be) - some of the better centers in the past have provided quite a lot of credit to their wingers for making them look good in interviews.

Teeder9 · Aug 18, 2014

bedsheetrubber92 said:
So I was watching a replay of an Avs-Kings game and noticed that predictably the kings used their last change to send out Kopitar to shut down Duchene. This led to Duchene having to take a bunch of faceoffs against Kopitar and I'd reckon that the average players Duchene/Crosby/Seguin/Giroux have to face off against are way better than average at faceoffs. This should lead to their faceoff percentage being much worse than it would be if they faced average competition.

Is there a magnificent record of faceoff match ups somewhere that I could use to run some sort of Elo rating system to determine how players rank? Has this sort of analysis already been done somewhere?

Thanks

EDIT:

Also, how exactly do they determine who wins a faceoff? Is it entirely dependent on the faceoff taker or does his team have to gain possession of the puck? When the puck is sent laterally often it becomes a battle between wingers. If one center has far stronger wingers than average it could largely inflate his numbers eh?

I'm confused by the premise. As far as I can tell, the reason anyone is good or bad is because of the competition faced, but you want to look for a way to make players like Duchene/Crosby ect look better than they are because they face better competition and it's not fair? I apologise if I read this wrong

Micklebot · Aug 18, 2014

Teeder9 said:
I'm confused by the premise. As far as I can tell, the reason anyone is good or bad is because of the competition faced, but you want to look for a way to make players like Duchene/Crosby ect look better than they are because they face better competition and it's not fair? I apologise if I read this wrong

He want's to adjust for quality of competition; seems reasonable. That said, I'd argue that faceoffs are one skill set that isn't strongly correlated to the overall skill of a player, so I'm not sure how much value this would provide.

diceman934 · Aug 18, 2014

bedsheetrubber92 said:
So I was watching a replay of an Avs-Kings game and noticed that predictably the kings used their last change to send out Kopitar to shut down Duchene. This led to Duchene having to take a bunch of faceoffs against Kopitar and I'd reckon that the average players Duchene/Crosby/Seguin/Giroux have to face off against are way better than average at faceoffs. This should lead to their faceoff percentage being much worse than it would be if they faced average competition.

Is there a magnificent record of faceoff match ups somewhere that I could use to run some sort of Elo rating system to determine how players rank? Has this sort of analysis already been done somewhere?

Thanks

EDIT:

Also, how exactly do they determine who wins a faceoff? Is it entirely dependent on the faceoff taker or does his team have to gain possession of the puck? When the puck is sent laterally often it becomes a battle between wingers. If one center has far stronger wingers than average it could largely inflate his numbers eh?

Possession determines who won or loss the draw when a clear winner is not determined. The unknown part of face-offs is intent of the center. Depending on alignment of your opponents on a face-off a purposeful loss of face-off can be used effectively by a team to create a opportunity to gain possession.

I know when I coach face-offs are tracked by win/loss and scrambled, then possession is added to each result. So winning a face-off and not gaining possession, is a win without possession. This depends on wingers and D man and at times the center, if the center does not contain the opponents Center and he steps around him and gets possession. Lots of variables to consider, faceoff won and lost are not as clear as the numbers indicate.

EVBetting Site · Aug 20, 2014

It's a very valid and interesting premise. The difficulty would be that you would literally have to watch every shift of every game in every season to compile the data.

And you would have to establish ground rules. Does every player start each season fresh? Everyone gets, say 1000 ELO to start the season and everything starts at game 1? Or should this start from their rookie season and carry over, since if you start off with an easy stretch of some poor faceoff competition, maybe you have an inflated number, especially in a stat that has so much variance in it already. Over 82 games, though, and hundreds of faceoffs, you would think that this would somewhat even out. Personally I would estimate that around 1000 faceoffs, you start to get real close to your true ELO.

This project has definitely caught my attention and I probably will begin to watch games from last season and start putting together the rankings with these basic ground rules:

I'm going to first tally up all players who have taken over 3500 faceoffs (combined) in their career (active players of course) and then distribute them with the assumption that 3500 faceoffs is enough to eliminate (reduce) variance enough to reasonably say where they fall in the NHL hierarchy. This will hopefully show what kind of distribution I will need to use. Then I'm going to apply an ELO type system for only this past season, starting with game 1 of the regular season.

The exact scoring system will have to be determined after I compile my list of faceoff-men who have qualified. Since hockey faceoffs are much more volatile than chess matches, with edges much, much smaller, a different scale will obviously have to be used.

Stat Guy · Aug 30, 2014

I've always been a fan of ELO ratings. I whipped up a formula and ran every faceoff through the formula starting in 2007 based on the NHL.com real time stat tracker.
My Macro takes about 12hrs to run for each season, so I have only done 2007&2008 so far as a test, but I will do all seasons from 2007 -2013 as I work out the kinks.

Below are the results sorted from 1st to worst for the 1st two seasons.
Starting rating is 1200...so that is the middle ground. Anyone over 1200 is "above average" The further above 1200, the better they are. The vise versa is true.

Because all players started at 1200, the first year run is pretty much a right off as they players are not correctly weighted. But as more and more seasons are run, the more accurate the rating will be. (ratings are carried over from one season to the next)

For those new to ELO. For an indepth description, try reading this...
In a nut shell, it is a ranking system that takes away points from the loser and gives them to the winner. How many points is determined by their ranking going into the matchup.
If a higher ranked player wins, he gains fewer points from the lower ranked player because he was "suppose" to win. The opposite is true. If the lesser ranked player wins, he will gain greater points.
This is an effective way to determine the "Level of competition". So a player who always goes up against the other teams top faceoff guy, may have a lower winning percentage, but will still have a high ELO rating because they gain more points for each win! (and lose less points for each loss)

Example. We pit the top ELO Bobby Holik at 1283 vs the worst ELO Richard Park at 1126.
If Park wins, he will gain 3.56 ELO points and Holik will lose 3.56 ELO points
If Holik wins, he will gain only 1.44 ELO points and Park would lose 1.44 ELO points.
Holik would have to win about 72% of the time to maintain his ELO rating versus Park.

Anyways, here are the results. I am a little ashamed that I can create an elo formula in excel, but I cant figure out how to format a table on the boards...sorry

Micklebot · Aug 30, 2014

Stat Guy said:
I've always been a fan of ELO ratings. I whipped up a formula and ran every faceoff through the formula starting in 2007 based on the NHL.com real time stat tracker.
My Macro takes about 12hrs to run for each season, so I have only done 2007&2008 so far as a test, but I will do all seasons from 2007 -2013 as I work out the kinks....

Would it make sense to feed the results from the 1st season back into the macro as the new starting point for each player, and keep cycling it until your results for year 1 normalize (I'm fully aware this would take a really long time, just more curious as to whether it would help increase the accuracy for early years)?

Also, might be cool to see how this ranking matches up against raw faceoff% over the same time period; if it's virtually identical, 12 hours might be more beneficial letting your computer run the SETI screensaver...

Anyways, here are the results. I am a little ashamed that I can create an elo formula in excel, but I cant figure out how to format a table on the boards...sorry

Column 1 name|Column 2 name
Column 1 row 1 data| Column 2, row 1 data
Column 1 row 2 data| Column 2, row 2 data...
[/table*]
^Remove*

Here it is fixed, btw you have PJ Axelsson and John Pohl in your dataset twice

Brainiac · Sep 2, 2014

Even if I think that faceoff ability is slightly over-rated, I do like your approach. Very interesting take on the subject and the results seem solid. Good work!

Stat Guy · Sep 3, 2014

Top 50 After 2009

Stat Guy · Sep 3, 2014

Bottom 50 after 2009

Search

Search

Faceoff Elo?

Steerpike

We are never give up

Caeldan

Whippet Whisperer

Teeder9

Free rent for Mo?

Micklebot

Moderator

diceman934

Help is on the way.

EVBetting Site

Registered User

Stat Guy

Registered User

Micklebot

Moderator

Brainiac

Registered Offender

Stat Guy

Registered User

Stat Guy

Registered User

Ad

Latest posts

Upcoming events

Ad

Ad