Burkie's Challenge Accepted: Lets Make An Analytics System!

[SebC]

Originally Posted by Freeway

Wins.

I'm not even sure if it's that... if the team is massively profitable despite losing... do they keep their jobs?

[MrMastodonFarm]

Originally Posted by SebC

I'm not even sure if it's that... if the team is massively profitable despite losing... do they keep their jobs?

No.

No.

Sigh.

No.

[EldrickOnIce]

Originally Posted by SebC

I'm not even sure if it's that... if the team is massively profitable despite losing... do they keep their jobs?

In Edmonton they do

[Itse]

Treliving needs to operate within the teams means, but I would assume he's not ultimately responsible for the team making money.

(To put it very crudely, I think Ken King is. Burke is somewhere in between.)

[PeteMoss]

Originally Posted by Rutuu

Goalies you need to figure out what drives sv% (since the difference between a 0.910 goalie and a 0.930 goalie is a playoff birth and a shot at Lord Stanley)...positioning, quickness (reaction time/athleticism) and puck direction (rebounds)?

1. Static positioning...where does the goalie position himself before the shot (Compare to an ideal model)

2. Dynamic positioning...where does the goalie position himself after the shot (in relation to the rebound...giving himself a chance at the second save...plus does he choose the optimal save type)
Ex: When in the butterfly does he lift the proper leg first after
Once again compare it to the ideal model

3. Where does the rebound go?...absorbed...or to a spot where it can be hammered in?

4. Add in a quickness component as well that has to do with reaction time and ability to go post to post.

Now you just need the ability to condense video footage of every goalie you want to evaluate and do the physics to figure out shot angles in a bunch of different rinks.

Or you do it like Feaster did and acquire a bunch of goalies for cheap and hope one emerges around the 0.920% mark for a season or two.

Goalies are hard to evaluate. Unless you are Hasek or behind a dominate defensive team (Tim Thomas/Rask) even the best goalies are inconsistent in terms of save percentage.

Varlamov is a top Veniza guy this year.. he's swung from .903 to .927 in terms of save percentage. Carey Price has gone from .905 to .927. Kiprusoff went from .903 to .933 (excluding his last season).

Makes you wonder if save percentage is that great of an indicator of goalie performance.

[Enoch Root]

Originally Posted by SebC

Burke and Treliving aren't tasked with improving the Flames advanced stats - their task is to improve the "bottom line" stats.

whoosh

[Coach]

Originally Posted by schooner

Economics degree here as well, but I was drunk, stoned or asleep for most the 4 years (alright to be honest 5.5 years, buts who's counting?) so I might have limited use.


On topic though, sure you can build a standard model like you would for any statistical problem, the biggest challenge is the quality of the data. And that is hard to work with in hockey. The integrity of the data is key to coming up with a robust correlation. Too many different leagues, coaches, rules, measures of success, statisticians recording data etc, to come up with a reliable predictor of success.

Are you....me?...

[Tinordi]

The big problem with corsi and the like is that it doesn't provide alot of ex ante insight for managers. Sure it can tell GMs that their teams are not good. It can't really tell you how to improve corsi. At least, it doesn't appear to me that acquiring high corsi rel players will improve your team corsi. So much of it is systems and coach as well.

But, this also confuses the issue with "advanced stats". Statistical analysis is beyond simply corsi or fenwick. It's about using all stats at your disposal to understand the game better. Typically this is about using event based statistics denominated by type of game (was it close) or by zone (which zone are you playing in?) and to understand how performance is affected by different variables.

The people who are most likely to benefit from using these stats are coaches not GMs.

[ricardodw]

After reading this thread my RGI stands up and my analysis that you don't win with 4 or more players having a low GRIT level holds up as the best practical application of secondary data to hockey management presented so far.

Of course there is not sufficient data being gathered and the quality is really suspect.

But with the minimal and suspect data available it was fairly easy to classify players as non-gritty..... there are some glaring anomalies in the Quality of data... Couture shows up as non-gritty and Butler one of the more gritty.

BUT perhaps that is the moneyball analogy... Jason Giambi was better by far than Scott Hatteberg but Hattenburg provided a much better contribution to winning for the salary he was paid. is Butler the equivalent to Hattenburg?? It would be interesting to see Butler's contribution on with a NHL team without the problem of the Flame teams he was on. He made much larger positive contributions on the Buffalo playoff team.

My RGI does does not help drafting as you can have up to 4 players with low grit. So I would say that the current Flames system is still full of low grit players.... If the Flames want to have a value pick like Petan in the 2nd round they have to know that they have to give up on one of Baertschi, Granlund (2nd round) or Gaudreau (4th round).

Intuitively Burke , Davidson and the best NHL brain thrust are already using the concepts of the RGI.

Burke knows that a team with Cammalleri, Hudler, Brodie, Baertschi, Backlund, Bouwmeester,Byron,Tanguay, Granlund and Gaudreau will not win anything no matter how big and truculent the rest of the team is or how great and skillful these smaller less gritty player play.

There should be no argument with this..... it is self evident.... except Feaster did not understand and stock piled these type of players well past the point where adding another one made the team worse. He also did not have an apparent strategy.... stockpile 3-4 small skilled players to fill the Flames small spots .... but then goes and signs Hudler long term to take the spot on a team that already had Tanguay and Cammalleri.

They are easy to add as the rest of league also understand that there is only room for 90 of these smaller skilled less gritty players in the NHL and hundreds of them playing hockey.

Examples Wellwood and Raymond Mason definitely have NHL skills but are border line on being in the top 100 non-gritty players.


I know that there this post will be derided and mocked but there have been no better analytic attempts made and I doubt that there will be.


The folks that support the CORSI have not come up with any way to improve a team using it .... no one , to my knowledge has gone out and said that the Flames need to get back Time Erixon because of his high Corsi or that we should go with Backlund as #1C because of his high Corsi and trade Monahon immediately because of his low CORSI. They might say that Backlund is more important to the Flames than he appears......but is Monahan a drag on the Flames?

[calumniate]

Think you have to include analysis of your own coaching and development group as well. A rookie can head off to Detroit and the Wings will typically get the max potential out of a player. Head to Edmonton and it's a different story.

On the player side, introduce DNA testing and analysis, or something of that nature

[DoubleF]

Let me try something silly. I want to focus on input that everyone discusses but never try to quantify.

Something like (For draft + minor)...

Spoiler!

(0-5)
Player base skill
x
(1 OR 2)
Professional maturity (NHL vs Minor grade; always lean 1 if uncertain)
+
(0-6)
fit of all (x teams/30 teams*30/5)
-
(0-5)
Spotlight profile (Balance out desirability and pressure)
+
(0-5)
team/positional need (0-1 /desired position(s )to fill * 5; No way you need to fill all positions... but if only need to fill 1 role and meets that role, then +5 bumps ranking)
-
(0-5)
handicaps (Size, attitude, play style, RH/LH etc. Max 2 per factor)


Max = 11, only because if the next guy really thinks he's that good, we'll borrow his vote in determining whether we're interested.

The reason for this calculation would be more like "sorting" of available draft candidates or UFA or something (But probably more on drafting). Multipliers are obviously something deemed perhaps more polarizing of the type of player will be churned out.

Player skill - No brainer
Professional maturity - NHL ready vs farm team training (Quickly polarizes players/splits the list in half for useful now vs prospects)
Fit of all - Useful as trade bait off ice/appreciating value player (Nod to BPA)
Spotlight - Should further lower individuals with low professional maturity down the list. Should be negated by the professional maturity above.
Positional need - Should remove players from list if we have no need for that role (Rank BPA by team needs)
Handicap - if players left over, what factors change the ranking of similar players


I feel like something like this is already done. Advanced sorting doesn't do much in terms of finding a good fit player. The inputs are also all subjective which means you could use this wrong or and get stupid outputs. (ie: Teams that think all players are skill 1000 and x2 for professional maturity) I wanted a number that was smaller purely on the idea that you should be able to count the reasons on both hands. If you have more than that by a significant amount, you're probably over analyzing the player.

Another could potentially be for established players:

Spoiler!

(0-5)
Player skill (on other team)
x
(0-2)
Effective use by team
+
(0-2.5)+(0-2.5)
Our ability to effective use + Our ability to fix
+
(0-6)
fit of all (x teams/30 teams*30/5)
-
(0-5)
Spotlight profile (Balance out desirability and pressure)
+
(0-5)
team/positional need (0-1 /desired position(s )to fill * 5; No way you need to fill all positions... but if only need to fill 1 role and meets that role, then +5 bumps ranking)
-
(0-5)
handicaps (Size, attitude, play style, RH/LH, injury prone etc. Max 1 per factor)

MAX = 6

Should be able to count on one hand a reason to chase someone more established. +1 to see if you hunt them down to keep them from other teams.

Many of these "values" we already talk about them. I tried to design the analysis to have polarizing values that somewhat eliminate hype or can really flush the player down the list if the player is not a necessary piece.

Maybe someone can also figure out a way to do a (balanced ranking + teams desired + residual value if walk) of a player for salary? Might be helpful for deals or figuring out how to negotiate a decent salary +/- figuring out extras and limits for "overpay" to know when to add or when to release. I mean, many posters already argue if players are worth between 3-6 mil. Someone is sure to figure out that calculation with inputs that are tweakable. Might want an actuary rather than stats person though.

Originally Posted by calumniate

Think you have to include analysis of your own coaching and development group as well. A rookie can head off to Detroit and the Wings will typically get the max potential out of a player. Head to Edmonton and it's a different story.

This seems pointless tbh. A team shouldn't be trading to keep good assets out of other teams hands. They should focus on having good assets they can use. I would assume this is over analyzing because it's a given. If a GM is traded though, the recalculation on players from the new team side should show this. The issue likely would be teams not inputting the players in the first place, or bogging down the players rankings with crap input values.

[Flames Draft Watcher]

Originally Posted by TheDebaser

2)Use tons of data from the entire league to create a kind of heatmap that could assign a shooting % from a given part of the ice. This would be ideal, but it would probably require programming knowledge. (If the data is even out there.)

The Flames actually do have this type of analysis already. The Pucks software shows where all shots come from. Using years worth of data you can assign % chances of scoring from certain areas on the ice. So they can modify SV% based on quality of shot and make a better corsi by taking into account quality of shot.

Had the chance to talk a bit with Chris Snow about this. What should be obvious is born out by the data. The majority of good chances are from the middle of the ice, from the slot all the way out to the middle of the point. But by far the statistically highest quality chances are from rebounds.

Several teams have modified corsi type numbers but are much more useful in that they aren't measuring all shots as the same. You need that extra step to get the numbers to be something more meaningful. Counting a shot from centre ice as the same quality as a shot from the slot 6 feet out makes corsi mostly worthless IMO. Hard for fans to do as you either need to subjectively rank shots as being of a certain quality or you need access to something like PUCKS to have the historical data of scoring % from every area of the ice.

[DoubleF]

What if... eliminating information was more beneficial? I mean, the amount of data must be insurmountable. Figuring out how to reduce "noise" would be helpful, no? My try on the earlier calculation tried to address this by pairing up + values and - values to reduce noise. (ie: High profile = high pressure; Fit vs team need etc). But it's still pretty complicated and useless. It's an attempt at a educated guess, sure. But still a guess.

Another issue is finding input data that is somewhat less subjective and more "clean". Skill on a scale of 1-10 might be a default idea, but it is pretty subjective. "How many teams want this guy" might be a more "objective" (dare I use this term?) input value.

For instance, let's say if we ranked Baertchi, I would likely expect ranges of 6-10 for skill across forums, across teams etc. Are we rating him as a player, or as a peer group? It's highly subjective. A 4 point difference is 40% variance. Now if you said "How many teams out of 30 would want Baertchi if he was on the market for 1 mil?" I could probably expect the number to be varied at most 4-5 teams of the whole 30. But again, what use is this info if it's still just a guess?

I would rather spend time eliminating info which every player and their peers already has. Everyone passes, everyone shoots etc. Sure, correlated numbers might show something... but you can't really determine if they will be any good on the team. It's an educated guess, but still a guess nonetheless.

Doing a radar chart on a team should probably fit somewhere. But speed for instance crushes stamina, while size would probably synergize? Again, not the most useful unless you can figure out the noise. It's not rock paper scissors.

[Mattman]

Originally Posted by Stupid

42?

OMG you are so STUPID

[Derek Sutton]

Somewhat on topic, somewhat off topic. Are stats kept on players who draw more penalties then others? I would be interested to know who has drawn the most penalties in the league this year, and what the trend is. Is it speedsters, agitators, little guys or divers? There could be a real advantage to targeting players who frequently draw more penalties then others. Then you add in a couple pp specialists and your goals per game jump, it should be that simple. You would also give up less goals by have the man advantage for much more time then other teams.

Could you imagine developing a system that would have you pass on say Sam Reinhart with the 4th overall pick in order to draft an unknown kid who fit your system better? (Come to think of it maybe Feaster had a system and that is why Jankowski was picked in the first round.) Billy Beane (Moneyball) had to do this due to the flawed process of the MLB draft, in this day and age the player rankings are pretty much common knowledge heading into the draft and fans would have horrible reactions to drafting players only because you have a chance to sign them. I know more and more teams are changing how they evaluate players and their scouting ways in general but at the top end of the draft they still take the "best" players.

When all is said and done, just draft kids with the best +/- ( I know, this contradicts my other though of putty together a team which would rely on its pp.).

[PeteMoss]

Here's the penalty stats (can't really vouch for its accuracy or how they figure it out):

http://www.behindthenet.ca/nhl_stati...42+43+44+45+46

[DoubleF]

Do you think there's an evaluation report somewhere saying how well the players may develop based on xyz development plans? I can't imagine all players are developed exactly the same. Coaches have different styles.

[indes]

Getting the input data is the hardest part as a lot of it is subjective. I think if you had a system that could map the puck and players distances from the puck at any given moment during a game you could measure some form of "impactness" (Indes' Impact Index if you will) I think with this system you could see how much of an impact a player had on any given play even if its not in the box score.