Analyzing Keeks' Rebound Rating System

[mcdavies]

Hi all...

My 11 year old son has attended Keeks' goaltending camp for the past few summers and both he as a participant and myself as a spectator have learned a ton! If anyone is looking for a super summer goalie camp, this is it!

At the conclusion of the camp this year, Keeks talked about using a rebound rating system for goalies. The premise being that the coach should be measuring rebound control as this is such a key component of a goaltender's game. I believe Keeks has posted his approach on here before, but as a refresher, each shot on goal is rated as follows:

1 - Goaltender makes the save and controls the rebound (covers up or passes off to a defender)

2 - Goaltender makes the save and steers the rebound into one of the corners

3 - Goaltender makes the save and the rebound goes straight back out (the goaltender remains square to the puck)

4 - Goaltender makes the save and the rebound goes off to one side or the other in front of the net (goaltender is not square to the puck)

5 - A goal is scored

I have been using this over the past few weeks for my son's team and the results are really interesting. I have attached screenshots of a spreadsheet I put together to track performance of the two goalies on the team (I couldn't figure out a way to attach the actual spreadsheet!). When I look at this, it provides good input and raises a few questions:

1. Note that the Goaltender B has a significantly better rebound rating than Goaltender A. This has a direct correlation to both save percentage and goals against average.

2. Question - what is considered a "good" average rating? I initially thought that a rating under 3 would be considered good. However, at this point Goaltender A is 2.61 and Goaltender B is 1.93...both significantly below 3. Having had time to look at actual results, I now believe that Goaltender B being below 2 is exceptional and Goaltender B has some work to do. However, I'm not sure...any thoughts?

3. Does it make sense to rate a 5 when a goal is scored? The system is not designed to assess saves vs. goals...seems to me that it may make more sense to eliminate goals and have a 4 point rating system. I also think that scoring 5 points when there is no rebound to assess significantly skews the average results. I'm considering moving to a 4 point system but would appreciate thoughts here as well.

4. The raw data seems to provide a better analysis than the average. I took the ratings and plotted them graphically (included in the spreadsheet). Although far more subjective, you can see just from the overview that Goaltender B has superior rebound control. Further, Goaltender A has far more instances of 3 and 4 rated rebounds...this gives an indication of what he needs to work on.

Overall, I find this system to be super. I can point out the importance of rebound control to the boys and show them what they need to work on. When combined with save percentage and GAA, this becomes a very complete stats package.

Input / feedback is always welcome...

Mark

[mcdavies]

Perhaps this version of the stats will be more legible...

[Law Goalie]

I see what you mean, but if you removed goals from the system, you should remove controlled rebounds too, since they are not, in fact, rebounds: the puck is either caught, trapped to the body, deflected out of play, or passed to a teammate. Any situation in which a rebound was not given out would be null, and the others would be rated 1, 2, 3.

The raw data is, as you say, more useful, but condensing it into an average is a handy way to gauge overall progress., I think.

[marvintpa]

Let's take a small but rather extreme example: Goalie A lets in 5 goals with no rebounds, and stops 10 shots with a gut trap leaving no rebound. Would he be superior to goalie B who let in 1 goal off of a rebound back to the player and stopped 10 with a gut trap? According to Keeks B is the superior goaltender (13points vs. 35), and if you drop the 5 for a goal then A is better (10 vs. 13).

The moral of the story is, generating a bad rebound is better than not generating any rebound at all.

At least that's the way I view what Keeks is trying to teach.

[The Wall33]

I am doing a lot of rebound tracking using this system and the average I am seeing is around 2.25 (NHL goalies - pre season).

I think that as long as you use the "5" for goals on a consistent basis then the comparison is the same for every goalie and there is no need to eliminate it.

One note: the "4" is a tricky one and is somewhat subjective. I have discussed this with Keeks recently and the intent of the "4" is any rebound that ends up in a high percentage scoring situation, especially if it is to the weak side.

A rebound to the weak side that clears the offensive player(s) is really a "2" not a "4"... I hope that makes sense?

The main point that Keeks made was to be consistent in the ranking system used. A "2" for one goalie must be a "2" for all goalies thereby ensuring consistent ranking...

Cheers.

Wellsie

[Lord of the Cage]

I think that Keeks' system is more than a rebound effectiveness calculator. It is a system that rates the performance, consistancy, and ability a goaltender has at first stopping the puck and secondly the goaltender's ability of controlling what the puck does after it hits them. Is it perfect? No. Does it weight the quality of the shot or scoring chance? No. Is a number 4 worse that a 2? Depends on the game situation. I can go on and on with situations from a game where there are going to be problems with this system. I think with this system, coupled with a schematic for rating shot/scoring chance quality gives a fair representation of a goaltender's ability to stop the puck and control the puck. G.A.A. is more a team stat and Save % is very basic and does not weight a dump in on net any different than a one timer in close where the goalie has a long way to travel.

It is pretty simple. The more a goaltender stops the puck and retains the puck the more control they possess and better success they will have no matter what level they play at or how good or bad the team is as a collective. The closer you are to having an average of 1 point the better the goaltender is at making saves and retaining full control of the puck. This is a goalie's job no matter what the situation. At least in the sense of net responsibilities.

When Keeks and Wellsie put on a clinic a while ago and Keeks was nice enough to have a Q and A he said something that stuck with me. He said the only one to trust out there is yourself because once you give up possession of control you put your fate and destiny in the hands of others. This is paraphrased, but that was the jist of it. I have a very competitive nature and try to keep this as well as other little things in the back of my mind when I am playing. I don't have the luxury of someone to track these numbers for me so I can review after a game so I just rhyme off this system in my head as I play. Keeps me focused and aware and most of all taking responsibility of my actions.

You have to calculate goals. That is the largest mistake a goaltender makes. No matter if a goal goes straight in or after a 3rd or 4th shot it is still the goaltender's responsibilty to stop the puck and control the puck. If a goaltender is racking up a lot of 5 point marks, or a lot of 3 and 4 point marks with a 5 at the end they are equally as bad in my opinion because the goaltender is not being effective at first stopping the puck and secondly retaining control which I rate as equally important as stopping a shot.

That is just my take on the position and my feelings that a goaltender has noone to blame when goals go in regardless of the situation other than the one staring back at them in the mirror.

Lord of the Cage

[efstratios1]

I always thought 5 was better than 1! Now you guys tell me....

[marvintpa]

As the original question appears to have been answered, I'll pose another that LOTC insipred. Shots off of rebounds would obviously also count in this scoring system. In those instances when you're fully half butterflied out to the side, you just stopped the puck with the lower pad area (near the boot break) and the opponent is just whcking away (2, 3 or 12 times, you know, before your defensemen realize the puck is in your own zone) what would those rebounds be scored as? The puck never gets more than 6" away from your pad. Just curious. It would appear to be a 3.

[Lord of the Cage]

marvintpa,

I would view your situation as a 3 on the initial shot and any shot after that where it is right infront as well.

In this particular situation it can be viewed from a couple different ways. The first is the goalie has to know of the threats even before they are threats. So with a fully extended initial save you have to know of the possible outcomes after the first shot. This situation could have been prevented with better stick use on the initial shot and a score of 2. Secondly if missed by the stick and only a few inches from the leg the goaltender should be aggressive and cover up sacraficing the hand to gain control of the puck. If the goaltender gains puck control before the opponent has a chance to get a whack at the puck it turns this 3 plus point situation into a 1 point score. With an in tight play like this on a second shot the player is just trying to jam the puck and your leg into the net so a solid low ice position and aggressive hand limits the number of chances for the player regardless if the Defenceman has lost their check or assignment. Furthermore instead of sliding your leg out to the side and cushioning the initial shot an aggresive leg movement into the shot could have produced a 3 or 4 higher in the defensive zone or even a safer less threatening area allowing for just the initial shot or a shot where you can regain position in time and get full control on the second shot chance. This also leads to knowing the abilities of your game and how your gear works for you. A harder pad with lively rebounds no matter the situation could have provided relief.

There are a lot of outcomes from a situation and as Keeks has said the only difference between some beer leagues and an NHLer is that the NHLer can see plays develop before they are present and anticipate the likelihood of the most probable outcome.

Lord of the Cage

[The Wall33]

Tom, the strict interpretation to your question is that if it is a shot on goal (ie. if teh goalie did not make the save and the puck would have entered the net) then it should be rated.

Notwithstanding the info JR provided (which I agree with) I would likely score each opf those 'quick whacks' as a "2" unless the puck eventually squeezed out to the weak side in which case it would be a "4"... and most likely, eventually, a "5"...

Wellsie

[Law Goalie]

That's the essence of the confusion. What Keeks worked out is not so much a rebound-rating system per se as it is a way of rating a goaltender on a per shot basis far more reliably than save percentage. The problem with save percentage is that (oh christ, I'm slipping into Kantisms again...) it's an attempt a priori to determine the likelihood of a goalie making a save, but it describes no specific save or save-patterns at all. Keeks system, on the other hand, is a purely a posteriori analytical tool: it describes the way the goaltender handled his save-situations (from highly positive to highly negative) both in general (as an average over a game or a season) or in specific (looking at the raw data in itself or for patterns).

Consequently, the real beauty of Keek's tool is that it can incorporate just about any other datum. You could add a time component to see how the goalie handled flurries of shots in short sequences or how he handled a shot after a long space, or whether he was a slow starter or a strong finisher in games. You could add something as simple as a 'PP' designator to abstract that class from the overall as necessary, and show how the goalie's rebound rating changed in those situations.

It's a damned brilliantly simple idea, I gotta say.

[efstratios1]

Stop using all those big fancy words Law, Wellsie won't understand you jks.

[psofmars]

Don't forget Keeks most known story - basically, if it goes in the net, there's "No Rebounds"

Go LSSU Lakers!!! sorry to Keeks, had to get my shot in.

I wish I could count shots, let alone keep stats. Sean keeps kind of a running feel for his rebound stats in his head. He's always liked Steve's point system. He can use it to show goalie-clueless coaches that he has some idea what he's trying to do.
Pam/Sue

[The Wall33]

Originally Posted by efstratios1

Stop using all those big fancy words Law, Wellsie won't understand you jks.

Wow... Sievie and I finally agree... Duncan, those are some damned fancy ass, high-falutin' words... When I send Keeks stuff and use fancy words he tells me to shut up and speak more plainly... so I say the same to you my friend!

One criticism I have of the "Keeks Sheet" is that it does not allow for recognition of extremely difficult saves vs. run of the mill saves.

For example: Goalie A faces 40 shots in a game but 15 of them are in the 'kill zone / house' and are very high quality chances. Goalie B faces 40 shots and only 5 are in the 'kill zone / house'... how do you distinguish who is the better goalie?

I had that situation when I was the goalie coach at Ryerson University. Our guys gave off a ton of shots in the kill zone and I adapted a modified ranking system that gave a lower rating to high percentage shots.

For example: a one-timed shot in the slot off a pass with a rebound to the corner would normally be a "2". I would put a circle around that shot and weigh it as a 1.5 because it was a great save off a very difficult shot.

Alternatively, a shot on net from the other end of the ice on a PK would normally be a "1" but I would put a square around it and weight it a 1.5 b/c it was a very simple save.

In conversations with Keeks he said not to worry about that as it all 'averages out' over an 82 game season, which is fine with goalies in the NHL, given the relative parity. In the minors, during a 22 game schedule, it is a different scenario.

One last point... Keeks has developed a new rating system which we/he is using right now. Nothing radical has changed, just a bit more detail and wrapping up of year-over-year numbers etc. I'm sure we/he will release it in due course, when it is fully refined.

Wellsie

[TartanBill]

I doubt that an average is much use. Comparing frequency distributions is probably more instructive.

The rank ordering is surely correct, but to evaluate the correct weightings requires a multiple regression. You dependent variable is actually #5, the total of 1 through 5 should sum to the total shots.

By plugging in the stats for a number of different goalies, you should be able to measure the weights and correlation, that is how much of the variation is explained by these types.