Here are some details:
Each goal starts out with a value of 1.00.
From this, value is added or subtracted according to the following protocol...
Goal Differential:
0 .2
1 .2
2 0
3 -.2
4 -.4
5 -.6
Period
1st 0
2nd 0
3rd .1
OT .3
Manpower
PP -.1
2 Man PP -.2
ES .1
SH .2
PS 0
Location
Home 0
Away .1
Opponent
Divisional .2
Conference 0
Other conf -.1
Opponent's win-loss record
Playoff team 0
Non-playoff team -.15
Opponent's defensive record
Stratified based on GAA: range .1 to -.1
Time of season
Early season 0
Late season 0.1
Special situations
EN Worth 0.5
Assists
First assist = value of associated goal
Second assist = value of first assist -.2
Each goal ends up with a calculated value.
The distribution is then statistically modified to zero the league average at 1.0, and to place the Clutch Factor into a meaningful and relevant range.
Undoubtedly the factors which we have selected are debatable. I look forward to comments about inclusions, exclusions, relative weighting.
Thanks.
7 comments:
Late season goals should not be worth any more than early season goals. If a player's team isn't in the playoff picture late in the season, scoring goals in later games is not more 'clutch'. In fact it would have been more clutch for him to have scored earlier in the season to keep his team in the race. See what i'm getting at? every win counts for 2 points, no matter when you get it. Clutch factor should be determined on the situation in-game, which you have otherwise done a good job evaluating. I think giving bonuses for divisional games is a good idea, because divisional games are more valuable. But early season divisional games matter just as much as late season divisional games. Also, please do not include a consideration of where a player's team is in the standings vs the opposing team (if you are considering adding this). Players in tight divisions (like say, the awful southeast division) would get a bigger clutch factor which seems a little cheesy. Just a request for what not to add to your system if you are considering it. Cheers.
Hi Matt,
Thanks for the feedback.
Regarding late season games, you may be right. I could make a case either way. As it stands, in the present stats, that factor has not been incorporated, as they've all been considered "early season" games so far.
Regarding the opposing team issue, we decided not to use any factor which takes the player's team, or teammates into consideration. Also, there is no factor which looks at "tight" divisions versus... uh... loose? divisions. However, we thought, and think that looking at the opposing team's strength gave some value to the equation. So if a player scored five goals against Tampa, that wouldn't allow him to pad his stats as much. But if those five goals game against Jersey, then maybe they'd be worth more. So to use your example, scoring against the southeast division, with most teams out of the playoff race, would give a player a lower, not higher, clutch factor.
What do you think?
SubPlot H
Yes I think opposing team's strength is probably a good idea to factor in. One problem is that early on in the season, teams are not separated by much in the standings yet and the good teams have not yet emerged. So you would not yet be able to evaluate strength of opposition as accurately as you could midway through the season. One way around this would be recalculating the strength of opposition factor for past goals scored based on current standings as you progress through the season. One other alternative is to base the strength of opposition factor on the opposing team's record in their previous ten games. This would have the benefit of gaging how hot an opposing team is at the point in the season the particular goal is scored at, as well as doing away with any need for recalculating strength of opp. factor for past goals scored as the season progresses. I think the best way is to factor in how hot the opposing team is as a part of clutchness. I'd love to help you guys out with this project, send me an email at mfoulger A T gee mail dot com if you want to bounce any ideas off me or would like any more suggestions.
Hey Matt,
Thanks for the feedback and the offer. Definitely will take advantage of your input as decisions need to be made.
Your point is a good one. Just so you know, the way we've chosen to assess the opposition strength, for this calculation, was to look at the how many points, and the GAA, at the all-star break. That way, if a player scores a bunch of his goals in the first few games of the year against a team which proves to be weak, the weakness will be revealed by the all-star break stats.
Make any sense?
First off, very cool. Where'd you get the data! I love trying to manipulate stats!
Me and a buddy were discussing and it doesn't appear to us that the "game situation" is factored in. For example: Ottawa is up 2-0 and Spezza scores to make it 3-0. Differential is 2 - no bonus. No problem. However, say Ottawa is down 2-0 and Alfy scores to make it 2-1. Differential is 2 - still no bonus. We feel that the "getting the team back in the game" factor is important.
Overall though, without factoring in everything and the kitchen sink, it looks like you nailed all the important criteria. Solid!
I really like your idea. It is great to see someone dissect individual performance in hockey like they do in other sports.
I have a few comments: I think weighing goals against divisional opponents more heavily is wrong. Ovechkin and Kovalchuk play in the worst division in the league and shouldnt be rewarded for it. Goals against non-conference powerhouses are arguably more important and players shouldnt be penalized for it.
I also think that opposing players on the ice at the time of the goal is pretty important. This adds another level of complexity to your analysis but it seems important.
Have you tried altering the values assigned to each category to see if the final rankings are consistent? I would be interested to know if Ovechkin routinely lands at the top.
Hey Guys,
Thanks for the feedback.
With regard to the differential, your point is a good one. Technically speaking, the differential we used referred to the score which was the RESULT of the goal scored. ie, if the goal in question made it 3-1, for example, then it would be called a diff of 2. If a goal was a tying goal, then the diff was zero. Make sense? What do you think?
Regarding the other comments, there is some validity in what you're saying, however, just the way the playoff system is set up, it strikes me that a divisional/conf game is worth more than a non-conf game. Still, it is true that teams in weaker divisions can pad their stats on weaker teams. That's where there are other modifiers, to do with the strength of the opposition. Not perfect, but an attempt...
I agree that there are other factors, such as the opposing players on the ice, which may be useful. Maybe for version 2.0.
Yes, we did play with the numbers and the weightings, and found that the rankings were pretty solid.
Thanks.
subplot h
Post a Comment