Starting Leafs 3yr Averages

Discussion in 'Toronto Maple Leafs' started by satyr9, Sep 26, 2013.

View Users: View Users
  1. satyr9

    satyr9 Registered User

    Joined:
    Sep 20, 2009
    Messages:
    258
    Likes Received:
    0
    Trophy Points:
    0
    So just farting around with my top 18 guys this morning and figured I'd share. The chart below shows projected minutes (with bias towards balance admittedly) and individual production based on each player's three year averages (what they've done the last three years pro-rated over this season's projected minutes). I'm not saying it's predictive and to the right I have my skimpy notes on why each one is optimistic, pessimistic, stable, or uncertain.

    Code:
    	          Minutes					Production					
    	       ES	PP	PK	MIN		G	A	Pt			
    Lupul	        15	3	0	18		34	43	77		Opt.	Health
    Kessel	16	3	0	19		33	40	73		Pes.	OldDataBias
    Kadri	        15	3	0	18		23	36	59		Uncertain	Sm.Smpl
    JVR	        16	1	1	18		28	25	53		Stable	
    Franson	17	4	1	22		10	38	48		Opt.	1yrSkew
    Kulemin	14	0	2	16		17	26	43		Opt.	OldDataBias
    Phaneuf	18	3	4	25		12	30	42		Stable	
    Bozak	        16	1	1	18		17	24	41		Pes.	OldDataBias
    Clarkson	15	2	1	18		26	14	40		Uncertain	NewTeam
    Boland	14	1	1	16		19	21	40		Uncertain	NewTeam
    Liles	        15	1	0	16		6	26	32		Opt.	OldDataBias
    Raymond	11	1	0	12		13	17	30		Uncertain	NewTeam
    Gardiner	17	1	1	19		6	22	28		Uncertain	Sm.Smpl
    Gunnarsson	18	0	2	20		4	20	24		Stable	
    Ranger	15	1	2	18		6	16	22		Opt.	OldDataBias
    McClement	8	0	4	12		9	9	18		Stable	
    Colborne	5	0	0	5		2	10	12		Uncertain	Sm.Smpl
    Orr	        5	0	0	5		3	2	5		Stable	
    
    So what do you think? Totals are 268 GF, which would be quite high, but it's also assuming perfect health so that's not really as optimistic as it looks. Is an excercise like this worth a damn? Would you weight years to avoid the old data bias (three years ago Liles, Kule, et al. were different players)? I didn't play with the numbers at all and showed why I think they're biased instead of trying to adjust the poorly predictive - IMO - numbers. Gotta say if we got this kind of production out of the bulk of the roster, I'd be pretty happy with how the season was going.
     
  2. DD03

    DD03 3D

    Joined:
    Mar 15, 2010
    Messages:
    21,733
    Likes Received:
    7
    Trophy Points:
    86
    Lupul won't out score Phil. I wish.
     
  3. satyr9

    satyr9 Registered User

    Joined:
    Sep 20, 2009
    Messages:
    258
    Likes Received:
    0
    Trophy Points:
    0
    Sure, I don't think he will either. All that shows is that Lupul was more productive per minute than Kessel for the last three years and projects, when healthy, to play enough minutes to have higher totals. But Phil's 64pt season three years brings down his production whereas Lupul has missed so many games. But I prefer to start with flawed numbers that I haven't tweaked even for obvious reasons. I do think it's interesting that taking all three years together Lupul is more productive than Phil. I also just realized I only included Lupul's Leaf time from that year, not both. His production in ANA would bring him down.

    edit: With ANA's 2010/2011 numbers you can downgrade Lupul to 31/42/73, almost identical to Phil, but still more productive 'cause he's doing it in a minute less per game.
     
  4. zeke

    zeke #freeVladdy

    Joined:
    Mar 14, 2005
    Messages:
    47,600
    Likes Received:
    5,833
    Trophy Points:
    201
    nice work.

    I'm a bit iffy on unweighted 3yr stats, though. They give players like Kuly and Bolland too much credit for his big year 3yrs ago, and they also penalize young players like Kadri and probably even Kessel too harshly, and give older players like Liles too much of a pass on declining stats.

    I wouldn't mind seeing some weighted 3yr numbers, or your per-minute take on 2yr numbers.
     
  5. The Missing Piece

    The Missing Piece What's Left?

    Joined:
    Sep 19, 2012
    Messages:
    1,318
    Likes Received:
    88
    Trophy Points:
    72
    Id be pretty happy with these stats. Id like to see Gardiner get some more pts than that.

    I assume orr/mclaren split that?
     
  6. Anth93

    Anth93 Registered User

    Joined:
    Jul 2, 2012
    Messages:
    3,017
    Likes Received:
    0
    Trophy Points:
    0
    How about instead we look at their numbers over the last two years while considering their zone stars and possession numbers?

    Boland isn't putting up 20+ goals and Kulemin sure has hell isn't scoring 40 points while getting 36% offensive zone faceoffs.
     
  7. zeke

    zeke #freeVladdy

    Joined:
    Mar 14, 2005
    Messages:
    47,600
    Likes Received:
    5,833
    Trophy Points:
    201
    percentage numbers like that get abused a little bit.

    Let's look at Player A.

    In year 1, Player A gets 500 offensive zone faceoffs, and 500 defensive zone faceoffs. 50% offensive zone faceoffs.

    In year 2, Player A gets 500 offensive zone faceoffs, and 250 defensive zone faceoffs. 75% offensive zone faceoffs.

    In year 3, Player A gets 500 offensive zone faceoffs, and 750 defensive zone faceoffs, 40% offensive zone faceoffs.


    In which year does he get the most offensive opportunity?
     
  8. scotto11

    scotto11 Registered User

    Joined:
    Sep 9, 2012
    Messages:
    29
    Likes Received:
    0
    Trophy Points:
    0
    i think you better give those a weighted average, maybe a 60-30-10
     
  9. Anth93

    Anth93 Registered User

    Joined:
    Jul 2, 2012
    Messages:
    3,017
    Likes Received:
    0
    Trophy Points:
    0
    Well of course, but that's if you think player A's zone starts actually increased by roughly 80% from year 2 to 3.

    That just simply doesn't happen.
     
  10. zeke

    zeke #freeVladdy

    Joined:
    Mar 14, 2005
    Messages:
    47,600
    Likes Received:
    5,833
    Trophy Points:
    201
    still, the percentage number is actually not the number we should be looking at.
     
  11. satyr9

    satyr9 Registered User

    Joined:
    Sep 20, 2009
    Messages:
    258
    Likes Received:
    0
    Trophy Points:
    0
    To refine raw data into a prediction or forecast you'll want to find a weighting system, but I can make an argument against options and if I do one thing, I need to account for many others.

    Say I use your 60/30/10 system. Now Lupul's production in 16 games is just one of many things given way too much weight, in general last year's short season gets too much weight that way IMO. But say I re-weight by percentage of total minutes first, then apply your 60/30/10 weights to years, the fact that last year was 48 games instead of 82 will basically give me a 45/45/10 weighting, which if I'd wanted that I could just do, but doesn't seem right either. And that's just the tip of the sword.

    Obviously this isn't a finished product. I was just goofing off at work with a spreadsheet and decided I'd post it for discussion.
     
  12. zeke

    zeke #freeVladdy

    Joined:
    Mar 14, 2005
    Messages:
    47,600
    Likes Received:
    5,833
    Trophy Points:
    201
    I like to play with these kinds of numbers quite a bit, and from a decade of doing it I find that that unweighted 3rd year causes all sort of problems.

    and as long as you weight the total stats first, instead of the per-game or per-minute stats, I don't think it causes any of the problems you think it might.
     

Share This Page

monitoring_string = "358c248ada348a047a4b9bb27a146148"