Ranking of Teams' Draft Success

[Micklebot]

Is there any way you could account for strength of the draft year? For example, Pit having the priviledge of drafting 1st in 2005 and 2nd in 2004 give them a significant advantage over say Edm who drafted first in 2010 and again in 2011 (yes, I know those two years are not included in the data set, just an example).

Would it be possible to somehow evalute how well they picked given what was actually available to them?

That said, I do there is room for improvement in my goalie and defensemen rankings. I REALLY want to us TOI/Game for defensemen, but can't find that in a format that wouldn't require me to manually search for every single defenseman's value.

What exactly do you mean by this?

Hockey reference has TOI and GP
http://www.hockey-reference.com/pla...p=gt&c4val=&threshhold=5&order_by=time_on_ice

You'd have to copy them out 100 players at a time until you got them all (or use a webcrawler or something), and then do a lookup from your draft position database, but that should do it, no?

edit: for what it's worth, when I clicked through, 1997 to present had 998 (10 pages) of Dmen who have played 1 min or more in the NHL. Idk if guys like Burns who has played both F and D was on the list though.

 

[Top 6 Spaling]

Is there any way you could account for strength of the draft year? For example, Pit having the priviledge of drafting 1st in 2005 and 2nd in 2004 give them a significant advantage over say Edm who drafted first in 2010 and again in 2011 (yes, I know those two years are not included in the data set, just an example).

I hadn't thought about that, but it's possible. I did calculate the averages, and the variances in average success wasn't huge (ever draft was between 7 and 9). I'll try to think of a way to incorporate that.

Would it be possible to somehow evalute how well they picked given what was actually available to them?



What exactly do you mean by this?

Hockey reference has TOI and GP
http://www.hockey-reference.com/pla...p=gt&c4val=&threshhold=5&order_by=time_on_ice

You'd have to copy them out 100 players at a time until you got them all (or use a webcrawler or something), and then do a lookup from your draft position database, but that should do it, no?

edit: for what it's worth, when I clicked through, 1997 to present had 998 (10 pages) of Dmen who have played 1 min or more in the NHL. Idk if guys like Burns who has played both F and D was on the list though.

For every stat I used, I could find it listed by draft year, which made copy and pasting super easy.. If I have time, I'll try to do a lookup function and make it work.

 

[Rants Mulliniks]

Upated top and bottom five teams with log regression used instead of mean:

Top 5
Montreal (2.719)
San Jose (2.428)
Toronto (1.940)
Detroit (1.916)
Pittsburgh (1.892)

Bottom 5
Phoenix (-2.872)
Tampa (-2.074)
Florida (-2.026)
Washington (-1.706)
Calgary (-1.361)

Toronto jumped all the way from 8 to 3, which chocked me. Buffalo falls out of the top 5. Washington takes the biggest tumble down to #4, and Tampa moves below Florida.

Despite what people think, Toronto has always been pretty decent at drafting. What they do well is producing serviceable NHLers. They don't produce stars. As some mentioned, they also struggle in trading things away. In over 30 years only 3 first round selections were still with the organization after 25 years of age (most gone well before this). None were there past 28.

I suspect the new drafting regime will change the "star" thing with their focus on skill and stats.

 

[Morbo]

Despite what people think, Toronto has always been pretty decent at drafting. What they do well is producing serviceable NHLers. They don't produce stars. As some mentioned, they also struggle in trading things away. In over 30 years only 3 first round selections were still with the organization after 25 years of age (most gone well before this). None were there past 28.

I suspect the new drafting regime will change the "star" thing with their focus on skill and stats.

Matt Stajan is the poster child of Leaf drafting in recent decades. 700 career games of meh.

 

[Rants Mulliniks]

Matt Stajan is the poster child of Leaf drafting in recent decades. 700 career games of meh.

Sure. When you consider over 50% of NHLers never play 100 games, picking one that plays 700 ain't half bad. In fact it puts him 657th all time. Very good chance he'll end up in the top 200 all time. Picked 57th overall, that's almost a homerun.

 

[r0bert8841]

I was wondering if you could show the median of each draft positions. I'm curious to see how different the value of each draft spot compared to nearby positions, and how the median changed as you move further into the draft.

 

[Top 6 Spaling]

Thanks for all of the input guys! Over the next few days, I'm going to add in TOI/G instead of PPG for defensemen, then using my fancy new logarithmic regression instead of means, I'll make a new post in the general NHL section.

I really do appreciate the recommendations and thanks. Let me know if anyone wants a breakdown of their specific team or anything.

I was wondering if you could show the median of each draft positions. I'm curious to see how different the value of each draft spot compared to nearby positions, and how the median changed as you move further into the draft.

I'll try to do this too! If excel had a "medianif()" function, this would be so much easier.

Let me see if SPSS can do this for me.

 

[r0bert8841]

I was under the impression you changed things to using the median to calculate the expectations out of each draft position. If you are using average instead I would like to see the average for each position. I don't want to make you go too far out of the way.

 

[Doctor No]

I'll try to do this too! If excel had a "medianif()" function, this would be so much easier.

You can do it with array formulas. For instance, put the numbers 1 through 10 in cells A1:A10.

Then put into cell A12:
=MEDIAN(IF($A$1:$A$10<=5,$A$1:$A$10,""))

(when you end the formula, don't hit ENTER, but hit CTRL-SHIFT-ENTER).

If you did it right, it should return 3.

 

[Top 6 Spaling]

I was under the impression you changed things to using the median to calculate the expectations out of each draft position. If you are using average instead I would like to see the average for each position. I don't want to make you go too far out of the way.

I did it with a logrithmic formula. I'd be happy to pull up a graph of that and show you.

You can do it with array formulas. For instance, put the numbers 1 through 10 in cells A1:A10.

Then put into cell A12:
=MEDIAN(IF($A$1:$A$10<=5,$A$1:$A$10,""))

(when you end the formula, don't hit ENTER, but hit CTRL-SHIFT-ENTER).

If you did it right, it should return 3.

Or find someone smart like this who can teach me

 

[Doctor No]

Eh, piece of cake.

Array formulas are great - and you can do a lot with them (for starters, you can replace MEDIAN with just about any other function). They're memory-intensive, so I wouldn't recommend them for huge workbooks.

 

[Micklebot]

Eh, piece of cake.

Array formulas are great - and you can do a lot with them (for starters, you can replace MEDIAN with just about any other function). They're memory-intensive, so I wouldn't recommend them for huge workbooks.

If you do it often, you could also create a user defined function using VBA; not sure but I think it would be less memory intensive.

Also, on the positive side, other people have already created the code so you don't need to work too hard just google it (though i'd be weary of using any code you don't fully understand, there is the potential for something malicious afterall)

Here is an example somebody smarter than me did;
http://www.bettersolutions.com/excel/EIK284/LN622911811.htm

I've played around with user defined functions in the past and they can be really helpful when there isn't a built in function.

 

[Top 6 Spaling]

UPDATES:

• TOI/G is now being used in leiu of PPG for defensemen
• The goalie success ratings have been completely revamped, weighing SV% more than they used to.
• I multiplied all success scores by a constant to make the highest rating equal 100 as a scale. So Crosby is 100 now, and so on.
• Still debating between using the average (I would probably change this to median) and regression line for success difference. My biggest issue with the regression line is that it doesn't emphasize how good the top is (Crosby is listed as the second best value pick of the decade because he outperformed the curve so badly...

Once I do some more work I'm going to make a big post on the main boards (or a blog if I can get someone to publish it...), let me know if there's anything else you want to see.

I'm also working on some way of representing the way average success changes as the draft goes on, as requested by rObert. Trying to find an interesting and applicable way of representing all of that data.

 

[barneyg]

Still debating between using the average (I would probably change this to median) and regression line for success difference. My biggest issue with the regression line is that it doesn't emphasize how good the top is (Crosby is listed as the second best value pick of the decade because he outperformed the curve so badly...

The issue with the median (haven't really thought this through -- this may not be an issue at all) is that beyond round 1 the median score is probably 0 or very close to it, because a lot of people don't make it at all. So except for an adjustment for very high picks all your calculations would revert back to something like "average score". If you have an issue with the regression line for picks 1-10 or something, you can fix that by adding a HIGHPICK dummy variable to the regression. The main reason why I suggested a regression was to avoid the "Rajamaki effect" you pointed out in your OP.

 

[Top 6 Spaling]

The issue with the median (haven't really thought this through -- this may not be an issue at all) is that beyond round 1 the median score is probably 0 or very close to it, because a lot of people don't make it at all. So except for an adjustment for very high picks all your calculations would revert back to something like "average score". If you have an issue with the regression line for picks 1-10 or something, you can fix that by adding a HIGHPICK dummy variable to the regression. The main reason why I suggested a regression was to avoid the "Rajamaki effect" you pointed out in your OP.

I think i'm gonna stick with the means. It isn't perfect by any means, but it's better than just the regression. Crosby being the 2nd best pick of the decade is pretty dumb.

Unless I averaged the mean and the regression to get a combo...Let's see.

Just ran it, and I liked it a lot. It helps mitigate the Rajamaki effect of using means and the Crosby effect of using regression. It's beautiful.

 

[Top 6 Spaling]

http://hfboards.mandatory.com/showthread.php?p=105232717#post105232717

Thanks for all the help guys! The suggestions helped quite a bit. I'm much happier with how the data looks now.

 

[syc]

Atlantic Division draft supremacy. Wow.

 

[Top 6 Spaling]

I was wondering if you could show the median of each draft positions. I'm curious to see how different the value of each draft spot compared to nearby positions, and how the median changed as you move further into the draft.

I just did a post about this. Mean instead of median because the median is 0 starting in, like, the third round...

http://hfboards.mandatory.com/showthread.php?p=105673777#post105673777

Never trade a top three pick. Ever.

 

[Henkka]

Great great work and nice to read these kind of things.

Could you do same thing based somehow on caphit value?

Who will draft most "salary value" in the league relatively against draft position?

Imo, caphits will determine very well player's level and overall value, and it compares forwards, defencemen and goalies quite equally.

So, all in all, caphit would be somehow included in that "success" formula. Or is it already there?

 

[Micklebot]

Great great work and nice to read these kind of things.

Could you do same thing based somehow on caphit value?

Who will draft most "salary value" in the league relatively against draft position?

Imo, caphits will determine very well player's level and overall value, and it compares forwards, defencemen and goalies quite equally.

So, all in all, caphit would be somehow included in that "success" formula. Or is it already there?

Caphit brings different problems. For example, is Keith a less succesful pick than Phaneuf, or on even grounds with Orpik? When a deal is signed can greatly affect the total, not to mention other issues like team discounts. Bad contracts like

 

[r0bert8841]

I just did a post about this. Mean instead of median because the median is 0 starting in, like, the third round...

http://hfboards.mandatory.com/showthread.php?p=105673777#post105673777

Never trade a top three pick. Ever.

Beautiful!!!

 

[Ail]

Really nice work man, enjoyed this.