The Data-Based Drafting Thread (what players would a Potato pick?)

[4Twenty]

Haha,its funny you mention that. I am in the final interview stage with a major team right now (not hockey.) Supposed to find out this week if I got the job, in which case I'll be moving to the US.

Good luck, hopefully you still grace us with your presence when you're a big wig.

 

[vancityluongo]

Haha,its funny you mention that. I am in the final interview stage with a major team right now (not hockey.) Supposed to find out this week if I got the job, in which case I'll be moving to the US.

Well deserved man, that's terrific.

 

[Dissonance Jr]

It does.

He would be ahead of Hughes if he were 18. The potato *loves* teenagers who produce in the Euro pro leagues. Is it over-compensating? Possibly.

It’s probably been discussed before, but I do wonder if NHL GMs are unduly biased against overagers. Like if Fagemo had been a 1st round pick in 2018 and then put up the season he’s had this year, most people would say he’s tracking extremely well. So you’d think that’d be an argument for picking him high this time around, rather than letting him slide. I guess we’ll see what happens.

 

[Melvin]

It’s probably been discussed before, but I do wonder if NHL GMs are unduly biased against overagers. Like if Fagemo had been a 1st round pick in 2018 and then put up the season he’s had this year, most people would say he’s tracking extremely well. So you’d think that’d be an argument for picking him high this time around, rather than letting him slide. I guess we’ll see what happens.

It's definitely an interesting case. His numbers in junior were pretty mediocre, worse than a lot of players who didn't get drafted (like Carl Wassenius,) but he moved up to the SHL and had a really good season. That doesn't happen very often. You wonder if he had a growth spurt or something like this.

It is probably the biggest reach that the potato has made yet. At least since it took Esa Lindell for the canucks 1st round pick in 2012.

 

[Melvin]

(without giving away the recipe) Are your outputs limited to NHL production? Linearly? Finland is in the midst of a golden generation of high-end talent, which could cause the league to appear stronger.

(edit:to be clear this is me just talking entirely outta my ass, and assuming when you talk about a league's strength you're referring to how its alumni do in the NHL)

Yes, you are right. And it's possible that this is the case.

One thing I need to consider doing is scrapping some of the older data. I have draft data going back to 2004. I like having as much data as possible but it's fair to argue that 2004 data may not be so relevant in calculating league ratings today. I have fiddled with trying different time periods and it doesn't make as much difference as you might think. I don't believe in calculating a different factor for every year (this is one of the problems I have with NHLe, ) but i have been trying to find some kind of middle ground.

I have a few ideas about this which I might look into prior to the actual draft, if I have time and am not too busy with moving, heh.

 

[vanuck]

Haha,its funny you mention that. I am in the final interview stage with a major team right now (not hockey.) Supposed to find out this week if I got the job, in which case I'll be moving to the US.

Wow, best of luck - you do great work. Don't forget your roots though!

 

[Hockeyphysio]

Thanks for this.

Curious where someone like Newhook sits in this model. Does the model have some limitations in regards to some of the Junior B leagues?

 

[Melvin]

Thanks for this.

Curious where someone like Newhook sits in this model. Does the model have some limitations in regards to some of the Junior B leagues?

Yeah these leagues have a very low factor. It's hard for any players from bchl/ajhl to rank very highly.

 

[Phenomenon13]

Are there any plans to incorporate goalies?

 

[Bleach Clean]

87th
He is 16th, literally just missed my arbitrary top-15 cut-off. He is tied with Raphaël Lavoie.

Why does Soderstrom rank so lowly when he's played a full season in the perhaps the toughest non-NHL league? In other words, what would he have been expected to produce to be ranked in the top15 this year?

Height factors are hard to figure, especially since a player's height can change and you have different sources, but historically there has been a significant enough difference between forwards listed at 5'9"* for the draft and those listed at 5'7" that the factor is different, and yes, Pelletier is not hit as hard as Caufield. The Q also still rates higher in my numbers than the USDP, despite people's annual insistence that the US program is better.

*Note: Important to distinguish "listed at" vs. whatever their real height is.

I tend to take the most corroborated estimate for height. I remember this came up during our Hughes discussion.

Does the Q rate higher than the USHL?

Because he didn't out-produce him? 79/56 > 78/60

Yeah, but it's pretty arbitrary. I use 30 games as a threshold, so if he were under 30 games he would rank lower. Because he managed 32 it doesn't affect him much. A lot of Euro players and US players only play 30-50 games so you can't ding players too much when they are in this range, although Turcotte is in the extreme low end of this range. Just one of those things.

Pavel Dorofeyev is an example of a player who would probably be in the top-15 if he had maintained his production for 30 games instead of 19.

Are you adding the USHL and USDP numbers for Boldy/Hughes? 57 + 27 is still 84 games for Zegras. Boldy has 89 games played.

Understood about the 30-50 game markers.

 

[Melvin]

Are there any plans to incorporate goalies?

I can't rule it out. But there are probably five or so items that rank higher on my to do list. I just don't care about goalies all that much.

 

[Melvin]

Why does Soderstrom rank so lowly when he's played a full season in the perhaps the toughest non-NHL league? In other words, what would he have been expected to produce to be ranked in the top15 this year?

Production is the backbone of the model. Everything else is just adjustments. When you start with just 7 points, it takes a lot of adjustments to move you up. I'm guessing he'd need about double that to be ranked highly.

Does the Q rate higher than the USHL?

Yes.

Are you adding the USHL and USDP numbers for Boldy/Hughes? 57 + 27 is still 84 games for Zegras. Boldy has 89 games played.

That's not how it works. The USHL numbers are a subset of the USDP numbers. They have not played that many games.

 

[Bleach Clean]

Production is the backbone of the model. Everything else is just adjustments. When you start with just 7 points, it takes a lot of adjustments to move you up. I'm guessing he'd need about double that to be ranked highly.



Yes.



That's not how it works. The USHL numbers are a subset of the USDP numbers. They have not played that many games.

Would it then be 57 and 61 games respectively?

14 points in 44 games is a tall task for a draft season defender in the SHL. Do you have Comparables that would suggest that this should be the conversion baseline?

 

[Melvin]

Would it then be 57 and 61 games respectively?

14 points in 44 games is a tall task for a draft season defender in the SHL. Do you have Comparables that would suggest that this should be the conversion baseline?

Yes. Matthew Boldy has played in 61 games. Of these, 28 were against USHL teams and the rest were against various competition. You can see this pretty clearly from his game log.

This is part of what makes evaluating US junior players a mess is that they face all kinds of random levels of competition. Like Jack Hughes scoring 12 points in 3 games against, like, Team Switzerland.

Re: Victor Söderström. His production is similar to that of Tim Erixon, Filip Westerlund, Erik Brannstrom. Giving him 12-14 points puts him more in line with Adam Larsson, David Rundblad. Still well behind Hedman or Dahlin, of course.

 

[Melvin]

Thanks for your work Melvin. How does the absolute ranking/score of the USHL players from this year compare to the score it gave for Boeser in the 2015 draft?

The ranking would be (USHL, non-USDP) :

1. Bobby Brink
2. Ronnie Attard
[Brock Boeser 2015]
3. Matias Maccelli
4. Robert Mastrosimone
5. Yegor Afanasyev
6. Shane Pinto

Also, someone asked about Newhook, he is ranked around 30.

 

[Tables of Stats]

I'd be curious to know if any of those projects that rank above analyzing draft eligible goaltenders are hockey related. You do exceptional work so I'm always hoping for more of it.

 

[Dudekun]

Tentative top-15 for 2019, should be pretty close to final, unless I find some kind of bug.

1. Kaapo Kakko, F, Liiga, 38 P in 45 GP.
2. Jack Hughes, F, USDP, 45 in 100.
3. Samuel Fagemo, F, SHL, 25/42.
4. Alex Turcotte, F, USDP, 59/32.
5. Arthur Kaliyev, F, OHL, 102/67.
6. Bowen Byram, D, WHL, 71/67
7. Ville Heinola, D, Liiga, 14/34
8. Jakob Pelletier, F, QMJHL, 89/65
9. Nathan Légaré, F, QMJHL, 87/68
10. Trevor Zegras, F, USDP, 78/55
11. Bobby Brink, F, USHL, 68/43
12. Dylan Cozens, F, WHL, 84/68
13. Brayden Tracey, F WHL, 81/66
14. Ryan Suzuki, F, OHL, 75/65
15. Thomas Harley, D, OHL, 58/68

Would imagine that Fagemo will be the potato pick for Vancouver; I think he's projected to go in the 3rd or something.

Love your project. If you have the time, could you see where the following overagers rank in your model?

Marc del Gaizo
Jere Innala
Samuel Bucek

I assume height will probably really hurt the first two, but these players seem intriguing as a stat watcher.

 

[Melvin]

Love your project. If you have the time, could you see where the following overagers rank in your model?

Marc del Gaizo
Jere Innala
Samuel Bucek

I assume height will probably really hurt the first two, but these players seem intriguing as a stat watcher.

Del Gaizo was 200 or so last year, but 389 this year. The height hurts him as well as just being a second-time-eligible NCAA player which in general is just very, very rarely a player worth drafting.
Innala I have 25th. Being a 21 y/o in Liiga doesn't hurt as much. He rates somewhat similarly to someone like Dominik Simon in 2015.

I actually don't have Bucek in my database this year. He was ranked around 100 in 2017 and I just didn't add his 2018 or 2019 seasons. I have about 500 players or so but I will end up missing some (there were probably 20 or so players drafted in 2018 who I didn't have in my database prior to the draft. Including Canucks pick Artyom Manukyan.)

 

[Dudekun]

Del Gaizo was 200 or so last year, but 389 this year. The height hurts him as well as just being a second-time-eligible NCAA player which in general is just very, very rarely a player worth drafting.
Innala I have 25th. Being a 21 y/o in Liiga doesn't hurt as much. He rates somewhat similarly to someone like Dominik Simon in 2015.

I actually don't have Bucek in my database this year. He was ranked around 100 in 2017 and I just didn't add his 2018 or 2019 seasons. I have about 500 players or so but I will end up missing some (there were probably 20 or so players drafted in 2018 who I didn't have in my database prior to the draft. Including Canucks pick Artyom Manukyan.)

Thanks, this is awesome stuff! Hoping Canucks are able to get Fagemo and Innala on day 2 of the draft.

 

[xtra]

What are you using to track this? As in how have you created this database(accumulated the data -manually?) and what program do you use to run the numbers?

 

[Melvin]

What are you using to track this? As in how have you created this database(accumulated the data -manually?) and what program do you use to run the numbers?

Python mostly. I have entered a lot of data manually (@vancityluongo helped and saw how painful it was hahaha) but I've since been able to automate the majority of it.

 

[xtra]

I assume that means you have written scripts to read the data you pull from the different websites?

Where do you get the data from? Just the generic websites for the leagues?

 

[vanuck]

I'm curious to know if this is based on a modified version of NHLe with adjustments for age, league and size etc.?

 

[Canadian Canuck]

Incredible work Melvin keep it up!

 

[Melvin]

I'm curious to know if this is based on a modified version of NHLe with adjustments for age, league and size etc.?

No, I don't use nhle but I calculate my own factors using a somewhat similar methodology that has a different goal. My factors are based specifically around the likelihood that a player producing x in league y will become an NHL player. I am not trying to calculate an equivalency. I also don't agree with calculating a sepearte one for each season as I believe the samples are too small, so I smooth them out.

And yes, same for age, height, etc. It's a very basic multiplicative calculation (which was the intention, hence it being a potato.) it's the baseline thst more complicated models should endeavour to beat.