NHL Draft Software: Final Rankings | Top200

[emptyNedder]

Really impressed--and as I have said I think your work will eventually be much better than the others who rank based on their personal viewings. Incorporating data is a great way to improve any decision by checking our biases.

The one thing that did jump out at me was how much higher Jenik is than Lauko. Their production is pretty similar: Jenik better in Hlinka, both had 6 points at U18 but Lauko had 3 goals and Jenik 0. Jenik is bigger but from everything I read Lauko is much better skater. Lauko played one league higher this past season and made WC 20 team while Jenik did not. So my question: Does Jenik rate significantly higher due to size and being 6 months younger?

Again thanks for the great work--it definitely influences how I view prospects.

 

[ProspectsFanatic]

Really impressed--and as I have said I think your work will eventually be much better than the others who rank based on their personal viewings. Incorporating data is a great way to improve any decision by checking our biases.

The one thing that did jump out at me was how much higher Jenik is than Lauko. Their production is pretty similar: Jenik better in Hlinka, both had 6 points at U18 but Lauko had 3 goals and Jenik 0. Jenik is bigger but from everything I read Lauko is much better skater. Lauko played one league higher this past season and made WC 20 team while Jenik did not. So my question: Does Jenik rate significantly higher due to size and being 6 months younger?

Again thanks for the great work--it definitely influences how I view prospects.

Yes, that is exactly it. Since they have 6 months difference comparing both of their statistics of this season is as valid as comparing Jenik this year stats with Lauko last year stats. If instead of comparing each season, you sort of average out what Lauko accomplished two seasons at the time you will get the idea why there is such clear separation between the two. You could also compare Jenik numbers with Chytil who is born in the same month and is panning really well as a 1st rounder, number wise Jenik compares extremely well.

 

[StatisticsAddict99]

Quick question, I’ve been following pretty much all your accurate data influenced rankings for over a year, I obviously like them as they follow many of my opinions from the view of these prospects, my analysis(it did the same last season as well) and they’re accurate in any point of the future so that said at any point in time(of the future) would you consider doing affiliated prospect evaluations?

 

[ProspectsFanatic]

Quick question, I’ve been following pretty much all your accurate data influenced rankings for over a year, I obviously like them as they follow many of my opinions from the view of these prospects, my analysis(it did the same last season as well) and they’re accurate in any point of the future so that said at any point in time(of the future) would you consider doing affiliated prospect evaluations?

Yes. In the future, I would like to build a website covering draft/prospects centered around the software.

 

[37 others]

If you have the data, I'd be interested in trying to create a machine learning model and see how that performs

It's a good idea, but you'll need lots of training data meaning you need unbiased grades for players' technique, skating skill, etc. that are qualitative data, not quantitative (unless the sheets can account for a lack of qualitative data). It's easy to get an answer to "how's Dahlin's skating right now", but it's much, much harder to get an answer to "how was Jiri Tylusty's skating before getting drafted".

 

[aigledefeu]

amazing work! I am curious, do you have some numbers for upcoming drafts? (2019 and 2020), I am curious how good did Lafrenière, hughes and few others did.

 

[mike14]

Really interesting project, major props for getting it all setup.
Have you run the data without inputting the more qualitative data (info from scouting reports etc), and if so how does it stack up as a list against your final one. Would be interesting to see what just 'stats' spits out

Also, how hard would it to be to take draft rankings (say pick 3 draft guides/services) and use their lists to support the model? Assign each ranking spot a value and use the three rankings (assuming each player was ranked by all 3 services) as part of the algorithm. I'm really interested in whether there would be any value (mainly money and time wise) in teams 'outsourcing' their drafting.

 

[atrud66]

It's a good idea, but you'll need lots of training data meaning you need unbiased grades for players' technique, skating skill, etc. that are qualitative data, not quantitative (unless the sheets can account for a lack of qualitative data). It's easy to get an answer to "how's Dahlin's skating right now", but it's much, much harder to get an answer to "how was Jiri Tylusty's skating before getting drafted".

I'd be relying upon ProspectFanatic's grades for each player's individual attributes, which I agree is subjective but only to an extent. Even if we were to use a grading system, let's say from 1-5, for each category, these categorical grades can be converted to one-hot encodings which can be used in machine learning models

 

[MasksAreForSlaves]

great work, I assume you've updated your rankings on draftwired?
http://draftwired.com:8888/apps/dwNHL

 

[37 others]

I'd be relying upon ProspectFanatic's grades for each player's individual attributes, which I agree is subjective but only to an extent. Even if we were to use a grading system, let's say from 1-5, for each category, these categorical grades can be converted to one-hot encodings which can be used in machine learning models

It's not the actual objectivity that's the problem, but it's the objectivity looking at older prospects (since you'd need their grades at the time of their draft). You'd need to go back many many years to get good training data, but the farther you go back, the harder it is to get the qualitative grades of prospects. How many years back does ProspectFanatic's grades go?

 

[Fogelhund]

Very interesting work, and thanks for sharing it. It will be interesting to see how the math develops over time for you.

 

[ProspectsFanatic]

It's not the actual objectivity that's the problem, but it's the objectivity looking at older prospects (since you'd need their grades at the time of their draft). You'd need to go back many many years to get good training data, but the farther you go back, the harder it is to get the qualitative grades of prospects. How many years back does ProspectFanatic's grades go?

Yeah I only have two years, we can't do it for now with the data I have, maybe in the future we could try. Not sure if the sample size will ever be big enough though.

 

[ProspectsFanatic]

great work, I assume you've updated your rankings on draftwired?
http://draftwired.com:8888/apps/dwNHL

It isn't me that posted my list, it is cool that you are using it though.

 

[MasksAreForSlaves]

It isn't me that posted my list, it is cool that you are using it though.

Oh, that's fine. Obviously your rankings are popular.

 

[ProspectsFanatic]

Really interesting project, major props for getting it all setup.
Have you run the data without inputting the more qualitative data (info from scouting reports etc), and if so how does it stack up as a list against your final one. Would be interesting to see what just 'stats' spits out

Also, how hard would it to be to take draft rankings (say pick 3 draft guides/services) and use their lists to support the model? Assign each ranking spot a value and use the three rankings (assuming each player was ranked by all 3 services) as part of the algorithm. I'm really interested in whether there would be any value (mainly money and time wise) in teams 'outsourcing' their drafting.

Yes, I could eliminate scouting evaluation, but it isn't really relevant since the algorithm relies on the scouting evaluations sort as an expected value to operate optimally, so the numbers you get won't be very representative. I do test it sometimes just to see, the list is very similar. Some players move a few ranks, usually, fast skaters drop because this is by far the attribute which I value the most, since fast skaters project the best even though they don't necessarily have the greatest statistics (like Formenton). The top-end in terms of forwards would stay the same. For defensemen Bouchard jumps up a bit, being not as good as a skater as the other defenseman. Bouchard had the best stats no doubt (excluding Dahlin), but I am not sure his game will translate at the NHL level as well as a player like Dobson.

Other rankings would be very easy to factor in, I used to incorporate the consensus rankings which I would convert into values of similar values to my overall scouting evaluations, and basically use those has an additional source/2nd opinion for my scouting evaluations. The thing is I don't have the time to produce consensus rankings, and it needs to be updated frequently to stay the most relevant, other HF members used to produce those lists, but it isn't the case anymore.

Outsourcing their drafting; they might as well hire me instead.

 

[mike14]

Outsourcing their drafting; they might as well hire me instead.

I guess that's my question. Would a team that's looking to cut costs (Ottawa, Arizona) actually be better off scrapping their scouting department and paying for an algorithm that they own and can continually update, as well as a few 'trusted' draft guides where they input the values just prior to the draft? The list that the program spits out is the list you go to the draft with and you just draft the next player available on your list.

Would love to see how a model like that stacks up against a full scouting department.

 

[ProspectsFanatic]

I guess that's my question. Would a team that's looking to cut costs (Ottawa, Arizona) actually be better off scrapping their scouting department and paying for an algorithm that they own and can continually update, as well as a few 'trusted' draft guides where they input the values just prior to the draft? The list that the program spits out is the list you go to the draft with and you just draft the next player available on your list.

Would love to see how a model like that stacks up against a full scouting department.

I see your reasoning, but on the flip side the entry draft is such a determining factor in a franchise success (which will to financial gains), it wouldn't be a department I would be inclined to eliminate investment in even on a very tight budget. But I mean yeah probably that the consensus rankings would have performed better than some teams in recent years if we were to do the comparison exercise. Either way, I guess it is human nature to believe that we can do better than our competitors and should rely on yourself or the people we trust to make important decisions, so far that reason I simply don't see it happening.

 

[mike14]

I see your reasoning, but on the flip side the entry draft is such a determining factor in a franchise success (which will to financial gains), it wouldn't be a department I would be inclined to eliminate investment in even on a very tight budget. But I mean yeah probably that the consensus rankings would have performed better than some teams in recent years if we were to do the comparison exercise. Either way, I guess it is human nature to believe that we can do better than our competitors and should rely on yourself or the people we trust to make important decisions, so far that reason I simply don't see it happening.

I can't see it happening either, but I'd love to see a side by side comparison over 5+ year drafting cycle!

 

[ProspectsFanatic]

amazing work! I am curious, do you have some numbers for upcoming drafts? (2019 and 2020), I am curious how good did Lafrenière, hughes and few others did.

Sorry just realized I never answered your question. Haven't had the time to enter players for the following drafts, but I am pretty sure both of them are projecting better than Dahlin/Svechnikov at the same age, both of them are going to be exceptional 1st overall picks if things continue progressing as they currently are.

 

[ProspectsFanatic]

Just listened to Bob Mckenzie's recent podcast. My list seems to be lining up well with Bob Mckenzie projections based on the infos he obtained from scouts.

• Dahlin clear cut #1, now Svech is clear cut #2 as well.
• Zadina and Tkachuk aren't clear cut #3 & #4 respectively anymore, it is now wide open with the other high-end picks for #3.
• Kotkaniemi is trending upwards, he is now ranked top5 pick for several teams. Bob said that it wouldn't be a stretch at all to see Montreal take him #3.
• Dobson is now the consensus best defensemen after Dahlin with Boqvist rounding the group behind Hughes/Bouchard.
• Most teams are ranking Merkley towards the tail end of the first round.
• Kravtsov is finally getting some recognition with at least one team even having him in the top5.

About Kavtsov, since others used to have Kravtsov in the lower parts of the first round it always affected negatively my perception of him (had a limited viewing of him so I factored other opinions), so I may reconsider how I evaluated him. That being said, I have always been high on him (had him higher than any other list I know before he exploded in the playoff, and he was in my top10 in my previous list).

So like I noted at the end of my list, you can expect some last minute tweaks to my list.

 

[Kevin Musto]

Just listened to Bob Mckenzie's recent podcast. My list seems to be lining up well with Bob Mckenzie projections based on the infos he obtained from scouts.

• Dahlin clear cut #1, now Svech is clear cut #2 as well.
• Zadina and Tkachuk aren't clear cut #3 & #4 respectively anymore, it is now wide open with the other high-end picks for #3.
• Kotkaniemi is trending upwards, he is now ranked top5 pick for several teams. Bob said that it wouldn't be a stretch at all to see Montreal take him #3.
• Dobson is now the consensus best defensemen after Dahlin with Boqvist rounding the group behind Hughes/Bouchard.
• Most teams are ranking Merkley towards the tail end of the first round.
• Kravtsov is finally getting some recognition with at least one team even having him in the top5.

But since others used to have Kravtsov in the lower parts of the first round it always affected negatively my perception of him (had a limited viewing of him so I factored other opinions), so I may reconsider how I evaluated him. That being said, I have always been high on him (had him higher than any other list I know before he exploded in the playoff, and he was in my top10 in my previous list).

So like I noted at the end of my list, you can expect some last minute tweaks to my list.

Yeah I was kind of shocked that you dropped Kravtsov. I always really liked how high you had him earlier.

 

[ProspectsFanatic]

Yeah I was kind of shocked that you dropped Kravtsov. I always really liked how high you had him earlier.

Yeah since I saw HockeyProspect having Kravtsov at #6, I am like why did I let other rankings that have him in the 20s(like FC at #29 or even ISS out of the 1st round) influence my opinion of him. He was one of the rare player I thought the software read must be flawed so I gave Kravtsov a notable negative bonus with reasoning "Lack of international exposure for comparison/Russian factor" which is just dumb in retrospect, he is flat-out good.

 

[RABBIT]

bravo, well done. thank you

 

[HisNoodliness]

Pretty awesome work. I'm excited to watch the draft and think your rankings are probably the most accurate ranking any of us have available.

 

[Peeps]

@ProspectsFanatic are you doing it this year?