Byron Bader draft rankings

[FlyguyOX]

I really enjoy his statistical approach to player success probabilities

 

[RefsIdeas]

He did indeed. And plugged it into his model.

He and his model are just so, so laughably bad. I get that he's not trying to provide a scouting report or anything - but time and time again little stuff like this shows that he doesn't know what he's talking about. He will defend his model over blatently obvious errors and omissions that anybody else could see is just him trying to downplay an accusation that his model may not be taking in certain things into account - then follow it up with "its proven to be more accurate than NHL scouts".

I have a very hard time taking much anything produced by him seriously. I'm positive there's other ways to go about an NHLe model more effectively.

 

[BerthMania]

Why are you auditing drafts from one and two years ago? Those results won't be known for a decade.


I asked him, this was his response:




So it seems he still has somewhat of an opinion baked into the production rankings.

Thanks for asking him the question!

I think his answer is pretty underwhelming though. He pretends he has found a very valuable tool to detect futur NHL stars (which he sells the access to for 25$/year), yet when a player ranks very high on said tool, but isn't consensus top pick he doesn't consider him.

Also, I find that weird considering that when we look at his lists from 2020 and 2021, he has lots of players that weren't considered first picks.

Lastly, he dismisses Dumais partly because of his size, yet puts a 5p8 defensemen at #8. Pretty weird in my view.

 

[Peasy]

Just saying, if this guy does heavily use stats to produce his rankings, the last two drafts would probably be the worse to look at in terms of his success rate considering how little some people played, especially last year.

And I would imagine stutzle ranked low due to not many players coming from the DEL so tough to draw conclusions just looking at stats. EP had stutzle at 7, i think Wheeler had him around there too. Same for Sanderson, his offensive numbers over the course of the entire season werent anything crazy, but he really turned it on in the second half.

 

[FlyguyOX]

Did he just add the USHL and USDP numbers from the elite prospects page without realizing USDP is just every game played?

Grimaldi only scored 39 as well

That's real funny if so

Wait, is this how EP works for every prospect? or just USNTDP kids? I had no idea they were the same games. I thought they were separate games, too. WTF why would EP do that??

 

[Static]

Just saying, if this guy does heavily use stats to produce his rankings, the last two drafts would probably be the worse to look at in terms of his success rate considering how little some people played, especially last year.

And I would imagine stutzle ranked low due to not many players coming from the DEL so tough to draw conclusions just looking at stats. EP had stutzle at 7, i think Wheeler had him around there too. Same for Sanderson, his offensive numbers over the course of the entire season werent anything crazy, but he really turned it on in the second half.

Even the consensus rankings are going to look wonky considering how little a lot of the prospects were scouted.

 

[Captain3rdLine]

What am i saying that isn't true? You're arguing his lists are worse and your evidence are lists that haven't even played out. What his models are effective at are stripping away biases that are placed on players. That's all they are, nobody argued otherwise. Just like a common scout's list, things should be taken in conjunction with other data to paint a full picture.

You don’t need wait 10 years to see that having Stutzle at 10 (would probably go #1 in redraft), Lundell at 18, or Schneider not ranked is stupid. All of those guys were consensusly ranked much higher and Bader said oh their numbers aren’t great so I can’t rank them highly. Or guys that aren’t even in the league yet such as Sanderson (14), Guhle (unranked), Quinn (20), and others are already poised to make him look more stupid over the next couple years.

 

[BondraTime]

Wait, is this how EP works for every prospect? or just USNTDP kids? I had no idea they were the same games. I thought they were separate games, too. WTF why would EP do that??

I mean, anyone who follows the players should realize the USDP doesn't play 70+ games a year, especially if you're inputting years of data into models. That's how they input the NTDP kids.

It is done in a way to see how each player does isolated in the USHL, as well as over the full season with NCAA and National games, outside of the Ivan Hlinka and U-18's. It's useful once known.

 

[Static]

You don’t need wait 10 years to see that having Stutzle at 10 (would probably go #1 in redraft), Lundell at 18, or Schneider not ranked is stupid. All of those guys were consensusly ranked much higher and Bader said oh their numbers aren’t great so I can’t rank them highly. Or guys that aren’t even in the league yet such as Sanderson (14), Guhle (unranked), Quinn (20), and others are already poised to make him look more stupid over the next couple years.

And what of the consensus rankings that will also look stupid? All you're doing is cherry picking because you don't agree with how the lists are curated. And if you don't like that approach, that's fine, but there is no perfect way to predict the future.

 

[Captain3rdLine]

Thanks for asking him the question!

I think his answer is pretty underwhelming though. He pretends he has found a very valuable tool to detect futur NHL stars (which he sells the access to for 25$/year), yet when a player ranks very high on said tool, but isn't consensus top pick he doesn't consider him.

Also, I find that weird considering that when we look at his lists from 2020 and 2021, he has lots of players that weren't considered first picks.

Lastly, he dismisses Dumais partly because of his size, yet puts a 5p8 defensemen at #8. Pretty weird in my view.

He has to do that and take consensus into account because it’s partly negates how bad his rankings will look. The more he takes the consensus into account the better his rankings will look.

And he hasn’t found a valuable way to do that. NHLe existed long before him and it doesn’t take his model to see that a player who has very good production might be worth taking earlier than you would take them just purely off their tools.

 

[FlyguyOX]

I mean, anyone who follows the players should realize the USDP doesn't play 70+ games a year, especially if you're inputting years of data into models. That's how they input the NTDP kids.

It is done in a way to see how each player does isolated in the USHL, as well as over the full season with NCAA and National games, outside of the Ivan Hlinka and U-18's. It's useful once known.

Been following prospects for about 6 years now and had no idea. Wow. Ok.

 

[Captain3rdLine]

And what of the consensus rankings that will also look stupid? All you're doing is cherry picking because you don't agree with how the lists are curated. And if you don't like that approach, that's fine, but there is no perfect way to predict the future.

None of the consensus rankings look remotely close to as stupid as his man. That’s the reality. There will be some steals and some busts because as you said you can’t predict the future. They’re trying to project how these kids will turn out. Using production alone will not yield good results. It’s a small piece of evaluating and projecting a prospect.

 

[Brobust]

Even the players don't get this butthurt over rankings.

 

[Static]

None of the consensus rankings look remotely close to as stupid as his man. That’s the reality.

It isn't! Using the last two drafts as any sort of evidence is ridiculous, most players haven't even reached the NHL.

 

[93LEAFS]

No I was responding to someone else’s post and I just think the guy gets way to much credit for what he’s doing and is taken too seriously. His rankings turn out bad pretty quickly.


And then there’s this today. For a guy who bases his rankings on pure production you would think he would at least make sure he puts the right numbers in. I’d guess from this that all of his numbers on USDP players are possibly messed up.


There’s absolutely an acceptable failure rate. But what you’re saying just isn’t true. It is very obvious that his rankings are much worse than anybody else’s. He’s not using any special stats or advanced analytics. He’s quite literally just looking at production man. Anyone can do that. Any half decent scout who looks at the numbers but also spends hours and hours watching and evaluating the prospects will come up with better rankings.

This is a joke, and anyone familiar on any level with the USNTDP needs to know that not every USNTDP game is equal. How you value a game vs Michigan in an exhibition is much different than a game vs a bottom tier opponent at the U-18's or a USHL game. I've seen basic versions of this model before, and a flaw that big is big enough to throw out anyone's rankings before you even get into double counting. Also, when 17 year old players have a season where they play 25% or so of their games internationally, those have to be figured out how to be weighted. I have no issue with people trying to co-relate production to NHL futures, but those list of flaws are acceptable from someone who is doing it as a hobby, not someone charging money for it.

 

[93LEAFS]

What am i saying that isn't true? You're arguing his lists are worse and your evidence are lists that haven't even played out. What his models are effective at are stripping away biases that are placed on players. That's all they are, nobody argued otherwise. Just like a common scout's list, things should be taken in conjunction with other data to paint a full picture.

This issue is, he double counted the USNTDP games, which throws all the measurements out the window. The USNTDP would be far the hardest to create a stat based counting measurements due to the uniqueness of their schedule. You would have to figure out adequate NHLe for 4 nation games, 5 nation games, U-18's and exhibition games vs NCAA opponents (and figure them out relative to conference you are playing). It's a clear flaw.

 

[BondraTime]

This is a joke, and anyone familiar on any level with the USNTDP needs to know that not every USHL game is equal. How you value a game vs Michigan in an exhibition is much different than a game vs a bottom tier opponent at the U-18's or a USHL game. I've seen basic versions of this model before, and a flaw that big is big enough to throw out anyone's rankings before you even get into double counting. Also, when 17 year old players have a season where they play 25% or so of their games internationally, those have to be figured out how to be weighted. I have no issue with people trying to co-relate production to NHL futures, but those list of flaws are acceptable from someone who is doing it as a hobby, not someone charging money for it.

When you have games against teams from the NAHL last year, USHL, NCAA Div3, NCAA Div 1 and international listed in the same USDP season, it's going to be completely out of whack. And adding double games and production from the USHL as well, woof.

 

[93LEAFS]

When you have games against teams from the NAHL, USHL, NCAA Div3, NCAA Div 1 and international listed in the same USDP season, it's going to be completely out of whack. And adding double games and production from the USHL as well, woof.

yeah, basically I don't mind complex attempts to count stats. But, I've done this myself with CHL players previously out of interest. Also, using raw NHLe is pointless because you need to figure out the conversion number for players who are 16-20 in grown men leagues, which is going to be much higher than SHL, KHLers in their mid 20s or older shifting between leagues. On an analytical basis their are very significant flaws to his methodology, and to charge money for it is comical.

 

[Static]

This issue is, he double counted the USNTDP games, which throws all the measurements out the window. The USNTDP would be far the hardest to create a stat based counting measurements due to the uniqueness of their schedule. You would have to figure out adequate NHLe for 4 nation games, 5 nation games, U-18's and exhibition games vs NCAA opponents (and figure them out relative to conference you are playing). It's a clear flaw.

Obviously that's a big problem, I'll ask him and see if he responds.

 

[Forge]

Why are you auditing drafts from one and two years ago? Those results won't be known for a decade.


I asked him, this was his response:




So it seems he still has somewhat of an opinion baked into the production rankings.

I honestly feel like this undermines his model

 

[FlyguyOX]

I honestly feel like this undermines his model

I've liked his stuff but this really does.

 

[Static]

I honestly feel like this undermines his model

It is a weird way of hedging that I hadn't seen before.

 

[93LEAFS]

Obviously that's a big problem, I'll ask him and see if he responds.

Yeah, any statistical model that tries to count all USNTDP games as equal is broken the second its released. He's basically saying I count U-18 games, but I can't count international games for someone like Slafkovsky because I can't adequately translate it to NHLe. This is a method that was legit develop in a slightly better fashion in Statshot which was released in 2016 for only CHL players. For a service that is trying to charge money for access, they are so far behind on purely statistical analysis they think they are leading the race.

 

[Static]

Obviously that's a big problem, I'll ask him and see if he responds.

 

[Poppy Whoa Sonnet]

Embarrassing but credit to the guy for not BSing here. I'd wonder how it changes his previous year rankings.