Prospect Info: The 2024-2025 Prospect Thread: Part 1: Skate or Die!

[Tables of Stats]

Models are useful for fans who don't really follow prospects and just want a rough idea of who are producing well in their draft year.

IMO best scouting is done, when you prioritize on hockey sense, and then the rest.

Even the people who put a whole draft list probably wouldn't have been able to watch every single prospect play.

The most informative way would be to watch the prospect yourself and iso watch them shift to shit and see how they process the game.

I'd agree with that. Models are pretty good at pointing out guys that are doing something flashy, but in that respect, they're not much better than an article by THN or Bob's yearly list which exist to fill roughly that same role. They're both prescriptive rather than predictive.

 

[Tables of Stats]

To end this whole model versus traditional scouting thing, I wonder which did a better job. Hockey's Future scouting reports from back before this place was just a forum or statistical models of a similar vintage?

 

[ManVanFan]

You're cherry-picking examples. I could point out times when the models have been wrong and a majority of scouts have been right and you'd dismiss that as "I already said that models aren't 100% perfect" but more importantly looking at any single data point is worthless for proving the efficacy of a model. The issue at hand is that you haven't shown that the models you like are good at what you claim them to be good at. You've cherry-picked a few examples, but that proves nothing, the proof of a model is showing that it's always above average, not that it gets the occasional pick correct. This is what you've failed to do.

I never said the model was even good. At no point. I posted someone's model about a Canucks prospect. I said I look at analytics as an easy way to look at progression on occasion. Then you few argued the validity of it. You asked for examples, I provided. If you don't like it, it's probably because you don't agree with it, not because it's wrong.

If you would like to keep tabs on who's correct and who's not. I've done this before with multiple others on another website. I'll tell you right now, those people did not do very well.

 

[Tables of Stats]

I never said the model was even good. At no point. I posted someone's model about a Canucks prospect. I said I look at analytics as an easy way to look at progression on occasion. Then you few argued the validity of it. You asked for examples, I provided. If you don't like it, it's probably because you don't agree with it, not because it's wrong.

If you would like to keep tabs on who's correct and who's not. I've done this before with multiple others on another website. I'll tell you right now, those people did not do very well.

Generally speaking, posting something and then arguing about how that something is useful is an endorsement of that thing. Thus in a very real way, you have said that you think the models you're posting about are good.

As for tracking who's right and who's not, that won't work with me. I don't have the desire to track an entire draft class. I look at prospects we're likely to take in the top few rounds, make a call, and leave it at that. If I'm wrong, I own it.

I was famously bullish on Gaudette for a hot minute a few years back and am more than willing to admit that was a poor take.

 

[ManVanFan]

To end this whole model versus traditional scouting thing, I wonder which did a better job. Hockey's Future scouting reports from back before this place was just a forum or statistical models of a similar vintage?

It was never a model versus traditional scouting thing except by the ones who don't believe in models. Lol.

 

[ManVanFan]

Generally speaking, posting something and then arguing about how that something is useful is an endorsement of that thing. Thus in a very real way, you have said that you think the models you're posting about are good.

As for tracking who's right and who's not, that won't work with me. I don't have the desire to track an entire draft class. I look at prospects we're likely to take in the top few rounds, make a call, and leave it at that. If I'm wrong, I own it.

I was famously bullish on Gaudette for a hot minute a few years back and am more than willing to admit that was a poor take.

I could have just posted that "this is how this model views this player".

That's fine, I've tracked lots of things before even 3rd, 4th round picks that people like. I just write them down. I filled a 100 page book over a 5 year period. Man, looking back years later was great. Meant to be fun and laugh but man the takes were hilarious and people argued how great they were at scouting even after all the terribly wrong takes and arguments. It was fun. One person could go back and laugh at another and ask "am I right about this player and this person wasn't" It was good.

I don't know what you mean by bullish. He had almost 100 NHL points now. 5th round pick that is great.

 

[Tables of Stats]

I don't know what you mean by bullish. He had almost 100 NHL points now. 5th round pick that is great.

I was putting a lot of effort into arguing that he was better than Greenway, Foegle, and Kunin at a time when all of them were young players still establishing themselves in the NHL. I didn't really rate him before his stint with the Canucks, so I can't claim a hit for liking him at the draft or as an AHL player. At this stage, even if he does prove to be better than those players when all is said and done it's clear that he's the kind of tweener that can score on bad teams that really doesn't hold much value and wasn't worth the time spent arguing about him.

 

[ManVanFan]

I was putting a lot of effort into arguing that he was better than Greenway, Foegle, and Kunin at a time when all of them were young players still establishing themselves in the NHL. I didn't really rate him before his stint with the Canucks, so I can't claim a hit for liking him at the draft or as an AHL player. At this stage, even if he does prove to be better than those players when all is said and done it's clear that he's the kind of tweener that can score on bad teams that really doesn't hold much value and wasn't worth the time spent arguing about him.

I did go look at his player values on the model compared to those others. LOL.

It's pretty hard to remember years down the road. That was the point of the book. Was for personal use but turned into everybody use.

Adam Gaudette was when I started to watch a lot of college hockey. He was fun to watch, he tried so hard every game. The guy had a will to win.

 

[MS]

Gaudette is a prime example of both nature vs. nuture and skills vs. person when looking at a prospect.

If Adam Gaudette had gone anywhere else but into the darkest Jim Benning-era Canucks with some of the worst prospect development in NHL history, there are probably a range of much more positive outcomes for him.

Similarly, if another player had Adam Gaudette's exact skillset but was less of a just generally weird dude, there are probably a range of better outcomes for that player.

Nothing in development is ever carved in stone and there are a huge range of outcomes that can come from any given point based on luck and opportunity and factors outside of just 'how good is this prospect at hockey at this specific time?'

 

[ManVanFan]

Gaudette is a prime example of both nature vs. nuture and skills vs. person when looking at a prospect.

If Adam Gaudette had gone anywhere else but into the darkest Jim Benning-era Canucks with some of the worst prospect development in NHL history, there are probably a range of much more positive outcomes for him.

Similarly, if another player had Adam Gaudette's exact skillset but was less of a just generally weird dude, there are probably a range of better outcomes for that player.

Nothing in development is ever carved in stone and there are a huge range of outcomes that can come from any given point based on luck and opportunity and factors outside of just 'how good is this prospect at hockey at this specific time?'

I'm reading this and I'm like damn, only if someone created a model of a range of outcomes, that could be continuously updated when factors change. If someone was to create something like that to give possibilities.

 

[Tables of Stats]

I'm reading this and I'm like damn, only if someone created a model of a range of outcomes, that could be continuously updated when factors change. If someone was to create something like that to give possibilities.

If somebody did, I'd hope that a database of their model's rankings would be available and that their methodology would be publicly available so people can understand the weights behind certain values and sanity-check things for themselves. I think one of the biggest issues with advanced stats and analytics in hockey is that everybody keeps things secret hoping their specific flavour of math will get picked up by a team somewhere for big money.

 

[MS]

I'm reading this and I'm like damn, only if someone created a model of a range of outcomes, that could be continuously updated when factors change. If someone was to create something like that to give possibilities.

It's almost like it's something you can't quantify in any meaningful way and that you can't put 'weird dude factor' or 'Jim Benning factor' or 'playing the wrong position factor' into some sort of mathematical formula. Even putting his play off the puck at lower levels into a model is something that none of the fanboy models attempt to accomplish.

Again : that there's a range of outcomes is common sense. That you can attempt to create a model for that range of outcomes is a notion worth looking at. Actually succeeding in creating a model that actually tells us anything of importance is something I have yet to see in any external model.

From the range of outcomes available when Gaudette won his Hobey Baker we probably got a 20th or 30th percentile result. But why we got that disappointing result is far beyond anything that could be answered with a model and requires in-depth understanding of the person, the player, the situation and the context that putting a few data points into a formula will never be able to replace.

In Gaudette's case, his numbers in the NCAA and the AHL were clearly projectable to the NHL. And in fact at various points (33 points in 59 games in 19-20, 13 goals already this year) he has actually shown that those skills and production can translate. The line between success and failure for this player is a very fine one and the reasons he has (mostly) failed are far more nebulous and intangible than anything a model could hope to answer.

 

[ManVanFan]

If somebody did, I'd hope that a database of their model's rankings would be available and that their methodology would be publicly available so people can understand the weights behind certain values and sanity-check things for themselves. I think one of the biggest issues with advanced stats and analytics in hockey is that everybody keeps things secret hoping their specific flavour of math will get picked up by a team somewhere for big money.

Revisiting NHLe part 2

Era-adjusted, more inclusive NHLe for everyone

thibaudchatel.substack.com

NHL Equivalency and Prospect Projection Models: Building the NHL Equivalency Model (Part 2)

A Few Minor Updates on an Established Framework

towardsdatascience.com

Chace has deleted all of his work off the internet after being hired by the Penguins. He had a massive report that the AHL had become a better league than the KHL. I chatted with him about it and how the AHL had become much better and the KHL had gone down massively but the AHL was still not above the KHL in viewings if he watched.

 

[ManVanFan]

It's almost like it's something you can't quantify in any meaningful way and that you can't put 'weird dude factor' or 'Jim Benning factor' or 'playing the wrong position factor' into some sort of mathematical formula. Even putting his play off the puck at lower levels into a model is something that none of the fanboy models attempt to accomplish.

Again : that there's a range of outcomes is common sense. That you can attempt to create a model for that range of outcomes is a notion worth looking at. Actually succeeding in creating a model that actually tells us anything of importance is something I have yet to see in any external model.

From the range of outcomes available when Gaudette won his Hobey Baker we probably got a 20th or 30th percentile result. But why we got that disappointing result is far beyond anything that could be answered with a model and requires in-depth understanding of the person, the player, the situation and the context that putting a few data points into a formula will never be able to replace.

In Gaudette's case, his numbers in the NCAA and the AHL were clearly projectable to the NHL. And in fact at various points (33 points in 59 games in 19-20, 13 goals already this year) he has actually shown that those skills and production can translate. The line between success and failure for this player is a very fine one and the reasons he has (mostly) failed are far more nebulous and intangible than anything a model could hope to answer.

If Adam Gaudette puts up shit numbers his entire career but he's weird? You would make a claim for him to be an NHLer?

 

[ManVanFan]

https://thehockeynews.com/news/secret-formula-analytics-in-the-nhl

 

[ManVanFan]

Methodology​

The methodology was simple: I wrote a program that pulled every single team's draft picks from 2007-2015 and put them alongside the players that TopDownHockey's model would have picked instead, ranked by likelihood of becoming a star. The metric for performance was WAR (of TDH's model) accumulated over the next seven seasons after being drafted, and then summing them all up to figure out who would have acquired more WAR if they were the ones drafting: the team, or the model. Some disclaimers before we get into the actual results:

• The reason I went with the following seven seasons is because that seems to be the standard for evaluating draft picks. I don't know why it was settled on seven exactly, but it makes some sort of sense: generally, if you haven't made the NHL within your first seven seasons you're not going to be any player of note (I think the only exception that I found was Nick Jensen, but I wasn't hunting for them so I definitely could've missed a few). Anything beyond seven and it probably hinges more on development and the environment than drafting anyways.
• I picked 2007-2015 for the drafts because the data only goes back to 2007, and the seventh season for 2015 draftees is 2021-22, which is the last season we have up-to-date WAR data for (someone tell TDH to update his WAR table for 2022-23).
• I picked ranking by star likelihood because that's what TDH uses when he posts on his Twitter. No other reason, really.
• Some people will inevitably have reservations over picking WAR to determine NHL success. It's not ideal, but there's no real better metrics out there. Points unfairly punish defensemen and GP tells nothing about actual effectiveness in the NHL. Besides, TDH has demonstrated that WAR is generally a better metric for determining winning than points anyways, so it's the best we have. If you're still not satisfied with using WAR, I'll post the spreadsheet that I have so you can make your own judgments over who made the better picks anyways (It's formatted awfully though—I'll trust that you'll power through it).
• TDH doesn't rank goalies or overage players, so I didn't either. Picks used on goalies were simply taken out of the dataset. There wasn't an easy way to determine overage players with a program though, so I left them in. Consider it like a freebie for the teams if their overage player actually became an NHL contributor.
• Another pick taken out was NYR's pick used on Alexei Cherepanov in 2007, as he tragically died very young, and thus I didn't think that was fair to include him in NYR's dataset. For what it's worth, TDH's model would've picked him first overall.
• Any player that played NHL games but posted a negative WAR within their first seven seasons was set to 0 WAR, as I didn't want to punish teams for having their players play in the NHL, even if they were a net negative on their team. Any player that played NHL games was probably a better draft pick than me, even if I would've had more WAR just by virtue of not playing.

Data​

I bolded all the teams that managed to outdraft NHLe.

Team	WAR	NHLe WAR	Difference
Anaheim Ducks	75.26248192	86.0578856	10.79540368
Arizona Coyotes	52.70552535	100.6280606	47.92253525
Boston Bruins	74.31326575	132.5093267	58.19606093
Buffalo Sabres	53.07765143	81.27886265	28.20121121
Calgary Flames	65.37393497	65.49411325	0.120178285
Carolina Hurricanes	78.17745991	94.85168494	16.67422503
Chicago Blackhawks	66.24949223	84.65136722	18.40187499
Colorado Avalanche	94.86944395	88.37525673	6.494187215
Columbus Blue Jackets	64.35772712	115.2004758	50.84274868
Dallas Stars	51.71521359	88.32291968	36.60770609
Detroit Red Wings	55.70576202	62.43467811	6.72891609
Edmonton Oilers	116.1647134	119.9187086	3.753995181
Florida Panthers	59.38695114	123.7786695	64.39171841
Los Angeles Kings	75.26115735	68.7038476	6.557309753
Minnesota Wild	58.96110164	68.69720687	9.73610523
Montreal Canadiens	45.95952079	68.56083485	22.60131406
Nashville Predators	69.58536049	80.73260342	11.14724293
New Jersey Devils	22.78388994	52.56784784	29.7839579
New York Islanders	81.273958	126.5243505	45.25039252
New York Rangers	41.78939637	37.26979607	4.519600298
Ottawa Senators	60.92148894	80.90309331	19.98160437
Philadelphia Flyers	43.4364581	53.59892426	10.16246616
Pittsburgh Penguins	36.22990234	60.32174721	24.09184487
San Jose Sharks	64.25186852	93.87956232	29.6276938
St. Louis Blues	77.45056404	101.9116382	24.46107412
Tampa Bay Lightning	113.515953	128.7486737	15.2327207
Toronto Maple Leafs	53.10183427	95.60597689	42.50414262
Vancouver Canucks	25.56315683	87.48578616	61.92262933
Washington Capitals	64.38664547	27.36114523	37.02550024
Winnipeg Jets	69.71173258	88.09308073	18.38134815

Results​

Perhaps surprisingly, only 4 NHL teams managed to outdraft the model. I'll get into one later, but I'll explain what made the model prevail over everyone else for the people that don't want to go into the sheet to look for themselves.

The model followed a high ceiling, low floor philosophy. It especially had a propensity for taking KHL players that often went undrafted in reality. This made 2008 a uniquely bad year for the model: it had Filatov ranked second, an undrafted player named Dmitri Klopov 4th, Kirill Petrov, who was picked in the third round, 5th, and so on. 2008 was a year where the model posted a WAR number of 0 compared to the vast majority of teams.

However, when it hit on the Russian players, it hit big. It ranked Tarasenko 1st and Panarin 8th in 2010, Kucherov 5th in 2011, and Kaprizov 8th in 2015. Those picks, along with players like Gaudreau (20th in 2011), Pastrnak (9th in 2014), Point (15th in 2014), and Kyle Connor (5th in 2015), is what allowed the model to pull ahead of most teams, even if it was actually much worse at picking players to make the NHL.

 

[ChilliBilly]

Oh My God. I just made the mistake of reading nearly 3 pages of the drivel that is this thread. Painful. Maybe 2 maybe 3 posts about prospects in the Canucks organization.

 

[Scumbag Frank]

Anyone who's done hiring for data analytics firms/departments has seen this type of guy in interviews many times, typically fresh out of university or very young.

1) Spend a lot of time superficially looking up different models just enough to gain a surface level understanding and name drop them all to try to impress
2) Appeal to authority by referencing the most buzzworthy models and treating them as the gold standard
3) Act condescending towards 'unsophisticated plebs' who don't also see the value in said models

But when you question him on what key factors fundamentally give each model its predictive power, and ask him to explain why that specific factor leads to correlation with the end result, he lacks substance and understanding and comes up empty

 

[bh53]

Oh My God. I just made the mistake of reading nearly 3 pages of the drivel that is this thread. Painful. Maybe 2 maybe 3 posts about prospects in the Canucks organization.

haha I read about 10 posts and just glanced through the rest. I was hoping there was lots of updates about our prospects to read about

 

[credulous]

But when you question him on what key factors fundamentally give each model its predictive power, and ask him to explain why that specific factor leads to correlation with the end result, he lacks substance and understanding and comes up empty

as someone who did this very successfully for 15 years this is absolutely hilarious

i am not taking a position on whether nhle or any other model is good but there's some real ignorance about how the whole "model" thing works on display in this thread. as soon as you start putting your finger on the scales you completely invalidate the whole point of having a model. note that this doesn't mean you can't use the model output in conjunction with subjective analysis to make a decision but expecting a model to incorporate subjective analysis is just a complete misunderstanding of what a model is and how it works

 

[JT Milker]

Imagine working at an investment bank and your model tells a client they should acquire some shit company, but when they ask for details on how specific balance sheet items are impacting your valuation you just say that’s subjective and don’t elaborate.

 

[Curm]

So anyway, I thought Mynio deserved more ice time at the WJC, he acquitted himself fairly well. Surely at least one of the LHD prospects, Mynio/EP/Kudryavtsev, will hit.

They need a stud center and goalie in the pipeline in the worst way.

 

[theguardianII]

So anyway, I thought Mynio deserved more ice time at the WJC, he acquitted himself fairly well. Surely at least one of the LHD prospects, Mynio/EP/Kudryavtsev, will hit.

They need a stud center and goalie in the pipeline in the worst way.

That is more hope than anything at this point. It might be 2+ years before anyone knows.

 

[ChilliBilly]

haha I read about 10 posts and just glanced through the rest. I was hoping there was lots of updates about our prospects to read about

yeah I tried to understand the arguments, but realized fairly quickly that they were just repeating themselves and insulting each other. What a waste of time. I scanned a lot of both of their comments.

 

[David Bruce Banner]

But would a Potato have drafted Mynio with that pick?