Draft Percentages

[Drunk Uncle]

Hopefully this is worthy of a new thread. I wasn’t 100% on whether I was going to share this but with the draft coming up soon I thought this would be timely.

I was really intrigued by some earlier posts and ongoing discussion focused around acquiring picks and the value of those picks if we do. The part that intrigued me most was whether it is worth trying to land a Brayden Point in the draft or whether you need to draft high in the draft (top 3).

Everything I have found written about draft success was centric to the number of games a player played post draft. That really doesn’t tell you much about how good the player is and from my perspective you need top 6 players (specifically top line players) with your draft picks because those are the hardest to trade for. To that end, I looked at the last 20 years of drafts and added PT/GP to the equation for forwards (data from hockey db). Using a career average of 0.78 PT/GP (64pts) as threshold from a top line player here are the odds of selecting a top 3 forward:

Top 3 pick: 44%
4-6 pick: 15%
7-15 pick: 3.5%
15-32 pick: 3.5%
2nd RD: 1.3%
3rd RD: 0.8%
4th RD: 0.3%
5th RD: 0.5%
6th RD: 0.3%
7th RD: 0%

Here are to odds of drafting a top 6 forward (0.59 PT/GP or 48 pts):
Top 3 pick: 73%
4-6 pick: 41%
7-15 pick: 25%
15-32 pick: 13%
2nd RD: 5%
3rd RD: 2%
4th RD: 1%
5th RD: 2%
6th RD: 1%
7th RD: 1%

I also built the data out to look at defense (using PT/GP and heavier weight on GP) and goalies (GP/season) and even went as far as evaluating every NHLs team draft performance. This looks at how many players a team drafted above the average PT/GP player at that pick position. I can also look at slices of years (not just 20). If you’d like to see something, let me know and I’ll try and grab the slice with that data. Only thing I'll say is that the data is

For me the results highlights that draft picks, especially late first rounders, are not as valuable as they are continually made out to be. What surprised me is that is basically a 50/50 chance of drafting a top line forward with a top 3 pick. I would have thought it would have been higher.

Others may draw different conclusions but what this meant for me as a Flames fan is, if the Flames choose not to re-build, I’m 100% ok trading our 12th overall if we can get a top 6 forward back. It also means there is zero chance I’m trading Tkachuk, Gaudreau or Monahan unless:
a) We are re-building
b) We get the best player back
c) They want out and force our hand

[Fighting Banana Slug]

I appreciate the effort on this and I agree with the basic conclusions. Even the "sure thing" first rounders are far from sure things.
Drafting 18 year olds is really difficult and requires a good dose of luck. While at some stages of a team's development, "more bullets in the chamber" is a prudent strategy, it doesn't guarantee success by any means. It is not a good strategy in my opinion where there is already talent in the 23-28 year old range on the team.

[powderjunkie]

Career average skews this way too high IMO.

In the last 20 years, only 59 guys have averages 0.78 or higher (500+ gp in that time). This includes a few players who played games before 2002 which may bump them up or down.

<0.78:
Vanek Parise Voracek Briere Nash Krejci Couture Perry Pacioretty Landeskog Monahan ROR Zibanejad Marleau J Carter O Jokinen

and many others. Most of the above had 5+ years as top-line players.


It's always hard to account for career trajectory in this kind of analysis. Aside from the greatest, most very good players follow a path resembling:

~2 years 3rd liner
~2 years 2nd liner
~5 years 1st liner
~3 years 2nd liner
~3 years 3rd liner

That would be a darn good 15 year career for any pick (other than maybe #1-3OA)


From a draft pick value standpoint, the first 8-12 years of a career are probably the most relevant, as most players will complete their lifecycle with a team in that time (getting traded at some point likely means positive value, too).

I wonder if it's possible to limit the averages to the first 10 years of a career? Or under age ~30?

[Drunk Uncle]

Originally Posted by powderjunkie

Career average skews this way too high IMO.

In the last 20 years, only 59 guys have averages 0.78 or higher (500+ gp in that time). This includes a few players who played games before 2002 which may bump them up or down.

<0.78:
Vanek Parise Voracek Briere Nash Krejci Couture Perry Pacioretty Landeskog Monahan ROR Zibanejad Marleau J Carter O Jokinen

and many others. Most of the above had 5+ years as top-line players.


It's always hard to account for career trajectory in this kind of analysis. Aside from the greatest, most very good players follow a path resembling:

~2 years 3rd liner
~2 years 2nd liner
~5 years 1st liner
~3 years 2nd liner
~3 years 3rd liner

That would be a darn good 15 year career for any pick (other than maybe #1-3OA)


From a draft pick value standpoint, the first 8-12 years of a career are probably the most relevant, as most players will complete their lifecycle with a team in that time (getting traded at some point likely means positive value, too).

I wonder if it's possible to limit the averages to the first 10 years of a career? Or under age ~30?

You are bang on and the career PT/GP is even more disingenuous to the rookies than the vets. I original had the number at >0.69 for top line players and, while I don't have the year by year stats for each player to take a players D+7 to D+12 (25yo to 30yo) stats and look only at that, I can lower PT/GP and look at those numbers:

For >0.69 (Top line guys) the percentages are (78 guys in this bucket now. Who wants the 12 top line bottom feeders anyway… they are in the 2nd line bucket now):
Top 3 pick: 54%
4-6 pick: 22%
7-15 pick: 10%
15-32 pick: 6%
2nd RD: 2.4%
3rd RD: 0.8%
4th RD: 0.3%
5th RD: 0.8%
6th RD: 0.3%
7th RD: 0.4%

Here are to odds of drafting a top 6 forward >0.54 PT/GP
Top 3 pick: 81%
4-6 pick: 54%
7-15 pick: 31%
15-32 pick: 16%
2nd RD: 6%
3rd RD: 3%
4th RD: 2%
5th RD: 0.8%
6th RD: 0.3%
7th RD: 0.4%

I draw the same general conclusions as my original post.

I also should have mentioned how surprised I was at how little value a 2nd and 3rd rounders hold. I’d be handing those out like candy for top 6 forwards. For example, the Arvidsson trade is basically Arvidsson for virtually no chance of drafting a player as good as Arvidsson.

[powderjunkie]

High-end forwards, especially C's, are especially hard to find the further down you go.

The Scheifeles and Kopitars and Getzlafs and Ahos and Bergerons and Points are still out there, so whether you're pick 8th or 80th, IMO you really have to priotize C.

While still rare, you're a lot more likely to find a great goalie, D, or winger in the later rounds (or undrafted).

[Drunk Uncle]

Originally Posted by powderjunkie

High-end forwards, especially C's, are especially hard to find the further down you go.

The Scheifeles and Kopitars and Getzlafs and Ahos and Bergerons and Points are still out there, so whether you're pick 8th or 80th, IMO you really have to priotize C.

While still rare, you're a lot more likely to find a great goalie, D, or winger in the later rounds (or undrafted).

I actually looked the D too using PT/GP. Admittedly using PT/GP is disingenuous to the Tanev’s, Scandella’s and Carlo’s of the world and overly generous to the Justin Schultz and Erik Gustaffson’s but when at the group in aggregate it does feel relatively accurate.

For >0.44 (Top 2 guys) the percentages are (57 guys):
Top 3 pick: 42%
4-6 pick: 25%
7-15 pick: 18%
15-32 pick: 6%
2nd RD: 4%
3rd RD: 2%
4th RD: 2%
5th RD: 2%
6th RD: 0.5%
7th RD: 0%

Here are to odds of drafting a top 4 D >0.29 PT/GP (143 guys):
Top 3 pick: 92%
4-6 pick: 35%
7-15 pick: 34%
15-32 pick: 19%
2nd RD: 12%
3rd RD: 5%
4th RD: 5%
5th RD: 5%
6th RD: 2%
7th RD: 1%

You do have a better chance to get a good D in the later rounds. It’s still really long odds though.

Here are the goalies based on GP/year. This one is more subjective however looking and the sample and who fits in the buckets most of the names you’d expect to be in the group are there.

Starter (>19 GP/YR. 37 guys in the bucket)
Top 3 pick: 100%
4-6 pick: 0%
7-15 pick: 22%
15-32 pick: 24%
2nd RD: 13%
3rd RD: 9%
4th RD: 5%
5th RD: 2%
6th RD: 3%
7th RD: 2%

Back-up goalies are always available so really no need to look at them.

[Drunk Uncle]

One more thing worth noting with the forwards: I also looked at the elite forwards (top 36 >0.85 PT/GP)

Here are the odds:
Top 3 pick: 35%
4-6 pick: 7%
7-15 pick: 3%
15-32 pick: 2%
2nd RD: 0.8%
3rd RD: 0.8%
4th RD: 0.3%
5th RD: 0.3%
6th RD: 0.3%
7th RD: 0%

Drafting guys like Point, Pastrnak, Gaudreau, Mark Stone, Kucherov, Aho or Marchand where they were drafted is like finding a unicorn. It's not a winning draft philosophy.

[Geeoff]

I think it's worth considering the salary cap.

-any established top 6 forward you acquire is likely to have a significant cap hit

-any successful team in a salary cap league has players outperforming their cap hit (such as players on ELCs)

-I know you only considered top 6 forwards important but look at how much cap space the Flames spent on the (terrible) bottom 6 this season. If we had not been throwing away draft picks, maybe we'd have cheaper and better options for the bottom 6.

Personally, I think these low percentages are an argument to draft more than 7 times a year, rather than less

[CASe333]

Originally Posted by Geeoff

Personally, I think these low percentages are an argument to draft more than 7 times a year, rather than less

I think both opinions are right to a certain extent.

As DrunkUncle noted it's foolish to trade a player who has consistently put up points for 2nd + 3rd draft picks. There just aren't enough of these players available to take on the risk of two magic beans turning into a comparable player. When trading a player like Arvidsson (0.62pt/GP + still young) you should always make sure you are getting a comparable player back in return rather than go for the magic bean approach or trading for a star past their prime. Using the %s in op times the amount of draft positions in that percentile results in just 3.57 players being drafted per year that meet the 0.78pt/GP threshold. For the 0.59pt/GP threshold on average only 11.1 players are drafted per year. Looking at the worst trades in NHL history its often a package of journeymen + later round draft picks traded for a superstar that looks really bad historically.

On the contrary, as pointed out in the last post if you can fill your third and fourth lines with younger players drafted after the first round you can save money to pay for multiple star players. Having multiple picks greatly increases odds of drafting a "unicorn" like Point or Marchand but also greatly increases odds of getting a decent 3rd-4th line support player. Having more picks is definitely a good thing so I would always support picking up more picks as long as you aren't giving up an early first round pick or bona fide star player.

For example Tampa Bay is icing two young fourth line players drafted late (Mathieu Josheph, age 19, drafted 120 in 4th round, salary 737,500 and Ross Colton age 22, drafted 118 in 4th round, salary 700,000). Cap circumvention arguments aside, having cheap young players to fill out their roster helps allow them to ice their stacked roster. I never understand why so many teams fill their bottom roster with expensive aging players that are way past their prime. Having one or two veterans is always good especially if they come cheap like a Corry Perry for 750K but some teams seem adamant in filling their bottom lineup with veterans who are almost always more expensive than a player on their entry contract.

[CF84]

This is a great thread!

I do think this year is going to be quite skewed however. I watched Makar during his draft year, and if he was playing during this shortened AJHL season, he would not have climbed up the draft rankings the way he did. Because of this, I believe it will be a great year for teams to find gems in every round of the draft.

On the flip side, I also think there may be more busts in the first rounds.

Here is a snippet from an article 8 games into the 2016-17 season:
BROOKS, AB (October 4, 2016) – "Brooks Bandits defenceman Cale Makar (Calgary, AB) has been named to the NHL Central Scouting Players to Watch List as a “B”-rated, or 2nd – 3rd round, prospect, it was announced today."
https://www.brooksbandits.ca/media-c...=2016&nID=9954

[ricardodw]

When you use this draft pick evaluations to show the lack of value in trading a top-6 forward 48 PT in a season then you also have to evaluate whether the 48 pt top-6 guy will remain at that level. Backlund might have qualified under your critera. Will he ever get 48 pts again?

Monahan would likely qualify as a top-3 guy in your scheme after his 84 pt season in 2018-19. Will he hit 60 again? Will he dropped off of your list of successful #6 overall. Will he remain a #6 OA pick that beat the odds?

Monahan put up a ho-hum 34-pt rookie season in D+1 when he could have been dominating the OHL that is bringing his career average down because the Flames didn't have any Centres at the time.

Erik Karlson got traded for prospects and picks and not likely he is the ppg D man ever again that SJ thought they were getting.

[Ped]

Originally Posted by CASe333

I think both opinions are right to a certain extent.

As DrunkUncle noted it's foolish to trade a player who has consistently put up points for 2nd + 3rd draft picks. There just aren't enough of these players available to take on the risk of two magic beans turning into a comparable player. When trading a player like Arvidsson (0.62pt/GP + still young) you should always make sure you are getting a comparable player back in return rather than go for the magic bean approach or trading for a star past their prime. Using the %s in op times the amount of draft positions in that percentile results in just 3.57 players being drafted per year that meet the 0.78pt/GP threshold. For the 0.59pt/GP threshold on average only 11.1 players are drafted per year. Looking at the worst trades in NHL history its often a package of journeymen + later round draft picks traded for a superstar that looks really bad historically.

That might be true in a vacuum, but we don't actually live in one. It's a salary-cap world in an expansion draft year, which makes trading a player like Arvidsson for picks a good deal, imo:

1) It frees up cap space in a flat-cap environment.
2) Arvidsson really does appear to be on the decline since his injury after the Bortuzzo cheap-shot
3) It frees up an expansion draft slot or at least gives the Preds assets, since it wasn't even sure they were going to protect Arvidsson anyway. The Athletic suggested they would go the eight skater route so they could protect Ekholm, Josi, Fabbro, Ellis, and Carrier on the blueline, and Scissons, Forsberg, and Kunin up front. Trading Arvidsson means the best player the Preds can lose is probably one of Jarnkrok, Johansen, or Duchesne, which isn't bad at all.

[Tacoman]

Thanks for this. One conclusion I'd also make is that it is not prudent to trade down in the NHL draft. The gaps between the different tiers in the first round are significantly bigger than any odds of additional second or third rounders get you in return.

[Drunk Uncle]

Hey All! I've really appreciated the feedback and comments on this and you've motivated me to find better stats. There is enough interest beyond myself that I'll share this when I get done. The goal of better stats will be to:

a) provide stats of players relative to their draft year ie. players D+7 to D+12 years
b) Include TOI/GP as a metric for D in addition to PT/GP to balance defensive D.
c) Get better goalie stats (not sure what I'm going to use). I'll probably decide what gives me the best x-section of buckets of goalies.

The good news is the way I built the model I should be able to swap out the stats without a large amount of pain.

I'll also aim to provide some results of how the Flames have drafted relative to the rest of the NHL.

The model already does this by fitting an exponential line of best fit to the Forward, D and G data and then measures whether a draft pick exceeded that line as a measure of a successful pick. I'd like to get the player metrics and philosophy more accurate and before providing that info. For the stats junkies I'll provide the R squared and STD DEV info too.

The model is sliceable by years as well so I can actually evaluate the draft performance of each of the Flames GMs over the last 20 years. Just a disclaimer that it will be really hard to tell anything about Treliving given that it usually takes 5-7 years after drafting a player to really see what you have (think more Mangiapane vs Gaudreau).

Having said that I'd be interested in others suggestions in what they think would add value and I'll consider using those too.

[dino7c]

Originally Posted by Tacoman

Thanks for this. One conclusion I'd also make is that it is not prudent to trade down in the NHL draft. The gaps between the different tiers in the first round are significantly bigger than any odds of additional second or third rounders get you in return.

Its irrelevant if you take the same guy at say 16 that you would have at 12...extra picks are extra picks

[Drunk Uncle]

Originally Posted by Tacoman

Thanks for this. One conclusion I'd also make is that it is not prudent to trade down in the NHL draft. The gaps between the different tiers in the first round are significantly bigger than any odds of additional second or third rounders get you in return.

I think it depends on where you are picking, how far down you are moving and what your tiers look like at your draft position. It's also worth noting other teams tier lists impact this too.

What I see is that past the first few picks you're really trying to figure out who will evolve from where they are currently into a future impact player. When a tier is defined in a draft list, their is agreement that the players in the tier are essentially equal. I think trading back for the extra pick is worth it within the same tier.

What this does say is that you better scrutinize your tier lists heavily to make sure that in fact the players you are putting together in a tier belong there because I do think trading back out of your tier is a mistake.

[gvitaly]

Is there a way to have a look at where members of the HHOF were drafted? I think it's one of the best indicators of franchise players. It essentially ignores all the players still playing though.

[Drunk Uncle]

Originally Posted by gvitaly

Is there a way to have a look at where members of the HHOF were drafted? I think it's one of the best indicators of franchise players. It essentially ignores all the players still playing though.

I only have the last 20 years of stats. All of the HHOF members were drafted prior to 2000.

I also think it's a pretty slippery slope looking at HHOF members only because longevity of career matters a lot so guys that were great players like Kent Nilsson for example would be ignored... and you shouldn't be ignoring a career 1.24 PT/GP player.

[Fighting Banana Slug]

Originally Posted by Drunk Uncle

Hey All! I've really appreciated the feedback and comments on this and you've motivated me to find better stats. There is enough interest beyond myself that I'll share this when I get done. The goal of better stats will be to:

a) provide stats of players relative to their draft year ie. players D+7 to D+12 years
b) Include TOI/GP as a metric for D in addition to PT/GP to balance defensive D.
c) Get better goalie stats (not sure what I'm going to use). I'll probably decide what gives me the best x-section of buckets of goalies.

The good news is the way I built the model I should be able to swap out the stats without a large amount of pain.

I'll also aim to provide some results of how the Flames have drafted relative to the rest of the NHL.

The model already does this by fitting an exponential line of best fit to the Forward, D and G data and then measures whether a draft pick exceeded that line as a measure of a successful pick. I'd like to get the player metrics and philosophy more accurate and before providing that info. For the stats junkies I'll provide the R squared and STD DEV info too.

The model is sliceable by years as well so I can actually evaluate the draft performance of each of the Flames GMs over the last 20 years. Just a disclaimer that it will be really hard to tell anything about Treliving given that it usually takes 5-7 years after drafting a player to really see what you have (think more Mangiapane vs Gaudreau).

Having said that I'd be interested in others suggestions in what they think would add value and I'll consider using those too.

Just drafting this e-mail explaining what you do is more effort than I would ever put in (let alone doing the actual work!). Thanks for the effort, it is an interesting project and I think will help many of us understand the true odds of success at the draft.