Statistical Analysis: 1st Round Draft Slot and GP

[MacOfNiagara]

I did this just to satisfy my own curiosity but thought I would share the data. As a Sabres fan we have discussed the importance of a high draft slot for most of the season.

These stats are based on the following:

1st round draft position from 1985-2000

It looks at the 1st round draft position in comparison to that players career games played (as one of many possible measures of the quality of an NHL player).

It then averages the number of games played for the draft position over the 15 year span.

Notes:
- Not all years had the same number of teams / draft slots. But it is an average so that should be mitigated a bit.
- Early career injuries, personal issues, etc., influenced some players GPs. But the law of averages helps mitigate that a bit too.
- I did not include the most recent years as some players take longer to make the NHL while others are thrust right into the NHL.

Take it with a grain of salt, and interpret the numbers as you wish.

Here are the raw numbers, followed by a chart:

Draft Slot Avg GP
1 959
2 855
3 894
4 580
5 751
6 504
7 661
8 511
9 521
10 581
11 586
12 341
13 504
14 547
15 288
16 324
17 321
18 384
19 358
20 364
21 318
22 303
23 366
24 272
25 342
26 252

Click to read more...

 

[Hynh]

I've always thought 4th was a hard spot to pick. Too high to go against the grain, too low to get the elites.

Also interesting is the sharp drop between 14th and 15th.

 

[Teeder9]

14th to 15th might be the spot between safe and homerun type picks.

 

[Brave Sir Robin]

Does this include goalies?

 

[Chieftans22*]

Is there anything like this for PPG or Points?

 

[TheStranger]

It's too bad we can't look at the players available then, knowing only their status coming into the draft, and see if people could draft better "BPA's" in each position.

 

[542365]

Pretty interesting. 14 to 15 is pretty interesting. 4 spot is a tough place to be.

 

[Mc5RingsAndABeer]

Good work. I'd be interested to see the stats for points as well (for forwards).

And perhaps separate out the old stuff because scouting has improved a lot since then.

 

[WarriorOfGandhi]

thanks for doing this OP, this is really cool, I've been wondering this stat myself. For those who have more (much more) time on their hands, creating additional position charts with points and sv% would also be very illuminating.

 

[NotABadPeriod]

I'd suggest adding a best fit curve (and r^2 value) to give a better picture of the correlation.

I think the issue with points (or ppg) is defense vs forwards: obviously the forwards are going to score more, and usually the first few slots tend to more likely be a forward than a D-man. Splitting them though might yield sample size issues.

 

[MacOfNiagara]

Started working on forwards only and points, but then needed to make a decision:

- Career points

or

- Avg PPG

Points per season would take too long as the database does not offer that stat on the same page and I would need to look them all up individually

eh, going to go with PPG as I would be more interested in that stat as far as assessing the quality of offensive forward

As mentioned by GrigsandGirgs there will be a bit more of an issue with sample size as some of the draft slots have less forwards than others.

 

[MacOfNiagara]

Forwards - ppg

Here is the re-run using only forwards and PPG average over career.


Notes:
- 1985-2000
- Did not want to drop to fewer years as it would just exacerbate the smaller sample size problem with using only forwards
- Dropped the last few slots (26-30) as the combination of small sample size and players not making the NHL left the sample size for those slots too low in my opinion to be even remotely useful.
- Made no adjustment for # of games played (therefore a PPG for a player only playing 3 career games is weighted equally with the PPG of someone playing 1000+ games). Reviewing the data as I went I think the overall effect was not too significant

Raw Data:
Draft Slot AVG PPG
1 0.82
2 0.73
3 0.60
4 0.41
5 0.60
6 0.50
7 0.40
8 0.43
9 0.34
10 0.44
11 0.48
12 0.36
13 0.39
14 0.33
15 0.34
16 0.31
17 0.26
18 0.26
19 0.29
20 0.28
21 0.31
22 0.40
23 0.38
24 0.34
25 0.34


Chart:

 

[Shoalzie]

I like how the non-lottery picks are around the same for GP and PPG...not nearly as much pressure to hit on those picks.

What's the bust factor at each spot...picks where the player hasn't played a single game in the league or hasn't been a regular?

Curious to see what the top 5 in the second round (31-35) compare to those non-lottery picks. Those picks are similar in value but playoff teams may have different approaches to the draft than the rebuilding teams.

 

[MacOfNiagara]

I like how the non-lottery picks are around the same for GP and PPG...not nearly as much pressure to hit on those picks.

What's the bust factor at each spot...picks where the player hasn't played a single game in the league or hasn't been a regular?

Curious to see what the top 5 in the second round (31-35) compare to those non-lottery picks. Those picks are similar in value but playoff teams may have different approaches to the draft than the rebuilding teams.

Obviously the definition of bust if pretty subjective. I didnt want to use 'a single game' because there were many 1st round picks where the player played only a handful of games (e.g. 3GP). To me that would be a bust.

So I used some subjective analysis. I counted as a bust any 1st rounder not playing at least 82 games.

Notes:

- I did not go through and look at the history of each player to see if there were any mitigating factors (other than not being good enough).
- I calculated % of picks at that slot that busted. Do to league expansion some of the lower picks (22-26) had less picks during the range of years calculated. So, I thought % of bust was better than simple total.

Raw numbers:
(Yikes at pick 15)

Draft Slot Bust%
1 0
2 0
3 0
4 13
5 7
6 27
7 20
8 27
9 20
10 13
11 27
12 47
13 40
14 13
15 67
16 33
17 53
18 27
19 47
20 20
21 60
22 36
23 40
24 20
25 22
26 33

Chart:

 

[Shoalzie]

That's a fair definition...you're right, bust is a subjective term. Might not really be games...it's more of an overall production compared to expectation.

 

[MacOfNiagara]

That's a fair definition...you're right, bust is a subjective term. Might not really be games...it's more of an overall production compared to expectation.

Agreed, but that would take way more time than I have.

 

[Bear of Bad News]

I think that games played is a reasonable proxy (although admittedly a coarse one).

If an NHL team is playing someone at the expense of other options, then they have value.

 

[schuckers]

You might want to take a look at this paper that we did on the value of NHL draft picks.

NHL Draft Pick Chart

 

[PSGJ]

If you ignore the noise the trend is pretty clear. The first 15 or so see a sharp decline, after that though it seems the difference is negligible.

So, If someone wants to trade up and offers you let's say pick 28 + for pick 18 then you should definitely say yes because those picks are pretty similar in value. Whatever that + is will be profit for you.

 

[reckoning]

Interesting thing about the drop-off at #15: two of those picks for the years of this study, Sakic and Kovalev, had very long and productive careers. They match up well against the top pair of any other number. But it's very much a weak group aside from those two.

It's no surprise that the trend is that higher picks turn out to be the better players, but I think it still comes down each teams player development. Getting an early pick is no guarantee of success. If it was, then the Atlanta Thrashers would've been a dynasty.

 

[Caeldan]

I find it interesting that not a single top 3 pick has played under 82 games?

Also that in any of these criteria... 4 always seems to be out of place for its position... and 10 seems to be a fairly strong pick (better than 4 even).

 

[PSGJ]

I find it interesting that not a single top 3 pick has played under 82 games?

Also that in any of these criteria... 4 always seems to be out of place for its position... and 10 seems to be a fairly strong pick (better than 4 even).

That's clearly randomness at play and is not something with predictive value. You have to focus on the overall curve of decline in value.