Is there a sophomore slump, or just probability?

[Steerpike]

So I've thought about this a little bit but not really done any number crunching.

Rookies are interesting because they don't have a very large body of work to estimate their average production from. Because of this small sample size, the overrating/underrating of rookies runs rampant. There are a lot of ramdomish factors that can influence a player's performance in a particular year. If everything randomly goes well, the kid has good point totals and high expectations. If things don't go well, the kid is "having a hard time breaking into the league" or whatever. Thus the high anomalies get the most hype and are the most likely to regress.

This turns into a lot of stories about successful rookies having a sophomore slump.

The rookies that were unlucky usually have a luckier sophomore year but people don't care about that story as much, good sophomores don't make as good of a story.

Click to read more...

 

[Doctor No]

My belief (basically the same as yours):

It's far more likely to have a rookie who is overachieving (for a rookie) than to have a rookie who is underachieving. If you underachieve as a rookie, then you're back riding the bus in Peoria.

The corollary is that when these rookies have a typical performance in year two, it's seen as a slump.

It's similar to the "Sports Illustrated Jinx" - to appear on the cover of SI, you likely just did something great, and that usually means that you performed beyond expectations. When you return to expectations, it's seen as a "jinx".

 

[Steerpike]

We could just look at total rookie points divided by total rookie games played and compare that to sophomore points divided by sophomore games played.

Another interesting thing would be to see if there really is a rookie wall. It seems like it could form from the same exact process.

 

[Doctor No]

Two problems I have with just comparing totals:

(1) Without a biased sample, we would expect the sophomore totals to be larger than the rookie totals.

(2) We seem to agree that our rookie sample is biased (in favor of the rookies).

Both of these statements work in opposing directions, and so I'm not sure what we would get out of comparing raw totals.

 

[Random Forest]

I find that, more often than not, a sophomore slump can be attributed to increased role.

Rookies almost always get sheltered minutes. Easy zone starts and low QoC makes it relatively easy to accumulate an impressive amount of points. But the next season, many players are relied on more heavily and put in less favorable situations. On the surface, the player suddenly doesn't appear as promising as he did the previous year even though "breaking even" should be considered to be progress.

That said, the slump only happens when individual player growth is outpaced by the increased role.

Players like Val Nichushkin and Sean Monahan, just as a couple examples, are primed to have sophomore slumps, imo, unless they take sizable steps forward in development over the summer.

EDIT: Don't mean to discount luck, either. I agree that perception and expectations for young players are often defined by luck. Monahan is a great example of this. Expectations for his sophomore year are VERY high because he enjoyed an extremely fortunate rookie season, and I do not think there is much chance he lives up to those expectations next year.

 

[Doctor No]

This is also a really good point - and would be something that I'd expect would be evident in the data.

 

[Caeldan]

Additionally for rookies, there's no 'book' on them yet.
But if they stand out enough, they'll get analysed enough that teams will know how to exploit their weaknesses in their sophomore year - which I think is more of an issue for goalies than forwards or defenders, but comes into play some as well.

I'd say that if we were to compare... maybe create three rookie 'buckets'? Have the overachievers, the underachievers, and the average/control group.

Then you could look at each group separately, compare the relative differences between the groups between years? Possibly extending the look into the third year to see if it's just a second year slump only, or if it's more of a general trend towards a mean?

 

[wgknestrick]

I find that, more often than not, a sophomore slump can be attributed to increased role.

Rookies almost always get sheltered minutes. Easy zone starts and low QoC makes it relatively easy to accumulate an impressive amount of points. But the next season, many players are relied on more heavily and put in less favorable situations. On the surface, the player suddenly doesn't appear as promising as he did the previous year even though "breaking even" should be considered to be progress.

That said, the slump only happens when individual player growth is outpaced by the increased role.

Players like Val Nichushkin and Sean Monahan, just as a couple examples, are primed to have sophomore slumps, imo, unless they take sizable steps forward in development over the summer.

EDIT: Don't mean to discount luck, either. I agree that perception and expectations for young players are often defined by luck. Monahan is a great example of this. Expectations for his sophomore year are VERY high because he enjoyed an extremely fortunate rookie season, and I do not think there is much chance he lives up to those expectations next year.

PP minutes and linemates would have much more of an impact on their totals IMO than zone starts/QoC. I think this is basically "fitting" a story where there most likely is just small sample effects clouding judgement. Rookies don't usually get tons of ice time/game either. I also have seen a TON of rookies sustain major injuries (Maatta/Hertl in their first year, so that is a HUGE consideration when evaluating.

I need names of players that this happened to, but can't think of any off the top of my head.

 

[boredmale]

I think in the case of many sophomores, other teams are better prepared to play against them, hence the drop in production.

 

[Alexistheman]

Aside from the fact that its an increased role. It has to do with how the coach now views the player. If he isnt producing he will be demoted in his sophomore season vs in his rookie year the coach living with his growing pains, which may result in more points but not necessarily the better season.

 

[roboninja]

I think it is just a natural attitude ahift. As a rookie, you are striving to make the team, doing everything you can. Once you had your fairly successful rookie year, it feels like you did it, you have arrived. The fire fades a bit, which is when those young players realize that to stay in the pros you need that fire at all times.

 

[Ola]

I think it is just a natural attitude ahift. As a rookie, you are striving to make the team, doing everything you can. Once you had your fairly successful rookie year, it feels like you did it, you have arrived. The fire fades a bit, which is when those young players realize that to stay in the pros you need that fire at all times.

Good thread, I think the phenomen is basically expained by the above.

You also have a point though for sure. When you are on the way up, you kind of fight the world. After you are established, you start to fight yourself more. This is definitely the reason for why many succsesful coaches got a bit of a tough love approch. Not many players are busy doubting themselves if they got a John Tortorella 3 feets behind them...

 

[Thesensation19*]

Players come into the league wanting to work hard and prove it. Some players start off great then get too comfortable and they go into a slump. Some are motivated to escape. Some are not

 

[Doctor No]

Given the forum that we're in, are there any proposals for how we might go about testing these theories?

 

[charlie1]

Given the forum that we're in, are there any proposals for how we might go about testing these theories?

How about starting with the question, "Is the sophomore slump real?"

That is, is performance, on average, more likely to decline in a player's 2nd year than it is in any other year?

A plot of (change in pts/game) vs. (year in the nhl) might be a good first cut.

 

[Dellstrom]

Hamilton had a pretty good rookie year, considering he had a lot of pressure on him and was 19 years old... His sophomore year was 1000x better than his rookie year.

Smith had 9 points in 37 games his rookie year in Dallas... Came over to Boston his sophomore season and put up 51.

I guess I agree that it's more likely if it's a rookie with a lot of hype, as they could have been overachieving.

 

[Caeldan]

Given the forum that we're in, are there any proposals for how we might go about testing these theories?

Well expanding on my earlier thought.
Let's look at say... the past 8 years of rookies?
Sticking with forwards and using points per game in rookie year as our measure of performance (with a minimum of say... 12 games played?)...
Separate into three groups:
Overachievers (over 1 sigma above average ppg for all rookies)
Underachievers (over 1 sigma below average ppg for all rookies)
Average (+/- 1 sigma of average ppg for all rookies)

Then look at each group in aggregate as a year to year change from first, second to third year?

 

[charlie1]

Well expanding on my earlier thought.
Let's look at say... the past 8 years of rookies?
Sticking with forwards and using points per game in rookie year as our measure of performance (with a minimum of say... 12 games played?)...
Separate into three groups:
Overachievers (over 1 sigma above average ppg for all rookies)
Underachievers (over 1 sigma below average ppg for all rookies)
Average (+/- 1 sigma of average ppg for all rookies)

Then look at each group in aggregate as a year to year change from first, second to third year?

How are you defining the sophomore slump? By the number of Overachievers that drop into the Underachievers/Average group?

In that case it would be heavily influenced by survivor bias. Many of the "underacheivers" will be eliminated in year two, which will increase the group average. Even if the remaining sophmore group put up the exact same pts/game many of the over achievers would drop into the "underachievers" and "average" players since you're only left with the right tail of the distribution.

I think the best way to test this is to look at (change in pts/game) from year 1 to year 2, and then compare that to (change in pts/game) for all other years. I expect this will follow an initial upward trend followed by a downward trend (when players pass their peak). If there is an initial dip in the trend on the sophomore year this would suggest a slump. No dip, no slump.

 

[Caeldan]

How are you defining the sophomore slump? By the number of Overachievers that drop into the Underachievers/Average group?

In that case it would be heavily influenced by survivor bias. Many of the "underacheivers" will be eliminated in year two, which will increase the group average. Even if the remaining sophmore group put up the exact same pts/game many of the over achievers would drop into the "underachievers" and "average" players since you're only left with the right tail of the distribution.

I think the best way to test this is to look at (change in pts/game) from year 1 to year 2, and then compare that to (change in pts/game) for all other years. I expect this will follow an initial upward trend followed by a downward trend (when players pass their peak). If there is an initial dip in the trend on the sophomore year this would suggest a slump. No dip, no slump.

Hm...
Or you could suggest that even for underachievers... not being able to survive is further evidence of the sophomore slump as you were able to fulfill a role in your rookie season that you couldn't repeat in your second. Unfortunately you might not get a third season.

But the separation idea comes from the concept that perhaps the sophomore slump is more exaggerated due to bias in seeing only the high profile examples. So I wanted to actually define the groups by their rookie season and see how they'd change. I'd expect the overachievers would have a dip in the second season but recover, the average group to stay relatively flat, while the underachievers might even have an increase year over year.
I think we need to somehow account for the failure rate in the second year, which you point out won't be caught in a pure points per game scheme - since if they're out of the league, they're not influencing it - and the ones who survive will generally have a higher ppg.

 

[charlie1]

Hm...
Or you could suggest that even for underachievers... not being able to survive is further evidence of the sophomore slump as you were able to fulfill a role in your rookie season that you couldn't repeat in your second. Unfortunately you might not get a third season.

But the separation idea comes from the concept that perhaps the sophomore slump is more exaggerated due to bias in seeing only the high profile examples. So I wanted to actually define the groups by their rookie season and see how they'd change. I'd expect the overachievers would have a dip in the second season but recover, the average group to stay relatively flat, while the underachievers might even have an increase year over year.
I think we need to somehow account for the failure rate in the second year, which you point out won't be caught in a pure points per game scheme - since if they're out of the league, they're not influencing it - and the ones who survive will generally have a higher ppg.

Ok, if we're only concerned with the overachievers, then I would identify them using your criteria, see where they rank in the league as rookies (in regards to say ppg), then ask if they, on average, maintain or increase their position in the league in the following year.

For example, we identify the overachievers based on being in the top 20% of their rookie class. We then see what their mean rank in the league was as a group (in terms of say, ppg), and we see whether their average rank increases or decreases in the following year.

Edit: The key here in my mind is that we are comparing each group (over/under/average achievers) to the league as a whole, rather than to just their own cohort which will suffer from survivor bias.

Edit again: And really we should look at pp60, since their playing time will likely increase their sophomore year.

 

[Caeldan]

Ok, if we're only concerned with the overachievers, then I would identify them using your criteria, see where they rank in the league as rookies (in regards to say ppg), then ask if they, on average, maintain or increase their position in the league in the following year.

For example, we identify the overachievers based on being in the top 20% of their rookie class. We then see what their mean rank in the league was as a group (in terms of say, ppg), and we see whether their average rank increases or decreases in the following year.

Edit: The key here in my mind is that we are comparing each group (over/under/average achievers) to the league as a whole, rather than to just their own cohort which will suffer from survivor bias.

Edit again: And really we should look at pp60, since their playing time will likely increase their sophomore year.

How do we account for dropouts though? Or do we assume that if they don't make it second year, they've likely had enough influence on the cohort to help prove the slump?

Or do we need to account for the number of players in a cohort and track that separately?

 

[charlie1]

How do we account for dropouts though? Or do we assume that if they don't make it second year, they've likely had enough influence on the cohort to help prove the slump?

Or do we need to account for the number of players in a cohort and track that separately?

Well I was mostly interested in the overachievers, since those are the only ones who could "slump" because you have to be good in the first place in order to have a slump. I don't expect there to be many dropouts in the overachievers group. Nor do I really expect there to be many dropouts in the average group.

For the underacheivers, there would be dropouts, and I don't have any good idea for how to deal with that bias. But again I'm not really interested in them since the slump only pertains to the overacheivers.

 

[Czech Your Math]

It shouldn't be all that complicated.

Take a fixed group of players (forwards only was a good suggestion IMO) which played at least X games in each of their first two seasons. X should be enough to give the sample some reliability (someone suggested 12 games... seems way too low... more like 40+).

Determine which metrics to use: adjusted PPG, adjusted GCPG, and PP60 each have their merits.

Take the median (or an average of the % change for the median Y... where Y is some fraction, like 1/3) for each metric used.

 

[charlie1]

That's essentially the same thing that that we just outlined, except we're explicitly identifying the group to track as being the overachievers.

It gets more complicated if you're trying to account for the survivor bias due to dropouts, but I don't think that is necessary since the sophomore slump only applies to overachievers (who generally do not drop out).

 

[Doctor No]

Take a fixed group of players (forwards only was a good suggestion IMO) which played at least X games in each of their first two seasons. X should be enough to give the sample some reliability (someone suggested 12 games... seems way too low... more like 40+).

One change to suggest - I'd remove the games played restriction in year two. If you keep it, then you'll miss players who disappeared entirely after their rookie season (the ultimate sophomore slump).