It took me way more time than I expected but ironically it could have been a very short post.
Frankly, the main point you can sum up in one sentence:
“fantasy NBA stars are valuable in snake draft not only because they offer the most stats and an unfair advantage but also they are… the most predictable group of players“.
In a longer version I’ll throw-in a couple of nuggets, I’ll explain how I’ve came up with this conclusion [be advised! This post contains some math ;-)] and how does it translate into different scoring systems.
Let’s start with the easier one, fantasy points and a typical 12-team 10-slots per team league. Again as an example I’ll use Yahoo’s Default Points Scoring [FGA (-0.45), FGM (1.0), FTA (-0.75), FTM (1.0), 3-pt Made (3.0), Point Scored (0.5), Rebound (1.5), Assist (2.0), Steal (3.0), Turnover (-2.0), Blocked Shot (3.0)].
I collected players’ statistics from dougstats.com, for every season in the last 10 years I sorted them from best to worst according to aforementioned formula [average per game not total].
Then I checked what players from every draft slot in Top120 did during next season and finally calculated average and standard deviation for a change in rating for each pick.
OK, if it sounds too complicated, here’s an example, according to Yahoo’s Default Points Scoring…
in 2009/10 3-rd best player was Jason Kidd who finished 20-th in 2010/11.
in 2008/09 3-rd best player was Dwyane Wade who finished 5-th in 2009/10.
in 2007/08 3-rd best player was Marcus Camby who finished 27-th in 2008/09.
in 2006/07 3-rd best player was Gilbert Arenas who finished 61-st in 2007/08.
in 2005/06 3-rd best player was Kevin Garnett who finished 1-st in 2006/07.
So for pick #3 in the last 5 years we have data points of -17 [because 20-3], -2, -24, -58, +2. Got it? In short it means “how would you do at each draft slot if you pick only by ranking from previous season?”.
I’ll spare you a whole table because it’s too long but I did the same thing as above for each pick from Top 120 for the last 10 years [you can find results in this file].
[FYI: outlier at #21 is thanks to Troy Murphy who was kidnapped by Nets last year and finished at staggering 336-th place!]
… and average…
What’s the point of those graphs and what does all this data tell us? IMHO…
1) Top5-8 picks are the easiest to predict!
So not only they offer more statistics [as I've mentioned in my post "why snake draft is unfair"] but there’s little surprise where guys like LeBron or Durant will finish. They may have a worse season than usual but even their worse seasons are still very good by league’s standards!
2) After 80-th pick it’s pretty much a crapshoot.
Which means you can draft there someone who will finish in the Top50 or outside of Top200 ;-)
Please note that “island of predictability” around 90-th pick. Those are usually “boring” guys – veterans like Battier or Ben Wallace who are drafted to fill up rosters!
3) Few draft slots regularly outperform their position from previous season.
I’m not gonna lie, I was very surprised by a scale of this phenomenon. Only 10 out 120 slots on average were better than in previous season? And most of them were “boring” veterans while none of them in the Top25? Isn’t it an argument NOT to base your ratings on previous year’s numbers?
OK, all above was about fantasy points… how does it change when we switch to rotisserie?
I used standard 9 categories for a 12-team 10 players-per-team league and values from basketballmonster.com.
Unfortunately they have only data for last 4 seasons…
Because of a smaller sample size there’s more noise here… but all conclusions are pretty much the same as above with one interesting twist: it’s easier to predict players in fantasy points than in rotisserie’s values!
On average for 120 picks standard deviation was higher by 5 picks.
How is it possible? I have one theory…
in rotisserie value has many dimensions while in fantasy points it has only one. So if a player adds one turnover and one assist per game in fantasy points his value will stay the same while in rotisserie it will change.