In a previous post, I looked at the benefits of hitting on a sleeper or missing on a bust. As I mentioned, I think one of the hardest things to define in that post was who qualifies as a sleeper and who qualifies as a bust. A lot of people rightfully questioned that, often debating that I was too loose on my sleeper definitions. I specifically defined sleepers as any player outside the starters at their position according to ADP, but a lot of people wanted a smaller set of "sleepers". So today, we are going back to sleepers. Specifically, we will look at how players who were called sleepers going into the draft season performed to get a sense of if we can predict who sleepers will be.
The Data
The main data source I am going to be using will be ESPN's annual sleeper posts between 2015 and 2020. I will also take ADP data and finish data from Asteroid Databases along with my draft database to determine championship odds. Using the ADP data, we can also define the opportunity cost of a player. In this instance, we will define the opportunity cost as the 2 players ranked 1 above and 1 below the sleeper at the same position. These should represent the players that were being debated instead of the sleeper. For example, one sleeper from last year was Cam Newton. The opportunity cost comparison for Cam Newton would therefore be Matthew Stafford (one positional ADP earlier) or Daniel Jones (one positional ADP later). Based on this we can answer whether "sleepers" really are more valuable than their peers.
Year-End Performance Ranking.
Box plots of how much a player's positional ranking changed between the draft and end of the year when broken by sleeper or alternative. Positive numbers mean a player finished better (say positive 2 for the 25th RB at the draft and who finished as the 23rd). ptype stands for player type.
When looking at the change in position ranking from ADP to the end of the year, the first thing that jumps out is that most players drop (with the median for RBs, WRs, and TEs all being less than 0 and therefore finished worse than their draft ranking). This is likely dragged down due to missed games either due to injury or starting as a backup before gaining the starting position (neither of which I controlled for here). More importantly, there is no difference between the sleeper (orange) compared to the other players at any position.
But that is looking at averages, we don't care about the average performance when it comes to sleepers. What we want are the hits. Specifically, we think that a sleeper has a better chance of becoming a top 12 player than a non-sleeper.
Top 12 Odds
Bar plots of the odds of a player finishing as a starter at their position. Error bars represent 95% confidence interval based on yearly variation in odds.
For the most part, this is the same story as before, with little difference between the sleeper pick and the players around it. The only standout point is in TE, where the sleeper's matched the player one pick earlier, but both had a better shot at finishing top12 than the next TE taken. The other interesting thing is how much better the odds of a starting QB or TE being a sleeper compared to a starting RB or WR. While not surprising, this just reinforces why RBs and WRs are more valuable
Again, this has the limitation of rewarding complete seasons, compared to a player that was a stud until they got hurt, or the rookie who didn't start for the first 6 games, but then was a top 5 player the rest of the year.
What About Short Term Starters?
To try to account for players that did not start the entire season, we can look at their weekly performance instead of year-end totals. Specifically, we will count the number of games where they finished as a top12 player at their position. Note, my sample size here is restricted to 2019 and 2020 only due to some data constraints.
A little noisier, but again, we are finding no significant differences between sleepers and their non-sleeper counterparts. Note, I expect this is just a quirk of the data, but the QBs did better the later they were drafted, though this result is by no means significant.
What about the Championship Odds of a Sleeper?
Finally, we can look at the odds of winning the championship when you draft a sleeper compared to a non-sleeper. This will be the smallest dataset yet, focusing on 2020 only.
Bar plot of the odds of winning your championship in 2020 when you drafted a sleeper. The Red line represents the baseline random odds of the dataset. Error bars represent 95% confidence intervals
This is kinda similar to the end-of-year starter odds. Basically no difference except maybe at Tight End as to whether you drafted a sleeper or someone else. Relatedly, all of these picks are close to your base odds, changing your championship odds by about 1% in either direction.
What if you miss on your sleeper pick?
Obviously, the fear with a sleeper is that they may never put up a good performance. So what happens if you miss on a sleeper. How do misses effect your championship odds? For this, we will define a miss as a player who finishes outside the top 18 for QBs and TEs or outside the top 36 for RBs and WRs.
Ok, so obviously missing on a pick will lower your chances at winning the championship. Importantly though, whether that pick was a sleeper or their counterparts does not really matter.
Limitations
There are a few limitations, namely to do with the dataset itself. First, I am using ESPN's consolidated sleeper and breakout player lists to define who are the sleepers. This may not be the best list, but at worst I think it is a representative list as they have 10 different people name a sleeper at each position. In reality, despite what many people say, ESPN is pretty good at analyzing fantasy football, at least on a weekly basis. These results also mirror what we see on a weekly basis for start/sit articles, regardless of the source. The bigger issue may be the small sample size, particularly on the weekly top12 finishes which only include 2019 and 2020, and on the championship odds which only include 2020. Expect both to be updated in the coming years as I gather more data.
Conclusions.
I think the main conclusion from this post and my last post on sleepers is that obviously hitting on a sleeper is valuable. At the same time, knowing who the sleepers are is really hard. Combine that with some of the value-based drafting results from this past year and I think you should feel comfortable drafting players you think are sleepers, knowing that they may or may not hit.
Questions? Comments? Let me know at ac@alexcates.com. Want to read more breakdowns like this? sign up for my newsletter. Finally, like what I do? Consider supporting me on buy me a coffee.
Comentarios