One of the things we don’t get in football that we do with other sports is sample size. Each team only has 16 regular season games, and much of the time there are coaching changes on a year to year basis, player injuries, etc. It makes calculating Standard Deviation a pain. Bales talks about smaller sample sizes in his article here.
For the purposes of this article, we are going to suspend all beliefs that we need a huge sample size to calculate Standard Deviation for NFL player data.
Note: This article is using DraftKings points and salaries.
William Sharpe meets DFS
I was a Finance Major and did my MBA with a concentration in Finance. I probably couldn’t tell you one thing that I learned pursuing any of my degrees. One thing that I do remember was a ratio in investment portfolio management called the “Sharpe Ratio,” developed by a dude named William Sharpe (I had to look up his first name). Simply put, this ratio talked about the extra return you get on a portfolio of stocks, for each additional unit of risk taken. In life, leaving all else equal, we should always be looking to get a desired return, at the lowest risk possible. This is what the Sharpe Ratio tells you – “Who is consistently outperforming their salary taking risk into account?” Instead of giving you a finance lesson – which I’d botch anyway – I’m going to go full WIIFM (what’s in it for me?). If William Sharpe played DFS, here’s what his formula might look like:
Sharpe’s DFS Equation Broken Down
1. Fantasy Points Above or Below Salary
This is Fantasy Labs’ Plus/Minus rating. The fact that Fantasy Labs has already calculated this is pretty much the only reason I’m able to write this article. This is a player’s average actual points minus salary-based expected points per game. So, a QB priced at $8,500 on DraftKings is expected to score 20 points (this is not the actual number). This means that on average, QBs priced at $8,500 points on DraftKings have scored 20 points. If Aaron Rodgers is priced at $8,500 and his projection is 25 points for the upcoming week, his “expected or projected” Plus/Minus is +5.0. He is expected to exceed production of QBs priced at the $8,500 level.
2. Standard Deviation of Player’s results
As I said in the first paragraph, NFL data is crazy. I went through and looked at all of the players and tried to eliminate some of the irrelevant data. For example, Antonio Brown put up a few duds this year without Ben Roethlisberger at QB. To me, those games are irrelevant when calculating the standard deviation of Antonio Brown. I think it’s widely known that he’s been one of the most consistent players in fantasy over the last two years. The only problem is that I’m not able to back that out of the Plus/Minus over the year. I tried to do this as much as I could in calculating each player’s Standard Deviation. It was time consuming, but I think we’re close.
Why a Player’s Sharpe Ratio is Huge for Cash Games and GPPs
When I think about what I’m looking for in cash games, I think about the words “consistency, value, predictability, etc.” I believe that is essentially everything that goes into a Sharpe Ratio’s calculation. If someone told me that they could give me something that consistently outperforms their salary, taking risk into account, would that not be a cash game dream come true? I think this is the answer to our dreams! This can also be a useful tool for GPPs. When would you NOT want to be rewarded for additional risk in tournaments? Easy: NEVER! Please remember that salary is a crucial aspect of a Sharpe Ratio. Fantasy Labs does an awesome job of already calculating the Plus/Minus for us. Why am I telling you this? Because these are the top DFS players, and not necessarily the top season-long fantasy players. There is a gigantic difference. If you’re reading this, I probably don’t have to tell you that. Do not go and pick up Brian Hoyer because I have him ranked the eighth QB overall this year.
Different Sharpes to Rule the World
I believe the cool thing about Sharpe ratios in fantasy is that we can use the current Sharpe ratio to evaluate the best DFS players throughout the year – sort of like a DFS player leaderboard. Best of all, we can use a “Forward Sharpe”, where we can see who have the best Sharpe ratios based on projections for the upcoming week, which is what we all care about – who is going to make me money this week. Here are lists of who may have made you money in previous weeks…
QB Sharpes
Umm, Jameis Winston? A solid DFS play all year? No way.
Remember how I talked about sample size? There’s a reason his standard deviation is way less than anyone else on the list – there is limited data on him this year. The limited data on Jameis has come with low variance. He has been pretty damn consistent this year. It will be awesome to see where he stands in a couple of weeks, but as of right now, for the price and risk attached to his salary on DraftKings, he looks to have been a solid play based on his risk. Carson Palmer has been an absolute beast this year – no surprise there. Tyrod is another surprise, but could also be attributed to a smaller sample size. Only time will tell.
RB Sharpes
I don’t think there are any surprises here. Devonta, even with his high risk, is still crushing. But how about Duke Johnson? I actually had to look at his game logs to confirm this, and yep, he’s exceeded salary expectations in 60% of his games this year at a relatively low risk. You can attribute his low risk to sample size, but I would have never thought that Duke Johnson was DFS relevant. This comes with caveats, however. I’d argue that Duke Johnson might be a good option if you need to fill out a cheap RB spot. BUT PLEASE….PLEASE do not build your lineups around him…yet.
WR Sharpes
This list is Shaaaaaaaaaaady! No Lesean McCoy pun intended obviously, because we are talking wide receivers. Willie Snead number one?! But think why this could be. He’s been cheap, he’s been awesome, and he’s been consistent. Then you start to think, duh. But seriously, is Willie Snead the first wide receiver you think of when it comes to DFS this year? Be honest. I’ve had an obsession with Crabtree this year. He would have been the only one that I guessed was on this list. Maybe Jarvis Landry, too.
TE Sharpes
Where is everyone’s favorite TE, Tyler Eifert? I say that because his ownership was nuts this past week. He is 11th. But why? He got crushed by variance. Here’s a little exercise – go to Fantasy Labs and check out Eifert’s game logs. His scores are all over the place. He and Greg Olsen have the highest Standard Deviation of all tight ends. At this point in the article, you could probably guess that Barnidge would be number one. You probably thought, who has been cheap, outperformed, at a semi-consistent rate? And you thought: Gary Barnidge.
Conclusions
There is so much to conclude about this. I like to throw out the caveats. As always, do not use this as a finite mechanism for selecting teams. I do that with everything I say – not because I’m not confident, or care about being wrong – but more because I don’t believe that there’s anything finite in fantasy, or in the prognostication of any future event. I’m feeling like a list is going to both work better and be easier here:
- This has been updated through week 9. I will make an effort to update the top 10 Sharpes on a weekly basis, and hopefully be able to deliver the top “projected” Sharpes for the current week.
- I basically had to go through each player’s game logs and weed out the logs (he said logs…in a Butthead voice) in which I thought would throw off the Standard Deviation of that player. That comes with two warnings – 1) there’s no way I caught them all, and 2) it will have my bias. I ruled out Antonio Brown’s weeks without Roethlisberger, but some may argue that I needed to keep those in.
- The obvious sample size argument. Enough said there.
I truly believe that these values are going to help in cash game player selection. An argument can be made about them being almost as helpful in GPPs. If you are looking for players that consistently outperform their expected production based on salary and the riskiness of the player, then I think the Sharpe ratio of a player will be exactly what you are looking for.
