NFL DFS: What Have You Done For Me Lately?

“What have you done for me lately?”

It’s not just a common refrain and hit song from the 80s by Janet Jackson (Miss Jackson, if you’re nasty); it’s a way for thinking that colors many of our most important decisions. It’s what psychologist and behavior economics call recency bias, or the tendency to think that trends and patterns we’ve observed in the recent past will continue into the future. There’s nothing wrong, per se, with weighting more heavily recent performance in your decision making, as long as those recent results are actually more predictive. The problem arises from our tendency to do this instinctively, whether or not it leads to better predictions.

I think most DFS players realize that you shouldn’t base your forecasts solely on the previous week’s results, knowing that a single game is too small of a sample. But what about two games? Three games? Maybe even the last five games? Now we’re entering the territory of the mythical “hot streak”. It seems reasonable to place a higher value on players who have been more productive over a number of recent games. But it that the case?
I attempted to answer that question by looking back at the historical data (2000-2014).

cole1
In the visualization above, there are two sets of data on each plot, one (red) representing players’ average fantasy points over Weeks 14-18, against fantasy points scored in the Divisional Round. The other (blue) data set uses average scoring from Weeks 1-13. The upward-sloping lines in each of the four positional graphs represent the regression estimate, or the line that best fits the relationship between the data on the X and Y axes.

You can see that there isn’t a whole lot of difference in the regression lines for recent performance (Weeks 14-18) and what a player did earlier in the season (Weeks 1-13). While it makes intuitive sense that you want players who are on a hot streak, there’s reason to believe that we could be overvaluing players with strong recent performance. It’s just as important for predicting Divisional Round performance to look at the first 13 weeks of the season.

I believe that recent performance has an advantage over older results in predictive power, all else being equal. But that advantage isn’t enough to offset a much larger sample. Even looking at a five-game streak — like we see above — isn’t going to be enough to outweigh the older, but more substantial, sample of 13 games to open the season.

Now we know that we shouldn’t overvalue recent streaks, and we can also assume that others may be doing that. Therefore, a good way to differentiate yourself from the crowd and find potentially undervalued players is to filter for those that have seen downticks in recent performances.

Quarterbacks

Name Team Pos Wks 1-13 Wks 14-18 Diff
Tom Brady NE QB 24.4 13.3 -11.1
Aaron Rodgers GB QB 20.5 12.9 -7.6
Carson Palmer ARI QB 20.8 14.9 -5.9
Ben Roethlisberger PIT QB 19.1 14.6 -4.5
Alex Smith KC QB 17.3 14.2 -3.1

 

It seems like a distant memory when “Angry” Tom Brady was laying waste to opposing defenses and hanging lopsided scores on real and fantasy scoreboards. We can’t be sure exactly why Brady and the entire Patriots’ offense has slowed recently. Was it the loss of Julian Edelman, who is set to return “full throttle” this week? Have defenses adjusted to the Patriots’ pass-heavy attack? Is uber-stud and favorite target Rob Gronkowski still hurting? The hard-to-accept truth is that we don’t really know the answer. More likely than not, we’ll think up a rationale that fits the results, not something that had predictive power in advance.

Brady was the top signal-caller through most of the regular season, but he’s likely not the favorite of most DFS players this week with quarterbacks who finished the season strong like Cam Newton and Russell Wilson on the slate. We all know that Brady can put up a three-plus touchdown performance any game, and this week we have the opportunity to use him while others are looking elsewhere.

Running Backs

Name Team Pos Wks 1-13 Wks 14-18 Diff
Charcandrick West KC RB 10.0 7.5 -2.5
Spencer Ware KC RB 10.4 8.2 -2.2
James Starks GB RB 10.8 10.3 -0.5

 

There aren’t a lot of recent underperformers to choose from at the running back position, and Charcandrick West probably isn’t the first name that comes to mind when you’re thinking about winning a tournament this week. That said, I’m starting to see the upside of plugging West into your lineups. First, he’s the primary pass-catching back on a team that has won 10 straight games, and many in dominating fashion. To put it lightly, game script has not been West’s friend. Spencer Ware has benefited with the Chiefs often playing with the lead, and Ware is perceived as the safer option. The thing is, the Chiefs might be in unfamiliar circumstances this week: playing from far behind. The team is facing the Patriots on the road as five-point underdogs. We know from his play earlier this season that West can produce with volume, and if the Chiefs get down earlier, he could see plenty of it.

Wide Receivers

Name Team Pos Wks 1-13 Wks 14-18 Diff
Martavis Bryant PIT WR 19.5 10.4 -9.1
Larry Fitzgerald ARI WR 19.5 11.7 -7.8
Jeremy Maclin KC WR 16.4 13.7 -2.7
Randall Cobb GB WR 13.9 11.5 -2.4
Brandon LaFell NE WR 8.8 6.9 -1.9

 

Larry Fitzgerald began the season strong, but started losing targets to John Brown and Michael Floyd later in the year. On top of that, the output of the entire Cardinals’ passing game slowed as the year progressed. I still think Fitzgerald could be a great contrarian play this week. You don’t really know which one of the three Cardinals’ receivers will be the main option this week, so DFS players could fall back on price being the top consideration. Fitzgerald is at least $1,000 more on both DraftKings and FanDuel, likely lowering his ownership. In a bigger slate, price would be more of a consideration, but this week you should be able to fit a pricier Fitzgerald into an otherwise solid lineup.

Martavis Bryant has struggled recently, not due as much to target competition as mistakes taking him off of the field. It’s an unpredictable week for the Steelers with Antonio Brown ruled out and an injured shoulder possibly limiting Ben Roethlisberger. Bryant shockingly looks like a relatively safe option in the receiving corps, and his big-play ability might be undervalued this week due to recent underperformance.

Tight End

Name Team Pos Wks 1-13 Wks 14-18 Diff
Greg Olsen CAR TE 18.1 9.9 -8.2
Richard Rodgers GB TE 13.3 6.4 -6.9
Owen Daniels DEN TE 12.2 6.3 -5.9
Rob Gronkowski NE TE 18.6 12.9 -5.7
Darren Fells ARI TE 10.7 5.8 -4.9

 

Most of the hype this week at the tight end position has surrounded Travis Kelce. We all want to know what the preseason favorite will do with Jeremy Maclin out of the lineup. Plus Kelce put up eight catches for 128 yards just last week.
While others are drooling over Kelce, Greg Olsen could be a little under the radar following a few lackluster weeks to end the season. Olsen averaged less than 50 yards and only scored one touchdown in Weeks 14-17. More likely than not, Olsen will exceed those numbers this week, and it’s possible he matches the numbers he put up the last time he faced the Seahawks in Week 7: seven receptions, 131 yards, and one touchdown.

cole2
 

Here are two examples of lineups I put together for DraftKings (left) and Fanduel (right). They both build around a core of the players I highlighted above, a defense that been relatively poor recently (Denver), and a rookie wide receiver that could have a breakout game if target-hog Maclin is out. I may be contrarian, but I’m not insane: David Johnson is also in both lineups.

Have fun building your own lineups using our Player Models, and remember that fading a hot streak can be just as profitable as riding one.

“What have you done for me lately?”

It’s not just a common refrain and hit song from the 80s by Janet Jackson (Miss Jackson, if you’re nasty); it’s a way for thinking that colors many of our most important decisions. It’s what psychologist and behavior economics call recency bias, or the tendency to think that trends and patterns we’ve observed in the recent past will continue into the future. There’s nothing wrong, per se, with weighting more heavily recent performance in your decision making, as long as those recent results are actually more predictive. The problem arises from our tendency to do this instinctively, whether or not it leads to better predictions.

I think most DFS players realize that you shouldn’t base your forecasts solely on the previous week’s results, knowing that a single game is too small of a sample. But what about two games? Three games? Maybe even the last five games? Now we’re entering the territory of the mythical “hot streak”. It seems reasonable to place a higher value on players who have been more productive over a number of recent games. But it that the case?
I attempted to answer that question by looking back at the historical data (2000-2014).

cole1
In the visualization above, there are two sets of data on each plot, one (red) representing players’ average fantasy points over Weeks 14-18, against fantasy points scored in the Divisional Round. The other (blue) data set uses average scoring from Weeks 1-13. The upward-sloping lines in each of the four positional graphs represent the regression estimate, or the line that best fits the relationship between the data on the X and Y axes.

You can see that there isn’t a whole lot of difference in the regression lines for recent performance (Weeks 14-18) and what a player did earlier in the season (Weeks 1-13). While it makes intuitive sense that you want players who are on a hot streak, there’s reason to believe that we could be overvaluing players with strong recent performance. It’s just as important for predicting Divisional Round performance to look at the first 13 weeks of the season.

I believe that recent performance has an advantage over older results in predictive power, all else being equal. But that advantage isn’t enough to offset a much larger sample. Even looking at a five-game streak — like we see above — isn’t going to be enough to outweigh the older, but more substantial, sample of 13 games to open the season.

Now we know that we shouldn’t overvalue recent streaks, and we can also assume that others may be doing that. Therefore, a good way to differentiate yourself from the crowd and find potentially undervalued players is to filter for those that have seen downticks in recent performances.

Quarterbacks

Name Team Pos Wks 1-13 Wks 14-18 Diff
Tom Brady NE QB 24.4 13.3 -11.1
Aaron Rodgers GB QB 20.5 12.9 -7.6
Carson Palmer ARI QB 20.8 14.9 -5.9
Ben Roethlisberger PIT QB 19.1 14.6 -4.5
Alex Smith KC QB 17.3 14.2 -3.1

 

It seems like a distant memory when “Angry” Tom Brady was laying waste to opposing defenses and hanging lopsided scores on real and fantasy scoreboards. We can’t be sure exactly why Brady and the entire Patriots’ offense has slowed recently. Was it the loss of Julian Edelman, who is set to return “full throttle” this week? Have defenses adjusted to the Patriots’ pass-heavy attack? Is uber-stud and favorite target Rob Gronkowski still hurting? The hard-to-accept truth is that we don’t really know the answer. More likely than not, we’ll think up a rationale that fits the results, not something that had predictive power in advance.

Brady was the top signal-caller through most of the regular season, but he’s likely not the favorite of most DFS players this week with quarterbacks who finished the season strong like Cam Newton and Russell Wilson on the slate. We all know that Brady can put up a three-plus touchdown performance any game, and this week we have the opportunity to use him while others are looking elsewhere.

Running Backs

Name Team Pos Wks 1-13 Wks 14-18 Diff
Charcandrick West KC RB 10.0 7.5 -2.5
Spencer Ware KC RB 10.4 8.2 -2.2
James Starks GB RB 10.8 10.3 -0.5

 

There aren’t a lot of recent underperformers to choose from at the running back position, and Charcandrick West probably isn’t the first name that comes to mind when you’re thinking about winning a tournament this week. That said, I’m starting to see the upside of plugging West into your lineups. First, he’s the primary pass-catching back on a team that has won 10 straight games, and many in dominating fashion. To put it lightly, game script has not been West’s friend. Spencer Ware has benefited with the Chiefs often playing with the lead, and Ware is perceived as the safer option. The thing is, the Chiefs might be in unfamiliar circumstances this week: playing from far behind. The team is facing the Patriots on the road as five-point underdogs. We know from his play earlier this season that West can produce with volume, and if the Chiefs get down earlier, he could see plenty of it.

Wide Receivers

Name Team Pos Wks 1-13 Wks 14-18 Diff
Martavis Bryant PIT WR 19.5 10.4 -9.1
Larry Fitzgerald ARI WR 19.5 11.7 -7.8
Jeremy Maclin KC WR 16.4 13.7 -2.7
Randall Cobb GB WR 13.9 11.5 -2.4
Brandon LaFell NE WR 8.8 6.9 -1.9

 

Larry Fitzgerald began the season strong, but started losing targets to John Brown and Michael Floyd later in the year. On top of that, the output of the entire Cardinals’ passing game slowed as the year progressed. I still think Fitzgerald could be a great contrarian play this week. You don’t really know which one of the three Cardinals’ receivers will be the main option this week, so DFS players could fall back on price being the top consideration. Fitzgerald is at least $1,000 more on both DraftKings and FanDuel, likely lowering his ownership. In a bigger slate, price would be more of a consideration, but this week you should be able to fit a pricier Fitzgerald into an otherwise solid lineup.

Martavis Bryant has struggled recently, not due as much to target competition as mistakes taking him off of the field. It’s an unpredictable week for the Steelers with Antonio Brown ruled out and an injured shoulder possibly limiting Ben Roethlisberger. Bryant shockingly looks like a relatively safe option in the receiving corps, and his big-play ability might be undervalued this week due to recent underperformance.

Tight End

Name Team Pos Wks 1-13 Wks 14-18 Diff
Greg Olsen CAR TE 18.1 9.9 -8.2
Richard Rodgers GB TE 13.3 6.4 -6.9
Owen Daniels DEN TE 12.2 6.3 -5.9
Rob Gronkowski NE TE 18.6 12.9 -5.7
Darren Fells ARI TE 10.7 5.8 -4.9

 

Most of the hype this week at the tight end position has surrounded Travis Kelce. We all want to know what the preseason favorite will do with Jeremy Maclin out of the lineup. Plus Kelce put up eight catches for 128 yards just last week.
While others are drooling over Kelce, Greg Olsen could be a little under the radar following a few lackluster weeks to end the season. Olsen averaged less than 50 yards and only scored one touchdown in Weeks 14-17. More likely than not, Olsen will exceed those numbers this week, and it’s possible he matches the numbers he put up the last time he faced the Seahawks in Week 7: seven receptions, 131 yards, and one touchdown.

cole2
 

Here are two examples of lineups I put together for DraftKings (left) and Fanduel (right). They both build around a core of the players I highlighted above, a defense that been relatively poor recently (Denver), and a rookie wide receiver that could have a breakout game if target-hog Maclin is out. I may be contrarian, but I’m not insane: David Johnson is also in both lineups.

Have fun building your own lineups using our Player Models, and remember that fading a hot streak can be just as profitable as riding one.