Avoiding Recency Bias
Recency bias is the tendency to think that trends and patterns we observe in the recent past will continue in the future. In reality a lot of these trends are really just random fluctuations and knowing how to avoid these are one of the most important things in daily fantasy sports.
If you perform a simple google search on the terms hot hand fallacy you will find hundreds of articles that have attempted to show that players actually get hot in games or that their hot streaks continue game to game. In reality it has been shown pretty conclusively this hot hand effect doesn’t really exist and the players perform approximately the same as their career or season averages in the near term.
The reason we can get tricked so easily by this phenomenon is because our brains are programmed to identify patterns in order to keep us safe. So whenever we see what we think might be a pattern we assume it is instead of random variations. One of the most famous stories about this is how Apple’s iTunes aren’t truly random because when Apple attempted to make them truly random users complained that they were hearing the same song multiple times in a row and that the system didn’t work. In fact, if any system was truly random you would have to hear the same song multiple times in a row quite often if you listened long enough.
There are two really important reasons why we need to avoid recency bias in our DFS selections
Player’s prices strongly reflect the performance in their last game. For example, Tom Brady’s price increased by $500 this week and Ben Roethlisberger’s increased by $600. Two weeks ago Doug Baldwin saw one of the biggest price increases of the season going all the way from $4,800 to $6,800 in one week! If these price increases are due to random variation we are going to be overpaying per point for these guys after their big games.
Players are owned a lot more heavily after their big game than before. This is a double negative especially in GPPs since these are two of our most key factors in determining values. For example, last week we saw Doug Baldwin’s usage rate increase all the way to his season high of 11% in GPP’s compared to 0% the week before when he was really a better play. All of a sudden his price was high and he was being picked a lot. Exactly what we want to avoid in GPP’s.
How do we avoid recency bias? The most important thing is to determine the reason the player played so well during his huge game (or down game). Is it something that is repeatable? If it’s not then I would recommend fading that player, especially in GPP’s. Let’s break down a couple of recent performances.
Early in the season Darren Sproles was the prime example of recency bias as he started the season on fire scoring a couple of really long touch downs. Long touchdowns, especially for running backs are very random. This is especially true when you are the secondary back on a team that has a star running back on their team. But after his big game against the Colts Sproles was one of the most highly owned players in all of Fanduel. If the player’s high score is due primarily to a couple of big plays I would highly recommend considering this random variation and not boosting their value for this.
Last week Tom Brady blew up against the hapless Bears defense. Brady threw for 354 yards on 35 attempts which is extremely efficient. His real explosion came from the fact that he passed for all 5 of the team’s scores. In an average game he would have had about 3 TDs instead of 5. Even so the main reason for Brady’s success is the return of Rob Gronkowski. Gronkowski had 9 receptions on 9 targets and turned them into 149 yards and three touchdowns. Because Brady is a completely different player when Gronkowski is healthy and Gronkowski is as healthy as he has been all year I expect Brady’s upward trend to continue. He won’t throw for 5 TD’s too often, and maybe not all rest of the year, but his price increase is warranted thanks to Gronkowski.
Last week saw another huge quarterback performance as Ben Roethlisberger saw an insane stat line of 522 yards and 6 TDs on 49 pass attempts! There are a couple of primary things that led to this explosion. First of all, Vontae Davis got injured very early in this game. The Colts game plan was to utilize Davis to cover Antonio Brown, one of the best WR’s in the whole NFL. With that part of the game plan destroyed the Colts didn’t really have a backup plan to contain Brown. But there are a lot of teams that can’t contain Brown. The second big thing is the emergence of a big wide receiver in Martavis Bryant. Up until the 28 minute mark of the Texans game the Steelers had been moving the ball with great efficiency but weren’t able to score TD’s. But now with a new big wide receiver that makes for an excellent red zone target combined with regression to the mean Roethlisberger’s TD rate should continue to increase. In this case, I am not ready to move all in on Roethlisberger but there are reasons to be optimistic.
Every week we want to seek out the true talent level of each player. We need to explore the real reason for the player’s success because just relying on the final stats is very dangerous. In doing so we want to avoid giving players credit for those things that aren’t really trend setting because it causes us to overpay for situations that aren’t real trends and to own players that are over owned. On the other hand if there is a valid reason for the change don’t hesitate to jump aboard or you may miss out on the next big money train like Golden Tate has provided us all year long.
Fanduel GPP plays
Below are my weekly GPP plays at Fanduel. Throughout the season I will be employing strategies mentioned in Week 1 about contrarian plays. My goal will be to highlight players owned less than 10% to give you a contrarian option unless the player is such a good value that you should play him anyway. But remember that if you think a player will score three times their salary at a higher rate than their ownership level they are still a good buy. It is also important to remember you don’t need to go contrarian on everyone one of your positions. Additionally I will take the approach outlined in my preseason article about the best approach to allocating money in my lineups. In this article, I showed that in general it was best to spend on studs at wide receiver and to a lesser extent quarterback and save your money at running back and tight end.
Brian Hoyer $6,200 1% - The Browns are one of the most run heavy teams this year so why would Hoyer be a good GPP quarterback? Simply because he gets the lucky draw of playing Tampa Bay this week. The Buccaneers are being passed on more heavily than any team in the league and teams are having great success in doing so as they are allowing 7.4 NYPA and 16 passing TD’s through only 7 games. Additionally, the Browns run game has been stifled the last two weeks averaging only 54 yards per game since Alex Mack’s season ending injury. This will force Hoyer to pass a lot more aggressively this week than he was asked to do earlier this year. Additionally, with Jordan Cameron out for the game Andrew Hawkins makes for a great stack option with so few passing options available.
Andre Ellington $7,700 7% - Teams have tended to be very pass happy against the Cowboys this year but that has partially been because of how well the Cowboys run game has dominated the game. With Tony Romo likely out the Cardinals will be able to stuff the box against Demarco Murray and force Brandon Weeden to beat them. On top of that the Cardinals have already been good at stopping the run allowing only 3.3 yards per attempt. This all adds up to giving the Cardinals a chance to establish the run and building a lead. The Cowboys have actually been pretty poor at stopping the run this year allowing 14% more than an average team. Add to that Ellington’s ability to be involved in the pass game and Stepfan Taylor’s injury which should make Ellington the goal line back and you get a great shot at a huge game.
The Giants Passing game – Rueben Randle $6,400 6%, Odell Beckham Jr Jr $6,600 13%, Larry Donnell $5,400 10%. The Giants passing game always makes for an interesting GPP play because of the inconsistency of Eli Manning. If good Manning shows up you could be in for a great effort. This week they get the Colts who just made the Steelers look like last year’s Broncos. A lot of that had to do with Vontae Davis’s injury and he is expected to be back. With Davis’s return I would tend to eliminate Randle but his low ownership, low price, and high target rate is going to leave him in play for me. Plus with any injury there is a chance that he reinjures himself by overcompensating or is just a step slower than normal and Randle is able to burn him for a couple of big plays. If Davis is able to shut down Randle then Beckham would be the primary beneficiary making him another highly volatile play. Finally, the Colts have had trouble stopping the TE this year which makes Donnell a great play as well. I expect a shoot out between the Colts and Giants on Monday night and want a piece of the Giants passing game. You can pair Manning with two of these guys or T.Y Hilton if you really want to go heavy on this game.
Zach Ertz $5,400 2% - Donnell is my preferred play at tight end but since I mentioned him above I will lean towards Ertz as my backup. Being owned at a rate of 2% for a player as explosive as Ertz is criminally low. I am almost always going to go with him at that price. Houston is basically average against the pass and tight ends. Ertz is also not playing a lot of snaps but his snap usage is very inconsistent. If this is one of the weeks where he is heavily used or even if he just catches a few TD’s in what should be a high scoring game going with Ertz to go against the crowd could pay huge dividends.