“The human understanding when it has once adopted an opinion (either as being the received opinion or as being agreeable to itself) draws all things else to support and agree with it.”
– Francis Bacon
It starts small.
You see a headline here. A tweet there.
Before doing any major DFS research your brain has already inherently started to formulate thoughts on trends, players, and teams to use.
Subconsciously you’re associating each one of these with past performance, current matchup, and how you can carve yourself an advantage.
Without even knowing it, you’re already confirming previous beliefs on how that particular player, team, etc. will perform before opening any spreadsheets or browsers.
Confirmation Bias is a fickle thing, my friend.
For those unaware, Confirmation Bias – in psychology and cognitive science terms – is a tendency to search for or interpret information in a way that confirms one’s preconceptions, leading to statistical errors. Dr. Renee Miller has done some tremendous work drawing parallels between these cognitive biases and the world of DFS.
The most dangerous part of this bias is the tendency to seek out “proof” that further backs up our preconceived notions, while discounting statistics or examples that do not support our ideas and projections.
In Week 10 this season, the Green Bay Packers faced the Detroit Lions at home as 11.5-point favorites. The Lions hadn’t won a game in Lambeau in their last 23 appearances there, dating back to 1991. This narrative-led history was trounced around all week as people found statistics confirming their play of James Starks as a big home favorite:
– The Lions gave up the third-most rushing yards and rushing touchdowns.
– Eddie Lacy was ruled out for this game, leaving Starks as the Packers’ primary rusher.
– Running backs that are 10-point favorites at home see an average of over three fantasy points boosted to their expected output.
These were all correct statements at the time, but were they just lazy analysis? The Packers had lost their last two games going into this matchup, while looking dreadful in each game. Their offense looked defunct and incapable of sustaining drives.
It took Aaron Rodgers a career-high 61 passing attempts to eventually find value in this 18-16 loss, as Starks faltered in this juicy matchup, finishing with just 43 rushing yards for zero scores. It took six catches by Starks to bail out his under-performing game and finish with a mediocre fantasy score.
How much does confirmation bias factor into our DFS research?
The Bad: Confirming Misleading Information
I’m very happy that we brought Matt Freedman on board here at FantasyLabs. I really enjoyed reading his first article Larry Fitzgerald, Postseason Success, and the Small Matter of “Representativeness. He approached a similar topic discussing how it was a popular move going into the Cardinals-Eagles Week 15 matchup to play Larry Fitzgerald.
In his six career games, Fitzgerald had averaged 6.3 receptions for 108.3 yards, 1.3 touchdowns, and 25.2 PPR fantasy points. He became a popular play that week based on the assumption that this trend would continue.
He ended the game with just three catches for 43 yards. Our tendency to blindly follow misleading information such as that has the potential (and very real possibility) to lead to disastrous results. Are six games over the course of 186 career games really indicative of what we could expect from Fitzgerald that week? No, of course they aren’t. And it was foolish to blindly accept that information and insert Fitzgerald into your lineups based on that information alone.
We can’t go into the week relying on other’s information to make our lineup decisions.
I see it all the time: statistics being manipulated to fit the author’s agenda to either shed light in a positive or negative way to enhance their arguments. We, as a smarter fantasy community, need to corroborate these statistics with more information to find out if they’re actionable data.
We don’t often take the time to find reasons to disprove our theories for inserting a player into our lineups. Make time. Don’t cling to the first bit of information that supports your claims — keep digging. You’ll have a much more efficient process when building lineups as a result.
The Good: Reaffirming Preconceived Notions
“Group think” is a dangerous thing.
It can lead to us settling on our earlier assertions without doing much investigating of our own. Twitter lends credence to a lot of this whenever we see a big time DFS player or analyst tout Player X for tonight’s matchup because of reasons Y and Z.
If we go into the week seeing these Y and Z reasons — and rely on them solely without our own research — we may fall victim to the aforementioned “bad” section listed above. However, if we find these statistics or trends for using Player X in our lineups lead us down a path to even more researching and reaffirming our choice to use them, we may uncover even better reasons to utilize that player.
Let’s use an example from last night’s NBA slate while it’s fresh in people’s minds.
Going into the night, I had wanted to play Westbrook on a short two-game slate with few players to select from. FantasyLabs NBA guru Justin Phan has produced terrific content and the Phan Model is where I begin most of my NBA lineup construction. His model had Russell Westbrook as the top-overall play of the night against New Orleans on FanDuel.
Confirmation bias achieved.
Let’s break it down a bit further as to the three reasons why I didn’t just lock Westbrook’s name then and there and move onto the next position.
1. Recent Play
I started by looking at how Westbrook is doing the past 10 games (Plus/Minus of +7.24), past month (+4.52), and on the season (+4.65). Westbrook has consistently played above expectation and was red-hot, exceeding his projected points in nine of the last 10 games. He entered the game with three triple-doubles in his last six games and the Thunder were projected to put up 115.5 points.
2. Matchup
The Pelicans have been slightly better than average against point guards, but nothing to cause trepidation. They’ve allowed opposing point guards to score 40.6 fantasy points per game but hadn’t face anyone of Westbrook’s caliber in quite some time. On a short slate like this we really just want to avoid major mismatches.
3. Upside
Westbrook had surpassed 50 fantasy points in seven of his last 11 games. He also leads the league in True Usage, an important statistic from an earlier article written by Bryan Mears on Why Usage Rate is a Misleading Stat in NBA DFS.
I found multiple reasons that backed up Westbrook as a smart play that were used in conjunction with Phan’s model. I was able to use his model to reaffirm my belief that Westbrook was a solid play as he ended up scoring 46.8 fantasy points on FanDuel, leading all point guards.
The Truth: It’s Out There
Good things in life are rarely handed out for free. If we want to build our bankrolls with any sense of stability and forward progress, we have to put in the work for each and every slate.
Determining which splits and statistics are actionable data or just noise can be the difference between a profitable DFS player and a very profitable DFS player.
The truth is out there and FantasyLabs’ tools, projections, and models can help you discover it.
