Approaching Fantasy Football: Stats, Rankings and Projections
August 13, 2015 | Chet
Since I’ve been entrenched in cranking out the Team Outlook pieces this summer, I wanted to get on my soapbox break up some of the monotony a touch by sharing some thoughts on fantasy football approach in general because I believe a lot of generalizations are just presumed by both the audience and analysts still today and fantasy football is a game that is still blossoming. I’m going to do a few of these to close the summer that hopefully culminate with an actual “How to Approach Your 2015 Draft” piece, but this one will be about three things that all have relation to each other and how I incorporate each one in to my fantasy acumen.
Data and Stats
If you look up the business definition of analytics, you’ll receive this answer:
Analytics often involves studying past historical data to research potential trends, to analyze the effects of certain decisions or events, or to evaluate the performance of a given tool or scenario. The goal of analytics is to improve the business by gaining knowledge which can be used to make improvements or changes.
As someone who obviously incorporates a lot of data into their approach, it’s important to maintain the goal of trying to find trends that help narrow down the target so we’re more accurate in our conclusions. No matter what kind of player you are, at the end of the day fantasy football is about accumulating stats.
Now, how you decipher data in a way to help you predict tallying up those stats is an entirely different story. As someone who uses a ton of data in their approach I would ballpark roughly 90 percent of data as more descriptive than predictive. Rust Cohle flat circles and everything, but a lot of analysis is done looking more so at yesterday in a vacuum. That doesn’t mean that descriptive data isn’t still useful and it shouldn’t be dismissed. Don’t throw out the baby with the bathwater or something of that sort, but applying how yesterday factors into tomorrow is how to make those descriptive components work for you. The way I use that descriptive data in relation to broad trends is in a similar fashion to this piece on making last season work for you this year.
We’re at a special time in fantasy football analysis as we have waves of new information crashing in like never before. More accurate and descriptive game charting from sites like Pro Football Focus, Football Outsiders and the work being done by Matt Harmon is the next bridge in bringing the worlds of analytics and real football to a collision. All of those sites (and many others) provide unique information and while some points of that data may still slide into a subjective area, they are rooted in actual points of fact from the football field. The initial issue with a lot of that data still being relatively fresh is that we just don’t have a lot of empirical evidence in place for its application to fantasy football, which can create some noise as these elements are being infused. Sifting through all of that data can be overwhelming (while also very enjoyable) as their relevance is hashed out, but it’s important to also be cognizant that a lot of times in fantasy football analytics we’re just finding new ways to say the same things.
I personally don’t care for a lot of offseason analysis since it promotes looking forward and backwards at seasons in totality when fantasy football is a week to week game. Projecting big picture outlooks is more accurate than fragmenting data, but how and why points are scored on a consistency basis are still prevalent in my approach.
When it comes to data that I care about, I start with weekly usage stats more than anything. Volume associated with pass attempts, targets, touches and snap counts is where I begin. Simply, I want players on the field and ones that are being used while on it. Then I’ll scale that initial usage into asking “how and where on the field is this player getting that volume?” That may be average depth of target, carries and targets near the end zone or based on down and distance, etc. After I’ve established those starting points, then I’ll get into what players are actually doing with that volume or lack thereof.
I start there because while rate stats are definitely useful, it’s important to understand their relationship with volume and usage. Inefficiency naturally inflates volume just as efficiency shrinks it. If we’re going to cite that Demaryius Thomas had 39 red zone targets in 2015 (which were 18 more than he’s ever had in a season) you need to be aware they happened because he was the worst he’s ever been at converting those targets. A large part of why Aaron Rodgers has never sniffed 600 pass attempts his career is because he’s so damn good on the attempts he does make. We all want uber-efficient players on our roster, but I still believe in pursuing volume and usage firsthand because it has more Velcro than sharpshooting rate stats.
That Thomas anecdote ties right into regression. For rate stats, it’s important to remember that a football season is already inherently a small sample in its entirety. Regression to the mean is an overused phrase in today’s fantasy lexicon and in football the recoiling speed of the spring is volatile. We’re all constantly looking to make a play on regression on a micro and macro level, but finding balance in overselling and underpaying in fantasy football is much tougher than it’s described.
Rate stats I hold in importance with volume are success rates per attempt for running backs over yards per carry, adjusted yards per attempt for quarterbacks since it factors in every possible outcome for the pass into the attempt and the target in a vice versa manner. Also yards per target and points per opportunity for receivers, tight ends and backs. I’ll add layers on top of that usage and efficiency on a case by case basis from there, but that’s where I begin and end a lot of my own decision making from a data stance.
I often go back and read this piece from James Todd once or twice a year because it covers this topic better than I can, but housing more information doesn’t always improve your decision making. It aids increasing the confidence level of your conclusion. Even Nate Silver has expressed often that your model should reach its conclusion with as few data points as necessary. This is something that always sticks with me when digging deep, but for something as disparate as fantasy football, I believe the basic pressure points from one conclusion to the next is a constantly moving target. Football is a sport where even the strongest correlation coefficients leave a lot of room for variation. Missing small is the goal, but a simple data point can mean accurately predicting one thing for Golden Tate and an entirely different one for Calvin Johnson. That’s a Pandora’s Box of sorts, but it’s important to know that adding information to the issue isn’t always going to lead to a more accurate decision. I’ll always strive to provide unique data for consumption, but how that is consumed and applied is entirely up to the player.
Rankings and Projections
As today’s fantasy football has become more statistically saturated than ever, it has also become a game that beginning to develop a crutch for rankings and projections, which is a scary endeavor to me because we lose some big picture focus that I still believe is relevant to playing successfully. There are players like John Paulsen, Mike Clay and others who are able to successfully tie rankings and projections into an actual draft structure, which is still very necessary for that accuracy translating into success as a player.
Also, if you need to take what I am about to say here with a grain of salt, by all means I will provide you with the ammunition to do so. I am terrible at ranking players when it comes to accuracy challenges. Don’t mistake a lack of a scored accuracy in that regard as a lack of accountability, which can be unintentionally done, but it is a weakness of mine as fantasy analyst and something to work towards improving on. I also am very comfortable with that being a weakness of mine for a number of reasons.
The main one is I just don’t value rankings and projections with a high importance when it comes to approaching fantasy football because it promotes consuming and playing the game in a vacuum. This applies to dynasty, seasonal and weekly formats. When it comes to ranking and how it is being treated by the majority of players today, I find it inherently counterproductive because it’s taking one of the most complex, context sensitive games and distilling that game in a strictly linear fashion and I’ve never been able to work around that. It’s just where the foundation of my playing style begins.
I’m still in belief that with so much saturation on individual player analysis (my own included), that strong ideas still separate league winners. Players and the game of real football are constantly in evolution, but the game of fantasy football rarely moves. Sure, “picking the right players” is the goal and can trump any strategy, but I don’t have the arrogance to state that I will always pick the right players. I won’t and you won’t.
First off, creating projections, models and rankings is fun as hell. I love it and enjoy doing the exercise. But also, I don’t really incorporate projections into my fantasy approach much beyond running the floor, median and ceiling projections to take into account things I may have missed in my initial analysis. This is important because I tack on team projections at the end of all my Team Outlooks. Those projections are what I deem the most probable full season scenarios, but again, I don’t have the ego to blindly follow them alone and neither should you. The content above them is what is important because they cover the range of those outcomes and where my conclusions ultimately lie over those numbers at the bottom.
I had a follower ask me why I considered DeSean Jackson a WR3 and Pierre Garcon a WR4 although I had them similarly projected overall for the season. It was a great question and further illustrates my point that the game isn’t played in a narrow void. We all want to crank out the lineup that scores the most points, but there are too many variables in play for me tackle accomplishing that linearly on a weekly level. A player like Jackson has consistently posted overall WR2 numbers, but I believe is far better served as your third receiver weekly than your second. Jackson may have those overall WR2 totals, but if he’s your second option weekly, then you’re going to have weeks in which the lean valleys really hurt your overall weekly performance.
This is where team structure comes into play in unison with those rankings and projections that often is overlooked by the average consumer because it’s not explained on a list and rankings are handled and approached differently by nearly all of those who create them. I love a guy like Jackson as my third receiver in the fifth round or beyond in which I’ve already taken two legit receivers to start (for this example, say I’m at the turn and went WR/WR to start), because I already have a naturally grounded floor at the position with an inherently high ceiling. In those circumstances, Jackson’s floor is irrelevant to me because I’m strictly trying to use him to throw dirt on graves. On the flip side, a guy like Garcon, who has a stable floor based on volume and usage, has weekly relevancy in wide receiver lineups cemented in volatility. If you went RB/RB to start then grabbed T.Y. Hilton and Kelvin Benjamin, a guy like Garcon matters in your lineup. In the end, my answer to the follower was that Jackson still has supreme relevance over a player like Garcon even with similar counting stats because his ceiling is far less replaceable than Garcon’s floor is.
That’s an anecdotal example, but one that applies infinitely throughout mapping out your initial draft strategy and submitting your weekly lineup that cannot be correctly conveyed and applied strictly through a list or projection. The ways we present rankings don’t tell the entire story. I may rank a running back 18th overall in an accuracy contest that the remainder of the community has ranked as the 28th overall back, but that doesn’t mean I would draft that player that highly simply because I won’t always have to based on average price point it takes to acquire that player.
If we’re holding what the industry has defined accuracy as into high account, even the very best in that regard resemble a hit rate close to Sex Panther. Relying on the confidence placed in vacuum projections is a large part of the reason why I have never subscribed to value based drafting. Tying my entire draft plan to a tent of statistics over a four month span leaves a lot of variance open on the table and can lead to a lot of hallow weeks. The most important part of any question is accounting for where you’re wrong and that is something amplified in any field of prognostication. My goal is to win weeks like everyone, but emphasizing overall points scored first isn’t as pressing as chipping away on the supply and demand, opportunity cost, replaceability and team structure elements of the game.
The future posts in this series won’t be a preachy and will be more strategy based, but I wanted to share some of my approach since I have been spoon feeding player analysis to start the summer. Overconfidence is a pitfall many owners stumble into and these three elements enhance that overconfidence trait the most.