Well I had a plan of what I wanted to write about this week and then I was struck by the same glaring discussion on my twitter timeline over and over again during the past two days. I'm taking it as a sign. The topic is about predicting player performance using statistics only or factoring in intangibles as well. It's MoneyBall vs. The Trouble with the Curve. Y'all know how we at RotoViz feel about this issue, right? So why bother exploring it again this week? Because the response and defense of a statistics based scientific approach to player evaluation can't just be "numbers are better and you're an idiot if you think otherwise".
Scientists strive to generate hypotheses that can be proven wrong. It's very difficult to prove something is always true, while it's often quite easy to prove something is not always true. The former requires you to have knowledge of every instance of the event, while the latter requires only one reliable instance that doesn't support the theory. The hypothesis to be tested is compared with a null hypothesis. Let's start with what intangibles might be worthy of considering in making our weekly predictions or lineup sit/start decisions. The ones I want to focus on involve motivation. I'll look at three cases of motivation that you might consider in determining player value: 1) being in a contract year, 2) playing a division rival, and 3) coming off a humiliating loss/poor individual performance.
I admit I can't tell you too much about the psychology of a team or an NFL player. The farthest I got in sports was varsity volleyball and softball, and for a weak school at that. So I don't know if a player's actual motivation, effort, or intent changes in any of those scenarios. Ultimately, though, that doesn't matter to us. What I can do is look at the situations that we predict may induce a change in motivation and look at the fantasy results in those situations. So regardless of whether being in a contract year actually influences the player's behavior or emotional state, we can evaluate if the situation has any effect on his performance from a fantasy standpoint.
1) The Contract Year. It comes up during every draft season, people write about who's due for a monster year owing to the fact that they hit free agency following the upcoming season. The motivation in this case: Money, arguably one of the biggest extrinsic motivators known to man. Play well, even beyond your potential, and a lucrative new contract awaits. The hypothesis generated by fantasy writers/players is that contract year players will have exceptional fantasy seasons. The null hypothesis is that they won't, and that could be framed as either they will perform similar to their past 2-3 seasons or that they will perform similarly to others in their position. There is simply no evidence to suggest that being in a contract year prevents injury related poor fantasy performance (Cutler, Vick, Bradshaw, Finley, Maclin to name a few) or idiocy related talent squandering (Britt). A couple of guys in the final year of their contract appear to be getting an opportunity to show their skills (Donald Brown, Ben Tate) and a couple are having the same kind of excellent years we've come to expect (Jimmy Graham, Eric Decker). Most players on this year's list are racking up pretty average fantasy stats in their contract year (MJD, McFadden, Boldin, DHB, James Jones, Brandon LaFell, Hakeem Nicks, etc). There is nothing to suggest that the null hypothesis is false in either of its iterations, nor to suggest that the interesting hypothesis of exceptional performance is supported. This contract year narrative will probably never die, but you can safely ignore it.
2) Division Rivalry. The storytellers of the NFL love their storylines, and rarely does one warm the hearts of loyal fan bases than the upcoming division rivalry tale. The theme is: Teams X and Y, division rivals, and players on Team X and Y, will play harder than usual, resulting in a close game. The null hypothesis states that the score of a division rivalry game will be no different, on average, than the scores of non-divisional games. As a fantasy player, you might expect (hypothesize) better performance from players on the lesser of the two rivals, for instance, Atlanta in last night's barn burner of a TNF game. You might also temper expectations of the favorite in a division game, believing that a tight game favors defense. So what does the data say about the score in a division game vs. a non-divisional game? The average score differential of a division rivalry is 9.2 (SD=6.9) over the past 7 weeks while the differential of non-divisional game is 12.1 (SD=9.6) points, a statistically non-significant difference. In other words, divisional games are decided by the same number of points, on average, as any other game. There is also no trend toward a difference in total points of divisional vs non-divisional games. Based on this data, I can't see letting a storyline like rivalry influence your fantasy lineup decisions.
3) The Bounce Back or Rebound Performance. This is an interesting one. The idea is that when we fuck up, we go all out to do everything we can to make up for it next time. The narrative is, he's due for a big game, he'll be looking to silence the critics in this one, after an embarrassing performance last week, so-and-so will be fired up and ready to run/throw/score touchdowns. The hypothesis is that following a particularly poor performance, athletes (teams) will bounce back with a great statistical day. I read a terrific piece a couple of years ago on this topic with respect to NBA players. It sticks in my mind whenever I'm deciding whether or not to roll with a really high priced guy coming off a dud game. The upshot is that the top players bounce back from bad days and resume their studly ways way more often than not.
Yesterday this tweet from Ryan Forbes appeared in my timeline showing the up and down season the New York Jets are having. They're due for a high scoring day, fellas. Kidding there, as you'll see. I looked at eleven games decided by a score of more than 20 points. In five cases, the team blown out went on to win the next week. In six cases, they lost the following week. Not great support for a team bounceback theory, despite the Jets season trend.
When it comes to individuals, I unfortunately don't have the manpower to investigate this on a player by player basis so I just used the PPR fantasy stats from one of my leagues for the top 25 scorers so far. QBs P Manning, Brees, and Wilson have had 4 instances of scoring low double digit points (11-16). In each case they bounced back for 23-32 fantasy points the following weeks. RBs Forte, McCoy, Lynch, and Moreno have at least doubled their dudliest fantasy totals the next week out a total of six times so far this year. WR Andre Johnson has had two bad weeks that he followed up with a strong showing, and DeSean Jackson is two out of three there. Several notables had their worst fantasy days just last week. Jamaal Charles, AJ Green, and Brandon Marshall will be looking to bounce back in week 12 (Moreno is in this group too, though it wasn't technically his worst outing). Others on my league's top 25 page have not had a bad outing, or they lost time due to injury. While certainly not comprehensive, at the highest level of fantasy football, it looks like the trend identified by John Paulsen for NBA stars holds true. Good defense happens. One bad game is likely just that, and I would start all of those guys with confidence this week.
Motivation is a fascinating subject, yet one that's difficult to study in a laboratory. Intrinsic motivation (feeling good, matching actions with values, achieving goals) is reported to be stronger than extrinsic motivation (money, pleasing others). Do players and coaches adjust their motivation and effort week to week? Do extrinsic factors like looking good on national TV, beating a division rival, or safeguarding a star reputation in the wake of a scrub performance influence effort or motivation? The answer to both questions is that it just doesn't matter if they do or not. Each and every one of us fantasy players, as much as we use and respect the numbers, is influenced by these motivational scenarios to some degree. Hopefully, these numbers show that we shouldn't be.
No comments:
Post a Comment