In the Third Person
Solid title, I think. Probably intriguing enough to get you to click. And if you’re a long-time reader of the site, that title also probably tells you that Freedman is back with a new Labyrinthian.
Or at least he will be if he can wrap up this lame-*ss introduction.
It’s October 2016
We are now in the 10th month of 2016. I can’t even remember where I lived last year, who my friends were, or what my job was. 2015 was a long time ago.
(I do, though, remember that I should tell you to check out our Trends tool.)
Up to this point in the season, a lot of daily fantasy sports analysis has relied on data from last year. For instance, in Weeks 1-4 it was really easy to say that rostering tight ends who were facing the Saints was a good idea, because the Saints last year allowed the most DraftKings and FanDuel points to the TE position. You could say that, and almost no one would bat an eyebrow. By the way, “bat an eyebrow” is my combination of “bat an eye” and “raise an eyebrow.” Thoughts?
Of course, we are now entering Week 5, and the Saints have allowed the 10th-fewest DK and FD points to TEs on the season, so . . .
I’m not saying that the Saints over the course of an offseason somehow learned how not to suck against TEs. In Week 4, they allowed rookie Hunter Henry to turn seven targets into four receptions, 61 yards, and a touchdown. The odds are decent that over the remainder of 2016 the Saints won’t be all that great against TEs — but at this point we probably don’t need to look at 2015 data to make that estimation. It’s enough for us to know that this year the Saints have allowed the most points in the league and fourth-most yards passing.
Again, I’m not saying that 2015 isn’t relevant. It is. It’s relevant this year that the Lions last year allowed the most TDs to opposing TEs. It’s relevant because this year’s team isn’t all that dissimilar from last year’s team: Teryl Austin still ‘coordinates’ the ‘defense,’ which still has many of the same ‘players.’
But last year’s stats no longer matter more than this year’s stats. In fact, they now matter less. What’s more relevant to this season: That last year the Lions allowed a league-high 12 TDs to TEs? Or that this year — in just four games — they have allowed a league-high six TDs to TEs?
Last year matters in that it reveals to us a significant portion of the path. But it now matters less than the road that we’ve been traveling for the last four weeks.
Week 4
So let’s talk about 2016. H*ll, let’s talk about just Week 4.
Fullback John Kuhn scored three TDs yesterday, as did wide receiver Michael Crabtree. The Steelers offense erupted with the return of Le’Veon Bell — and everyone but him scored a TD. Julio Jones followed up perhaps the worst game of his career with one of the best receiving games in NFL history. The Patriots got shut out at home for the first time since 1993. And the Cowboys are 3-1.
Things fall apart in ways that we never expect — in ways that many DFS players do not take into account when creating their lineups.
We know that Julio is a great receiver, but (before this weekend) we failed to acknowledge that he had a real chance of getting 300 yards receiving in a game.
We know that the Steelers offense is great with Le’Veon, but we seem to forget that the Steelers have already shown us that it’s possible for them to score a sh*tload of points with Le’Veon on the field and without him actually scoring any of the points. See October 26, 2014.
We know that Kuhn scored three TDs in a single game against the Giants in 2010, that he scored five TDs over the final five games of that year, and that the next year he actually led all f*cking Packers running backs in TDs — we know that Kuhn has essentially made an NFL career out of being a goal-line vulture — if there were a vulture Hall of Fame, Kuhn might be a first-ballot candidate — and even then we still act as if he never has any chance of scoring a TD — and then when he does score we act as if a great probabilistic injustice has been committed against not only Mark Ingram, our DFS lineups, and our bankrolls but also the immutable laws of mathematics.
We get caught up in the lottery paradox. We act as if no one wins the lottery.
The Lottery
People win the lottery. It happens. I can get my head around that. It’s possible — even probable — that with enough lottery tickets out there one of them (and maybe even a few of them) will have the winning combination of random numbers. We can all see how it happens. Someone will win.
And yet we also tend to think of all people who buy lottery tickets as destined losers. Why is that?
Because we know that the odds of any one ticket being a winning ticket are infinitesimally small. Almost nil. And so, in practice, we act as if the odds actually are nil.
We create a spread between the odds and the way we act on those odds. We create opportunities that others can arbitrage.
In theory, the difference between 0.01 and 0.009 is almost nothing. Millions of people treat those numbers as if they are the same. And then a few people in the extreme make millions of dollars by capitalizing on the spread between the odds and the manner in which the market of millions of people act upon those odds.
Too many DFS players treat highly unlikely fantasy performances as if they are actually impossible. And they treat these performances as impossible in bulk. And then each week when a few unlikely events occur, they are amazed and frustrated and outraged at the fantasy gawds . . . and yet still hopeful that next time they can be the lucky ones to win the fantasy lottery.
And that’s their mistake. They treat as a game of chance that which is a game of skill. They believe that different randomness is all that’s needed for them to win the non-lottery of a guaranteed prize pool.
The Dutch Book
I could quote for you stuff that I’ve read on Wikipedia, but I don’t know why I’d do that when instead I could quote (at length) a passage from Nassim Nicholas Taleb’s Black Swan:
Subjective probability was formulated by Frank Plumpton Ramsey in 1925 and Bruno de Finetti in 1937. The take on probability by these two intellectual giants is that it can be represented as a quantification of the degree of belief (you set a number between 0 and 1 that corresponds to the strength of your belief in the occurrence of a given event), subjective to the observer, who expresses it as rationally as he wishes under some constraints. These constraints of consistency in decision making are obvious: you cannot bet there is a 60 percent chance of snow tomorrow and a 50 percent chance that there will be no snow. The agent needs to avoid violating something called the Dutch book constraint: that is, you cannot express your probabilities inconsistently by engaging in a series of bets that lock in a certain loss, for example, by acting as if the probabilities of separable contingencies can add up to more than 100 percent.
If you are a book-maker, you want to create your book in such a way so that bettors can’t lock in certain profits merely by taking action on all sides of the event in question. If you’re a bettor, you wouldn’t want to bet on all possible outcomes associated with an event if doing so would cause you to lose money regardless of the outcome. Those are basically the two sides of the Dutch Book coin.
Basically, the Dutch Book constraint mandates that your probabilistic views be coherent. Even though you view the world subjectively, you must do so in a way that doesn’t inherently make you a loser.
But too many DFS players employ Dutch Book logic.
Let’s say that there’s a small and unimpressive No. 3 WR on a slow-paced run-heavy team. This No. 3 WR doesn’t see a lot of snaps each game, and he’s not targeted all that often. Let’s say that before the season starts you project him to finish the year with just two TDs. Fine. That works.
Now let’s say that it’s Week 1. How many TDs are you going to project him to score? Zero. The same in Weeks 2-4. And the same all the way to Week 17. Each week for the entire season you will predict that this guy will score zero TDs even though you expect him to finish the season with two TDs. That‘s the fantasy football equivalent of the Dutch Book.
It’s Never Just One Bet
OK. It’s just one guy. It’s just one piddly two-touchdown guy. By the way, “piddly” is totally a word I got from Mama Lynn, so . . . sorry about that.
Anyway, it might seem like I’m making a big deal out of nothing. I’m talking about just one guy who’s seemingly insignificant. But here’s the thing:
- It’s never just one guy.
- Big profits are made on what many people consider insignificant.
I’m not talking about one guy, one projection, one action, one lineup, or one slate. I’m talking about a perspective.
You might think, “Hey, if this guy scores a TD in two random weeks, I’ll be 100 percent correct all of the other weeks.” In the words of Dwight Schrute: “False.” Technically that was just a word, but whatever.
With Dutch Book probabilistic thinking, you’ll be wrong (and ultimately unprofitable) 100 percent of the time.
In any given slate, you’re not making just one projection on one player. You’re contemplating tens of thousands of outcomes for hundreds of players. If you are making incoherent and imprecise conjectures for an entire slate of players, then you will always be wrong in the aggregate even when you are right on a number of players.
The First Rule
You can be a pretty decent DFS player simply by having an awareness of market dynamics. Last week, Julio Jones on FD was the cheapest he’d been in more than a year, his price had dropped $800 in one week, and he had a 100 percent Bargain Rating. He was the highest-rated WR in the Bales Player Model. You didn’t need an oracle to tell you that he was someone to roster in Week 4 (. . . although an Oracle did tell you). A simple understanding of the DFS market can serve you pretty well.
But if you want to be more than an average DFS player, at some point you need to start thinking about what will happen in the actual games. It’s not enough to know how the market values a player. You also must know what the player is actually worth that week. You must evaluate him. And the first rule of evaluation is to have accuracy and precision. Have perspectives that are coherent. Avoid the Dutch Book.
If you build your own models on FantasyLabs — and I encourage you to do so — you’ll probably notice when experimenting that the best models are those that heavily weight our Projected Plus/Minus metric. Why is that?
It’s because that metric measures both intrinsic worth and market value. It moves beyond market considerations and toward the knowledge that is almost invaluable. And Projected Plus/Minus is important because it relies on our projections. And through four weeks of the NFL season, our projections have been pretty f*cking good.
We didn’t predict that Kuhn would score three touchdowns in Week 4. But we did headline last week’s QB Breakdown with Matt Ryan, so . . . you know.
We’re not always going to be right, but I like our chances of being right often enough in high-leverage situations. Why?
We don’t have a Dutch Book.
———
The Labyrinthian: 2016, 89
This is the 89th installment of The Labyrinthian, a series dedicated to exploring random fields of knowledge in order to give you unordinary theoretical, philosophical, strategic, and/or often rambling guidance on daily fantasy sports. Consult the introductory piece to the series for further explanation. Previous installments of The Labyrinthian can be accessed via my author page.
