We often use the term “lottery ticket” when discussing fantasy football. Implicit in the idea that a player is a lottery ticket is the notion that they have a low cost and wild potentially unknowable upside.
But in real life lottery tickets are horrible investments. Some call them a tax on the poor. Yet if you applied the advice often given in fantasy football circles, lottery tickets would be your only real life investments.
Let’s think about how that advice usually appears.
- “I took Player X with in the third round. If he bounces back to his 2012 form, I just won my league.”
- “Player Y’s value will never be lower due to [insert factor of uncertainty. . .potentially a murder investigation] you should trade for him now. That’s how fantasy leagues are won.”
- “Player Z could miss the first four weeks of the season. But even 12 weeks of Player Z will be enough to win your league if you get him in the fourth round.”
The primary problem that exists with the statements above is that the market is probably already taking into consideration the piece of information that the discount is predicated upon.
In example #1, what expectation do you have that the player could return to prior form?
In example #2, everyone knows about the factor of uncertainty. The discount is wholly due to that factor. You need to have some informational advantage over the market for it to be better than a lottery ticket.
In example #3, there are two problems. First, you don’t have any more information than the broad market as to when the player will return. In fact everyone (including you) might be overrating the potential for the player to return on the timetable given. But there’s an additional problem – and this problem is rampant in fantasy analysis – it’s that the expectation for the player after they’ve returned from the injury is not 100 percent of remaining games. It’s maybe 90 percent of remaining games and that assumes that the injury involved doesn’t have some increased risk of re-injury.
The roster moves that are advocated by numbers one through three above are what I would call blind risk seeking. If gambling is what you like to do, then it’s totally fine to craft your strategies around moves like that. But it’s no different than walking up to a roulette wheel and placing a chip on the double zero space. The payout if your number hits is 35 to 1. You could turn $100 into $3500 by doing it. Your internal rate of return on that bet is going to be huge if it hits. But it’s still negative from a risk-adjusted standpoint. If you do it every spin for a few days straight you’re going to lose. But even if you eliminated the rake in this example, the bet is still just neutral.
The kinds of bets we’re looking for on RotoViz should contemplate the idea of risk, but they shouldn’t be blindly risk seeking. We should be looking for bets where the market is assigning a five percent chance of something happening and we can use a mix of logic and numbers to figure out that the true odds are closer to a 10 percent chance. We’re still going to lose nine times out of the 10 times we make the bet. Then when it does hit we’re going to get 20X our investment back. Or maybe the market is assigning a 25 percent chance and we believe the true odds are closer to 40 percent. Those are the kinds of margins we’re trying to squeeze out.
If some examples of these types of bets help, here is what I wrote about Josh Gordon last year:
If you offered me a chance to acquire a player for basically a fair price based on the prior year’s production, but that player had a range of outcomes that could be wildly better than the prior year outcome, I would take it. In fact I will be taking it with Gordon. You’re basically paying market for Gordon’s 2012 production when you take him at the mid-30s among WRs. Normally you might look at that and say that the player might be just as likely to underpeform their draft spot as they would be to overperform it. But if you adjust the odds by making them conditional upon using that pick on a 22-year-old player, and also conditional upon using that pick on a player with an elite Physical Score, then the odds change. Now you have an asymmetric bet, which is what we’re always looking to make in fantasy football.
That is perhaps a cherry-picked example. Actually, you can delete the word “perhaps” from the preceding sentence.
This year I think an example could be Geno Smith. The market is assigning him essentially no chance of being good. Some part of the market’s valuation is tied to the potential that Michael Vick could supplant Smith at some point in 2014. That’s fair enough. But even if you consider that possibility, the cost to draft Smith is incredibly low. He doesn’t even show up in FF Calculator’s ADP. But it’s not just the fact that Smith has a low ADP that makes him attractive. It’s the fact that because he picks up points with his legs, his upside might be really high. He averaged 23 fantasy points/game over the final four weeks of 2013, even while he faced the generally good CAR and CLE defenses along with the average MIA defense over that stretch.
I don’t want to make this post about Smith because Justin Winn has already made a very good case for the guy. But I do want to point out that the only way we can get ahead over the long run is to make bets where we have a positive expectation. We can’t just do it by making blind risk seeking bets.
That’s especially true in fantasy football where it’s easy to forget that each league actually requires your time to manage. Even if you have a positive ROI, you still have to figure out how much of your time it costs to get there. It’s totally possible that you end the season with a plus in the money column but you would have been better off flipping burgers.
Another way to think about the idea I’m talking about might be to ask yourself what your unfair advantage is. If it’s just having the fortitude to make high variance fantasy moves, that’s not good enough. If that’s all that was required, anyone could do it–and they could apply it more broadly than just fantasy football. They could buy penny stocks, they could buy lottery tickets, they could play keno, they could enter raffles, they could see every flop with suited connectors.
To get ahead over the long run you have to be able to look at all of the information in front of you, figure out where the market’s valuation is wrong, and then be relatively sure that your assessment of the probabilities is more correct.
To return to the above straw man examples I used, it’s definitely possible that you could make those moves and still be making the correct risk-adjusted bet. Specifically, #3 might sound very close to last year’s case for Rob Gronkowski. I’m not criticizing people who drafted Gronk last year and in fact I had him in a few leagues. But the math that will get you to make a correct decision as to Gronk’s value is not “Does he have a really wide range of outcomes? Y/N” – it has to be more than that.
The math to get you to make a correct decision in that case has to include alternative options, like other players you could select with your 4th round pick. It has to include an estimate of the range of games missed that Gronk might incur. It has to include the potential that he plays at some reduced level. It has to include a plan for that roster spot while Gronk is out of the lineup. All of these things can be estimated1, but my point is that the math is more than just the total amount of risk and then deciding that if it’s a lot of risk, it makes a great strategy.
The overriding consideration should always be the “risk-adjusted” part and not just the “risk” part.
- and based on a Twitter discussion I had with @adamharstad recently, I may take a shot at some sensitivity analysis to estimate them (back)