Use the Wisdom of Crowds to Draft the Best Fantasy Football Team


Use the Wisdom of Crowds to Draft the Best Fantasy Football Team

It’s that time of the year—when fantasy football fans take to making their draft picks. But rather than relying on blind intuition, utilizing the wisdom of crowds will put your picks above the experts almost every time.

In his remarkable book, “The Wisdom of Crowds” James Surowiecki tells the story of Francis Galton, a British scientist who in 1906 stumbled across a competition at a local fair. The goal was to guess the weight of an ox.

“Eight hundred people tried their luck. They were a diverse lot. Many of them were butchers and farmers, who were presumably expert at judging the weight of livestock, but there were also quite a few people who had, as it were, no insider knowledge of cattle. …

“[Galton] turned the competition into an impromptu experiment. …[A]mong other things, he added all the contestants’ estimates, and calculated the mean of the group’s guesses. That number represented, you could say, the collective wisdom of the Plymouth crowd. …

“Galton undoubtedly thought the average guess of the group would be way off the mark. After all, mix a few very smart people with some mediocre people and a lot of dumb people, and it seems likely you’d end up with a dumb answer. But Galton was wrong. The crowd guessed that the ox, after it had been slaughtered and dressed, would weigh 1,197 pounds. After it had been slaughtered and dressed, the ox weighed 1,198 pounds. In other words, the crowd’s judgment was essentially perfect.”

Here we’ll talk about using the wisdom of crowds when drafting your fantasy football team. The result? First, a better team—you’ll see below how fantasy football crowd beat ESPN’s top expert over the past three seasons. But it’s also substantially less work. No need to examine A.J. Green’s redzone target conversion percentage to see how it stacks up against Dez Bryant. No more tiers or formulas. Instead, you’ll be able to feel confident about your draft with 20 minutes of preparation.

Utilizing the wisdom of crowds does not mean polling thousands of people on every player in the NFL. Instead, we have effectively the same thing in results from thousands of actual fantasy drafts. The key metric is each player’s average draft position (ADP). This is exactly what it sounds like—if Adrian Peterson is the #1 pick in 500 leagues and #2 in another 500, his ADP is 1.5. The lower a player’s ADP, the more valuable fantasy owners collectively view him. We can view ADP as a consensus on many factors, e.g. off-season development, supporting cast, injury risk—even “I shouldn’t draft him here, I can probably wait and get him later.” It’s all priced in.

The hypothesis: ranking players according to their ADP gives us a cheatsheet that’ll rival any expert. To test it, let’s look at how ADP stacks up against the world’s most famous fantasy expert in Matthew Berry of ESPN.

The Crowd vs ESPN’s Matthew Berry

We’ll start by taking the difference between Berry’s rankings and ADP rankings for each position. So in the table below, last year Berry had Andrew Luck as the 12th best QB. The “crowd” had him at 9 (i.e. he was the 9th QB drafted on average) for a difference of 12 – 9 = 3. The larger the difference, the more the crowd valued a player relative to Berry.

Here are a few more players to get the gist of it:

Player Team Position ADP ESPN Diff
Luck, Andrew IND QB 9 12 3
Manning, Eli NYG QB 13 15 2
Manning, Peyton DEN QB 3 3 0
Newton, Cam CAR QB 5 5 0
Palmer, Carson ARI QB 20 17 -3

One way of checking who did better on average is to look at the relationship between this difference and actual points scored. If larger differences (ESPN minus ADP) are associated with players scoring more points on average, than that means the crowd’s judgment is better than Berry’s. If ESPN minus ADP is associated with fewer points, than Berry wins. We can find this relationship using a simple linear regression.

So what’s the verdict? Total, over three years (2011-2013) times four (QB, RB, WR, TE) positions, the crowd beat Berry 4-1. The rest were too close to call (technically, the relationship between ESPN minus ADP and points wasn’t statistically significant from zero). One way of looking at this: ADP does as well as, or better, than Berry 92% of the time. Not bad for a strategy that takes a fraction of the time to implement. Want to check out the analysis yourself? You can download the data and code here.

We don’t mean to pick on Berry, who seems like a cool guy and a good case-study in making things happen for yourself. He just happens to be the most prominent expert out there (plus his pre-season rankings from the past few years are helpfully still up online).

And obviously, just because this strategy beats Berry (by one metric) doesn’t necessarily mean it’ll beat every expert every year. These were the only ones tested. We would be willing to bet however, that many expert rankings would fare similarly. Really this should make any fantasy draft expert a bit nervous, because the strategy works on so many levels. It’s more accurate, takes less time to implement, and is free.

With that, let’s get into the nuts and bolts of implementation.

Get the Average Draft Position Data

The first step is getting our ADP data. There are a few options here. We personally like (that’s where the data that beat Berry above come from) because it includes results from actual drafts (a lot of ADP data comes from practice drafts, which isn’t as helpful – we want our crowd to have a real investment in their picks) and you can filter on things like league size and scoring system. So, as close to your draft as possible (we want to be using the most up to date information available) go to the MFL website.

Enter in your league parameters in the drop down boxes at the top. The important parameters to consider are:

Scoring System: Default is All Scoring Systems but choose to look at just leagues with either PPR Scoring Systems (point per reception) or Non-PPR Scoring Systems depending on your league rules. Most leagues on use PPR scoring, so that will likely give us rankings that aren’t as noisy, but enough leagues do Non-PPR scoring that that’s is okay to use too.

Start Date: This is important: restrict your analysis to the most recent available that’ll give you a big enough sample size (more is better but anything over, say, 75ish will work). We’re constantly getting new information—Kendall Hunter is out for the season, LeSean McCoy is looking great in camp—that might be helpful to incorporate into player rankings. The crowd will adjust accordingly, but we need to make sure we’re using the most recent drafts.

League Size: Most leagues on have 12 owners. That’s great news if you’re in a 12 person league yourself, but can mean that sample sizes get kind of small if you’re not. Luckily, this doesn’t matter as much. Just set this to 12 Franchise Leagues if you’re in a 12 person league; otherwise don’t feel bad about leaving it on All League Sizes.

Now you’ve got your cheatsheet. Either print it out by position, or for the time-crunched, select Non-IDP Positions in the position dropdown box and print out the whole thing at once.

During the Draft

As your draft progresses, your opponents will be making picks based on a mixture of personal opinion, ‘expert’ rankings, and blind panic. They’ll often draft players earlier than their ADP, which—by definition—means you’ll be able to draft other players later than their ADP. This is how you find value.

In general, ADP works better in selecting players within a given position (e.g. picking one RB over another) rather than selecting players between positions (e.g. picking some RB vs some other WR). The reasons for this are twofold:

First, ADP can change for certain positions depending on league parameters. For example, WRs will go higher in leagues that start three of them vs a league that starts two. Because our ADP doesn’t make a distinction between these types of leagues, the wisdom of crowd can get a little murky on inter-position comparisons.

Just as importantly, the value of a given position to you changes with the flow of the draft. If you’ve already drafted Aaron Rodgers and Ben Roethlisberger, for example, it might not make sense to draft Phillip Rivers, even if he does have the best ADP of anyone left.

Bottom line: In general, decide on a position, then use ADP to select the best player at that position. Like everything, there are exceptions. You might want to consider drafting the highest player available (regardless of position) when exceptional value is available, your draft depth is relatively balanced across positions, or you’re relatively new to fantasy football and having some trouble deciding amongst positions on your own. (If you are brand new to fantasy football, one addition to the “decide on a position, draft the player with the highest ADP” rule: don’t draft a Defense and K too early—in fact, it’s perfectly okay if these are your last two picks).

Finally, the wisdom of crowds is not meant to turn you into a fantasy robot. Drafting purely based on ADP is tough to beat consistently and requires very little preparation, but there’s room for independent thinking. You’re free to use this however you want, anything from a straight checklist to an anchor that keeps your draft generally on track.

Think a straight check-list is boring? Maybe, but you would beat have Berry—a guy who by all accounts is very good at his job and spends thousands of hours each year preparing—more often than not over the past three years.

Even if you’re completely sold on ADP, there’s still room for some flexibility. Often, ADP’s are very close (e.g. Giovani Bernard’s 17.81 and Montee Ball’s 17.91). In that case, if you have a strong preference a player—even if he has a slightly worse ADP—then go for it! Odds are the difference isn’t statistically significant anyway, and that’s part of the fun of fantasy football.

After the Draft and Next Steps

The wisdom of crowds gives us a best initial guess at player rankings coming into the season. This assures us of an above average draft; now we just have to keep the momentum rolling.

Things inevitably change once the season starts. Some players do better than expected, some worse. In short, observing actual games gives us data, and the results affect decisions about who to start, add, drop or trade for going forward. Unfortunately, that also means our ADP pre-draft rankings inevitably get out of date.

Ideally, leagues would have new drafts after every single week. That way we could update our weekly in-season rankings using the wisdom of crowds. Since this isn’t likely to happen any time soon, let me offer a next best alternative.

Bayesian Fantasy Football is specifically designed for this very in-season problem. We start with an initial set of pre-season rankings (based on ADP) and update them as data (game results) trickles in. While in some sense this is what every fantasy player does, we formalize the process using Bayesian inference, which is branch of statistics concerned with updating beliefs using new data. These ideas have enjoyed an explosion in popularity recently, in large part because of people like Nate Silver, whose tremendously successful 2012 political predictions involved Bayesian methods.

Although they can help if you find yourself a little lost as the season goes on, Bayesian based in-season rankings are not required to dominate your draft. As we’ve seen, that part requires very little time and absolutely no money. In about five minutes, you can incorporate the wisdom of crowds and get a jump on everyone else in your league and most experts. Happy Drafting.

Nate and Matt are the creators of Bayesian Fantasy Football, where they provide weekly, statistically-derived player rankings. Get in touch with questions or comments on their site.

Image adapted from otsill (Shutterstock) and kentoh (Shutterstock).

Want to see your work on Lifehacker? Email Andy.