Skill, Luck, and ADP

Thoughts on how simply being average can be a path to success in fantasy football.

One of the fun things about participating in a seasonal hobby like fantasy football for a long enough time is you start to get a feel for the natural rhythms. And one of my favorite rhythms is the annual ritual starting sometime in November of debating how much of fantasy success is “skill” vs. how much is “luck”.

Now, the polite answer to the question is “when you’re winning it’s mostly skill, and when you’re losing it’s mostly luck”. Playoff teams want to brag about their brilliant moves, while non-playoff teams prefer to gripe about their bad breaks.

But if you want to be rude and actually take a shot at answering the question, you need to first define what, exactly, we mean by “skill” and “luck”. I usually use the following definition: your final record equals .500 + (your skill factor) + (your luck factor).

Over a single league, the luck factor dominates. But if we could theoretically get a person to play in a million leagues, the luck factors would largely offset and their overall record would closely approximate their skill level.

(There’s a bit of noisiness here, too. Certain years can be exceptionally kind or unkind to certain strategies— two years ago was a terrible time to draft RBs early, while last year was a phenomenal time to. Ideally, you’d have someone play millions of leagues over millions of years to truly isolate genuine skill.)

But for now, let’s accept this basic frame. Final Record = (.500) + (skill) + (luck), and as the number of leagues increases, luck offsets more and more until at an arbitrarily large sample size, all that’s left is skill.

Platonic Idealism and ADP

I’m going to go off on a quick tangent, but it’s leading somewhere. Plato posited something called the “Theory of Forms”. Essentially the idea is that what we see in our lives is just a flawed reflection of the perfect ideal of that thing.

We might see an equilateral triangle, but it’s not a Platonic Ideal triangle, because the sides are not perfectly straight and the angles aren’t exactly 60 degrees when you carry them out to infinite decimal places.

Average Draft Position, or ADP, is a good illustration of this theory. The idea behind ADP is to get a perfect representation of where a player is actually being drafted. But the reality is a bit messy.

Consider our in-house ADP data, for instance. Footballguys does its best to provide an accurate reflection of average draft position, but it’s impossible to get a single perfect result. Le’Veon Bell is drafted second overall according to RTSports and Fantasy Football Calculator, but on MyFootballLeague, Bell goes third and it is Ezekiel Elliott who goes second.

There are many different league hosts, all of which display players in a different order, which impacts ADP. There are many different scoring systems, all of which impact ADP. There are many different league types, which impacts ADP. Quarterback ADP derived from casual work leagues will greatly overestimate how high quarterbacks are going to be drafted in industry insider leagues, for instance.

And even if we get past all of those issues, ADP remains a lagging indicator. It tells you where players have been drafted recently, which is not necessarily where they’re going right now. If Le’Veon Bell tears his ACL right this moment and you’re drafting five minutes from now, it doesn’t matter whether his ADP is second overall or third.

So ADP is a simple concept in theory, but can be messy in reality. It’s still a tremendously powerful tool, which is why so much fantasy football analysis centers around it. “This player is underrated according to ADP, this player is overrated, rookie running back ADP is too high because of Ezekiel Elliott’s success last year”, and so on.

But imagine for a second that when I talk about ADP I’m not talking about the flawed reality of ADP that we see on a day to day basis. Imagine someone had access to the Platonic ideal of ADP, that they had access to a resource that told them exactly where players were actually being drafted at that very second, on that site, in that scoring system.

And imagine that they didn’t just have Platonic ideal average draft position. Imagine during the season they had a source that gave them Platonic ideal average start/sit advice, and Platonic ideal average waiver advice.

In short, imagine that they were the Platonic ideal of an average player. Let’s name this hypothetical player “Plato”, just to make it easy. Every decision Plato makes at every moment is perfectly average in a way that’s only achievable in theory.

Average Performance and Success

Let’s go back to our frame that (Record) = (.500) + (skill) + (luck). Growing the sample size so that Plato is playing in a million leagues causes that to converge to (Record) = (.500) + (skill).

If all of Plato’s decisions were Platonic ideal average decisions, (with some basic failsafe rules like “don’t draft five quarterbacks in your first eight picks” to approximate actual real-world average drafting patterns), then in theory that should negate his skill factor entirely. He is neither skilled nor unskilled, he is instead perfectly average.

In that case, we should expect our formula to simplify to (Record) = (.500). Over a million leagues, we’d intuitively expect our perfectly average player to have perfectly average results.

But we’d be wrong. (Feel free to mentally insert “record scratch” sound effect here.)

This hypothetical Platonic ideal of a perfectly average player would have a winning record. This isn’t something I believe. It’s not something I’m pretty sure of. In fact, there might not be anything in fantasy football that I am more certain of: “do the best thing according to a perfectly ideal average of what all fantasy players are doing” is winning advice. Period.

Okay. But Why? And How?

Our baseline assumption is that our hypothetical average fantasy football player is equally likely to wind up with every single player. He’s not pursuing a specific strategy, he doesn’t have specific favorites, he’s just taking whoever has the highest ADP. If he plays exclusively in 10-team leagues, he should pretty much own every player in 10% of his leagues.

So if David Johnson has another massive year and carries a ton of teams to titles, our hypothetical owner should benefit 10% of the time. But if Antonio Brown is this year’s title-winner, instead, our hypothetical owner again benefits 10% of the time.

Okay, but every owner in a 10-team league has a 10% shot at the title by default, so our hypothetical owner having a 10% shot no matter who has a good year isn’t exactly a point in his favor.

But there is one key effect that’s been overlooked so far. Our hypothetical owner isn’t actually equally likely to own all players. He’s actually slightly more likely to own players who are drafted higher than players who are drafted lower.

For instance, consider the player with the #1 overall ADP. 10% of the time, our hypothetical owner will draw the #1 pick and draft him. But 10% of the time, he’ll draw the #2 pick, and some unknown percentage of those times, a different player will go #1 overall, leaving Plato to take the guy with the #1 overall ADP. So he’ll own the guy with the top ADP more than 10% of the time.

Basically, Plato is getting all of the same players as everyone else, and in the worst case he’s getting them at the exact same cost as everyone else, (if he’s drafting them right at their ADP). But the vast majority of the time, he’s getting them at a discount after they fall.

This is the equivalent of getting extra high-round picks. Sometimes he’s using his 4th round pick to draft a 3rd round player. Sometimes he’s using his 8th round pick to draft a 6th round player. And since, overall, higher-drafted players are better than lower-drafted players, this leaves him with a stronger roster than the average fantasy owner.

There is one kink in this reasoning; some players have more variation in their ADP than others. Per Footballguys ADP, Devonta Freeman currently goes as the 11th player drafted on average, while Jordan Howard goes as the 12th. In theory, Plato’s “average” team should wind up with more Freeman shares than Howard shares.

But if Freeman theoretically was drafted with the 11th pick in every single draft, while Howard went anywhere from 6th to 18th, he’d instead get a lot more Howard shares. The only way he’s getting Freeman is if he has pick 11 (10% of the time), and the first 10 picks are exactly the 10 picks predicted by ADP, (less than 100% of the time).

a single reach during that span means someone else falls and Plato takes someone other than Freeman at pick 11, leaving him with Devonta Freeman on fewer than 10% of his teams overall.

This should theoretically only matter if the players who have a lower variance in ADP were more likely to have good fantasy seasons. But there’s no compelling reason to believe that should be the case, so absent evidence to that effect, I remain quite positive that Plato would have a lifetime winning record in fantasy football.

Average is the New Best

In fact, I think “just draft the best player remaining by Platonic ideal ADP, (plus some basic rules like don’t draft three quarterbacks with your first five picks)” might just be the single best strategy in all of fantasy football.

By this, I don’t mean that at the end of the year this is the strategy that will produce the best record. It probably won’t. If there are 1000 strategies out there, and I add this one to the mix, there might be 300 strategies that prove to be more successful.

The problem is that we don’t know which 300 strategies those are. Not ahead of time, at least. 2017 could be a great year for a WR-heavy strategy like “ZeroRB”, just like 2015 was. Or it might be a terrible year for a WR-heavy strategy like “ZeroRB”, just like 2016 was.

But every year is a great year for “strive for the Platonic ideal of average decision-making”. It probably isn’t the best strategy, but there isn’t a single strategy I’m more certain is a good strategy.

If I were giving a million people fantasy football advice and my life literally depended on them collectively having a good season, this is the advice I would give.

And even if you don’t want to slavishly tie yourself to ADP, I think there’s a really key takeaway from this all. Sometimes reaching on players, (drafting a player when there are more valuable players left on the board, per ADP), winds up proving to be the correct decision. Any owner who “reached” on David Johnson was handsomely rewarded last year.

But collectively, the total outcome of reaching on players is negative. For every success there’s enough failures to more than offset. Everyone thinks that the reaches they’re making are the “good” reaches, but evidence tells us they’re more likely to be wrong than they are to be right.

There’s nothing wrong with having a strategy, or with having players that you love and players that you hate. But at the end of the day, remember that there are very, very few certainties in fantasy football.

But one of those certainties is that the easiest, surest way to make a profit is to grab the value that presents itself to you when it presents itself to you. It turns out that just taking the money and running is a skill, too.

More articles from Adam Harstad

See all

More articles on: Strategy

See all

More articles on: Timeless

See all