Sean Vostinar spacer
Lab Test: The Snake Draft

| More
More articles from Sean Vostinar

The Problem
Now and then in the FantasySharks.com forum, a topic will come up about the imbalance in the traditional snake draft. In one of these, Agenda42, my friend and fellow Über Leaguer (we had it first, ESPN!), took a sampling of one of our drafts as an example of how the top draft positions tend to get a disproportionate amount of value. In particular, that reversing the order in round two doesn’t compensate for the overwhelming value of the top picks.

Most of us sense this intuitively. The top two or three picks are almost always money in the bank, and big money at that. Top runningbacks in most formats, they’re consistent, have the best position value by a wide margin, and are very high scorers. Outside of these, the remaining players all tend to score less and have some defect, like injury history, a youngster in the wings, or the Cincinnati Police Department. After the first few owners get the cream of the crop, everyone else picks from these other guys where you have to actually think about risk versus reward.

This disparity makes intuitive sense, but we can do better – more than half a billion times better. Especially in a presidential election year, opinions demand facts, and for this study I’ve made some. Using suggestions from The Tank and my own ideas, I built a draft simulator to find disparities in different draft orders and possible alternatives.

The Geekly Details: May cause irritation to eyes and skin
There are as many different ways to simulate a draft pick as there are ways John Madden can say “the highest scoring team wins,” but we’re not looking for the best draft strategy. We’re looking for problems with the draft orders themselves. So long as we use a variety of draft strategies and players, and our simulated owners are consistent with each other in each draft, the biases will show themselves however the selections are made.

The draft board is based on Average Draft Positions (ADP) from MyFantasyLeague.com for the last 7 years, with each board capped off with random infrequently drafted players. Each draft test has 12 rounds and 12 teams, with about 25,000 drafts per test, giving us a nice, low margin of error. Players are valued using typical scoring with a starting lineup of 1 QB, 2 RB, 3 WR, 1 TE, and 1 PK. Team defense scoring varies widely, so we’ll skip them.

The draft strategies vary with how far an owner could reach for a player; half at random, half using a simple weighting scheme that favors picks higher in the ADP ranking. The owners’ reach varied in the number of rounds, from one and a half rounds to anyone on the board. One and a half rounds respects the fact we all have different valuations, but also the reality that if you’re really that far off from the consensus you should either wait and hope or trade down.

The Tests
All told, I performed about 4 million draft simulations for this article, amounting to over half a billion draft picks (actually, I mostly slept while my computer simulated the drafts, but I told it what to do). Using the different draft strategies described, I tested three different types of drafts:

Random Orders
Any good study starts with random tests. Drafts were run with 12 randomizations (every round), 6 (every other round), 4 (every other other round), and 3 (every – ah forget it).

Traditional Snake
Any good The old standby. And yes, in one set of tests with a loose draft strategy, the one and only Mr. J. J. Stokes did get nabbed at 1.01 – 51 times.

Modified Snake
Because the contention is mostly around the value in the early first round, most were Slanted in favor of the teams at the end of the original order, giving team 12 as many as 6 straight round-opening picks.
 
All else being equal, we’re looking for the draft with smallest gaps between the best and worst teams, and the fairest odds to have the best or worst team.

Snakes on a 2-Dimensional Plane

This first chart shows the average difference between the best and worst teams with ADP-weighted picks. Lower is better, and we can clearly see that the Traditional Snake (blue) gets worse the looser the owners get, confirming what we expected. We also see the best performer is the Single-Slant Snake draft (yellow), the modified snake with the reversed 3rd round {1-12, 12-1, 12-1…} and snakes thereafter. Any more consecutive slanting into rounds 4, 5, etc. clearly tips the scales too far. The same patterns were less pronounced but confirmed by the unweighted drafts (not shown).


This next chart depicts a sense of just how often a particular position is favored, like the first seat in the Traditional Snake, introducing some results from the random drafts for comparison.

We use random drafts to produce an ideal average to compare more typical results. You can clearly see from the gap between the Tradtional Snake (blue) and those tiny Randomized draft bars, that we have a long way to go. We again see our Modified Snakes generally outperform the Traditional Snake, but are still nowhere near the idealized random drafts.

A look at the detail behind the chart reveals that our Double-Slant Snake (green) goes too far. The extra two rounds simply take the advantage away from the front of the draft and give it to the end. On the other hand, the Single-Slant Snake (yellow) has no clear best seat, but still mostly favored the front of the draft.

From these charts, we can see that the Single-Snake Slant draft is an improvement, reducing the value gap and bias in favor of the front of the draft, but by no means the Holy Grail of draft orders we're looking for.

Normal is Bad
Finally, we turn to the randomized drafts. The average value gaps were very small, as much as 15 times smaller than the Traditional Snake, and the average draft position bias approached non-existent, but there’s a bigger problem.

At the risk of stirring grade school flashbacks, a bunch of random draft orders looks like something on every algebra teacher’s wall: A bell curve. In a Traditional Snake, everyone has an average pick of 6.5 (e.g. 1.01 and 2.12). In Randomized Drafts, a team can be first in every randomized round, for an average pick of 1.0! They call this curve a normal distribution and, here in the Stat Lab, normal is bad.

Among other things, this says that about 1 in 4 drafts (outside the yellow) will have owners crying foul because one or a few wound up on one end of the order or the other more often than everyone else. 1 in 20 will put someone at an extreme every time. The average is no worse than the Traditional Snake but how much better or worse is going to vary widely. The results of the random drafts are useful for comparison, but definitely not for your draft, unless your league happens to draft a few thousand times each season.

Conclusion
The Traditional Snake draft is inherently biased, favoring the teams at the front of the draft and hurting those at the end, growing with how loose your opponents are with their picks. The #1 and 2 spots are where you want to be, and be ready to work the wire if you draw #11. Our modified snakes were an improvement, but either went too far or not far enough, and don’t even think about random drafts. Still, the Single-Slant Snake is a slight improvement if you want to reduce the advantage of the front of the draft. This is encouraging, and at least warrants more research.

In the mean time, there are a few ways to work around the problems with the Traditional Snake. You can turn the problem into a pity prize, setting your draft order to the reverse of last year’s standings. You can turn it into a reward, maybe for whoever drafted the fastest last year. You can shrink the draft advantage with waiver advantages in reverse of the order. You could dump drafts altogether and go with an auction. Or you can recognize the problem, suck it up, and laugh that much harder when you win drafting out of the 11 spot.

There's a lot more data than I can show here, and you might have ideas for other draft orders. If you’ve got ideas for new draft orders to test, questions, or have feedback on this article, stop into the Article Discussions Forum at FantasySharks.com, or join the discussion on new ways of drafting in the Main Tank.