Your NFL Wins Pool Draft Order Needs Some Adjusting

Tweet Follow @eldo_co

---

This story was updated on August 16, 2019 to reflect further improvements to the NFL wins pool draft order. The updated version is even more balanced. To skip down to the new 10-person draft order, click here. To skip down to similarly balanced draft orders for pools with nine, eight, seven, or six players, click here.

​How to improve the "Bill Simmons draft order"

I was in my hometown a couple of weeks ago for an afternoon family event, so I floated a text to a couple of my buddies who live nearby to see what they were up to that night. As it turned out, a bunch of them were meeting at a bar a few hours later for their "fantasy football draft" — or at least what remained of it. 

After years on life support, their 14-year old fantasy league of 10 childhood friends — all dudes from the HS class of 2001, a year above me — had finally succumbed to the fact that we're in our mid-30s, most of the guys have little kids, and only Johan and Dan actually give a shit. (And Josh, but he'd never admit it.) 

So then what the hell were they doing there? And why show up at your friends' fantasy draft when they at least need to half-focus and having a couple of drinks / making fun of their picks / playing the role of court jester probably gets old by the draft's third round? My buddy McArdle's text revealed the answer: "We're doing a wins pool so the draft will be five minutes long. So you should come by if you're around."

Ah. A wins pool. That sounds nice. The rules? Ten participants each pick three different NFL teams — using a unique draft order, which is the focus of this article — and whoever has the three with the most combined wins at the end of the regular season wins the pool. (The same team can't be picked twice.) 

Most importantly, wins pools require little or no preparation and zero time or attention during the season.[1] My own hometown fantasy league flirted with folding for five-ish years in a row too, each time inching closer to death, twice requiring that we grab a semi-random when one of the originals dropped out. We also ended up looking to a wins pool as a low-key way to keep some semblance of camaraderie alive.

---

The draft is easy — but the draft order is not entirely straightforward

If you sat down with nine friends to start a standard 10-person NFL wins pool, it wouldn't take long to realize that a traditional snake draft doesn't work — at least not fairly. Ten people making three picks across 30 NFL teams means three draft rounds, which is an odd number. Player 1 with the 1st overall pick would draft 1st, 20th, and 21st, while Player 10 with the 10th overall pick would draft 10th, 11th, and 30th.

The sum of Player 1's draft positions is 41, and her average pick is 13.7. Meanwhile, Player 10's sum is 51, and his average is 16.7. There'd be a big advantage to picking early and big disadvantage to picking late.

---

The Bill Simmons draft order is fine — but we can do better

​To remedy the snake draft issue — and extol the virtues of the wins pool upon America, which he's done successfully — Bill Simmons shared a wins pool drafting system on Grantland in 2012. Simmons did not create the draft order, but it has more or less become known as the "Bill Simmons draft order." Here it is:

THE BILL SIMMONS METHOD
Player 1 — 1, 20, 26
Player 2 — 2, 16, 29
Player 3 — 3, 13, 30
Player 4 — 4, 18, 25
Player 5 — 5, 15, 27
Player 6 — 6, 19, 22
Player 7 — 7, 11, 28
Player 8 — 8, 17, 21
Player 9 — 9, 14, 23
Player 10 — 10, 12, 24

Simmons concedes that he has "no idea how the creator came up with these numbers." But the order seems balanced enough — and in the end, NFL wins are random enough — that there isn't much reason to wonder. If you're curious, the Simmons wins pool draft order attempts to make the sum (or average) of each player's draft position as equal as possible. Every sum is 46 or 47, and every average is 15.3 or 15.7.

---

Last year was an improvement — but we can do even better than that

​The "Simmons method" is better than the snake draft. But as originally detailed here in 2018, we were able to create a "fairer" wins-pool draft by basing the draft order on the sums and averages of historical win patterns, rather than the sums and averages of draft positions. It turns out that we can do even better.

Down in the comments section, readers "Brian Golden" and "Harrison" suggested looking at how many games NFL teams typically win based on their preseason Vegas win-total lines. The idea is that when we all sit down for our NFL wins pool drafts, those win-total over/unders are the best market indication of how each team will perform, and that a logical wins pool draft order should therefore generally follow suit.

Put another way, if wins pool drafts occur in descending order of preseason win-total lines — meaning the team with the highest win total gets picked 1st, the team with the 2nd-highest total gets picked 2nd, and so forth — then we could look back at how many games the team with the highest, 2nd-highest, 3rd-highest win totals (etc.) usually wins. A fair wins pool draft order should balance those projected wins.

I was a little hesitant at first, in large part because I wasn't convinced that people actually drafted their wins-pool teams with win-total lines in mind. After all, there's only a modest correlation between preseason win-total over/unders and actual wins (0.46 since 2002). They tell us much more about how good a team was in the previous season (0.75) than do about how good they'll be in the upcoming season. 

But the comments stuck with me, so I decided to dive back in. I don't have look-through to hundreds or thousands of wins-pool draft results — if anyone does, let me know — but among the samples I analyzed, there was a 90% correlation between preseason win-total lines and wins-pool draft results. Consciously or not, people draft as preseason win totals suggest they should. Brian and Harrison's suggestion holds up.

 

---

How to improve your NFL wins pool draft order — balancing projected wins

Based on trends since 2002 — the start of the 32-team era in the NFL — the team with the highest preseason win-total over/under wins 12.3 regular-season games. The team with the 2nd-highest preseason win-total line wins 11.0 games, the #3 preseason team wins 10.2 games, and down from there.

Overall, the top 30 NFL teams in terms of preseason win-total lines — the 30 that would be selected in a logical 10-person wins pool draft — go on to win 245.4 games. So to create a fair draft, we should structure the draft order so that each participant expects to win an average of 24.5 games across their three teams. 

Under the Simmons method, the person with the first overall pick (#1, #21, #26) has a major advantage. They project to garner 12.3 wins (#1), 6.9 wins (#21), and 6.6 wins (#26) — good for an expected total of 25.9 wins, or 1.4 wins above the equitable baseline. The person with the second overall pick (#2, #16, #19) also has a sizable advantage. In fact, every single player except Player #9 is at an advantage or a disadvantage.

Anything can happen, of course, and these discrepancies aren't going to make or break your draft. But by reshuffling the deck based on the methodology described above, we can create an NFL wins pool draft order in which every player projects to win almost exactly 24.5 games. Here's how the revised order looks:

Due rounding, some numbers will not appear to sum

Under the new arrangement, all advantages and disadvantages are eliminated, and every participant's baseline is within 0.1 wins of the 24.5-win average. The standard deviation between the 10 wins-pool participants falls from 0.758 wins to 0.096 wins — a major decrease that greatly evens the playing field.

Here's a stand-alone summary with downward changes in red and upward changes in blue:

​THE ELDORADO METHOD
Player 1 — 1, 28, 30
Player 2 — 2, 21, 24
Player 3 — 3, 18, 22
Player 4 — 4, 17, 20
Player 5 — 5, 15, 23
Player 6 — 6, 14, 26
Player 7 — 7, 11, 29
Player 8 — 8, 16, 19
Player 9 — 9, 13, 25
Player 10 — 10, 12, 27

I made it a point to still start the draft with a normal #1 through #10 order, making all of the changes elsewhere in the draft. There are other ways to arrive at a similar point — or perhaps to even flatten down the remaining 0.1-win discrepancies — but this order has produced the lowest standard deviation so far.[2]

---

The win-totals curve — and why Player 1's last two picks aren't as bad as they look

​Although I feel confident that this method produces the fairest draft order yet, I worry that it might harbor some obstacles to adoption — specifically when it comes to the optics of Player 1's picks. After all, it looks pretty shitty to have the 1st pick but then not pick again until the end of the draft, in spots #28 and #30.

But based on trends, the NFL team with the highest preseason win-total line projects to win 12.3 games, the team with the 28th-highest over/under projects to win 6.5 games, and the team with the 30th-highest over/under projects to win 5.9 games — good for a total of 24.6 wins between the three selected teams, which is essentially the spot-on projected average across all players. (It's actually a shade advantageous).

Having the 28th and 30th picks isn't all that bad because preseason win-total lines aren't super accurate and the wins curve is really flat in certain areas. The 28th pick might sound bad on paper, but as you can see in the chart below, the actual wins it yields aren't much different from picking elsewhere in the 20s.[3] 

That's not the case at the very front end of the curve. There's a large premium to having the first overall pick — if you take the team with the highest win-total line, results are more consistent and you're reliably looking at double-digit wins. The best way to offset that is to also give Player #1 the 28th and 30th picks.  

In the following chart, you can see how many regular-season games NFL teams win (y-axis, out of 16 games) based on their preseason win-total over/under ranking (x-axis, out of 32 teams) before the season.

So when you've hit the point in your fantasy football league when a couple of people are auto-drafting, a couple more are forgetting to set their rosters, only two or three are working the waiver wire, and two or three more think that the guys who were good five years ago are still good (usually me), it could be time to hang your glory on the wall and switch to an NFL wins pool. Or just do one anyway because they're fun.

 

---

Don't have 10 players? — don't snake draft — these draft orders balance projected wins

NINE PLAYERS
Player 1 — 1, 26, 27
Player 2 — 2, 18, 25
Player 3 — 3, 16, 21
Player 4 — 4, 14, 20
Player 5 — 5, 12, 23
Player 6 — 6, 11, 22
Player 7 — 7, 10, 24
Player 8 — 8, 15, 17
Player 9 — 9, 13, 19

Every player is within 0.1 projected wins of the average (25.2) except Player 1, who projects to win 25.4 games. But with the first overall pick and last two overall picks (#26, #27), this is the closest we can get Player #1 to the projected average. A traditional snake draft would only go back and forth 1.5 times, which is inherently unfair and does not account for the nuances of the wins curve. The revised draft order reduces the standard deviation from 0.8 to 0.1 projected wins. 

EIGHT PLAYERS
Player 1 — 1, 17, 19, 32
Player 2 — 2, 16, 18, 31
Player 3 — 3, 13, 23, 29
Player 4 — 4, 12, 21, 28
Player 5 — 5, 10, 24, 25
Player 6 — 6, 9, 22, 26
Player 7 — 7, 11, 20, 27
Player 8 — 8, 14, 15, 30

Every player is within 0.06 projected wins of the average (31.9). A traditional eight-person snake draft would go back and forth two full times — which may be tempting — but because of the uneven win-projections curve, it would create a major advantage for Player #1 (0.8 projected wins above average), some advantage for Player #2 (0.2), and disadvantages for all others. The revised draft order reduces the standard deviation of projected wins from 0.33 to 0.04.

SEVEN PLAYERS
Player 1 — 1, 19. 20, 22
Player 2 — 2, 13, 21, 24
Player 3 — 3, 11, 18, 27
Player 4 — 4, 9, 17, 25
Player 5 — 5, 10, 15, 28
Player 6 — 6, 8, 16, 23
Player 7 — 7, 12, 14, 26

Every player is within 0.04 projected wins of the average (33.3). A traditional seven-person snake draft would also go back and forth twice, but because of the uneven win-projections curve — and the fact that spots #28-32 aren't there to offset the front end of the curve — it would create an enormous advantage for Player #1 (2.2 projected wins above average), a large advantage for Player #2 (1.0), and major disadvantages for all others. (If you snake draft with seven players, Player #1 starts with a 3.3 projected-win advantage over Player #7.) The revised draft order reduces the standard deviation of projected wins from 1.23 to 0.03. 

SIX PLAYERS
Player 1 — 1, 12, 20, 23, 30
Player 2 — 2, 14, 16, 22, 26
Player 3 — 3, 11, 17, 19, 25
Player 4 — 4, 9, 13, 24, 27
Player 5 — 5, 7, 15, 18, 28
Player 6 — 6, 8, 10, 21, 29

Every player is within 0.02 projected wins of the average (40.9). A traditional six-person snake draft would only go back and forth 2.5 times here, which is inherently unfair and doesn't account for the nuances of the wins curve. The revised draft order reduces the standard deviation from 1.49 to 0.01 projected wins.

In all of the scenarios above, I made it a point to give each player a first-round pick. In certain instances, you can reduce the remaining (negligible) imbalances even further if you don't hold yourself to that. I'm sure there are other ways to arrive at similar results. If you want me to run this for five, four, or three players, please comment down below!

---

Footnotes

[1] The only instance I can think of in which an NFL wins pool might require some of your attention during the season would be if you were close to winning toward the end of the season and wanted to consider hedges. But that's optional.

[2] If we had enough historical data, we could probably begin to factor in other variables, like how often the team with the highest win-total over/under is actually taken first (and so on and so forth), plus other variance in draft-order results and actual wins. Somebody out there could probably then use these probabilities and deviations to create an even fairer draft.

[3] I suppose the same flatness and randomness argument could be spun back in defense of the Simmons method. As mentioned above, a few spots in something this random is not going to make or break your draft. But the goal here is to create the fairest draft possible with the data we have. And given the premium that exists for the player with the 1st overall pick, trends suggest that coupling the 1st pick with the 28th and 30th picks best balances and equalizes projected wins.

---

The main data sources for this article were pro-football-reference.com and sportsoddshistory.com. Data was compiled and analyzed by ELDORADO. All charts and graphics herein were created by ELDORADO.

---