We’re trying something new on FUN tonight. 10 teams will be competing in a live IRI auction draft and drafting their own alliances. Each team will receive ₪100 to spend and create their own alliance with as many teams as possible.
You can even play along! After the draft, you can pick your own alliance paying the same prices that were set live with your own ₪100. This draft form will open shortly after the draft is completed and you will have until 3pm ET Thursday to submit your picks.
The draft was a lot of fun to watch. It was pretty clear to me that early teams in the draft were being overvalued, and teams later in the draft were being undervalued, so I set out to show this.
Here is a plot of all teams’ Elo versus their draft value:
Elo clearly has some shortcomings as a summary statistic of a team’s value for this draft, most notable in my mind is the lack of accounting for the additional RPs, but it’ll do fine as a starting point.
Next, let’s define a team’s overvalue as their draft value minus their predicted value according to the trendline above. Note that I did try other fits to the Elo vs draft value graph, as well as using Elo rank versus draft value. The best fit according to R^2 was an exponential fit using Elo rank, but the R^2 was only 0.482, so I stuck with the linear fit on Elo value for simplicity.
According to this definition of overvalue, at the extremes we have 2590 who was undervalued by 20 FUN dollars, and 5406 who was overvalued by 22 dollars.
We can then take this definition of overvalue and compare it to the draft order to get the following graph:
The first few teams drafted on average were overvalued by about 5 dollars, and the last few teams were undervalued by about 5 dollars. Essentially none of the first 12 teams selected were undervalued, and none of the last 20 teams selected were overvalued (excepting 1706 who was just a money burn). Don’t burn your money early in auctions people.
Woah now! Don’t burn you money early in this auction format. Nomination processes and roster sizes play massive roles in auction strategy. For the FUN auction, the nomination order was random (which led to teams like 1619, 2056, and 2590 looming as late pick-ups) and there was no limit on roster size (you could buy as many team’s as your budget permitted). The strategy a fantasy owner uses would change significantly if there was a roster cap or if there was a different method to nominate teams for bids.
This so very true, it completely depends on the format. We based ours off of the show that the software was originally intended for (Night Attack <> ) and talking to the developer they said that the format that they use works well for live shows and production value.
In other auction drafts, or if RNG put all the heavy hitters in front it might not have been a good strategy. We will be doing this again in the future (Chezy Champs anybody), so it will be interesting to see how that dynamic changes on each live draft. I’m just happy the software worked in amazing fashion for a live show :).
To build on this, I’ve done a ton of fantasy football auctions and even under the same exact format (scoring, league size, roster size…), the values can significantly vary. Sometimes the first few players go for way too much but other times they are amazing values as people are still figuring out values.
Just as a follow-up if anyone still cares. Here is a plot of every team’s draft position compared to their “value” which I’ve defined as the points the team earned divided by their cost.
The four teams with the highest “value” by this metric were 1024, 2590, 2771, and 708, all of whom were selected in the second half of the draft. The positive slope still holds even if you remove these four outliers.