23 Aug 2013
I've gotten a few very technical e-mails this year detailing problems that people had where they don't like the results that come out of the auction functionality in KUBIAK. I just wanted to say a little bit about that.
The ranks in KUBIAK are not necessarily built to exactly equal pure value over baseline. They are built to try to simulate what rankings look like when people really draft in real leagues. That's why you have things like the waiver adjustment to move kickers and defenses down the list. We also know that it doesn't fully add up to the correct amount of money to be spent at the auction. This is because of another change we make to try to make our auction values do more to reflect how players will be valued in real leagues, attempting to model the fact that (Auction Bid / Roto Value) tends to curve upwards in most auction leagues with the best players going for a premium. I'm sure there is a way to do this and make the money work out correctly. Unfortunately, I've never figured it out myself.
The equations are also built to cut the listed values for players who based on the projections should be taken for bench roles. Fantasy football doesn't work like fantasy baseball. Not every player you draft will end up scoring points for you. So in almost all fantasy football auctions, bench players are discounted. In almost all auctions, a good number of players at the end of the auction will go for $1 even though their real values are more like $4 or $5. We want to list things in such a way that it corrects for these issues.
Of course, we can't know how much discount is given to bench players in your league. That gets to the biggest reason why delivering perfect auction values is impossible. We simply cannot determine exactly what you should bit for players in your league. Understanding the bidding patterns and the aggressiveness of the other players in your league is far more important than having the best preseason projections, and there is nothing that Football Outsiders can do to help with that.
Now, all that being said, we know that the auction functionality can be improved. We are certainly open to ideas from readers on how to improve the equations to get closer to an even more realistic model of auction values. However, there are two issues with this.
1) We very specifically don't want to create a model that displays things solely based on a very simple equation that just looks at how many "points over baseline" are in the league, how much money is in the league, and then just divides the first by the second. That won't end up helping people. It will tell them to spend $5 on a backup defense.
2) Unfortunately, the time to make major changes in the KUBIAK application is February, not August. Good ideas now need to be stored away until the 2013 season is over.
If you disagree with how we try to model auctions, go ahead and have at us in the comments. We're open to suggestions.
18 comments, Last at 14 Sep 2013, 4:47am by kavin
How much do we tend to know after five weeks? Bill Connelly compares five-week data to full-season data to find out if we should be worried about TCU and Baylor.