4 ms·

I have real world experience with a lot of different auction designs being deployed, including many variations of the VCG auction. The VCG auction is only good

by Taek 2y ago

I have real world experience with a lot of different auction designs being deployed, including many variations of the VCG auction. The VCG auction is only good in theory. In practice, most people don't know how much they actually value a thing, and when they are trying to play the VCG game they typically underbid and later have deep remorse. That's the most common mistake, but I've seen plenty of other mistakes as well. In other words, VCG auctions tend not to work well because the bidders often don't have a strong enough grasp over game theory nor a deep enough understanding of how valuable an asset is to them, and therefore the VCG usually generates suboptimal outcomes. In my experience, the auction that seems to work the best is the Ebay auction, where you get immediate feedback if you underbid, but you aren't allowed to see how high the other person is willing to go. Instead, you have to talk yourself into taking a risk on being left holding the bag if you choose to push the price up.

JumpCrisscross 2y ago
> VCG auctions tend not to work well because the bidders often don't have a strong enough grasp over game theory nor a deep enough understanding of how valuable an asset is to them, and therefore the VCG usually generates suboptimal outcomes May I hazard a guess that you worked on low-value recurring auctions? You're describing unsophisticated bidders unwilling to expend search costs. For them, yes, "the seller has an interest in providing participants with as much information as possible about the object’s value" both before bidding starts and during it, the latter due to the impact to bids being less than the overcoming of search costs. > the auction that seems to work the best is the Ebay auction, where you get immediate feedback if you underbid, but you aren't allowed to see how high the other person is willing to go Another comment highlights why this doesn't work for high-value items [2]. (They also assume bona fide bids, which isn't a problem if you can filter out shill bids.) [1] https://www.nobelprize.org/prizes/economic-sciences/2020/popular-information/ https://www.nobelprize.org/prizes/economic-sciences/2020/pop... [2] https://news.ycombinator.com/item?id=42389681 https://news.ycombinator.com/item?id=42389681
- Taek 2y ago
  > May I hazard a guess that you worked on low-value recurring auctions? It varies. In all cases, the value was below $20,000, in many cases the auction was non-recurring and >$1,000 (for example, specific web domains). I've seen enough smart people fumble a $1,000+ VCG auction that I'd be nervous to assume bidding would be any better at $100 million. I've also seen enough investors fumble $10m+ investment deals (not auctions though) to feel comfortable asserting that a higher value auction does not necessarily imply that the participants will be better at auction theory.

by Taek 2y ago

JumpCrisscross 2y ago

> VCG auctions tend not to work well because the bidders often don't have a strong enough grasp over game theory nor a deep enough understanding of how valuable an asset is to them, and therefore the VCG usually generates suboptimal outcomes May I hazard a guess that you worked on low-value recurring auctions? You're describing unsophisticated bidders unwilling to expend search costs. For them, yes, "the seller has an interest in providing participants with as much information as possible about the object’s value" both before bidding starts and during it, the latter due to the impact to bids being less than the overcoming of search costs. > the auction that seems to work the best is the Ebay auction, where you get immediate feedback if you underbid, but you aren't allowed to see how high the other person is willing to go Another comment highlights why this doesn't work for high-value items [2]. (They also assume bona fide bids, which isn't a problem if you can filter out shill bids.) [1] https://www.nobelprize.org/prizes/economic-sciences/2020/popular-information/ https://www.nobelprize.org/prizes/economic-sciences/2020/pop... [2] https://news.ycombinator.com/item?id=42389681 https://news.ycombinator.com/item?id=42389681

Taek 2y ago
> May I hazard a guess that you worked on low-value recurring auctions? It varies. In all cases, the value was below $20,000, in many cases the auction was non-recurring and >$1,000 (for example, specific web domains). I've seen enough smart people fumble a $1,000+ VCG auction that I'd be nervous to assume bidding would be any better at $100 million. I've also seen enough investors fumble $10m+ investment deals (not auctions though) to feel comfortable asserting that a higher value auction does not necessarily imply that the participants will be better at auction theory.