vickrey auction with single duplicate approximates
play

Vickrey Auction with Single Duplicate Approximates Optimal Revenue - PowerPoint PPT Presentation

Vickrey Auction with Single Duplicate Approximates Optimal Revenue Hu Fu Sikander Randhawa UBC UBC Chris Liaw (UBC) EC 19, June 2019 Setting bidders, single item # ~ # & ~ & # , & ,


  1. Vickrey Auction with Single Duplicate Approximates Optimal Revenue Hu Fu Sikander Randhawa UBC UBC Chris Liaw (UBC) EC โ€˜19, June 2019

  2. Setting โ€ข ๐‘œ bidders, single item ๐‘ค # ~๐บ # ๐‘ค & ~๐บ & ๐‘ค # , ๐‘ค & , ๐‘ค ' independent ๐‘ค ' ~๐บ '

  3. Bulow and Klempererโ€™s Theorem Second price (Vickrey) auction Revenue-optimal auction รผ Simple and prior-free โœ˜ Complex auction รผ Efficient allocation โœ˜ Requires prior knowledge โœ˜ May have poor revenue รผ Maximizes revenue William Vickrey Roger Myerson

  4. Bulow and Klempererโ€™s Theorem Second price (Vickrey) auction Revenue-optimal auction รผ Simple and prior-free โœ˜ รผ Efficient allocation โœ˜ รผ Maximizes revenue โœ˜ Theorem. [Bulow, Klemperer โ€˜96] Given ๐’ i.i.d. bidders, the second price auction with one additional bidder, from the same distribution, yields at least as much revenue as the optimal auction with the original ๐’ bidders. [assuming value distributions are โ€œregularโ€] Second price auction Optimal auction โ‰ฅ

  5. Bulow and Klempererโ€™s Theorem Second price (Vickrey) auction Revenue-optimal auction รผ Simple and prior-free โœ˜ รผ Efficient allocation โœ˜ รผ Maximizes revenue โœ˜ Theorem. [Bulow, Klemperer โ€˜96] Given ๐’ i.i.d. bidders, the second price auction with one additional bidder, from the same distribution, yields at least as much revenue as the optimal auction with the original ๐’ bidders. [assuming value distributions are โ€œregularโ€] Q: Is a similar result true when distributions are not identical? It does not work to choose an arbitrary bidder and recruit a copy. E.g., what if only Mario has a high value for mushroom?

  6. A non-i.i.d. version of BK Theorem. [Hartline, Roughgarden โ€™09] Given ๐’ independent bidders, the second price auction with ๐’ additional bidders, one from each given distribution, yields at least half as much revenue as the optimal auction with the original ๐’ bidders. [assuming value distributions are โ€œregularโ€] Second price auction ยฝ โ‹… Optimal auction โ‰ฅ

  7. A non-i.i.d. version of BK Theorem. [Hartline, Roughgarden โ€™09] Given ๐’ independent bidders, the second price auction with ๐’ additional bidders, one from each given distribution, yields at least half as much revenue as the optimal auction with the original ๐’ bidders. [assuming value distributions are โ€œregularโ€] Theorem. [Bulow, Klemperer โ€˜96] Given ๐’ i.i.d. bidders, the second price auction with one additional bidder, from the same distribution, yields at least as much revenue as the optimal auction with the original ๐’ bidders. [assuming value distributions are โ€œregularโ€] Two key differences: 1. Recruits ๐’ bidders instead of one . Approximation is necessary . 2. Revenue is approximately optimal. Better than ยพ is impossible. Q: How many bidders suffice for second price to be approximately optimal? Q: Can we recruit fewer than ๐’ additional bidders? What about one bidder?

  8. Main Result Theorem. [Fu, L., Randhawa โ€˜19] Given ๐’ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐œต(๐Ÿ) fraction as much revenue as the optimal auction for the original ๐’ bidders [assuming value distributions are โ€œregularโ€] . Recruit one of these. OR OR Second price auction ๐›(๐Ÿ) โ‹… Optimal auction โ‰ฅ

  9. Main Result Theorem. [Fu, L., Randhawa โ€˜19] Given ๐’ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐œต(๐Ÿ) fraction as much revenue as the optimal auction for the original ๐’ bidders [assuming value distributions are โ€œregularโ€] . Remark. Techniques can be extended to show that for auctions with ๐‘™ identical items and ๐‘œ unit-demand bidders, a ๐‘™ + 1 th price auction with ๐‘™ additional bidders yields at least ฮฉ(1) fraction of optimal revenue.

  10. Main Result Theorem. [Fu, L., Randhawa โ€˜19] Given ๐’ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐œต(๐Ÿ) fraction as much revenue as the optimal auction for the original ๐’ bidders [assuming value distributions are โ€œregularโ€] . Remark. Techniques can be extended to show that for auctions with ๐‘™ identical items and ๐‘œ unit-demand bidders, a ๐‘™ + 1 th price auction with ๐‘™ additional bidders yields at least ฮฉ(1) fraction of optimal revenue. Up to an approximation , BK theorem extends to non-i.i.d. setting with the same number of recruitments.

  11. Additional results Theorem. Suppose there are ๐Ÿ‘ independent bidders. Recruiting a copy of each bidder and running a second price auction yields at least ยพ fraction of revenue of the optimal auction with original ๐Ÿ‘ bidders. [assuming value distributions are โ€œregularโ€] Improves on the ยฝ-approximation and is tight . [Hartline, Roughgarden โ€™09] To prove this, we make a connection between the second-price auction with recruitments and Ronenโ€™s โ€œlookahead auctionโ€. En route, this gives a new proof of Hartline and Roughgardenโ€™s ยฝ-approximation result.

  12. Proof sketch of main result Theorem. [Fu, L., Randhawa โ€˜19] Given ๐’ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐œต(๐Ÿ) fraction as much revenue as the optimal auction for the original ๐’ bidders [assuming value distributions are โ€œregularโ€] . Theorem would be true if: 1. Second price for original bidders is approximately optimal. 2. Some bidder has high value with high probability. โ€ข Via a reduction to Bulow-Klemperer Theorem. Lemma. Given n distributions, at least one of the following must be true: revenue of 2 nd price auction is ฮฉ 1 โ‹… ๐‘ƒ๐‘„๐‘ˆ ; or 1. 2. some bidder ๐‘— has value ฮฉ 1 โ‹… ๐‘ƒ๐‘„๐‘ˆ with probability ฮฉ(1) . [assuming โ€œregularityโ€] Rev. of optimal auction.

  13. Overview of approach Lemma. Given n distributions, at least one of the following is true: revenue of 2 nd price auction is ฮฉ 1 โ‹… ๐‘ƒ๐‘„๐‘ˆ ; or 1. 2. some bidder ๐‘— has value ฮฉ 1 โ‹… ๐‘ƒ๐‘„๐‘ˆ with probability ฮฉ(1) . [assuming โ€œregularityโ€] Overview of approach: 1. We consider the โ€œex-ante relaxationโ€, allowing us to decouple interaction amongst the bidders. 2. โ€œRegularโ€ distributions have nice geometric properties which we exploit on a per-bidder basis. Ex-ante relaxation is common technique to obtain upper bounds. [e.g. Alaei et al. โ€˜12; Alaei โ€˜14; Alaei et al. โ€™15; Chawla, Miller โ€˜16; Feng, Hartline, Li โ€™19]

  14. ๏ฟฝ ๏ฟฝ Sketch of lemma Lemma. Given n distributions, at least one of the following is true: revenue of 2 nd price auction is ฮฉ 1 โ‹… ๐‘ƒ๐‘„๐‘ˆ ; or 1. 2. some bidder ๐‘— has value ฮฉ 1 โ‹… ๐‘ƒ๐‘„๐‘ˆ with probability ฮฉ(1) . [assuming โ€œregularityโ€] Suppose that case 2 does not hold, i.e. ๐’’ ๐’‹ = Pr ๐‘ค ? โ‰ฅ # & โ‹… ๐‘ƒ๐‘„๐‘ˆ โ‰ค # & for all ๐‘— . Using properties of regularity & geometry of โ€œrevenue curvesโ€ we show that C ๐’’ ๐’‹ โ‰ฅ 1 ? Simple Fact. Suppose we flip ๐‘œ coins, where coin ๐‘— has prob. of heads ๐’’ ๐’‹ โ‰ค # & and โˆ‘ ๐’’ ๐’‹ โ‰ฅ 1 . Then at least two coins are heads with probability ฮฉ(1) . ?

  15. Information requirements for recruitment Theorem. [Fu, L., Randhawa โ€˜19] Given ๐’ independent bidders, and assuming โ€œmild distribution knowledgeโ€ , there is an algorithm that decides a bidder to recruit so that the second price auction with an additional copy of that bidder yields at least ๐œต(๐Ÿ) fraction as much revenue as the optimal auction for the original ๐’ bidders. [assuming value distributions are โ€œregularโ€] In the paper, we give some examples of distribution knowledge which are sufficient for recruitment.

  16. Conclusions & Open Questions โ€ข We showed that recruiting a single bidder and running 2 nd price yields revenue which is at least # #D of optimal revenue. โ€ข Can this approximation be improved? โ€ข Impossible to do better than โ‰ˆ 0.694 . โ€ข If we recruit ๐‘œ bidders, best approximation is ยฝ and better than ยพ is impossible. โ€ข For ๐‘œ = 2 , the ยพ is tight. โ€ข Q: What is the tight approximation ratio for this setting?

Recommend


More recommend