Vickrey Auction with Single Duplicate Approximates Optimal Revenue Hu Fu Sikander Randhawa UBC UBC Chris Liaw (UBC) EC โ19, June 2019
Setting โข ๐ bidders, single item ๐ค # ~๐บ # ๐ค & ~๐บ & ๐ค # , ๐ค & , ๐ค ' independent ๐ค ' ~๐บ '
Bulow and Klempererโs Theorem Second price (Vickrey) auction Revenue-optimal auction รผ Simple and prior-free โ Complex auction รผ Efficient allocation โ Requires prior knowledge โ May have poor revenue รผ Maximizes revenue William Vickrey Roger Myerson
Bulow and Klempererโs Theorem Second price (Vickrey) auction Revenue-optimal auction รผ Simple and prior-free โ รผ Efficient allocation โ รผ Maximizes revenue โ Theorem. [Bulow, Klemperer โ96] 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 โregularโ] Second price auction Optimal auction โฅ
Bulow and Klempererโs Theorem Second price (Vickrey) auction Revenue-optimal auction รผ Simple and prior-free โ รผ Efficient allocation โ รผ Maximizes revenue โ Theorem. [Bulow, Klemperer โ96] 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 โregularโ] 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?
A non-i.i.d. version of BK Theorem. [Hartline, Roughgarden โ09] 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 โregularโ] Second price auction ยฝ โ Optimal auction โฅ
A non-i.i.d. version of BK Theorem. [Hartline, Roughgarden โ09] 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 โregularโ] Theorem. [Bulow, Klemperer โ96] 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 โregularโ] 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?
Main Result Theorem. [Fu, L., Randhawa โ19] Given ๐ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐ต(๐) fraction as much revenue as the optimal auction for the original ๐ bidders [assuming value distributions are โregularโ] . Recruit one of these. OR OR Second price auction ๐(๐) โ Optimal auction โฅ
Main Result Theorem. [Fu, L., Randhawa โ19] Given ๐ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐ต(๐) fraction as much revenue as the optimal auction for the original ๐ bidders [assuming value distributions are โregularโ] . Remark. Techniques can be extended to show that for auctions with ๐ identical items and ๐ unit-demand bidders, a ๐ + 1 th price auction with ๐ additional bidders yields at least ฮฉ(1) fraction of optimal revenue.
Main Result Theorem. [Fu, L., Randhawa โ19] Given ๐ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐ต(๐) fraction as much revenue as the optimal auction for the original ๐ bidders [assuming value distributions are โregularโ] . Remark. Techniques can be extended to show that for auctions with ๐ identical items and ๐ 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.
Additional results Theorem. Suppose there are ๐ 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 bidders. [assuming value distributions are โregularโ] Improves on the ยฝ-approximation and is tight . [Hartline, Roughgarden โ09] To prove this, we make a connection between the second-price auction with recruitments and Ronenโs โlookahead auctionโ. En route, this gives a new proof of Hartline and Roughgardenโs ยฝ-approximation result.
Proof sketch of main result Theorem. [Fu, L., Randhawa โ19] Given ๐ independent bidders, there exists one bidder such that the second price auction with an additional copy of that bidder yields at least ๐ต(๐) fraction as much revenue as the optimal auction for the original ๐ bidders [assuming value distributions are โregularโ] . Theorem would be true if: 1. Second price for original bidders is approximately optimal. 2. Some bidder has high value with high probability. โข 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 ๐๐๐ ; or 1. 2. some bidder ๐ has value ฮฉ 1 โ ๐๐๐ with probability ฮฉ(1) . [assuming โregularityโ] Rev. of optimal auction.
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 ๐๐๐ ; or 1. 2. some bidder ๐ has value ฮฉ 1 โ ๐๐๐ with probability ฮฉ(1) . [assuming โregularityโ] Overview of approach: 1. We consider the โex-ante relaxationโ, allowing us to decouple interaction amongst the bidders. 2. โRegularโ 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. โ12; Alaei โ14; Alaei et al. โ15; Chawla, Miller โ16; Feng, Hartline, Li โ19]
๏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 ๐๐๐ ; or 1. 2. some bidder ๐ has value ฮฉ 1 โ ๐๐๐ with probability ฮฉ(1) . [assuming โregularityโ] Suppose that case 2 does not hold, i.e. ๐ ๐ = Pr ๐ค ? โฅ # & โ ๐๐๐ โค # & for all ๐ . Using properties of regularity & geometry of โrevenue curvesโ we show that C ๐ ๐ โฅ 1 ? Simple Fact. Suppose we flip ๐ coins, where coin ๐ has prob. of heads ๐ ๐ โค # & and โ ๐ ๐ โฅ 1 . Then at least two coins are heads with probability ฮฉ(1) . ?
Information requirements for recruitment Theorem. [Fu, L., Randhawa โ19] Given ๐ independent bidders, and assuming โmild 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 ๐ต(๐) fraction as much revenue as the optimal auction for the original ๐ bidders. [assuming value distributions are โregularโ] In the paper, we give some examples of distribution knowledge which are sufficient for recruitment.
Conclusions & Open Questions โข We showed that recruiting a single bidder and running 2 nd price yields revenue which is at least # #D of optimal revenue. โข Can this approximation be improved? โข Impossible to do better than โ 0.694 . โข If we recruit ๐ bidders, best approximation is ยฝ and better than ยพ is impossible. โข For ๐ = 2 , the ยพ is tight. โข Q: What is the tight approximation ratio for this setting?
Recommend
More recommend