Pith. sign in

REVIEW 3 major objections 6 minor 33 references

Auctions Meet Bandits: An Empirical Analysis

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper tries to establish that in a real sponsored-search auction where conversion rates are learned online, the exploration policy that maximizes platform revenue is much more aggressive than the policy that maximizes efficiency, and t

desk verdict Fresh empirical result on exploration as a pricing tool in auction-bandit settings; the headline 32% gain rests on a boundary prior and an untested behavioral invariance, but the qualitative mechanism is solid. read the letter →

arxiv 2508.21162 v1 pith:TVNJU7QG submitted 2025-08-28 cs.GT

Mohammad Rashid , Omid Rafieian , Soheil Ghili This is my paper · ORCID
classification cs.GT MSC 91B2662L05
keywords auctionsmulti-armedbanditsThompsonSamplingsponsoredsearchcold-startexplorationmarketthicknessrevenue-efficiencytradeoffqualityscoreprior
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that in a combined auction-and-bandit problem, the platform's revenue-maximizing exploration policy is far more aggressive than the efficiency-maximizing policy, and that the gap is driven by the auction side, not by learning alone. Using impression-level data from an Asian app store's Thompson Sampling second-price auction, it compares revenue and efficiency across different priors on new advertisers' conversion rates and finds the revenue-optimal prior is 0.1 while the efficiency-optimal prior is 0.002, yielding about 32% more expected revenue. The reason is that a high entrant prior moves the entrant into second place, compressing the score gap and raising the price the winner pays. This 'entrant-second' channel is absent in pure bandit problems. The paper also finds that customizing the prior by keyword and using a UCB-style algorithm can improve revenue and ease the revenue-efficiency trade-off.

What carries the argument

The central object is the Thompson Sampling Second-Price (TS-SP) auction mechanism, in which each advertiser's quality score q is drawn from a Beta distribution with prior parameters (α0, β0) for new entrants, updated only through observed wins and conversions, and in which the highest quality-adjusted bid b·q wins and pays the second-highest quality-adjusted bid per conversion. The mechanism's key working part is the entrant-second channel: an inflated entrant prior elevates the entrant's score to second place, compressing the gap between the top two scores and raising the payment extracted from the winner. This gap is operationalized as market thickness, defined as the average ratio of the

What would settle it

Run or observe the same sponsored-search auction with the entrant prior set to 0.1 in a field A/B test, and measure whether incumbents' bids fall in response to the higher cost-per-conversion caused by inflated entrant scores; a negative bid response would mean the simulated 32% revenue gain is overstated and the revenue-optimal prior would be lower.

Watch

Extended reading notes

Core claim

The paper's central claim is that the revenue-maximizing prior mean for an entrant's conversion rate in a Thompson Sampling second-price auction is 0.1, while the efficiency-maximizing prior mean is 0.002, and that switching from the efficiency-optimal to the revenue-optimal prior raises expected revenue by roughly 32%. The efficiency-optimal prior corresponds to the solution of a weighted bandit problem, where exploration only serves to learn the entrants' true quality. The revenue-optimal prior is far higher because it does additional work: it pushes the entrant's quality-adjusted score into second place, closing the gap between the first and second scores and thereby increasing the second

Load-bearing premise

The load-bearing assumption is that advertisers maximize per-period utility and bid truthfully under every counterfactual prior, including the revenue-maximizing prior of 0.1, even though the paper's test for strategic bid shading only covers the observed prior of 0.1.

Editorial extensions

If this is right

  • Platforms setting cold-start priors should expect the auction side to push optimal exploration well beyond what pure bandit learning would recommend, with the revenue-optimal prior roughly 50 times the efficiency-optimal prior.
  • The extra revenue from a higher entrant prior comes primarily from auctions where the entrant finishes second, not from entrants winning impressions, so revenue gains can be achieved without large misallocations in thin markets.
  • Customizing the prior by keyword shifts the revenue-efficiency Pareto frontier outward, yielding about a 5% revenue gain over the best uniform policy even in thick markets, and creating near-vertical segments of the frontier where revenue rises at negligible efficiency cost.
  • UCB-style exploration raises revenue relative to TS-SP at the same observed exploration level, because UCB maintains a positive quality-score bias for non-winners that raises second prices.
  • Because the exploration prior functions like an implicit first-price payment mechanism inside a declared second-price auction, high priors are a potential regulatory and transparency issue for auction platforms.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If a platform adopted the revenue-optimal prior in practice, incumbents might eventually learn to shade bids in response to inflated entrant scores, a behavior the paper only tests under the observed prior; modeling that dynamic response would likely lower the optimal prior below 0.1.
  • The entrant-second channel should generalize to any auction market where the seller learns a quality attribute of new entrants, so cold-start priors in markets such as cloud computing, freelance labor, or content platforms could similarly act as hidden pricing tools.
  • The near-vertical parts of the customized Pareto frontiers suggest a testable 'free lunch' policy: identify ex ante the keywords with strong incumbents and large second-score gaps, and raise priors only there, capturing revenue with almost no efficiency cost.
  • A transparency regulation requiring platforms to disclose their quality-score prior parameters would let outsiders estimate how much of the prior is learning-driven versus revenue-driven; comparing disclosed priors to empirically estimated entrant conversion rates would provide a direct audit.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper studies sponsored-search auctions at a large Asian app store, where the platform allocates a single sponsored slot through a second-price auction with quality scores learned by Thompson Sampling with a Beta(1,9) prior. Using impression-level data with full bidder information, the authors estimate advertiser valuations per conversion (average bids) and conversion rates, then simulate counterfactual TS-SP mechanisms that vary the prior mean for entrant quality scores. They report an efficiency-maximizing prior mean of 0.002 and a revenue-maximizing prior mean of 0.1, with about 32% higher revenue under the latter, and attribute the gap to entrants finishing second and raising the winner's payment. They also analyze UCB-style exploration, decompose revenue by entrant rank and market thickness, propose keyword-specific prior customization, and provide a budget-cap robustness check.

Significance. If robust, this would be valuable first empirical evidence on the combined auction--bandit problem, showing that revenue-maximizing exploration can substantially exceed efficiency-maximizing exploration and identifying a concrete mechanism (entrant-second price pressure) that is absent from pure bandit problems. The paper's strengths include an unusually rich dataset with losing bids, a clearly specified simulation algorithm (Algorithm 1), a useful revenue decomposition, and a candid acknowledgment of limitations in footnote 5 and Section 9. The central quantitative claim, however, is currently overstated because the purported revenue optimum sits at the boundary of the policy space, and because the behavioral assumption underlying the simulation is only validated at the observed prior, not at the policies that would be needed to establish a true optimum. These issues are load-bearing for the headline 32% figure but are addressable through reframing and additional analysis.

major comments (3)
  1. [§6.1.1, Figure 8b, footnote 5] The paper states that 'the revenue-maximizing prior mean is 0.1' and uses this to claim a 32% revenue gain over the efficiency-maximizing prior of 0.002. Footnote 5, however, explicitly says revenue is expected to increase further above 0.1 and that the range was truncated because sufficiently high priors would create bid-shading incentives. Thus 0.1 is not identified as an optimum; it is the largest prior for which the authors are willing to maintain Assumption 1. The 32% figure is therefore a lower bound on the gain at an arbitrary truncation point, not the gain from the revenue-optimal policy. This wording affects the abstract, Section 1, and Section 9. The authors should reframe the result as a restricted-policy lower bound or extend the model to cover the actual optimum.
  2. [§5.2.1, §5.3 (Algorithm 1), §9] Assumption 1 and Proposition 1 (truthful bidding, ba,k,t = va,k,t) are the core input to Algorithm 1, which feeds these bids directly into allocation and payment rules. The empirical support for this assumption (Table 2 and Appendix A) is estimated only under the observed prior mean of 0.1. Footnote 5 and Section 9 acknowledge that at sufficiently high priors incumbents could have dynamic incentives to shade bids, letting entrants win so their scores are revised downward. If the true revenue-optimal prior lies above 0.1, the simulated revenue at that optimum would be computed under exactly the behavioral regime the authors doubt. This is load-bearing because the headline revenue results are conditional on an untested behavioral invariance for the policies needed to establish the optimum. The paper should either provide evidence on bidding behavior under more exploratory priors or substan
  3. [§5.3, Algorithm 1] The counterfactual simulation fixes the set of competing ads Ak,t as observed in the data. A higher entrant prior could change entry and participation decisions—attracting more entrants or altering incumbents' incentives to participate—which would feed back into the entrant-second revenue channel. Since the managerial recommendations in Section 8 are stated in terms of changing priors globally, the fixed-participation assumption should be explicitly discussed as a limitation or tested in a robustness check that varies the bidder set.
minor comments (6)
  1. [§4.1.1] 'installment' should be 'installation' (e.g., 'click/installment').
  2. [Figure 8] The x-axis is on a log scale but the captions do not say so; add axis labels and a note about the log scale to improve readability.
  3. [§7.1, Definition 3] Clarify whether t in log(t+1) is the global time index or a keyword-specific impression index; the exploration bonus's decay behavior depends on this.
  4. [§5.2.1, Eq. (7)] The abbreviation CPC is defined as 'average cost-per-conversion,' but CPC usually stands for cost-per-click; consider using a less ambiguous term such as CPA or spell it out.
  5. [References] Auer et al. (2002) and Agrawal and Goyal (2012) are listed in the references but not cited in the body; either cite them in Section 2 or remove them.
  6. [§5.2.2] The unbiasedness claim for the sample analogue estimator depends on random assignment; Appendix B shows IPS gives similar results, but this supporting evidence should be mentioned more prominently in the main text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: counterfactual priors are simulation outputs, not fitted targets; the boundary caveat is a validity limitation, not a circular reduction.

full rationale

The paper's derivation chain is: (i) under Assumption 1, truthful bidding, advertiser valuations are recovered as average bids (Eq. 8); (ii) conversion rates are recovered from observed wins and conversions (Eq. 9); (iii) Algorithm 1 uses these mechanism-invariant primitives to simulate TS-SP counterfactuals under different Beta priors; (iv) revenue and efficiency are computed from the simulated allocation and payment rules. The 'revenue-maximizing prior' (0.1) and 'efficiency-maximizing prior' (0.002) are outputs of this simulation, not parameters fitted to those revenue/efficiency targets. The reported 32% revenue gain is therefore a model counterfactual, not an identity forced by construction. Footnote 5 and Section 9 explicitly disclose that revenue is increasing at the upper end of the tested range and that the range is truncated to preserve the truthful-bidding assumption; this is an identified-set/robustness limitation, not a circular step. The self-citations (Rafieian 2020; Rafieian and Yoganarasimhan 2021) appear only in literature reviews and sampling conventions and are not load-bearing for the main results. No uniqueness theorem or posited functional form is imported from the authors' prior work. The central quantitative claims are thus self-contained conditional on the stated behavioral and identification assumptions, and no equation reduces to its own input.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central results rest on estimated primitives (mu, v) and on behavioral invariance of bids across counterfactual priors. The paper does not introduce new physical or theoretical entities. The main unverified assumptions are behavioral and are tested only under the status quo prior.

free parameters (3)
  • mu_a_k (conversion rate per ad-keyword) = estimated from data via Eq. 9
    The sample analogue estimator in Eq. 9, estimated from six months of data, directly determines conversion probabilities in the simulation, and thus both revenue and efficiency. Its accuracy is load-bearing.
  • v_a_k (valuation per conversion per ad-keyword) = estimated as average bid via Eq. 8
    Valuations are assumed equal to bids under Proposition 1 and are used as the bids in all counterfactual simulations. Any measurement error in this estimate propagates into every simulated revenue and efficiency number.
  • alpha0 (prior shape parameter) = fixed at 1 for all policies
    The paper only varies beta0 and keeps alpha0=1 in all counterfactuals. This restricts the family of Beta priors and is a modeling choice, not derived from data.
assumptions (6)
  • domain assumption Assumption 1: Bidders maximize per-period expected utility.
    In SS5.2.1, this behavioral assumption is the foundation for Proposition 1 (truthful bidding). The paper tests it only under the status quo prior; it is assumed to hold under all counterfactual priors.
  • domain assumption Assumption 2: Advertisers' private valuations do not change over time.
    SS5.2.1; needed to use time-averaged bids as valuations. Violations would bias the counterfactual bids.
  • domain assumption Assumption 3: Advertisers' conversion rates are fixed over time.
    SS5.2.2; standard in bandit literature and needed for the Bernoulli simulation draws. Short horizons make it plausible, but it excludes nonstationarity.
  • standard math Proposition 1: Truthful bidding (bid = valuation) is optimal in a second-price auction with quality scores.
    Derived from Assumption 1 and the single-slot second-price payment rule; standard dominant-strategy argument extended to quality-adjusted scores.
  • domain assumption The set of participating bidders per impression is fixed as observed (no entry/exit response to changed priors).
    Algorithm 1 exogenously takes Ak,t from data; a different prior might induce advertisers to enter or exit, which is not modeled.
  • domain assumption No dynamic manipulation of quality scores through bids (no deliberate loss to lower future entrant scores).
    SS9 acknowledges this possibility for more exploratory policies, but the counterfactuals assume truthful bids regardless of prior level.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Auctions Meet Bandits: An Empirical Analysis." pith.science (2026). https://pith.science/paper/TVNJU7QG

@misc{pith2026250821162,
  author       = {Pith},
  title        = {Pith review of: Auctions Meet Bandits: An Empirical Analysis},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TVNJU7QG}},
  note         = {Machine review of arXiv:2508.21162}
}
read the original abstract

Sponsored search positions are typically allocated through real-time auctions, where the outcomes depend on advertisers' quality-adjusted bids - the product of their bids and quality scores. Although quality scoring helps promote ads with higher conversion outcomes, setting these scores for new advertisers in any given market is challenging, leading to the cold-start problem. To address this, platforms incorporate multi-armed bandit algorithms in auctions to balance exploration and exploitation. However, little is known about the optimal exploration strategies in such auction environments. We utilize data from a leading Asian mobile app store that places sponsored ads for keywords. The platform employs a Thompson Sampling algorithm within a second-price auction to learn quality scores and allocate a single sponsored position for each keyword. We empirically quantify the gains from optimizing exploration under this combined auction-bandit model and show that this problem differs substantially from the canonical bandit problem. Drawing on these empirical insights, we propose a customized exploration strategy in which the platform adjusts the exploration levels for each keyword according to its characteristics. We derive the Pareto frontier for revenue and efficiency and provide actionable policies, demonstrating substantial gains for the platform on both metrics when using a tailored exploration approach.

Figures

Figures reproduced from arXiv: 2508.21162 by the authors.

Figure 1
Figure 1. Sponsored position allocation through the auction used by the app store [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Posterior mean of quality score over time for one trajectory [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Empirical Cumulative Density Function (ECDF) of Impression Count for Entrants [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Empirical CDF of the Difference Between Average Conversion Rates of Entrants and [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Histogram of the Market Thickness for Keywords [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Proportion of Each Place Won by an Entrant [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Revenue Share Based on Market Thickness and the Role of Entrants [PITH_FULL_IMAGE:figures/full_fig_p017_7.png]
Figure 8
Figure 8. Figure 8: (a) Efficiency and (b) Revenue Across Priors (Log Scale). [PITH_FULL_IMAGE:figures/full_fig_p027_8.png]
Figure 9
Figure 9. Figure 9: Points Represent Efficiency-Revenue Achieved if Using Different Priors [PITH_FULL_IMAGE:figures/full_fig_p028_9.png]
Figure 10
Figure 10. Figure 10: Revenue and allocation efficiency under revenue-maximizing and efficiency-maximizing [PITH_FULL_IMAGE:figures/full_fig_p029_10.png]
Figure 11
Figure 11. Figure 11: Decomposition of Total Revenue in Revenue-Maximizing and Efficiency-Maximizing [PITH_FULL_IMAGE:figures/full_fig_p030_11.png]
Figure 12
Figure 12. Figure 12: Pareto Curve for Customized and Uniform Priors Across Thickness Groups [PITH_FULL_IMAGE:figures/full_fig_p031_12.png]
Figure 13
Figure 13. Figure 13: Revenue-efficiency tradeoff for UCB (red) and TS(blue). Labels represent prior mean [PITH_FULL_IMAGE:figures/full_fig_p034_13.png]
Figure 14
Figure 14. Figure 14: Winner and nonwinner quality score bias the total payment by advertiser a on keyword k and day d, the cap for (a, k) is: B¯ a,k = max d Sd(a, k) We assume that (i) any daily budget, if present, is weakly above the largest observed daily spent for that pair within our …
Figure 15
Figure 15. Figure 15: Revenue (left) and efficiency (right) vs. prior mean under the baseline, unconstrained [PITH_FULL_IMAGE:figures/full_fig_p036_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 33 canonical work pages

  1. [1]

    The price of truthfulness for pay-per-click auctions

    Devanur, Nikhil R and Kakade, Sham M. The price of truthfulness for pay-per-click auctions. In Proceedings of the 10th ACM conference on Electronic commerce, 99--106, 2009

    work page 2009

  2. [2]

    Truthful mechanisms with implicit payment computation

    Babaioff, Moshe and Kleinberg, Robert D and Slivkins, Aleksandrs. Truthful mechanisms with implicit payment computation. Journal of the ACM (JACM), 62(2), 1--37, 2015

    work page 2015

  3. [3]

    Improved online learning algorithms for ctr prediction in ad auctions

    Feng, Zhe and Liaw, Christopher and Zhou, Zixin. Improved online learning algorithms for ctr prediction in ad auctions. In International Conference on Machine Learning, 9921--9937, 2023

    work page 2023

  4. [4]

    Internet advertising and the generalized second-price auction: Selling billions of dollars worth of keywords

    Edelman, Benjamin and Ostrovsky, Michael and Schwarz, Michael. Internet advertising and the generalized second-price auction: Selling billions of dollars worth of keywords. American economic review, 97(1), 242--259, 2007

    work page 2007

  5. [5]

    Position auctions

    Varian, Hal R. Position auctions. international Journal of industrial Organization, 25(6), 1163--1178, 2007

    work page 2007

  6. [6]

    Sponsored search auctions

    Lahaie, S \'e bastien and Pennock, David M and Saberi, Amin and Vohra, Rakesh V. Sponsored search auctions. Algorithmic game theory, 1, 699--716, 2007

    work page 2007

  7. [7]

    Revenue analysis of a family of ranking rules for keyword auctions

    Lahaie, S \'e bastien and Pennock, David M. Revenue analysis of a family of ranking rules for keyword auctions. In Proceedings of the 8th ACM Conference on Electronic Commerce, 50--56, 2007

    work page 2007

  8. [8]

    A dynamic model of sponsored search advertising

    Yao, Song and Mela, Carl F. A dynamic model of sponsored search advertising. Marketing Science, 30(3), 447--468, 2011

    work page 2011

Show all 33 references
  1. [9]

    A structural model of sponsored search advertising auctions

    Athey, Susan and Nekipelov, Denis. A structural model of sponsored search advertising auctions. In Sixth ad auctions workshop, 15, 5, 2010

    2010

  2. [10]

    Nonparametric estimation of sponsored search auctions and impact of Ad quality on search revenue

    Kim, Dongwoo and Pal, Pallavi. Nonparametric estimation of sponsored search auctions and impact of Ad quality on search revenue. Management Science, 2025

    2025

  3. [11]

    What makes them click: Empirical analysis of consumer demand for search advertising

    Jeziorski, Przemyslaw and Segal, Ilya. What makes them click: Empirical analysis of consumer demand for search advertising. American Economic Journal: Microeconomics, 7(3), 24--53, 2015

    2015

  4. [12]

    Monetizing online marketplaces

    Choi, Hana and Mela, Carl F. Monetizing online marketplaces. Marketing Science, 38(6), 948--972, 2019

    2019

  5. [13]

    Correcting audience externalities in television advertising

    Wilbur, Kenneth C and Xu, Linli and Kempe, David. Correcting audience externalities in television advertising. Marketing Science, 32(6), 892--912, 2013

    2013

  6. [14]

    Incorporating hidden costs of annoying ads in display auctions

    Stourm, Valeria and Bax, Eric. Incorporating hidden costs of annoying ads in display auctions. International Journal of Research in Marketing, 34(3), 622--640, 2017

    2017

  7. [15]

    Reserve prices in internet advertising auctions: A field experiment

    Ostrovsky, Michael and Schwarz, Michael. Reserve prices in internet advertising auctions: A field experiment. Journal of Political Economy, 131(12), 3352--3376, 2023

    2023

  8. [16]

    Revenue-optimal dynamic auctions for adaptive ad sequencing

    Rafieian, Omid. Revenue-optimal dynamic auctions for adaptive ad sequencing. Working paper: Cornell University, 2020

    2020

  9. [17]

    Online advertising: Heterogeneity and conflation in market design

    Levin, Jonathan and Milgrom, Paul. Online advertising: Heterogeneity and conflation in market design. American Economic Review, 100(2), 603--607, 2010

    2010

  10. [18]

    Buy-it-now or take-a-chance: Price discrimination through randomized auctions

    Celis, L Elisa and Lewis, Gregory and Mobius, Markus and Nazerzadeh, Hamid. Buy-it-now or take-a-chance: Price discrimination through randomized auctions. Management Science, 60(12), 2927--2948, 2014

    2014

  11. [19]

    Targeting and privacy in mobile advertising

    Rafieian, Omid and Yoganarasimhan, Hema. Targeting and privacy in mobile advertising. Marketing Science, 40(2), 193--218, 2021

    2021

  12. [20]

    Cold start to improve market thickness on online advertising platforms: Data-driven algorithms and field experiments

    Ye, Zikun and Zhang, Dennis J and Zhang, Heng and Zhang, Renyu and Chen, Xin and Xu, Zhiwei. Cold start to improve market thickness on online advertising platforms: Data-driven algorithms and field experiments. Management Science, 2022

    2022

  13. [21]

    Characterizing truthful multi-armed bandit mechanisms

    Babaioff, Moshe and Sharma, Yogeshwer and Slivkins, Aleksandrs. Characterizing truthful multi-armed bandit mechanisms. In Proceedings of the 10th ACM conference on Electronic commerce, 79--88, 2009

    2009

  14. [22]

    An empirical evaluation of thompson sampling

    Chapelle, Olivier and Li, Lihong. An empirical evaluation of thompson sampling. Advances in neural information processing systems, 24, 2011

    2011

  15. [23]

    Optimal auction design

    Myerson, Roger B. Optimal auction design. Mathematics of operations research, 6(1), 58--73, 1981

    1981

  16. [24]

    Bandit algorithms

    Lattimore, Tor and Szepesv \'a ri, Csaba. Bandit algorithms. Cambridge University Press, 2020

    2020

  17. [25]

    Optimal nonparametric estimation of first-price auctions

    Guerre, Emmanuel and Perrigne, Isabelle and Vuong, Quang. Optimal nonparametric estimation of first-price auctions. Econometrica, 68(3), 525--574, 2000

    2000

  18. [26]

    Nonparametric approaches to auctions

    Athey, Susan and Haile, Philip A. Nonparametric approaches to auctions. Handbook of econometrics, 6, 3847--3965, 2007

    2007

  19. [27]

    Counterspeculation, auctions, and competitive sealed tenders

    Vickrey, William. Counterspeculation, auctions, and competitive sealed tenders. The Journal of finance, 16(1), 8--37, 1961

    1961

  20. [28]

    A generalization of sampling without replacement from a finite universe

    Horvitz, Daniel G and Thompson, Donovan J. A generalization of sampling without replacement from a finite universe. Journal of the American statistical Association, 47(260), 663--685, 1952

    1952

  21. [29]

    Confidence intervals for policy evaluation in adaptive experiments

    Hadad, Vitor and Hirshberg, David A and Zhan, Ruohan and Wager, Stefan and Athey, Susan. Confidence intervals for policy evaluation in adaptive experiments. Proceedings of the national academy of sciences, 118(15), e2014602118, 2021

    2021

  22. [30]

    Off-policy evaluation via adaptive weighting with data from contextual bandits

    Zhan, Ruohan and Hadad, Vitor and Hirshberg, David A and Athey, Susan. Off-policy evaluation via adaptive weighting with data from contextual bandits. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2125--2135, 2021

    2021

  23. [31]

    Credible auctions: A trilemma

    Akbarpour, Mohammad and Li, Shengwu. Credible auctions: A trilemma. Econometrica, 88(2), 425--467, 2020

    2020

  24. [32]

    Finite‐time Analysis of the Multiarmed Bandit Problem

    Auer, Peter and Cesa‐Bianchi, Nicol \`o and Fischer, Paul. Finite‐time Analysis of the Multiarmed Bandit Problem. Journal of Machine Learning Research, 3, 215--256, 2002, https://jmlr.org/papers/v3/auer02a.html

    2002

  25. [33]

    Analysis of Thompson Sampling for the Multi‐armed Bandit Problem

    Agrawal, Shipra and Goyal, Navin. Analysis of Thompson Sampling for the Multi‐armed Bandit Problem. In Advances in Neural Information Processing Systems 25, 239--247, 2012, https://papers.nips.cc/paper/2012/hash/6aca97005c68f1206823815f66102863-Abstract.html

    2012

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.