Pith. sign in

REVIEW 3 major objections 4 minor 55 references

AV-AIVAT: 74x Cheaper Agent Evaluation with Certified Anytime-Valid Stopping in Imperfect-Information Games

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read AIVAT corrections plus anytime-valid confidence sequences let agent evaluation stop the moment evidence suffices, keeping the guarantee intact: a median 74x reduction in hands needed at ±1 BB in HUNL, with exact certificates in Leduc.

desk verdict A solid anytime-valid variance-reduction paper whose advertised 74x HUNL certificate is exactly what the paper itself admits: asymptotic and descriptive, not exact finite-sample — but the components are real and the transparency is exemplary. read the letter →

arxiv 2608.06362 v1 pith:SBCF5OZ5 submitted 2026-08-06 cs.GT cs.AIcs.CLcs.LGcs.MA

Boning Li , Yu Chen , Longbo Huang This is my paper
classification cs.GTcs.AIcs.CLcs.LGcs.MA MSC 62L1062L1262F2591A40
keywords AIVATconfidencesequencesanytime-validinferencevariancereductionimperfect-informationgamesagentevaluationoptionalstoppingheads-upno-limithold'em
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

The paper solves a practical matching problem: deciding which of two agents is stronger by playing games is expensive, and nobody knows in advance how many games the verdict will require. Its proposal is to let the value function that powers AIVAT corrections be refit online during the evaluation — as long as it is fixed before the hand it scores — so variance reduction and sequential monitoring work together. On 71,439 paired heads-up no-limit hold'em hands across 15 configurations, the corrected stream has median 54x lower variance than the raw stream, and at the nominal 95% level with a ±1 big blind target it stops after a median 74x fewer hands under the asymptotic confidence sequence. Exact finite-sample certification is deliberately separated from that asymptotic screen: it needs an independently justified bound on corrected payoffs, which the paper derives structurally for Leduc hold'em (|Y| ≤ 117) but not for HUNL, where the declared 200 BB bound is taken from observed data and the EB-CS run is labeled descriptive.

What carries the argument

The load-bearing mechanism is the predictable AIVAT interface feeding two confidence sequences with different guarantees. At each of a fixed finite set of chance and evaluated-agent decision nodes, the evaluator must know the conditional action kernel p_{t,h}, must fix the enablement decision S_{t,h} before seeing the outgoing action, and must use a value function v_t built only from hands 1..t−1; under these conditions the correction telescopes into conditionally mean-zero summands, giving E[C_t | F_{t−1}] = 0 and hence E[Y_t | F_{t−1}] = µ. That single identity licenses refitting the value model during the run and plugs the corrected stream into the cited betting-based empirical-Bernstein construction, which becomes an exact confidence sequence once a data-independent bound B_Y on |Y_t| is supplied. The second mechanism is the width floor: because the EB-CS bets are capped at 1/2 and all numerator terms are nonnegative, the payoff-scale half-width obeys w_t ≥ 4B log(2/α)/t for every realization, so an exact interval converts a variance gain into earlier stopping only when the declared bound is tight relative to σ²/ε (Lemma 1, Theorem 1). The efficient primary interval is the asymptotic CS, whose width tracks realized variance, and Theorem 2 guarantees that a past-only value learner with sublinear variance regret reaches the oracle-value stopping horizon with no asymptotic delay.

What would settle it

Two checks would settle the claims. First, find any legal Leduc trajectory whose corrected payoff exceeds 117 chips, which would break the structurally derived certificate, or any AIVAT-corrected HUNL hand in the corpus whose absolute value exceeds the declared 200 BB, which would invalidate the descriptive EB-CS replay. Second, run the AsympCS as its own monitor on a zero-effect stream for 50,000 hands and count exclusions of zero: exclusion rates near the nominal 5% would support the asymptotic screen, while rates far above it would confirm that finite-horizon calibration, not the asymptotic guarantee, is its real operating characteristic.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that predictability is the only condition that matters for valid variance reduction in sequential agent evaluation: if the value function v_t used for hand t is measurable with respect to information available before hand t, the conditional action kernel is known at every enabled correction node, and correction enablement is fixed before the action is seen, then the AIVAT correction C_t has conditional mean zero (Proposition 1). The corrected payoff Y_t therefore keeps the target mean, and the bounded Empirical-Bernstein confidence sequence, rescaled by any independent almost-sure bound B_Y on |Y_t|, is an exact (1−α) time-uniform interval. The paper then shows on its HUNL corpus that this interface delivers a median 54x variance reduction and a median 74x reduction in hands needed to reach the ±1 BB target under the asymptotic CS, while the exact certificate's speed is governed by a deterministic width floor 4B log(2/α)/t set by the declared bound and the bet cap. It proves this in Leduc, where the tree structure yields the analytic bound |Y| ≤ 117 and the exact interval runs within about 1% of its floor; in HUNL the bound is taken from observed maxima, so that replay is presented as descriptive rather than certified.

Load-bearing premise

The exact finite-sample certificate stands on a bound on corrected payoffs that is justified independently of the data; the paper has such a bound only for Leduc (|Y| ≤ 117), while in HUNL the 200 BB bound is taken from observed maxima and the resulting EB-CS is labeled descriptive rather than certified.

Editorial extensions

If this is right

  • Evaluations of expensive interactive agents can be run until the evidence is in and stopped at a data-dependent time with the stated level intact, replacing fixed budgets that either overpay or underpower.
  • At roughly $0.07–$0.30 per evaluated hand for LLM poker agents, a median 74x reduction in hands to a ±1 BB verdict turns a budget question into a routine one for the asymptotic screen.
  • A published early-stopping claim is recheckable by a third party from the corrected payoff prefix and the stop metadata alone; in the constructed zero-effect audit, checking either CS at the reported stop screens out essentially all of the 61% false claims that a continuously monitored fixed interval produces.
  • Exact finite-sample certification is achievable in games whose structure supplies a payoff bound: in Leduc, |Y| ≤ 117 makes the EB-CS exact and its realized width sits within about 1% of the deterministic floor.
  • Refitting the value function on past hands only is asymptotically free: under sublinear variance regret, the stopping time matches the oracle-value benchmark while exact EB-CS validity continues to hold.

Reading between the lines

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

  • The 74x factor belongs to the asymptotic screen, not the exact certificate: on the same HUNL data the descriptive EB-CS replay shows only a 1.37x ratio, so claims built on the headline number should be read as screening-grade evidence, not certified.
  • The single step that would promote the headline result from descriptive to certified is a sample-independent almost-sure bound on corrected HUNL payoffs, analogous to the Leduc |Y| ≤ 117 certificate; the paper's own release protocol names exactly this requirement.
  • The width-floor analysis implies a design target: a confidence sequence whose range dependence adapts to the realized payoff scale would convert more of the 54x variance gain into earlier exact stopping, which is the direction the paper flags as its sharpest open problem.
  • The continuous-monitoring audit generalizes beyond poker: any leaderboard or benchmark that inspects a fixed-sample interval repeatedly is vulnerable to the demonstrated 61% false-positive effect, and publishing the monitoring rule with a time-uniform bound before the run is the corresponding fix.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper introduces AV-AIVAT, a protocol that combines AIVAT variance reduction with confidence sequences to enable anytime-valid early stopping in imperfect-information game evaluation. It proves that past-only updates of the value function preserve the conditional mean-zero property of AIVAT corrections (Proposition 1), giving an exact Empirical-Bernstein confidence sequence once an independent almost-sure bound on the corrected payoff is available. An asymptotic CLT-based CS (AsympCS) is used as the efficient primary stopping rule, and the paper characterizes when variance reduction converts into earlier stopping: a bet-capped EB-CS has a deterministic width floor proportional to the declared bound, while an asymptotic width benchmark shows the usual variance-ratio scaling. Empirically, on 71,439 paired HUNL hands from 15 LLM-agent configurations, AIVAT reduces variance by a median 54x and the AsympCS stopping-time ratio is a median 74x at the ±1 BB target; the HUNL EB-CS runs are labeled descriptive because no independent corrected-payoff bound is available. A structural bound is derived for Leduc, giving an exact finite-sample certificate there. The paper also proposes a release protocol for rechecking early-stopping claims and demonstrates the dangers of continuous monitoring with fixed-sample intervals.

Significance. The paper makes a valuable methodological contribution by drawing a clear distinction between asymptotic screening and exact certification, by proving that predictable online value functions preserve AIVAT validity, and by quantifying the width floor that limits how much variance reduction becomes earlier stopping for the EB-CS. The empirical corpus is substantial, and the paper is unusually transparent: it explicitly labels the HUNL EB-CS results as descriptive, reports finite-horizon calibration rates above nominal, and provides a detailed release protocol. The exact Leduc certificate is a genuine finite-sample anytime-valid result. However, the headline '74x cheaper with certified anytime-valid stopping' overstates the strength of the HUNL guarantee, since the 74x ratio is produced by the asymptotic screen and the HUNL exact certificate is not established. This is a framing and scope issue that is fixable within the manuscript's aims, but it is load-bearing for how the contribution is presented.

major comments (3)
  1. [Abstract, Eq. (5), Table 1, Appendix D.2] The title and abstract present 'Certified Anytime-Valid Stopping' as the headline contribution, but the 74x stopping-time reduction is measured under AsympCS (Eq. 5), whose guarantee is only asymptotic (Proposition 2). Appendix D.2 states that the HUNL corrected stream has no analytic ceiling; the declared B = 200 BB is the observed raw maximum rounded upward, with the largest observed corrected |Y| = 153.79 BB. Consequently the EB-CS rows in Table 1 are explicitly descriptive, and the exact anytime-valid certificate is established only for Leduc (BY = 117, Appendix D.4). The paper is transparent about this in Sections 6 and 9, but the title and abstract do not carry the qualification. This mismatch is load-bearing for the central claim and should be resolved by either supplying an independent HUNL bound or reframing the headline as an asymptotic-screen result.
  2. [Table 3, Section 6.1] The AsympCS screen's finite-horizon exclusion rates exceed the nominal 5% level: P0 reports an average 7.1067% (12.43% for AIVAT-only streams) and the A2 standalone false-positive rate is 10.4%. The abstract's phrase 'At the nominal 95% level' appears immediately before the 74x claim, inviting the reader to treat the stopping rule as having 95% coverage at the observed horizons. The paper should state explicitly that AsympCS is an approximate screen with these measured finite-horizon rates, rather than implying that the 74x ratio is obtained under a guaranteed 95% confidence level.
  3. [Proposition 1, Section 8] Proposition 1's exact EB-CS validity is conditional on an independently justified almost-sure bound BY on the corrected payoff. The paper derives such a bound only for Leduc; for HUNL no such bound is proved, as acknowledged in Appendix D.2 and Table 1's footnote. The release protocol in Section 8 correctly requires an independent bound for exact claims, but the Contributions section and the abstract should match that standard by explicitly limiting exact certification to settings where an analytic or structural bound is available. As written, a reader may reasonably infer that the HUNL experiments enjoy the same exact anytime-valid certificate as Leduc.
minor comments (4)
  1. [References] Reference [22] cites Peter Grünwald's discussion comment as the source of the empirical-Bernstein confidence sequence, but the construction in Eq. (4) originates from Waudby-Smith and Ramdas's paper on estimating means of bounded random variables by betting. The citation should be corrected to the primary source.
  2. [Eq. (4)] The notation 'log 2 α' is easily misread as log(2α); since the text defines it as log(2/α), the paper should use log(2/α) consistently in displayed equations.
  3. [Figure 2] Figure 2 labels the descriptive B = 22 and loose B' = 200 curves, but the analytic BY = 117 floor anchors are mentioned only in the text; adding them to the figure would make the comparison immediate.
  4. [Throughout] The name 'AIVAT' is frequently typeset as 'AIV AT' with a space; this should be corrected to a single token for consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's formal claims are derived from explicit conditions and cited external confidence-sequence theorems, while its headline speedups are measurements rather than fitted predictions.

full rationale

We walked the paper's derivation chain. Proposition 1 derives E[C_t | F_{t-1}] = 0 from the stated pre-action conditional-kernel and predictability assumptions; it does not assume its conclusion, and the exact EB-CS validity is imported from the published e-process theorem of [22] after verifying boundedness and constant conditional mean, which is independent support rather than a self-referential loop. Proposition 2 and Theorems 1 and 2 are likewise asymptotic or finite-sample statements proved from explicit regularity conditions. The observed 54x variance reduction and 74x AsympCS stopping-time ratio are empirical summaries of the 71,439-hand corpus, not outputs of a parameter fit that is then relabeled as a prediction. The HUNL EB-CS row is explicitly labeled descriptive because B = 200 BB is taken from the observed maximum rather than from an independent bound, so no fitted quantity is presented as a certified prediction; the gap between the headline 'certified' claim and the actual HUNL bound is a limitation in the strength of the evidence, not a circular reduction. Self-citations appear only as data provenance (PokerSkill corpus) and related work, and none of them carries the load of the validity theorems. We therefore find no step in which a claimed result is equivalent by construction to its inputs.

Assumptions & free parameters 4 free parameters · 8 assumptions · 0 invented entities

The central claims rest on the predictability and boundedness of value functions, known action kernels, game-structural payoff bounds, and the regularity conditions of the two confidence-sequence constructions. No fitted parameter appears inside the theorems; the empirical free parameters are tuning choices and observed-maximum bounds that the paper labels descriptive where exactness is not established.

free parameters (4)
  • AsympCS tuning parameter rho (via topt = 1000 default) = rho^2 = 2 log(1/alpha) / topt, topt = 1000
    The AsympCS half-width (5) depends on rho; a fixed value is chosen by calibration and frozen before the evaluation. This choice affects stopping times and finite-horizon false-positive rates.
  • HUNL corrected-stream payoff bound B_Y = 200 BB (P0) = 200 BB, rounded up from observed maximum absolute corrected payoff 153.79 BB
    Used to rescale the EB-CS for the AIVAT-corrected HUNL streams; not independently justified, so the resulting EB-CS is descriptive rather than exact (Appendix D.2).
  • Leduc descriptive corrected bound B = 22 (original E1) = 22 chips, rounded up from observed maximum
    Replaced the registered raw bound 13 when corrected samples exceeded it; the reported 0% EB-CS miscoverage at this bound is descriptive (Appendix D.3). The structural rerun uses BY = 117.
  • Observed-max descriptive bounds in E2 EB coverage and A2 entries = varies per replication, ceiling of observed maximum absolute payoff
    The E2 coverage calculation and each A2 simulated entry declare bounds from observed data, so their EB-CS summaries are descriptive resampling outcomes, not exact coverage certificates (Appendix D.3, D.5).
assumptions (8)
  • domain assumption The evaluator knows the conditional action kernel p_{t,h} at every correction node before the action, and enablement S_{t,h} is fixed before seeing the action.
    Needed for E[C_t | F_{t-1}] = 0 in Proposition 1 and equation (3). Violations bias the correction. Table 6 asks for logged kernels, but for black-box LLM agents this kernel may be unverifiable.
  • domain assumption Raw hand payoffs X_t are independent and identically distributed with fixed mean mu and are bounded in [-B_X, B_X].
    Section 3.1 states i.i.d. hands under a fixed agent and opponent; both confidence sequences and the empirical corpus analysis use this. Real LLM agents may adapt or exhibit serial dependence.
  • domain assumption The value functions v_t are uniformly bounded (||v_t||_infty <= V) and F_{t-1}-measurable.
    Boundedness gives |Y_t| <= B_X + 2KV and the EB-CS range; predictability gives the conditional mean-zero property. See Proposition 1 and equation (3).
  • domain assumption At most K correction nodes occur on any trajectory.
    Used to bound corrected payoffs and in Lemma 2. For Leduc the tree enumeration verifies a small K for opponent transitions; for HUNL no such structural count for the corrected stream is given.
  • domain assumption For AsympCS: conditional Lindeberg condition, relative consistency of the sample variance, and t^{-1} V_t converging to a finite nonzero limit hold almost surely.
    Proposition 2 inherits asymptotic validity from Waudby-Smith et al. [47]. Finite-horizon false-positive rates above nominal show these asymptotic conditions are not exact at the stopping times used.
  • standard math The published e-process theorem for the predictable plug-in EB-CS applies to the rescaled bounded stream with constant conditional mean and predictable bets.
    Proposition 1, Appendix A.3 relies on the cited e-process result; the paper does not re-derive it. The citation points to a discussion article rather than the primary source.
  • domain assumption For HUNL raw payoffs, |X_t| <= 200 BB is structural because seats have 200 BB stacks and rebuys are forbidden.
    Used to justify the raw-stream bound in P0; it does not extend to the corrected stream, as the paper notes.
  • domain assumption For Leduc, the structural corrected-payoff bound BY = 117 follows because telescoping leaves at most four opponent transitions, each bounded by 26 chips, verified by enumerating the complete legal tree.
    Needed for the exact EB-CS certificate in Leduc. The enumeration is described in Appendix D.4 but is not fully reproduced in the preprint.

how reviews work

0 comments
Cite this review

Pith. "Pith review of AV-AIVAT: 74x Cheaper Agent Evaluation with Certified Anytime-Valid Stopping in Imperfect-Information Games." pith.science (2026). https://pith.science/paper/SBCF5OZ5

@misc{pith2026260806362,
  author       = {Pith},
  title        = {Pith review of: AV-AIVAT: 74x Cheaper Agent Evaluation with Certified Anytime-Valid Stopping in Imperfect-Information Games},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SBCF5OZ5}},
  note         = {Machine review of arXiv:2608.06362}
}
abstract

Deciding which of two agents is stronger means playing games until skill outweighs luck, and every game costs money, model inference, or expert time. Since the number of games needed is unknown, fixed-budget evaluations either keep paying after the result is settled or stop before the agents can be told apart, while naive optional stopping with an ordinary confidence interval invalidates the stated level. We make such an evaluation stop as soon as its evidence suffices, with the guarantee intact. The Action-Informed Value Assessment Tool (AIVAT) reduces variance in imperfect-information games through conditional mean-zero corrections, by a median $54\times$ across 15 LLM agent configurations spanning 71,439 paired Heads-Up No-Limit Hold'em (HUNL) hands, but does not say when to stop. We combine AIVAT with continuously monitored Confidence Sequences (CSs) into anytime-valid AIVAT (AV-AIVAT), whose online value model learns only from past games so that no game scores its own correction. At the nominal 95\% level and a target precision of $\pm1$ Big Blind, raw outcomes need a median $74\times$ as many hands as AIVAT-corrected outcomes to stop under the Asymptotic CS (AsympCS). Exact finite-sample certification uses the Empirical-Bernstein CS (EB-CS), which needs an independently justified bound on corrected payoffs. We establish such a bound structurally for Leduc hold'em and characterize a width floor set by the CS's bet cap and that bound, which governs how much of a variance gain becomes earlier stopping; the descriptive HUNL EB-CS runs show a median $1.37\times$ stopping-time ratio. AV-AIVAT turns variance reduction into efficient, auditable early stopping while separating asymptotic screening from exact certification, so an evaluation can stop the moment its evidence suffices and hand a third party everything needed to recheck the verdict at that very stopping time.

Figures

Figures reproduced from arXiv: 2608.06362 by the authors.

Figure 1
Figure 1. Run-level heterogeneity in the 15 HUNL evaluations. Each point is one configuration. [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗
Figure 2
Figure 2. EB-CS width curves on Leduc under different declared bounds. The loose [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. E2 per-epoch variance under the controlled construction. Predictable past-only learning [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Injected-effect experiment. A first-crossing fixed CI often claims success before a sequential [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]
Figure 5
Figure 5. Figure 5: Post hoc entry-length sensitivity. Error bars are Wilson 95% intervals from the archived [PITH_FULL_IMAGE:figures/full_fig_p033_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 48 canonical work pages

  1. [1]

    Poker, chance and skill.Unpublished Manuscript, pages 1–17, 2007

    Noga Alon. Poker, chance and skill.Unpublished Manuscript, pages 1–17, 2007

    work page 2007

  2. [2]

    A tool for the direct assessment of poker decisions.ICGA Journal, 29(3):119–142, 2006

    Darse Billings and Morgan Kan. A tool for the direct assessment of poker decisions.ICGA Journal, 29(3):119–142, 2006

    work page 2006

  3. [3]

    Strategy evaluation in extensive games with importance sampling

    Michael Bowling, Michael Johanson, Neil Burch, and Duane Szafron. Strategy evaluation in extensive games with importance sampling. InProceedings of the 25th international conference on Machine learning, pages 72–79, 2008

    work page 2008

  4. [4]

    Heads-up limit hold’em poker is solved.Science, 347(6218):145–149, 2015

    Michael Bowling, Neil Burch, Michael Johanson, and Oskari Tammelin. Heads-up limit hold’em poker is solved.Science, 347(6218):145–149, 2015

    work page 2015

  5. [5]

    Regret-based pruning in extensive-form games.Advances in neural information processing systems, 28, 2015

    Noam Brown and Tuomas Sandholm. Regret-based pruning in extensive-form games.Advances in neural information processing systems, 28, 2015

    work page 2015

  6. [6]

    Baby tartanian8: Winning agent from the 2016 annual computer poker competition

    Noam Brown and Tuomas Sandholm. Baby tartanian8: Winning agent from the 2016 annual computer poker competition. InIJcAI, pages 4238–4239, 2016

    work page 2016

  7. [7]

    Strategy-based warm starting for regret minimization in games

    Noam Brown and Tuomas Sandholm. Strategy-based warm starting for regret minimization in games. InProceedings of the AAAI Conference on Artificial Intelligence, volume 30, 2016

    work page 2016

  8. [8]

    Reduced space and faster convergence in imperfect- information games via pruning

    Noam Brown and Tuomas Sandholm. Reduced space and faster convergence in imperfect- information games via pruning. InInternational conference on machine learning, pages 596–604. PMLR, 2017

    work page 2017

Show all 55 references
  1. [9]

    Safe and nested subgame solving for imperfect-information games.Advances in neural information processing systems, 30, 2017

    Noam Brown and Tuomas Sandholm. Safe and nested subgame solving for imperfect-information games.Advances in neural information processing systems, 30, 2017

    2017

  2. [10]

    Superhuman ai for heads-up no-limit poker: Libratus beats top professionals.Science, 359(6374):418–424, 2018

    Noam Brown and Tuomas Sandholm. Superhuman ai for heads-up no-limit poker: Libratus beats top professionals.Science, 359(6374):418–424, 2018

    2018

  3. [11]

    Solving imperfect-information games via discounted regret minimization

    Noam Brown and Tuomas Sandholm. Solving imperfect-information games via discounted regret minimization. InProceedings of the AAAI Conference on Artificial Intelligence, volume 33, pages 1829–1836, 2019

    2019

  4. [12]

    Superhuman ai for multiplayer poker.Science, 365(6456): 885–890, 2019

    Noam Brown and Tuomas Sandholm. Superhuman ai for multiplayer poker.Science, 365(6456): 885–890, 2019

    2019

  5. [13]

    Depth-limited solving for imperfect- information games.Advances in neural information processing systems, 31, 2018

    Noam Brown, Tuomas Sandholm, and Brandon Amos. Depth-limited solving for imperfect- information games.Advances in neural information processing systems, 31, 2018

    2018

  6. [14]

    Deep counterfactual regret minimization

    Noam Brown, Adam Lerer, Sam Gross, and Tuomas Sandholm. Deep counterfactual regret minimization. InInternational conference on machine learning, pages 793–802. PMLR, 2019

    2019

  7. [15]

    Combining deep reinforcement learning and search for imperfect-information games.Advances in neural information processing systems, 33:17057–17069, 2020

    Noam Brown, Anton Bakhtin, Adam Lerer, and Qucheng Gong. Combining deep reinforcement learning and search for imperfect-information games.Advances in neural information processing systems, 33:17057–17069, 2020. 20

    2020

  8. [16]

    Aivat: A new variance reduction technique for agent evaluation in imperfect information games

    Neil Burch, Martin Schmid, Matej Moravcik, Dustin Morill, and Michael Bowling. Aivat: A new variance reduction technique for agent evaluation in imperfect information games. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 32, 2018

    2018

  9. [17]

    Baseline: practical control variates for agent evaluation in zero-sum domains

    Joshua Davidson, Christopher Archibald, and Michael Bowling. Baseline: practical control variates for agent evaluation in zero-sum domains. InAAMAS, volume 13, pages 1005–1012, 2013

    2013

  10. [18]

    Low-variance and zero-variance baselines for extensive-form games

    Trevor Davis, Martin Schmid, and Michael Bowling. Low-variance and zero-variance baselines for extensive-form games. InInternational Conference on Machine Learning, pages 2392–2401. PMLR, 2020

    2020

  11. [19]

    Poker is a skill.Gaming Law Review, 12(1): 31–36, 2008

    Michael A DeDonno and Douglas K Detterman. Poker is a skill.Gaming Law Review, 12(1): 31–36, 2008

    2008

  12. [20]

    Actor-critic policy optimization in a large-scale imperfect-information game

    Haobo Fu, Weiming Liu, Shuang Wu, Yijia Wang, Tao Yang, Kai Li, Junliang Xing, Bin Li, Bo Ma, Qiang Fu, et al. Actor-critic policy optimization in a large-scale imperfect-information game. InInternational conference on learning representations, 2021

    2021

  13. [21]

    Endgame solving in large imperfect-information games

    Sam Ganzfried and Tuomas Sandholm. Endgame solving in large imperfect-information games. InProceedings of the 2015 International Conference on Autonomous Agents and Multiagent Systems, pages 37–45, 2015

    2015

  14. [22]

    Peter Gr¨ unwald. Proposer of the vote of thanks to waudy-smith and ramdas and contribution to the discussion of ‘estimating means of bounded random variables by betting’.Journal of the Royal Statistical Society Series B: Statistical Methodology, 86(1):28–30, 2024

    2024

  15. [23]

    Time-uniform, nonparametric, nonasymptotic confidence sequences.The Annals of Statistics, 49(2):1055–1080, 2021

    Steven R Howard, Aaditya Ramdas, Jon McAuliffe, and Jasjeet Sekhon. Time-uniform, nonparametric, nonasymptotic confidence sequences.The Annals of Statistics, 49(2):1055–1080, 2021

    2021

  16. [24]

    Slumbot nl: Solving large games with counterfactual regret minimization using sampling and distributed processing

    Eric Jackson. Slumbot nl: Solving large games with counterfactual regret minimization using sampling and distributed processing. InAAAI Workshop on Computer Poker and Incomplete Information, 2013

    2013

  17. [25]

    Measuring the size of large no-limit poker games.arXiv preprint arXiv:1302.7008, 2013

    Michael Johanson. Measuring the size of large no-limit poker games.arXiv preprint arXiv:1302.7008, 2013

    2013 arXiv

  18. [26]

    Accelerating best response calculation in large extensive games

    Michael Johanson, Kevin Waugh, Michael Bowling, and Martin Zinkevich. Accelerating best response calculation in large extensive games. InIJCAI, volume 11, pages 258–265, 2011

    2011

  19. [27]

    Evaluating state-space abstractions in extensive-form games

    Michael Johanson, Neil Burch, Richard Valenzano, and Michael Bowling. Evaluating state-space abstractions in extensive-form games. InProceedings of the 2013 international conference on Autonomous agents and multi-agent systems, pages 271–278, 2013

    2013

  20. [28]

    Postgame analysis of poker decisions

    Morgan Hugh Kan. Postgame analysis of poker decisions. 2007

    2007

  21. [29]

    Heuristic pathologies and further variance reduction via uncertainty propagation in the aivat family of techniques.arXiv preprint arXiv:2605.14261, 2026

    Juho Kim and Tuomas Sandholm. Heuristic pathologies and further variance reduction via uncertainty propagation in the aivat family of techniques.arXiv preprint arXiv:2605.14261, 2026

    2026 arXiv

  22. [30]

    Value func- tions for depth-limited solving in zero-sum imperfect-information games.Artificial Intelligence, 314:103805, 2023

    Vojtˇ ech Kovaˇ r ´ ık, Dominik Seitz, Viliam Lis` y, Jan Rudolf, Shuo Sun, and Karel Ha. Value func- tions for depth-limited solving in zero-sum imperfect-information games.Artificial Intelligence, 314:103805, 2023. 21

    2023

  23. [31]

    Monte carlo sampling for regret minimization in extensive games.Advances in neural information processing systems, 22, 2009

    Marc Lanctot, Kevin Waugh, Martin Zinkevich, and Michael Bowling. Monte carlo sampling for regret minimization in extensive games.Advances in neural information processing systems, 22, 2009

    2009

  24. [32]

    Efficient online pruning and abstraction for imperfect information extensive-form games

    Boning Li and Longbo Huang. Efficient online pruning and abstraction for imperfect information extensive-form games. InThe Thirteenth International Conference on Learning Representations, 2025

    2025

  25. [33]

    Agents that certify their own exploits: Confidence-scheduled restricted responses for safe opponent exploitation, 2026

    Boning Li and Longbo Huang. Agents that certify their own exploits: Confidence-scheduled restricted responses for safe opponent exploitation, 2026. URL https://arxiv.org/abs/2607. 28520

    2026

  26. [34]

    Effective, efficient, and general information abstraction for imperfect-information extensive-form games.arXiv preprint arXiv:2605.10900, 2026

    Boning Li and Longbo Huang. Effective, efficient, and general information abstraction for imperfect-information extensive-form games.arXiv preprint arXiv:2605.10900, 2026

    2026 arXiv

  27. [35]

    Real-time parallel counterfactual regret minimization.arXiv preprint arXiv:2605.19928, 2026

    Boning Li and Longbo Huang. Real-time parallel counterfactual regret minimization.arXiv preprint arXiv:2605.19928, 2026

    2026 arXiv

  28. [36]

    Rl-cfr: improving action abstraction for imperfect information extensive-form games with reinforcement learning

    Boning Li, Zhixuan Fang, and Longbo Huang. Rl-cfr: improving action abstraction for imperfect information extensive-form games with reinforcement learning. InProceedings of the 41st International Conference on Machine Learning, pages 27752–27770, 2024

    2024

  29. [37]

    Correlated chance sampling for monte carlo counter- factual regret minimization, 2026

    Boning Li, Yu Chen, and Longbo Huang. Correlated chance sampling for monte carlo counter- factual regret minimization, 2026. URL https://arxiv.org/abs/2607.27035

    2026 arXiv

  30. [38]

    Pokerskill: Llms can play expert-level poker without training or solvers.arXiv preprint arXiv:2605.30094, 2026

    Boning Li, Baoxiang Wang, and Longbo Huang. Pokerskill: Llms can play expert-level poker without training or solvers.arXiv preprint arXiv:2605.30094, 2026

    2026 arXiv

  31. [39]

    Escher: Eschewing importance sampling in games by computing a history value function to estimate regret

    Stephen Marcus McAleer, Gabriele Farina, Marc Lanctot, and Tuomas Sandholm. Escher: Eschewing importance sampling in games by computing a history value function to estimate regret. InThe Eleventh International Conference on Learning Representations

  32. [40]

    Refining subgames in large imperfect information games

    Matej Moravcik, Martin Schmid, Karel Ha, Milan Hladik, and Stephen Gaukrodger. Refining subgames in large imperfect information games. InProceedings of the AAAI Conference on Artificial Intelligence, volume 30, 2016

    2016

  33. [41]

    Deepstack: Expert-level artificial intelligence in heads-up no-limit poker.Science, 356(6337):508–513, 2017

    Matej Moravˇ c ´ ık, Martin Schmid, Neil Burch, Viliam Lis` y, Dustin Morrill, Nolan Bard, Trevor Davis, Kevin Waugh, Michael Johanson, and Michael Bowling. Deepstack: Expert-level artificial intelligence in heads-up no-limit poker.Science, 356(6337):508–513, 2017

    2017

  34. [42]

    Gto wizard benchmark.arXiv preprint arXiv:2603.23660, 2026

    Marc-Antoine Provost, Nejc Ilenic, Christopher Solinas, and Philippe Beardsell. Gto wizard benchmark.arXiv preprint arXiv:2603.23660, 2026

    2026

  35. [43]

    Game-theoretic statistics and safe anytime-valid inference.Statistical Science, 38(4):576–601, 2023

    Aaditya Ramdas, Peter Gr¨ unwald, Vladimir Vovk, and Glenn Shafer. Game-theoretic statistics and safe anytime-valid inference.Statistical Science, 38(4):576–601, 2023

    2023

  36. [44]

    Variance reduction in monte carlo counterfactual regret minimization (vr-mccfr) for extensive form games using baselines

    Martin Schmid, Neil Burch, Marc Lanctot, Matej Moravcik, Rudolf Kadlec, and Michael Bowling. Variance reduction in monte carlo counterfactual regret minimization (vr-mccfr) for extensive form games using baselines. InProceedings of the AAAI Conference on Artificial Intelligenc...

    2019

  37. [45]

    Student of games: A unified learning algorithm for both perfect and imperfect information games.Science Advances, 9(46):eadg3256, 2023

    Martin Schmid, Matej Moravˇ c ´ ık, Neil Burch, Rudolf Kadlec, Josh Davidson, Kevin Waugh, Nolan Bard, Finbarr Timbers, Marc Lanctot, G Zacharias Holland, et al. Student of games: A unified learning algorithm for both perfect and imperfect information games.Science Advances, 9...

    2023

  38. [46]

    Solving large imperfect information games using cfr+.arXiv preprint arXiv:1407.5042, 2014

    Oskari Tammelin. Solving large imperfect information games using cfr+.arXiv preprint arXiv:1407.5042, 2014

    2014 arXiv

  39. [47]

    Time- uniform central limit theory and asymptotic confidence sequences.The Annals of Statistics, 52 (6), 2024

    Ian Waudby-Smith, David Arbour, Ritwik Sinha, Edward Kennedy, and Aaditya Ramdas. Time- uniform central limit theory and asymptotic confidence sequences.The Annals of Statistics, 52 (6), 2024

    2024

  40. [48]

    Abstraction pathologies in extensive games.AAMAS (2), 2009:781–8, 2009

    Kevin Waugh, David Schnizlein, Michael H Bowling, and Duane Szafron. Abstraction pathologies in extensive games.AAMAS (2), 2009:781–8, 2009

    2009

  41. [49]

    Learning a value analysis tool for agent evaluation

    Martha White and Michael H Bowling. Learning a value analysis tool for agent evaluation. In IJCAI, pages 1976–1981, 2009

    1976

  42. [50]

    Douzero: Mastering doudizhu with self-play deep reinforcement learning

    Daochen Zha, Jingru Xie, Wenye Ma, Sheng Zhang, Xiangru Lian, Xia Hu, and Ji Liu. Douzero: Mastering doudizhu with self-play deep reinforcement learning. Ininternational conference on machine learning, pages 12333–12344. PMLR, 2021

    2021

  43. [51]

    Pokerbench: Training large language models to become professional poker players

    Richard Zhuang, Akshat Gupta, Richard Yang, Aniket Rahane, Zhengyu Li, and Gopala Anumanchipalli. Pokerbench: Training large language models to become professional poker players. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 26175–26182, 2025

    2025

  44. [52]

    Optimal unbiased estimators for evaluating agent performance

    Martin Zinkevich, Michael Bowling, Nolan Bard, Morgan Kan, and Darse Billings. Optimal unbiased estimators for evaluating agent performance. InProceedings of the National Conference on Artificial Intelligence, volume 21, page 573. Menlo Park, CA; Cambridge, MA; London; AAAI Pr...

    1999

  45. [53]

    Regret mini- mization in games with incomplete information.Advances in neural information processing systems, 20, 2007

    Martin Zinkevich, Michael Johanson, Michael Bowling, and Carmelo Piccione. Regret mini- mization in games with incomplete information.Advances in neural information processing systems, 20, 2007. A Proofs of Predictable Validity We prove Propositions 1 and 2. Fix the finite gam...

    2007

  46. [54]

    Contain” means that the interval still contains zero at the monitor’s selected stop; “Asymp FP

    implies 2 ε/a≤ 1/2, so all masses are positive; a≤B places the common support inside [−B, B]. 26 If either expected stopping time is infinite, the claimed lower bound is immediate. Otherwise, at the stopping time decide P+ when the reported interval center is positive and P− o...

  47. [55]

    Frozen and oracle corrected streams are identical in this matched control, so there is no identifiable oracle variance gap

    S1 pairs the CFR +-2000 policy with the uniform opponent model used by the frozen evaluator. Frozen and oracle corrected streams are identical in this matched control, so there is no identifiable oracle variance gap. S2 pairs CFR +-2000 with CFR +-20; S3 reverses these policy-...

    2000

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.