REVIEW 3 major objections 5 minor 9 references
Win Probabilities, Hand Sizes, and Game Duration Analysis in the Bhikar-Sawkar Card Game
T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash
Pith's one-line read The paper claims that shuffling each won hand in Bhikar-Sawkar breaks all cycles, so every game ends, and reports Monte Carlo estimates of durations, hand sizes, and win probabilities across 25 configurations.
desk verdict A clean but modest Monte Carlo study of a never-analyzed card game, whose headline termination guarantee is asserted, not proved. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing mechanism is the reshuffle-on-win rule: a player who wins a hand collects the central pile, shuffles it randomly, and appends it to the bottom of their private deck. This single stochastic step separates Bhikar-Sawkar from deterministic Beggar-My-Neighbour, where won cards are appended without shuffling and cycles can persist. In the paper's argument the reshuffle is what guarantees termination and creates the heavy-tailed duration distribution the Monte Carlo study then measures.
What would settle it
Run the simulation code with cycle detection that records the full ordered state of every player's deck and the central pile after each shuffle; if any state repeats before a winner is declared, the termination claim is false for that configuration. Alternatively, an exhaustive search over the finite reachable states for a small case such as two players and one deck would either find a winner from every state or exhibit a reachable cycle.
Extended reading notes
Core claim
The paper's central discovery, stated on its own terms, is that introducing a reshuffle whenever a player wins a hand makes Bhikar-Sawkar generically terminating: unlike deterministic Beggar-My-Neighbour, which has known non-terminating sequences, the reshuffle destroys the exact card-order cycles that would otherwise repeat forever. In $10^{6}$ simulations per configuration, every game ended, with the longest observed game reaching 26,322 turns for four players and five decks; the paper is careful to call this a rare sample outcome, not an upper bound. The simulation statistics then imply that typical games are much shorter than extreme ones, that hand sizes peak below five cards with heavy tails at larger deck counts, and that small-deck games show slight positional asymmetries that wash out as the number of decks increases.
Load-bearing premise
The load-bearing premise is that shuffling a won hand always destroys every possible cycle, so no sequence of play can repeat forever; the paper supports this only by noting that all its simulated games ended, not by a proof.
Editorial extensions
If this is right
- All 25 simulated configurations terminate, so the reported duration, hand-size, and win-probability distributions describe a finite process rather than a possibly divergent one.
- A single-deck game can end in as few as 52 turns, when no hand is won; minimum durations grow with deck count and player count.
- For larger deck counts, hand-win counts and game-win probabilities become effectively uniform across players, so the winner of a randomly played game is close to a fair lottery.
- Maximum durations rise with deck count but peak at an intermediate player count, indicating a non-monotonic relationship between players and extreme game length.
- Because duration distributions are heavy-tailed, the expected game length is much lower than the extreme tail values; rare long games dominate variance but not typical play.
Reading between the lines
- The paper leaves implicit that a formal proof of termination would need to show the shuffled process is a finite-state random process that must eventually reach a winner; the simulation evidence alone does not rule out rare, never-ending paths of measure zero.
- The observed uniformity at large deck counts suggests that fairness questions about starting-player advantage disappear once enough decks are used, while the small-deck asymmetries may repay a closer analytic look.
- The paper's finding that the longest observed game occurs at an intermediate player count invites a follow-up question it leaves open: whether the expected or extreme duration is maximized at a specific ratio of players to decks, and whether that optimum reflects a structural property of the shuffle process.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper reports a Monte Carlo simulation study of the Bhikar-Sawkar card game, a stochastic variant related to Beggar-My-Neighbour. For every combination of N in {2,3,4,5} players and K in {1,2,3,4,5} decks, the author simulates 10^6 games and reports the observed maximum and minimum game durations (in turns), distributions of game durations and hand sizes, and player-level win probabilities. The central theoretical claim is that the rule of shuffling the central pile after each won hand prevents non-terminating cycles, so that every game ends with a single winner. The simulation code is publicly available. The paper is descriptive: no analytical model or statistical error analysis is provided.
Significance. If the termination property and the simulation methodology are sound, the paper provides a useful descriptive benchmark for a stochastic card game that is otherwise difficult to analyze analytically. Its strengths are the large, systematic simulation grid (25 configurations, 10^6 games each), the absence of fitted free parameters, and the release of reproducible code. The empirical distributions, especially the heavy-tailed duration distributions, could be of interest to researchers studying stochastic game processes. However, the significance is tempered by two issues: the termination guarantee is asserted rather than proved, which conditions the interpretation of the extreme-value statistics, and the absence of error bars or statistical tests makes several comparative claims (e.g., about an 'optimal' number of players for long games) unsupported.
major comments (3)
- [Section III and Section V] The load-bearing claim that the reshuffling mechanism 'prevents such cycles, thereby ensuring that every game eventually terminates with a single winner' is asserted without proof. The game is a finite Markov chain (finite card configurations plus random shuffles), and a finite Markov chain can contain closed recurrent classes that never reach the absorbing state of one remaining player. Randomizing the order of a won pile does not by itself rule out such classes. The only support offered is that all sampled games in the 10^6-per-configuration simulations terminated, but a finite sample cannot establish almost-sure termination. To make the reported maxima and duration distributions well-defined properties of the game rather than of games that happened to stop, the author should either prove that from every reachable state there is a positive-probability path to absorption and no closed class avoids it, or explicitly condition the results on observed termination and state that the termination guarantee remains open. The disclaimer in Section V that maxima are not absolute upper bounds partially mitigates overinterpretation, but it does not replace the missing proof.
- [Section IVA and IVB] The reported maxima and the comparative statements about game durations have no error bars, confidence intervals, or other measures of sampling variability. This is especially problematic because the duration distributions are heavy-tailed (as shown in Figure 2), so sample maxima are highly sensitive to the number of simulations. For example, the observation in Section IVA that the longest game (26322 turns) occurs for N=4, K=5 and not for the maximum number of players is presented as a finding, yet with a single simulation run per configuration it could easily be a sampling artifact. The author should provide bootstrap or other uncertainty estimates for the maxima and, if the 'optimal number of players' claim is retained, support it with a statistical test rather than an anecdotal comparison.
- [Section IVB] The paper describes the duration distributions as 'heavy-tailed' but then states 'we do not characterise these distributions.' A visual inspection of count histograms with linear axes is not sufficient to establish heavy-tailedness. To make this a substantive result, the author should provide a quantitative characterization, for example by comparing the empirical tail to a power-law or lognormal model, reporting a tail-index estimate, or showing log-log or complementary CDF plots for all configurations. Without this, the heavy-tailed claim remains a qualitative impression rather than a quantified property.
minor comments (5)
- [Section I] The phrase 'despite being governed by entirely deterministic rules' contradicts the paper's own description of the game, which includes shuffling after every won hand and is later described as 'fundamentally non-deterministic' in the conclusion and abstract; this should be corrected.
- [Figure 2] The figure caption uses D for the number of decks, whereas the text and other figures use K; please standardize the notation.
- [Figure 1 caption] The caption says 'maximum and minimum number of games' but should say 'maximum and minimum number of turns'; as written it suggests the plot shows counts of games.
- [Throughout] There are several spacing and typographical errors, such as 'Bhikar-Sawkarcard game' in the abstract and 'Bhikar-Sawkarintroduces' in Section III; a thorough proofread is recommended.
- [Section IVD] The statement that win probabilities become 'more uniform' as decks increase is made without any formal measure of uniformity or significance test; adding e.g. a chi-square or entropy-based measure would strengthen the claim.
Circularity Check
No significant circularity: the paper is a forward Monte Carlo study with no fitted parameters, and the unsupported termination assertion is not a circular reduction.
full rationale
The paper's results are obtained by direct Monte Carlo simulation of a fully specified stochastic game. The inputs (number of players N, number of decks K, reshuffling rule) are independent of the outputs (max/min turns, hand-size distributions, win probabilities). There are no fitted parameters, no quantity defined in terms of another quantity that is then 'predicted', and no calibration step. The central claim in Section III that reshuffling 'prevents such cycles, thereby ensuring that every game eventually terminates' is an unproved mathematical assertion, not a circular derivation: it is not justified by citing the paper's own results, and it is not equivalent to any fitted input or constructed statistic. The simulation statistics are empirical observations, and the paper itself cautions in Section V that observed maxima are not absolute upper bounds. The only self-citation is to the author's own simulation code ([7]), which is not load-bearing and does not substitute for an argument. Absence of a rigorous termination proof is a correctness or completeness concern, but not an instance of circular reasoning under the stated criteria.
Assumptions & free parameters
assumptions (3)
- domain assumption The Bhikar-Sawkar rules as described in Section II are the actual game rules used in the simulation.
- domain assumption The shuffling after each hand is uniformly random and independent.
- domain assumption Every game terminates under these rules.
Cite this review
Pith. "Pith review of Win Probabilities, Hand Sizes, and Game Duration Analysis in the Bhikar-Sawkar Card Game." pith.science (2026). https://pith.science/paper/H6V475OQ
@misc{pith2026250603680,
author = {Pith},
title = {Pith review of: Win Probabilities, Hand Sizes, and Game Duration Analysis in the Bhikar-Sawkar Card Game},
year = {2026},
howpublished = {\url{https://pith.science/paper/H6V475OQ}},
note = {Machine review of arXiv:2506.03680}
}
read the original abstract
We present a Monte Carlo simulation study of the Bhikar-Sawkar card game, a non-deterministic game structurally similar to the classic Beggar-My-Neighbour, which is fully deterministic. Although both games share a common setup, key differences in their rules, particularly the reshuffling of cards after each won hand in Bhikar-Sawkar, introduce stochasticity and significantly increase the space of possible game evolutions. This inherent randomness raises a range of interesting statistical questions regarding the duration of the game, the hand-winner distributions, and the probability of winning the game for a given player. These questions are systematically investigated through large-scale simulations across multiple game configurations.
Figures
Figures from the paper (2 more)
Reference graph
Works this paper leans on
-
[1]
Setup: TheBhikar-Sawkargame is played by N players seated in a circle, facing each other. The game uses K standard decks of 52 playing cards each, which are thoroughly shuffled together to form a single combined deck. A player is randomly selected to begin the distribution (designated as Player 1), and the cards are dealt one by one in a clockwise directi...
-
[2]
The act of playing a card counts as a turn
Game Flow: The game begins with Player 1 placing the top card of their private deck face-up in the center of the table. The act of playing a card counts as a turn. The next player in the clockwise direction (Player 2) then plays the top card of their own deck on top of the central pile. At each turn, the current player’s card is compared to the card immed...
- [3]
-
[4]
R. A. B. P.Eng., in2016 ASEE Annual Conference & Exposition, 10.18260/p.27135 (ASEE Confer- ences, New Orleans, Louisiana, 2016) https://peer.asee.org/27135
-
[5]
A. Madsen, L. Nielsen, and F. Jensen, inProceedings of the 11th International Florida Artificial Intelligence Research Symposium Conference, edited by Cook, Diane J. (ed.) (AAAI Press, United States, 1998) pp. 435–439, probSy - A System for the Calculation of Probabilities in the Card Game Bridge ; Conference date: 19-05-2010. 12
work page 1998
-
[6]
M. M. Paulhus, The American Mathematical Monthly106, 162 (1999), https://doi.org/10.1080/00029890.1999.12005024
arXiv 1999
-
[7]
A non-terminating game of beggar-my-neighbor,
B. Casella, P. M. Anderson, M. Kleber, R. P. Mann, R. Nessler, W. Rucklidge, S. G. Williams, and N. Wu, “A non-terminating game of beggar-my-neighbor,” (2024), arXiv:2403.13855 [math.CO]
-
[8]
E. L. Lakshtanov and A. I. Aleksenko, Problems of Information Transmission49, 163 (2013)
work page 2013
Show all 9 references
-
[9]
Monte carlo simulation code for bhikar-sawkar card game,
M. Durve, “Monte carlo simulation code for bhikar-sawkar card game,” (2025). 13
2025
Reviewed August 7, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.