Beyond the briscola advantage: a Monte Carlo dominance test for deterministic strategies in two-player Briscola Game

Piero Giacomelli

arxiv: 2605.17043 · v2 · pith:GK4FBH2Pnew · submitted 2026-05-16 · 💻 cs.GT

Beyond the briscola advantage: a Monte Carlo dominance test for deterministic strategies in two-player Briscola Game

Piero Giacomelli This is my paper

Pith reviewed 2026-05-20 15:28 UTC · model grok-4.3

classification 💻 cs.GT

keywords BriscolaMonte Carlo simulationdeterministic strategiestrick-taking gamestrump advantagestrategy dominancecard game simulation

0 comments

The pith

Two refined deterministic strategies in Briscola win more often than greedy play regardless of trump cards received.

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

The paper tests the popular idea that Briscola is mostly decided by who draws more cards of the trump suit. It compares a basic greedy strategy against two rule-based improvements: one that hoards trump cards and one that counters based on visible information. Through a million simulated games per matchup, the refined strategies show higher win rates even when the opponent holds more trumps. A sympathetic reader would care because the results separate the effect of strategy from the luck of the deal in a well-known game. The work uses statistical checks to measure how much each factor contributes to the outcome.

Core claim

Two deterministic rule-based refinements of the naive greedy policy—a briscola-hoarding policy and a public-information counter policy—dominate the greedy baseline irrespective of trump luck, as shown by higher win rates in round-robin Monte Carlo tournaments of one million games per pairing, confirmed through statistical tests that isolate strategy effects from briscola-count imbalance.

What carries the argument

The round-robin Monte Carlo tournament of 10^6 games per strategy pairing, analyzed with Wilson confidence intervals, Bonferroni-corrected binomial tests, and logistic regression on strategy pair and signed briscola-count imbalance.

If this is right

Refined strategies raise win probability even in games where the opponent receives more trump cards.
Logistic regression can separate the contribution of strategy choice from the trump-count difference.
Deterministic hoarding and counter policies produce reliably higher scores than simple greedy selection.
The outcome depends on both strategy and trump imbalance, but strategy remains detectable after accounting for the deal.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same simulation approach could test strategy dominance in other two-player trick-taking games where deal luck is often blamed for losses.
Players who adopt these hoarding or counter rules might improve their results in actual matches beyond what the deal alone predicts.
Extending the model to include imperfect information or opponent modeling could reveal further performance gaps.

Load-bearing premise

The simulation correctly implements the exact Briscola rules for dealing, trump selection, legal plays, and trick resolution, and the three strategies run without coding errors.

What would settle it

An independent re-implementation of the simulation that finds no consistent win-rate advantage for the refined strategies over the greedy baseline when the number of trumps is held fixed.

Figures

Figures reproduced from arXiv: 2605.17043 by Piero Giacomelli.

**Figure 2.** Figure 2: Empirical G1 win rate as a function of the signed briscola imbalance ∆briscola = BG1 − BG2 , for each of the nine ordered matchups (rows index G1’s policy, columns G2’s). Grey ribbons: Wilson 95% confidence intervals. Dashed guides mark ∆briscola = 0 and Pr(G1 wins) = 0.5. Bins with fewer than 50 games are omitted. — where G1 wins 50% of non-tied games — drifts with the policy pair: when G1 = πG and G2 = π… view at source ↗

read the original abstract

Briscola is a traditional Italian trick-taking card game whose simplest form is played by two players. Popular folklore credits victory almost entirely to the player who is dealt more cards of the trump suit (the so-called \emph{briscola}), so that the game would be a near-deterministic function of the deal. We test this folklore against a pre-registered alternative, namely that two deterministic rule-based refinements of the naive greedy policy -- a briscola-hoarding policy $\stratH$ and a public-information counter policy $\stratC$ -- dominate the greedy baseline $\stratG$ irrespective of trump luck. To this end we run a round-robin Monte Carlo tournament of $10^{6}$ simulated games across the nine ordered pairings of $(\stratG,\stratH,\stratC)$, retaining approximately $1.08\times 10^{5}$ non-tied games per pairing, and we analyse the resulting outcomes through Wilson confidence intervals, a Bonferroni-corrected pairwise binomial test, and a logistic regression of the game outcome on the strategy pair and on the signed briscola-count imbalance, so as to quantify the relative contribution of strategy and trump luck. We close with a reproducibility appendix that makes the simulation, the random seed and the analysis script fully deterministic.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Monte Carlo runs show two rule-based strategies beating greedy in Briscola after adjusting for trump imbalance, but the additive logistic model does not fully confirm the advantage holds at every imbalance level.

read the letter

The key point is that this work uses large-scale simulation to show that two specific deterministic strategies outperform the basic greedy approach in Briscola, and the statistical analysis points to a real edge beyond just getting better trumps. The new element is the combination of briscola-hoarding and public-information counter policies tested head-to-head against greedy in a round-robin setup with a million games. They apply standard tools like Wilson intervals for proportions, Bonferroni for multiple comparisons, and logistic regression to separate the strategy effect from the signed imbalance in briscola count. The reproducibility appendix stands out because it locks down the random seed and shares the analysis script. This setup does a good job of addressing the folklore that the game is mostly about the deal. By including the imbalance covariate, the regression quantifies how much strategy contributes on top of luck. The pre-registered alternative and the focus on non-tied games keep things focused. One soft spot is in the regression specification. It models outcome as a function of strategy pair plus imbalance without interaction terms. That means the estimated strategy advantage is assumed constant across all levels of trump imbalance. If the true advantage is larger when imbalance is high or low, the current model does not directly test that. Stratified win rates or marginal effects at different imbalance values would strengthen the claim that dominance holds irrespective of trump luck. The stress test note flags this, and it seems worth checking in the full results. Implementation details matter here too. The strategies are rule-based and deterministic, which is good for reproducibility, but any error in coding legal plays or trump selection could bias the outcomes. Since they provide the full setup, a referee could verify that. This paper fits readers working on Monte Carlo studies of imperfect-information games or on folk card games. It is narrow in scope but the methods are transparent. I think it deserves peer review. The core simulation and analysis are solid enough to benefit from external checks on the code and perhaps requests for more granular breakdowns of the results.

Referee Report

1 major / 1 minor

Summary. The paper claims that in two-player Briscola, two deterministic rule-based refinements of the naive greedy policy—a briscola-hoarding policy (stratH) and a public-information counter policy (stratC)—dominate the greedy baseline (stratG) irrespective of trump luck. This is tested via a pre-registered round-robin Monte Carlo tournament of 10^6 simulated games, retaining ~1.08×10^5 non-tied games per pairing, analyzed with Wilson confidence intervals, Bonferroni-corrected pairwise binomial tests, and logistic regression of game outcome on strategy pair plus signed briscola-count imbalance, with a reproducibility appendix ensuring deterministic simulation and analysis.

Significance. If the central dominance result holds after refinement, the work provides a substantive empirical challenge to the folklore that Briscola outcomes are near-deterministic functions of the trump deal. The large-scale forward simulation, pre-registered analysis plan, use of standard external statistical tools (Wilson intervals, Bonferroni correction, logistic regression), and fully deterministic reproducibility appendix are clear strengths that enhance credibility in the game-theory and computational game-playing literature.

major comments (1)

[Abstract and analysis section] Abstract and analysis section: the logistic regression is specified with main effects only for strategy pair and signed briscola-count imbalance. This additive model assumes the strategy effect is constant across imbalance values and does not test for interactions or evaluate marginal effects at different imbalance quantiles. Because the headline claim requires dominance 'irrespective of trump luck' (i.e., after conditioning on imbalance), the current specification leaves open the possibility that the reported advantage is an artifact of particular imbalance ranges; stratified win rates or an interaction term would be needed to substantiate the claim.

minor comments (1)

[Reproducibility appendix] Reproducibility appendix: while the deterministic seed and script are welcome, explicit pseudocode or decision tables for stratH and stratC would further aid independent verification of the rule-based implementations.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thoughtful and constructive review. The suggestion to strengthen the logistic regression analysis is well taken, and we outline our planned revisions below.

read point-by-point responses

Referee: [Abstract and analysis section] Abstract and analysis section: the logistic regression is specified with main effects only for strategy pair and signed briscola-count imbalance. This additive model assumes the strategy effect is constant across imbalance values and does not test for interactions or evaluate marginal effects at different imbalance quantiles. Because the headline claim requires dominance 'irrespective of trump luck' (i.e., after conditioning on imbalance), the current specification leaves open the possibility that the reported advantage is an artifact of particular imbalance ranges; stratified win rates or an interaction term would be needed to substantiate the claim.

Authors: We agree that the current main-effects specification does not explicitly test for heterogeneity in the strategy effect across levels of briscola-count imbalance. To directly address the claim of dominance irrespective of trump luck, we will revise the analysis to include an interaction term between strategy pair and signed imbalance. We will also add stratified win-rate tables by imbalance quartiles, allowing readers to inspect marginal effects at different ranges. Updated results, tables, and a brief discussion of the interaction coefficients will be incorporated in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

No circularity: results rest on forward Monte Carlo simulation of game rules plus standard external statistical tests

full rationale

The paper generates outcomes via independent forward simulation of the exact Briscola rules across 10^6 games for each strategy pairing, then applies Wilson intervals, Bonferroni-corrected binomial tests, and an additive logistic regression to quantify strategy versus imbalance effects. No parameter is fitted on a subset and renamed as a prediction; the logistic model serves only for post-hoc decomposition rather than generating the dominance claim. The simulation is rule-based and deterministic given the seed, with no self-citations, ansatzes, or uniqueness theorems invoked to close the argument. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that the simulation correctly encodes Briscola rules and that standard statistical procedures can isolate strategy effects; no free parameters or new entities are introduced.

axioms (1)

domain assumption Briscola game mechanics including dealing, trump assignment, legal moves, and trick resolution are accurately implemented in the Monte Carlo simulator.
All dominance conclusions depend on the simulation matching the real game; any mismatch would invalidate the comparison to the folklore claim.

pith-pipeline@v0.9.0 · 5757 in / 1279 out tokens · 61817 ms · 2026-05-20T15:28:29.611680+00:00 · methodology

Beyond the briscola advantage: a Monte Carlo dominance test for deterministic strategies in two-player Briscola Game

Core claim

What carries the argument

If this is right

Where Pith is reading between the lines

Load-bearing premise

What would settle it

discussion (0)