We study a class of random-turn resource-allocation games. In the Trail of Lost Pennies, a counter moves on $\mathbb{Z}$. At each turn, Maxine stakes $a \in [0,\infty)$ and Mina $b \in [0,\infty)$. The counter $X$ then moves adjacently, to the right with probability $\tfrac{a}{a+b}$. If $X_i \to -\infty$ in this infinte-turn game, Mina receives one unit, and Maxine zero; if $X_i \to \infty$, then these receipts are zero and~$x$. Thus the net receipt to a given player is $-A+B$, where $A$ is the sum of her stakes, and $B$ is her terminal receipt. The game was inspired by unbiased tug-of-war in~\cite{PSSW09} from 2009 but in fact closely resembles the original version of tug-of-war, introduced~\cite{HarrisVickers87} in the economics literature in 1987. We show that the game has surprising features. For a natural class of strategies, Nash equilibria exist precisely when $x$ lies in $[\lambda,\lambda^{-1}]$, for a certain $\lambda \in (0,1)$. We indicate that $\lambda$ is remarkably close to one, proving that $\lambda \leq 0.999904$ and presenting clear numerical evidence that $\lambda \geq 1 - 10^{-4}$. For each $x \in [\lambda,\lambda^{-1}]$, we find countably many Nash equilibria. Each is roughly characterized by an integral {\em battlefield} index: when the counter is nearby, both players stake intensely, with rapid but asymmetric decay in stakes as it moves away. Our results advance premises [HV87,K12] for fund management and the incentive-outcome relation that plausibly hold for many player-funded stake-governed games. Alongside a companion treatment [HP22] of games with allocated budgets, we thus offer a detailed mathematical treatment of an illustrative class of tug-of-war games. We also review the separate developments of tug-of-war in economics and mathematics in the hope that mathematicians direct further attention to tug-of-war in its original resource-allocation guise.
翻译:----
我们研究了一类随机轮流分配资源的游戏。在失落的一分钱中,一个计数器在整数集 $\mathbb{Z}$ 上移动。每次,Maxine 投注 $a\in [0, \infty)$ 而 Mina 投注 $b\in [0, \infty)$。然后,计数器 $X$ 按 p= $a/(a+b)$ 的概率向右移动。如果 $X_i$ 在这个无限次数的游戏中 $\to -\infty$,那么 Mina 获得 $1$,而 Maxine 获得 $0$;如果 $X_i \to \infty$,那么这些收益为 $0$ 和 $x$。因此,给定玩家的净收益是 $-A+B$,其中 $A$ 是她的赌注之和,$B$ 是她的终端收入。这个游戏源于 2009 年引入的无偏对抗掉铁环,但实际上与 1987 年经济学文献中引入的对抗掉铁环的原始版本~[HarrisVickers87] 非常相似。我们展示了这个游戏具有令人惊异的特性。对于一个自然的策略类,当 $x$ 在 $[\lambda,\lambda^{-1}]$ 中时 Nash 均衡存在,其中 $\lambda \in (0,1)$。我们表明 $\lambda$ 与 $1$ 非常接近,证明 $\lambda \leq 0.999904$ 并清晰地展示 $\lambda \geq 1 - 10^{-4}$ 的明确数值证据。对于 $[\lambda,\lambda^{-1}]$ 中的每个 $x$,我们找到了可数的 Nash 均衡。每个 Nash 均衡大致上被一个整数的“战场”指数所表征:当计数器附近时,两个玩家投注强烈,但是投注随着它远离而快速但是不对称地衰减。我们的结果推进了资金管理和激励结果关系的前提条件 [HV87, K12],而这些前提条件可能适用于许多玩家赞助的赌注博弈。随着对分配预算游戏的伴随处理 [HP22],我们因此提供了一个关于对抗掉铁环游戏的具体数学处理。我们还回顾了对在经济学和数学中对抗掉铁环的分离发展,希望数学家在其原始的资源分配形式下进一步关注对抗掉铁环。