刘昊辰
(2025-10-27 14:21):
#paper Strongly Solving 2048 4×3. 本文由日本东京大学研究者提出,成功强解了 2048 游戏的 4×3 变体(2048₄ₓ₃),核心关键技术是基于 ”年龄(age)”(定义为棋盘上所有方块数字之和)对状态空间进行划分 —— 状态与后续动作后的过渡态(afterstate)年龄保持不变,过渡态到新状态时年龄因新增方块(2 或 4)增加 2 或 4,据此可分阶段枚举状态并控制内存占用;同时采用Elias-Fano 编码实现状态的紧凑存储,将约 4.4TiB 的原始存储需求压缩至 1.4TiB(最优玩法专用存储仅需 300GiB)。研究结果显示,最常见初始状态(两个 2 方块,年龄 4)的最优策略期望得分为50724.26,可到达状态数与过渡态数分别为1.15×10¹²和7.40×10¹¹,且验证了 “生成大数字方块(如 2048)时难度显著提升” 等玩家直觉。下载地址:https://arxiv.org/pdf/2510.04580
arXiv,
2025-10-06T08:31:59Z.
DOI: 10.48550/arXiv.2510.04580
Strongly Solving 2048 4x3
Tomoyuki Kaneko,
Shuhei Yamashita
Abstract:
2048 is a stochastic single-player game involving 16 cells on a 4 by 4 grid,<br>where a player chooses a direction among up, down, left, and right to obtain a<br>score by merging two tiles with the same number located in neighboring cells<br>along the chosen direction. This paper presents that a variant 2048-4x3 12<br>cells on a 4 by 3 board, one row smaller than the original, has been strongly<br>solved. In this variant, the expected score achieved by an optimal strategy is<br>about $50724.26$ for the most common initial states: ones with two tiles of<br>number 2. The numbers of reachable states and afterstates are identified to be<br>$1,152,817,492,752$ and $739,648,886,170$, respectively. The key technique is<br>to partition state space by the sum of tile numbers on a board, which we call<br>the age of a state. An age is invariant between a state and its successive<br>afterstate after any valid action and is increased two or four by stochastic<br>response from the environment. Therefore, we can partition state space by ages<br>and enumerate all (after)states of an age depending only on states with the<br>recent ages. Similarly, we can identify (after)state values by going along with<br>ages in decreasing order.
Related Links: