Papers
Topics
Authors
Recent
Search
2000 character limit reached

Three-Move Subtraction Games

Updated 2 February 2026
  • Three-move subtraction games are impartial combinatorial games defined by a fixed three-element subtraction set, exhibiting rich periodic and algebraic structures.
  • Algorithms such as the Sprague–Grundy function and the PTFN procedure efficiently classify winning positions and reveal underlying recurrence relations.
  • The analysis uncovers explicit formulas, connections to Beatty sequences, expansion properties, and open problems in multi-dimensional game segmentation.

A three-move subtraction game is an impartial combinatorial game where two players alternately remove a fixed, allowable number of tokens from one or several heaps. When the subtraction set contains exactly three elements, the analysis reveals rich periodic and algebraic structure. The study of these games encompasses single- and multi-dimensional settings, periodicity, explicit formulae for winning positions, expansion properties, and deep connections to Beatty sequences and combinatorial game theory.

1. Formal Definition and Structure

A classic three-move subtraction game on a single heap is played with a fixed subtraction set S={s1,s2,s3}S=\{s_1,s_2,s_3\}, where 1s1<s2<s31\leq s_1<s_2<s_3. Players alternate moves; each move consists of subtracting exactly one sSs\in S from the heap, provided sns\leq n, for heap size n0n\geq0. The player unable to move wins under normal play convention. Every position nn is recursively classified as a P\mathcal{P}-position (previous player wins with best play) or an N\mathcal{N}-position (next player wins) via structural recurrences rooted in Zermelo's principle: all terminal positions are P\mathcal{P}, and any N\mathcal{N}-position admits a move to a P\mathcal{P}-position, while all moves from a P\mathcal{P}-position lead to N\mathcal{N}-positions (He et al., 2012).

Extending to dd-dimensional games, a position xN0d\mathbf{x}\in\mathbb{N}_0^d allows subtractions by vectors sSN0d\mathbf{s}\in S\subset\mathbb{N}_0^d with coordinatewise constraint sx\mathbf{s}\leq\mathbf{x}. Normal play, periodicity, and recursive outcome classification generalize accordingly (Larsson et al., 2023).

2. Algorithms for Solving Three-Move Games

Central algorithms for identifying winning positions are the Sprague–Grundy function and the PTFN ("P-position To Find N-position") procedure. The Sprague–Grundy function assigns to each heap nn the nim-value

G(n)=mex{G(ns):sS, sn},G(n) = \mathrm{mex}\{ G(n-s) : s\in S,\ s\leq n \},

where mex\mathrm{mex} denotes the minimum excluded value (Ho, 2012). P\mathcal{P}-positions are those with G(n)=0G(n)=0.

In contrast, the PTFN algorithm employs a $0$–$1$ bit array X[0n]X[0\dots n] indicating P\mathcal{P}-positions ($0$) and N\mathcal{N}-positions ($1$). The recurrence for three-move games (with a<b<ca<b<c) is

X[n]={0,n<a, 1,s{a,b,c}: X[ns]=0, 0,otherwise.X[n]= \begin{cases} 0, & n<a,\ 1, & \exists\, s\in \{a,b,c\}:\ X[n-s]=0,\ 0, & \text{otherwise}. \end{cases}

Equivalently,

X[n]=1    (¬X[na])(¬X[nb])(¬X[nc]),X[n]=0    X[na]=X[nb]=X[nc]=1.X[n]=1 \iff (\neg X[n-a]) \lor (\neg X[n-b]) \lor (\neg X[n-c]), \qquad X[n]=0 \iff X[n-a]=X[n-b]=X[n-c]=1.

The algorithm runs in O(nS)O(n|S|) time and uses O(n)O(n) space, faster and more space-efficient than Sprague–Grundy when S=3|S|=3 (He et al., 2012).

3. Periodicity and Explicit Formulas

Every fixed finite subtraction set yields an ultimately periodic sequence for both X[n]X[n] and G(n)G(n). In numerous three-move families, one observes alternating gap patterns among P\mathcal{P}-positions and closed-form formulas. For instance, in S={1,3,4}S=\{1,3,4\}, P\mathcal{P}-positions occur at P={0,2,7,9,14,16,21,}P=\{0,2,7,9,14,16,21,\dots\} with gaps repeating 2,5,2,5,2,5,2,5,\dots and closed form

p2k=7k,p2k+1=7k+2(k0).p_{2k}=7k, \quad p_{2k+1}=7k+2 \qquad (k\ge0).

Analogous patterns emerge for various arithmetic and algebraic templates, such as S(a,b,a+b)S(a,b,a+b) and S(1,a,b)S(1,a,b), with the period and pre-period determined by explicit parameter-dependent formulas (Ho, 2012).

Moreover, in two-dimensional settings, periodicity is maintained along rows and columns. Theorem: Every finite two-dimensional set is eventually row- and column-periodic, with outcome-type repeating after some TT once past an initial transient x0x_0 or y0y_0 (Larsson et al., 2023).

4. Connections to Beatty Sequences and Generalizations

A significant class of three-move games produces P\mathcal{P}-positions as complementary pairs of Beatty sequences: P={(nα,nβ)}P=\{(\lfloor n\alpha\rfloor,\lfloor n\beta\rfloor)\}, with α,β\alpha,\beta irrational and 1/α+1/β=11/\alpha+1/\beta=1. Fraenkel and modified subtraction games use mex-style recurrences

an=mex{ak,bk:k<n},bn=f(an)+bn1+(anan1),a_n = \mathrm{mex}\{ a_k, b_k : k < n \}, \qquad b_n = f(a_n) + b_{n-1} + (a_n - a_{n-1}),

where ff encodes move constraints, possibly derived directly from Beatty differences (2208.00041).

A necessary and sufficient "Beatty-complementarity inequality" 2minfmaxf12\min f-\max f\ge1 governs which pairs arise. The Relaxed Wythoff family generalizes this further, showing that for any complementary Beatty pair, one can construct a simple three-move game whose P\mathcal{P}-positions are exactly the pair, providing a solution to the Fraenkel problem and a modification of the Duchêne–Rigo conjecture. Examples include the classical golden ratio case and parameterized tt-Wythoff families.

5. Expansion Properties and Rigidity

The notion of an expansion set SexS^{\text{ex}} formalizes which additional moves can be added to SS without altering the nim-sequence. The precise characterization:

  • sSexs\in S^{\text{ex}} iff G(n+s)G(n)G(n+s)\neq G(n) for all nn in one period.
  • For larger ss, periodicity governs inclusion: sSexs\in S^{\text{ex}} iff spSexs-p\in S^{\text{ex}}, with pp the period.

The expansion set can be explicitly computed for families (e.g., S={1,8,11,27}S=\{1,8,11,27\} yields Sex={1,8,11}{13,20,27}+19S^{\text{ex}} = \{1,8,11\} \cup \{13,20,27\}^{+19}) (Ho, 2012).

A central open conjecture asserts that ultimately bipartite games (those with tail nim-sequence 0,1,0,1,0,1,0,1,\dots) are non-expandable; no proper superset of SS can maintain the same nim-sequence. This conjecture is proven for several families, supported by combinatorial obstructions and explicit counterexamples for expansion by odd moves above sks_k.

6. Multi-Dimensional Three-Move Games and Segmentation Phenomena

Three-move rulesets in two or more dimensions reveal "mosaic" and "slice" structures in the game-board segmentation. Symmetric-additive and arithmetic-additive families admit complete solutions via segmentwise coloring algorithms (termed "coloring automaton" Editor's term). In the symmetric-additive case S={(a,b),(b,a),(a+b,b+a)}S=\{(a,b),(b,a),(a+b,b+a)\}, P\mathcal{P}-positions fill segments defined by rational slopes, recursively closed under specific translation vectors.

Experimentally, for many small three-move sets, the first quadrant decomposes (modulo a finite "corner") into finitely many semi-infinite slices, each bounded by rational slope lines. The coloring automaton precisely identifies P\mathcal{P}-positions with periodic patterns within each segment (Larsson et al., 2023).

Open conjectures propose that every three-move dd-dimensional game admits segmentation into finitely many semi-infinite slices solved by such automata, and that kk-move games attain at most k+1k+1 slices. Computational complexity is open: two-move games are solvable in linear time, but the complexity for three moves remains unresolved.

7. Open Problems and Directions

Classification of the periodicity and nim-sequence behavior for all three-move sets is incomplete; several family subcases (e.g., S(a,b,a+b)S(a,b,a+b) with even kk) lack explicit solution. The full structure of expansion sets and rigidity for general sets needs further exploration (Ho, 2012). In higher dimensions, a natural necessary and sufficient condition ensuring solvability by a coloring automaton remains open, as does the problem of computational complexity for decision queries in three-move games.

The inverse construction for Beatty pairs, the periodic mosaic phenomenon, and outcome segmentation remain central research themes, with implications for combinatorial game theory and algebraic combinatorics (2208.00041, Larsson et al., 2023).


Table 1: Summary of Key Algorithmic and Structural Features

Aspect Single-Heap S=3|S|=3 Multi-Dimensional
Solution Recurrence PTFN or mex/Grundy Coloring Automaton, mex
P\mathcal{P}-position pattern Ultimately periodic Row/column periodic, slices
Explicit formulas Exists for many families Segment-wise constructions
Expansion set Computable, conjectural rigidity Yet to be classified
Beatty connection Extensive via recurrences Present in variants

The theory of three-move subtraction games forms a nexus of combinatorial periodicity, algebraic structure, extremal rigidity, and open-ended complexity, with explicit solutions available for key families and deep connections to number theory and game dynamics.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (4)

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to Three-Move Subtraction Games.