Documentation

CombinatorialGames.Literature.OnFactorisationsOfLeftDeadEnds

On Factorisations of Left Dead Ends #

This module records the formalisation of Alfie Davies. On factorisations of Left dead ends.

2. Preliminaries #

def OnFactorisationsOfLeftDeadEnds.definition_2_2 {G : Type (u + 1)} [Form G] (A : G → Prop) (hA : A ≤ Form.IsShort) :
G → Prop

Given a set of games $\mathcal{A}$, define the universal closure of $\mathcal{A}$, denoted $\mathcal{D}(\mathcal{A})$, as the minimal universe containing $\mathcal{A}$ (with respect to set inclusion).

This is Davies (Definition 2.2 on p. 6).

Equations
Instances For

    TODO: Proposition 2.3

    If $\mathcal{U}$ is a universe, and $G$ and $H$ are Left dead ends, then $G \ge_{\mathcal{U}} H$ if and only if the following hold:

    1. if $G \cong 0$, then $H \cong 0$; otherwise,

    2. for every $G^R \in G^{\mathcal{R}}$, there exists some $H^R \in H^{\mathcal{R}}$ such that $G^R \ge_{\mathcal{U}} H^R$.

    This is Davies (Theorem 2.4 on p. 7).

    TODO: Corollary 2.5 (multiversality of comparison)

    In a monoid $M$, we call a non-invertible element $m \in M$ a molecule if $m$ is atomic but not an atom.

    This is Davies (Definition 2.6 on p. 9).

    Equations
    Instances For

      3. Properties of the monoid #

      The monoid of Left dead ends $\mathcal{L}$ is reduced.

      This is Davies (Theorem 3.1 on p. 10).

      If $G$ and $H$ are Left dead ends with $G \ge H$, and $k$ is an integer satisfying $0 \le k \le \tilde{b}(G)$, then, for any run $(G_0, G_1, \ldots, G_k)$ of $G$ of length $k$, there exists a run $(H_0, H_1, \ldots, H_k)$ of $H$ of length $k$ such that $G_i \ge H_i$ for every $0 \le i \le k$.

      This is Davies (Corollary 3.2 on p. 10).

      If $G$ is a combinatorial game, then we will say that a (Left, respectively Right) option $G'$ of $G$ is good if it is not strictly dominated by another (Left, respectively Right) option of $G$.

      This is Davies (Definition 3.3 on p. 11).

      Equations
      Instances For

        If $G \ge H$ and $G^R$ is a good option of $G$, then some (good) option $H^R$ of $H$ has $G^R \ge H^R$.

        This is the $\ge$ case of Davies (Corollary 3.4 on p. 11).

        If $G = H$ and $G^R$ is a good option of $G$, then some (good) option $H^R$ of $H$ has $G^R = H^R$.

        This is the $=$ case of Davies (Corollary 3.4 on p. 11).

        theorem OnFactorisationsOfLeftDeadEnds.lemma_3_5 {A : GameForm → Prop} {g h j : GameForm} (hgA : A g) (hhA : A h) (hj : GameForm.LeftDeadEnd.IsPocancellativeIn A j) (hge : (g + j) ≥m A h + j) :
        g ≥m A h

        If $\mathcal{U}$ is a universe, $G, H \in \mathcal{U}$ are Left dead ends, and $J$ is a pocancellative element of $\mathcal{U}$ satisfying $G + J \ge H + J$, then $G \ge H$.

        This is Davies (Lemma 3.5 on p. 12).

        theorem OnFactorisationsOfLeftDeadEnds.lemma_3_6 {A : GameForm → Prop} {g h j : GameForm} (hgA : A g) (hhA : A h) (hj : GameForm.LeftDeadEnd.IsCancellativeIn A j) (heq : (g + j) =m A h + j) :
        g =m A h

        If $\mathcal{U}$ is a universe, $G, H \in \mathcal{U}$ are Left dead ends, and $J$ is a cancellative element of $\mathcal{U}$ satisfying $G + J = H + J$, then $G = H$.

        This is Davies (Lemma 3.6 on p. 12).

        The monoid of Left dead ends $\mathcal{L}$ is pocancellative.

        This is Davies (Theorem 3.7 on p. 12).

        If $G$ and $H$ are Left dead ends with $n \cdot G = n \cdot H$ for some $n > 0$, then $G = H$.

        This is Davies (Proposition 3.8 on p. 13).

        TODO: Corollary 3.9 (the group of differences of $\mathcal{L}$ is torsion-free)

        TODO: Open Problem 3.10 (the isomorphism type of the group of differences)

        If $G$ is a combinatorial game and $(G_0, G_1, \ldots, G_k)$ is a run of $G$ such that $G_k \cong 0$, then we say the run is terminal.

        This is Davies (Definition 3.11 on p. 14).

        Equations
        Instances For

          If $G$ is a short game, then we define the set of terminal lengths of $G$, denoted $T(G)$, as the set of all $n$ such that there exists a terminal run of $G$ of length $n$.

          This is Davies (Definition 3.12 on p. 14), taken here in the equivalent recursive form of Lemma 3.13.

          Equations
          Instances For

            $n \in T(G)$ if and only if there exists a terminal run of $G$ of length $n$.

            This is Davies (Definition 3.12 on p. 14), recovered as a theorem from the recursive definition of terminalLengths.

            If $G$ is a non-zero Left dead end, then $T(G) = \bigcup (T(G') + 1)$, where $G'$ ranges over all options of $G$.

            This is Davies (Lemma 3.13 on p. 14).

            If $G$ is a short game, define the race of $G$, denoted $r(G)$, as $r(G) := \min(T(G))$.

            This is Davies (Definition 3.14 on p. 15).

            Equations
            Instances For

              A Left dead end is an integer if it is isomorphic either to $0 := \{\cdot \mid \cdot\}$ or to $\overline{n} := \{\cdot \mid \overline{n - 1}\}$ for some $n \ge 1$ (where $\overline{0} := 0$).

              This is Davies (p. 15).

              Equations
              Instances For

                A Left dead end $G$ is an integer if and only if $|T(G)| = 1$. In particular, $G = \overline{n}$ if and only if $T(G) = \{n\}$.

                This is Davies (Lemma 3.15 on p. 15).

                If $G$ is a Left dead end, then the following are equivalent: $G = 0$; $G \cong 0$; $0 \in T(G)$; $r(G) = 0$; $b(G) = 0$.

                This is Davies (Lemma 3.16 on p. 15).

                If $G$ and $H$ are Left dead ends with $G \ge H$, then $r(G) \ge r(H)$.

                This is the race half of Davies (Corollary 3.18 on p. 16).

                If $G$ and $H$ are Left dead ends with $G \ge H$, then $b(G) \le b(H)$.

                This is the birthday half of Davies (Corollary 3.18 on p. 16).

                If $G$ and $H$ are Left dead ends with $G \ge H$, and $k$ is an integer satisfying $0 \le k \le \tilde{b}(G)$, then, for any run $(G_0, G_1, \ldots, G_k)$ of $G$ of length $k$, there exists a run $(H_0, H_1, \ldots, H_k)$ of $H$ of length $k$ such that $T(G_i) \subseteq T(H_i)$ for every $0 \le i \le k$.

                This is Davies (Corollary 3.19 on p. 17).

                If $G$ and $H$ are Left dead ends, then $T(G + H) = T(G) + T(H)$, a Minkowski sum.

                This is Davies (Proposition 3.20 on p. 18).

                If $G$ and $H$ are Left dead ends, then $|T(G + H)| \ge \max(|T(G)|, |T(H)|)$, with equality if and only if at least one of $G$ and $H$ is an integer.

                This is Davies (Corollary 3.21 on p. 18).

                If $G$, $H$, and $K$ are non-zero Left dead ends with $G = H + K$, then $b(H) < b(G)$.

                This is Davies (Lemma 3.22 on p. 19).

                The monoid of Left dead ends $\mathcal{L}$ is a BF-monoid. In particular, if $G$ is a Left dead end, then the length of each factorisation of $G$ is at most $r(G)$; that is, $\overline{\mathsf{L}(G)} \le r(G)$.

                This is Davies (Lemma 3.24 on p. 19).

                4. In search of a prime #

                If $G$ is a Left dead end, then $\overline{1} + G' \ge G$ for every option $G'$ of $G$.

                This is Davies (Lemma 4.1 on p. 20).

                If $G$ is a Left dead end, then $\overline{1} + G = \{\cdot \mid G\}$.

                This is Davies (Proposition 4.2 on p. 21).

                If $G$ and $G'$ are Left dead ends, then $G$ has exactly one good option (equal to) $G'$ if and only if $G = \overline{1} + G'$.

                This is Davies (Proposition 4.3 on p. 21).

                The game $\overline{1}$ is prime in the monoid of Left dead ends $\mathcal{L}$.

                This is Davies (Theorem 4.4 on p. 21).

                5. Flexibility #

                If $G$ is a Left dead end, then we define the flexibility of $G$, written $f(G)$, recursively as $f(G) = 1 + \max(f(G'))$, where $G'$ ranges over all options of $G$, when $G$ is not an integer, and as $f(G) := 0$ otherwise.

                This is Davies (Definition 5.1 on p. 22).

                Equations
                Instances For

                  If $G$ is a Left dead end, then we say an option $G'$ of $G$ is versatile if $f(G') \ge f(G) - 1$.

                  This is Davies (Definition 5.3 on p. 23).

                  Equations
                  Instances For

                    If $G$ is a Left dead end, then $f(G) \le b(G)$, with equality if and only if $G = 0$.

                    This is Davies (Lemma 5.4 on p. 24).

                    If $G$ is a Left dead end, then we say a run $(G_0, G_1, \ldots, G_n)$ of $G$ of length $n > 0$ is flexible if $G_i$ is not an integer for $i < n$. We define all runs of length $0$ to be flexible.

                    This is Davies (Definition 5.6 on p. 25).

                    Equations
                    Instances For

                      A flexible run is maximal if it ends at an integer; that is, if it cannot be extended to a longer flexible run.

                      This is Davies (Definition 5.6 on p. 25).

                      Equations
                      Instances For

                        If $G$ is a Left dead end, then $f(G)$ is equal to the length of a longest flexible run of $G$.

                        This is Davies (Lemma 5.7 on p. 25).

                        If $G$ is a Left dead end, then we define a racing option of $G$ to be an option $G'$ such that $r(G') = r(G) - 1$.

                        This is Davies (Definition 5.8 on p. 25).

                        Equations
                        Instances For

                          If $G$ is a Left dead end, then we define a stalling option of $G$ to be an option $G'$ such that $b(G') = b(G) - 1$.

                          This is Davies (Definition 5.9 on p. 25).

                          Equations
                          Instances For

                            If $G$ is a non-integer Left dead end, then $f(G + \overline{n}) = f(G) + n$ for all $n \ge 0$.

                            This is Davies (Proposition 5.10(1) on p. 26).

                            If $G$, $H$, and $K$ are Left dead ends such that $H$ is non-zero, $K$ is not an integer, and $H + K \ge G$, then $b(H) < f(G)$.

                            This is Davies (Proposition 5.11 on p. 27).

                            If $G$ is a Left dead end with more than one good option, then, for any non-zero Left dead ends $H$ and $K$ with $H + K = G$, it holds that $b(H), b(K) < f(G)$.

                            This is Davies (Corollary 5.12 on p. 28).

                            If $G$ is a Left dead end with $f(G) = 2$, then either $G$ has exactly one good option (i.e. $\overline{1} \mid G$), or $G$ is an atom.

                            This is Davies (Proposition 5.14 on p. 28).

                            6. Lengths of factorisations #

                            If $G$ is a Left dead end, then $\overline{\mathsf{L}(G)} \le \overline{\mathsf{L}(G')} + 1$ for every good option $G'$ of $G$.

                            This is Davies (Lemma 6.1 on p. 30).

                            If $G$ is a Left dead end with exactly one good option $G'$, then $\overline{\mathsf{L}(G)} = \overline{\mathsf{L}(G')} + 1$.

                            This is Davies (Theorem 6.2 on p. 30).

                            7. Uniqueness of factorisations #

                            If $G$ is a Left dead end with a good option $H$ such that $\overline{\mathsf{L}(G)} = \overline{\mathsf{L}(H)} + 1$, then

                            1. $H$ is a racing option of $G$; and

                            2. there exists a unique atom $K$ such that $G = H + K$, and $K$ is terminable.

                            This is Davies (Proposition 7.1 on p. 32).

                            If $G$ is a Left dead end with a good option to an atom $H$, then either $G$ is an atom, or $H$ is a racing option of $G$ and there exists a unique (terminable) atom $K$ such that $G = H + K$; in particular, $G$ is uniquely factorisable.

                            This is the "in particular" of Davies (Corollary 7.2 on p. 33).

                            If $G$ is a Left dead end with good options to distinct atoms $H$ and $K$, then either $G$ is an atom, or $H + K$ is the unique factorisation of $G$.

                            This is Davies (Corollary 7.3 on p. 33).

                            If either of $H$ or $K$ is not terminable (i.e. if $r(G) \ne 2$), then $G$ is an atom.

                            This is the "in particular" of Davies (Corollary 7.3 on p. 33).

                            If $G := G_1 + \cdots + G_n = H_1 + \cdots + H_m =: H$ are factorisations of a Left dead end $G$, and there is a good option of $G$ playing on the component $G_1$ to some prime-factorisable $G'_1$, then there exists an index $j$ such that $G_1 = H_j$.

                            This is Davies (Lemma 7.5 on p. 34).

                            If $G := G_1 + \cdots + G_k$ is a factorisation of a Left dead end such that, for every $1 \le i \le k$, there exists a good option of $G_i + \cdots + G_k$ on the component $G_i$ to some prime-factorisable $G'_i$, then $G$ is uniquely factorisable.

                            This is Davies (Theorem 7.6 on p. 35).

                            Reading a factorisation as $G_1 + \cdots + G_k$: for every $1 \le i \le k$, there exists a good option of $G_i + \cdots + G_k$ on the component $G_i$ to some prime-factorisable $G'_i$.

                            This is the hypothesis of Davies (Theorem 7.6 on p. 35).

                            Equations
                            Instances For

                              If $G + H$ is a sum of Left dead ends, and $G' + H$ is a good option of $G + H$, then $G'$ is a good option of $G$.

                              This is Davies (Lemma 7.7 on p. 35).

                              If $S \subseteq \mathcal{L}$ is a set of atoms such that every good racing option of every $G \in S$ is prime-factorisable, then every $S$-factorisation is uniquely factorisable.

                              This is Davies (Theorem 7.8 on p. 36).

                              If $S \subseteq \mathcal{L}$ is a set of atoms such that every good stalling option of every $G \in S$ is prime-factorisable, then every $S$-factorisation is uniquely factorisable.

                              This is Davies (Theorem 7.9 on p. 36).

                              If $G$ is a Left dead end with a good prime-factorisable option $H$, then $G$ is uniquely factorisable.

                              This is Davies (Proposition 7.10 on p. 37).

                              If $G$ is a Left dead end with a good option $\overline{n}$, then either $G = \overline{n + 1}$, or else $G$ is an atom.

                              This is Davies (Proposition 7.11 on p. 37).

                              If $G$ is a Left dead end, then, for every $n \in \mathbb{N}_1$, $G'$ is a good option of $G$ if and only if $G' + (n - 1) \cdot G$ is a good option of $n \cdot G$.

                              This is Davies (Lemma 7.13 on p. 38).

                              If $G$ and $H$ are Left dead ends, where $G$ is uniquely factorisable, and $H$ is prime-factorisable, then $G + H$ is uniquely factorisable.

                              This is Davies (Lemma 7.14 on p. 38).

                              If $G$ is a Left dead end that is an atom with a good prime-factorisable option, then $G$ is a strong atom.

                              This is Davies (Lemma 7.15 on p. 38).

                              If $G$ is a Left dead end with a good prime-factorisable option, then $n \cdot G$ is uniquely factorisable for all $n$.

                              This is Davies (Theorem 7.16 on p. 39).

                              8. Final remarks #

                              Every good option of a minimal counterexample to unique factorisation is a molecule.

                              This is part of the Section 8 discussion in Davies (p. 39).

                              Two factorisations of a minimal counterexample to unique factorisation share no atom.

                              This is part of the Section 8 discussion in Davies (p. 39).

                              A. Similarities with the impartial monoid #

                              TODO: Lemma A.1

                              E #