Consider the following infinite perfect information game with two players (the name I gave in the title of the post it totally made up): at each round $i \in \omega$, player $\mathrm{I}$ picks a natural number $x_i$ and a $\boldsymbol{\Delta}_2^0$ subset of the Baire space $\omega^\omega$ denoted by $A_i$ such that $A_{i-1}\subseteq A_i$ for all $i$; player $\mathrm{II}$ then picks a boolean value $y_i \in \{0,1\}$.


So at the end of the game player $\mathrm{I}$ will have built an infinite sequence $x = (x_n)_{n \in \omega}$ and an increasing chain (wrt set inclusion) of $\boldsymbol{\Delta}_2^0(\omega^\omega)$ sets $(A_n)_{n \in \omega}$ whilst player $\mathrm{II}$ will have built a boolean sequence $y = (y_n)_{n \in \omega}$ (an element of the Cantor space). Player $\mathrm{II}$ wins the play if at least one of these conditions is satisfied:

  • $\bigcup_{n \in \omega} A_n \not\in \boldsymbol{\Delta}_2^0(\omega^\omega)$
  • The sequence $y$ is eventually constant, i.e. $\exists k \ \forall n \ge k \ y_n = y_k$. Moreover The sequence $y$ is eventually equal to $1$ if and only if $x \in \bigcup_{n \in \omega} A_n$. Intuitively we require player $\mathrm{II}$ to "guess" whether the real played by $\mathrm{I}$ will or won't belong to the set $\mathrm{I}$ is building.

Now I'm wondering whether player $\mathrm{II}$ has a winning strategy in this game, or if the game is determined, and, eventually, under which hypotheses.

In a simpler game, in which player $\mathrm{I}$ does not keep changing the sets $A_i$, player $\mathrm{II}$ has a winning strategy (see, for this result and a wider discussion, this paper by Raphael Carroy Playing in the first Baire class, specifically Proposition 3.10).

Any idea? Thanks


1 Answer 1


Player I has a winning strategy: First play a singleton $A_0=A_1=\ldots=\{z_0\}$, for some real $z_0$, and the $x_n$'s consistent with $z_0$, until player II plays their first 1, if they ever do. After they play a 1 at stage $n$, continue with $A_{n+1}=A_{n+2}=\{z_0\}$, but make $x_{n+1}$ inconsistent with $z_0$, and proceed in this way until player II plays their first 0 after this stage, say at stage $m$. Then play $A_{m+1}=\{z_0,z_1\}$ with some $z_1$ where $(x_0,x_1,\ldots,x_m)\subseteq z_1$, and keep playing $A_{m+2}=A_{m+3}=\ldots=\{z_0,z_1\}$ and the $x_i$'s consistent with $z_1$ until player I again plays a 1. Then continue playing $\{z_0,z_1\}$ but play the $x_i$'s inconsistent with $z_1$, until player II next plays a $0$ at stage $m_1$, etc. The $A_n$'s and $\bigcup_{n<\omega}A_n$ are easily boldface-$\Delta^0_2$, and it's easy to see that player I wins.

(In the first edit I said something suggesting that every countable set is boldface-$\Delta^0_2$, but that's false, as e.g. the rationals are not boldface-$\Pi^0_2$, as they are dense but not comeager. If $\bigcup_{n<\omega}A_n$ is infinite above, note that its complement is the union of an open set with a singleton (the singleton contains the limit of the $z_n$'s), which is therefore boldface-$\Sigma^0_2$.)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.