I have a question concerning the proof of this theorem in Kechris' "Classical Descriptive Set Theory." The rules of the game i as follows: Let $X$ be a nonempty perfect Polish space with compatible complete metric $d$. Fix a basis $\{V_n\}$ of nonempty open sets for $X$. Given $A\subseteq X$, consider the game $G^*(A)$, where Player 1 (I) plays pairs of basic open sets (with certain requirements, but they are not necessary for my question), while Player 2 (II) moves by choosing one of those open sets, I plays two open sets from the set that II chose, and so on... (p. 149, Kechris).

Now, in the proof of $\Rightarrow$, Kechris does as follows: Assume $\sigma$ is a winning strategy for II. Given $x\in A$, call a position \begin{align*} p=((U_0^{(0)},U_1^{(0)}), i_0,...,(U_0^{(n-1)},U_1^{(n-1)}),i_{n-1}), \end{align*} $i_j\in \{0,1\}$, "good" for $x$, if it has been played according to $\sigma$ and $x\in U_{i_{n-1}}^{(n-1)}$. He then concludes that for each $x\in A$ there is a maximal good $p$ for $x$. Now, if $p$ is maximal good for $x$, then \begin{align*} x\in A_p =\{y\in U_{i_{n-1}}^{(n-1)}:\forall \,\, \text{legal} \,\, (U_0^{(n)},U_1^{(n)}), \,\, \text{if i is what}\,\, \sigma\,\, \text{requires II to play next, then}\,\, y\not \in U_i^{(n)}\}. \end{align*} He then claims that $A_p$ contains at most one point, and this is what I cannot convince myself of! Can anyone help me? :)

  • $\begingroup$ Kechris explains why $A_p$ contains at most one point in the proof ("Now notice that $A_p$ contains at most one point, since..."). What part of this explanation did you not understand? $\endgroup$ – Alex Kruckman Jan 9 '19 at 19:46
  • $\begingroup$ Yeah. I don't get, why he assumes that $y_i\in U_i^{(n)}$. $\endgroup$ – MLS Jan 9 '19 at 20:18

Suppose for contradiction that $A_p$ contains more than one point. So we have distinct points $y_0\neq y_1$ in $A_p\subseteq U^{(n-1)}_{i_{n-1}}$. Now because we're in a Polish space, we can find basic open neighborhoods $y_0\in U^{(n)}_0$ and $y_1\in U^{(n)}_1$, each with diameter less than $2^{-n}$, such that $\overline{U^{(n)}_0}\cap \overline{U^{(n)}_1} = \emptyset$ and $\overline{U^{(n)}_0\cup U^{(n)}_1} \subseteq U^{(n-1)}_{i_{n-1}}$.

So it is legal for Player I to play $(U^{(n)}_0, U^{(n)}_1)$, and the strategy $\sigma$ requires Player II to respond with $i\in \{0,1\}$. If $i = 0$, then $y_0\in U^{(n)}_i$, contradicting $y_0\in A_p$. And if $i = 1$, then $y_1\in U^{(n)}_i$, contradicting $y_1\in A_p$.

  • $\begingroup$ That really helped me - thank you so much! $\endgroup$ – MLS Jan 9 '19 at 20:34
  • $\begingroup$ @MLS You're welcome! Instead of thanking me in a comment, you can upvote my answer (if you think it's a good answer) and/or accept it (if it settles the question). $\endgroup$ – Alex Kruckman Jan 9 '19 at 20:40
  • $\begingroup$ I'm afraid I can't upvote, I'm too new here. $\endgroup$ – MLS Jan 9 '19 at 20:42
  • $\begingroup$ @MLS Ah, that's right, I forgot about the reputation restriction on voting. $\endgroup$ – Alex Kruckman Jan 9 '19 at 20:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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