18
$\begingroup$

A set $X\subseteq\mathbb{R}$ is strong measure zero if, for every sequence $(\epsilon_i)_{i\in\mathbb{N}}$ of positive reals, there is a sequence $(I_i)_{i\in\mathbb{N}}$ of open intervals covering $X$ such that $\mu(I_n)<\epsilon_n$.

Looking at various computability theoretic questions around strong measure zero (see e.g. https://math.stackexchange.com/questions/1446602/anti-random-reals), I've run into two possible strengthenings; I'm curious what is known about them.

For the first, we demand that the cover depend on the sequence one bit by bit:

  • $X\subseteq\mathbb{R}$ is strategically strong measure zero if player II has a winning strategy in the following game: on move $n$, player I plays a positive real $\epsilon_n$, and player II plays an interval $I_n$ with $\mu(I_n)<\epsilon_n$; and player II wins if the $I_n$ form a cover of $X$.

For the second, we merely ask that the cover depend on the sequence continuously. Let $Eps$ be the set of infinite sequences of positive reals, and let $Int$ be the set of infinite sequences of open intervals in $\mathbb{R}$, each topologized as usual.

  • $X\subseteq\mathbb{R}$ is continuously strong measure zero if there is a continuous $F$ from $Eps$ to $Int$ such that, for every $f\in Eps$, $F(f)$ is a cover of $X$ and $\mu(F(f)(n))<f(n)$.

Clearly strategically strong measure zero implies continuously strong measure zero implies strong measure zero; consistently all strong measure zero sets are countable (this is Borel's conjecture), so no nonimplications can be proved over ZFC. My question is whether we can say anything else; in particular,

Are any other implications provable over ZFC+A, where A is some reasonable axiom which does not imply Borel's conjecture?


Note that ZFC proves that there are continuum many continuously strong measure zero sets $S_r$ such that every continuously strong measure zero set is contained in one of the $S_r$. This seems likely to not be the case in general for strong measure zero sets in ZFC, so I suspect that the situation is not entirely trivial.

EDIT: To clarify, this informal argument merely suggests that not every strong measure zero set is continuously strong measure zero; as Andreas' answer below shows, it's quite likely that every continuously strong measure zero set is countable, which would be triviality in the other direction.

FURTHER EDIT: See "Nicely" strong measure zero sets for a continuation of this question.

$\endgroup$
  • $\begingroup$ You will probably be interested in F. Galvin's, Indeterminacy of point-open games $\endgroup$ – Paul Plummer Sep 27 '15 at 3:11
  • 3
    $\begingroup$ I don't think you want $Eps$ to be topologized as usual. It would be more interesting to use Baire space instead of $Eps$ and let $f \in \omega^\omega$ correspond to the sequence $\epsilon_n = 2^{-f(n)}$, for example. The issue is that $Eps = (0,\infty)^\omega$ is connected. If $x \in X$ and $X$ is csmz via $F$, we can define $H_x:Eps \to \omega$ by $H_x(f) = \mu n[x \in F(f)(n)]$. Since $\omega$ is discrete and $H_x$ is continuous, $H_x$ must be constant. If $x \neq y$, then $H_x$ and $H_y$ can't have the same constant value and it follows immediately that $X$ is countable. $\endgroup$ – François G. Dorais Sep 27 '15 at 13:35
  • $\begingroup$ @FrançoisG.Dorais That's a good point - this is what I get for not paying attention to the details! $\endgroup$ – Noah Schweber Oct 12 '15 at 3:47
20
$\begingroup$

Strategically strong measure zero is equivalent to countable. To prove the nontrivial direction, suppose $X$ is strategically strong measure zero and $s$ is a winning strategy for player II. Consider the tree of finite plays in which player I plays only rational numbers and player II follows $s$. For any partial play $p$ in which player I is to move next (i.e., the length of $p$ is even), say that an element $x\in X$ is killed at $p$ if, no matter what (rational) move player I makes next, player II's response (using $s$) is an interval that covers $x$.

I claim first that, at any $p$, at most one $x$ is killed. The reason is that, if $x\neq x'$, then player I could play an $\epsilon<| x-x'|$ and then player II's response would be an interval too short to cover both $x$ and $x'$.

Since the tree of partial plays is countable (that's why I restricted player I to rational $\epsilon$'s), only countably many $x$'s are killed anywhere in the tree.

To finish the proof, I claim that every element of $X$ is killed somewhere in the tree. Suppose, toward a contradiction, that $x$ is a counterexample. Then, at the initial position, player I can move so that II, using $s$, doesn't immediately cover $x$ (because $x$ isn't killed at the initial position). Then, player I can again move so that II doesn't immediately cover $x$ (because $x$ isn't killed at the position after II's first move). continuing in this way, player I can ensure that II never covers $x$. Since $x\in X$, this contradicts the assumption that $s$ is a winning strategy for II.

$\endgroup$
  • $\begingroup$ This is very nice! It doesn't obviously extend to the continuously strong measure zero case, though, does it? The natural game to play in that case would be: player I plays epsilons, player II plays intervals, but is allowed to pass. But in that game, many reals can be killed at a given stage, so the analysis seems to break down. Or am I missing something? $\endgroup$ – Noah Schweber Sep 26 '15 at 23:36
  • $\begingroup$ Something I missed at first glance, which I'm mentioning here because it might be interesting to others: this argument is essentially the same as the proof that if Empty has a winning strategy in the perfect set game for $X$, then $X$ is countable. $\endgroup$ – Noah Schweber Sep 27 '15 at 1:34
  • $\begingroup$ Thank you very much for this answer! I've accepted, and asked a new question expanding on the rest of this question: mathoverflow.net/questions/220687/…. $\endgroup$ – Noah Schweber Oct 12 '15 at 4:18

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.