The site's wording does not say which random walk is meant, and the answer depends on the law: for the walk whose steps are $(1,0)$ and $(0,1)$ with probability $1/2$ each, no site is visited twice, so $\lvert F(n)\rvert=n+1$ equals $r$ only at $n=r-1$ and the probability is $0$ for every $r\ge3$, while for the simple random walk it is $1$ at $r=3$. The poser's own text leaves the law open as well: the setup of Section 6.1 of [Va99], printed p. 11, which precedes Erdős and Révész's item 6.77 on printed p. 12, says only "a random walk on $\mathbb{Z}^2$", so the ambiguity is already in the poser's text and the site's wording copies it. The change inserts "simple" before "random walk", the walk whose independent increments are uniform on $(\pm1,0),(0,\pm1)$; nothing else changes, and the time-zero visit is counted as the site counts it. The evidence is the literature's statement of the question as Erdős and Révész's. Hao, Li, Okada and Zheng [HLOZ24], Section 1, display (1.2), state it for discrete-time simple random walk on $\mathbb{Z}^d$ for every $d\ge1$, apart from their theorem's range $d\ge2$. Tóth [To01], Section 1, printed pp. 484–485, independently states Erdős and Révész's favorite-site question for simple symmetric random walk on $\mathbb{Z}$. The site credits both papers, each about simple random walk, with the answer. The correction does not rest on the texts in which Erdős and Révész raised the question (1984, 1987 and 1991, cited by both papers). No result about any other law is recorded. The standing judges this precise Statement.