Conjecture on arbitrarily oscillating win-probability derivatives

Let \normalfontWin(v,w;p){\normalfont{\operatorname{Win}}}(v,w;p) denote the probability that the binary word vv occurs before the binary word ww in an iid Bernoulli(p)(p) sequence.

Oscillation conjecture. There exists a constant C>0C>0 such that, for every positive integer kk, there is a pair of binary words (v,w)(v,w) with lengths at most exp(Ck)\exp(Ck) for which

ddp\normalfontWin(v,w;p)\frac{d}{dp}{\normalfont{\operatorname{Win}}}(v,w;p)

changes sign at least kk times as pp ranges over (0,1)(0,1).

Win probabilities are rational functions of pp, with endpoint values constrained to {0,12,1}\{0,\frac12,1\}. The conjecture predicts that despite this algebraic structure, derivatives of win-probability functions can exhibit arbitrarily many sign changes using pairs whose word lengths grow only exponentially in the number of changes. The source presents this as supported by heuristic arguments and computer checks, with no resolution given.

Sources & referencesView supporting material

Primary source

Mathew Drexel, Xuanshan Peng and Jacob Richey, “Word length, bias and bijections in Penney's ante”, arXiv:2409.19195 (2024).

Progress summary

Never refreshed

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Solutions 0

No solutions have been posted yet.