Upper bounds on the smallest size of a complete cap in PG(N,q), Nge3, under a certain probabilistic conjecture

In the projective space $\mathrm{PG}(N,q)$ over the Galois field of order $q$, $N\ge3$, an iterative step-by-step construction of complete caps by adding a new point on every step is considered. It is proved that uncovered points are evenly placed on the space. A natural conjecture on an estimate of the number of new covered points on every step is done. For a part of the iterative process, this estimate is proved rigorously. Under the conjecture mentioned, new upper bounds on the smallest size $t_{2}(N,q)$ of a complete cap in $\mathrm{PG}(N,q)$ are obtained, in particular, \begin{align*} t_{2}(N,q)<\frac{\sqrt{q^{N+1}}}{q-1}\left(\sqrt{(N+1)\ln q}+1\right)+2\thicksim q^\frac{N-1}{2}\sqrt{(N+1)\ln q},\quad N\ge3. \end{align*} A connection with the Birthday problem is noted. The effectiveness of the new bounds is illustrated by comparison with sizes of complete caps obtained by computer in wide regions of $q$.