Retirement threshold with multiplicative capture risk (source code)

= Retirement threshold with multiplicative capture risk

Consider a finite-horizon decision problem with current nonnegative wealth $w$. Retirement preserves $w$. Action $i$ captures the player with probability $p_i\in(0,1]$ and then pays zero; with probability $q_i=1-p_i$ it survives and adds nonnegative reward $A_i$, with finite conditional mean $a_i$. The same action distributions remain available at each surviving step, independent of current wealth and previous history.

For finitely many actions put
$$
H=\max_i\frac{q_i a_i}{p_i}.
$$
If $w<H$, the maximizing action followed immediately by retirement has expectation $q_i(w+a_i)>w$. If $w\geq H$, retirement is optimal for every finite remaining horizon. For a proof, induction gives the sharper bound $V_n(z)\leq\max(z,H)$. Indeed, for $z\leq H$, $\max(z+A_i,H)\leq H+A_i$, so its continuation expectation is at most $q_i(H+a_i)\leq H$. For $z\geq H$, surviving wealth stays above $H$; induction reduces continuation to $q_i(z+a_i)\leq z$.

Thus the threshold is independent of the positive remaining horizon; equality allows retirement, possibly tied with continuing. The assumptions matter. A zero-capture action with strictly positive mean reward makes continuing strictly preferable at every finite wealth when a step remains, so its case cannot be represented by division by $p_i=0$.