Fig. 5.

An exemplary decision tree (above) with sensitivity analysis (below) for three strategies (going to $\mathsf{c}\mathsf{1}$, $\mathsf{c}\mathsf{2}$, and $\mathsf{c}\mathsf{3}$ represented with thick solid, dashed, and dotted line, respectively). The horizontal axis denotes the deviation in the probability of going to the upper node ($\mathsf{t}\mathsf{1}$, $\mathsf{t}\mathsf{3}$, and $\mathsf{t}\mathsf{5}$, respectively) independently for each strategy. $\mathsf{c}\mathsf{1}$ is optimal for baseline probabilities, shaded regions illustrate boundaries for stability (dark), $P_{max,\mathit{\epsilon }}$-optimality (medium), and $P_{min,\mathit{\epsilon }}$-optimality (light). It is $P_{\text{mode},\mathit{\epsilon }}$-optimal for all the deviations (mode-tending requires maximizing $\mathit{\Delta }p$). Gray horizontal lines drawn to help see where expected payoffs equate (observe that the expected payoff of the decision to choose c3 temains constant for $\mathit{\Delta }p>20\%$ as probability of t6 is equal to $20\%$). ${\mathit{\epsilon }}_1\approx 3.63\%$, ${\mathit{\epsilon }}_2=8\%$, and ${\mathit{\epsilon }}_3\approx 26.67\%$