I'm playing a game that has an optimization problem built into it that I am struggling to determine the correct solution for.
In this game, you have a probability to win determined by the function $f(x)$ posted below (e.g., when $x = 300$ you have a 20% chance of success, at $x = 500$ you have a 50% chance).
$$f(x) = \begin{cases} 0 & \text{if } x < 300 \\ (x - 300) \cdot 0.0015 + 0.2 &\text{if } 300 \le x < 500 \\ (x - 500) \cdot 0.001 + 0.5 &\text{if } 500 \le x < 1000 \\ \end{cases}$$
$x$ generally increases linearly with time $t$, i.e. $\Delta x = \Delta t$. You can attempt your first chance to win this game anytime $x >= 300$. However, if you fail, you cannot try again until $x = g(x') + x'$ where $x'$ is the value of $x$ that you last attempted to win with. Additionally there is penalty imposed in that $x$ increases at half the rate as it did before, i.e., $\Delta x = \frac 12 \Delta t$, until you have reached the new $x$ that you can (optionally) try again with, i.e., $x = g(x') + x'$.
$$g(x) = \frac{1000 - x}{10 \cdot 0.86}$$
Ultimately, the goal is to win the game with the minimal amount of time $t$. Given these constraints, how should I best approach solving at what $x$ should I first start attempting to win the game?
I am struggling with how to best approach this problem and the only solution I can think of is to code and run a large number of simulations where I first attempt at $x = 300, x = 301, ...$ and use the large volume of data to find the smallest value of $t$.
Variance: I note that an individual's tolerance for variance does come into play. For now I'm assuming that the player does not mind a high variance but I suppose I should also consider the case where a player wishes to have low vairance.