The model in the paper accounts for that, having both "good luck" events that double the guy's income on a successful "talent" check, and "bad luck" events that

The model in the paper accounts for that, having both "good luck" events that double the guy's income on a successful "talent" check, and "bad luck" events that unconditionally halve income. The different runs show that luck weights a lot more than talent in the final outcome, thought there are strategies that alleviate this problem.