In this thesis report, we delve into a contest game within game theory, where agents’ risk-taking capacity, rather than effort, becomes the pivotal variable. High-quality performance in the game is associated with a higher probability of leading to superior scores through the use of risky strategies. The application of this con test game to labor markets prompts considerations for both employees and employers. From the perspective of employees, understanding the balance between risk-taking and payoff can be helpful in decision-making. On the employer’s side, the efficiency of selection mechanisms becomes a critical factor. We assess selection efficiency by examining the winning rates of high-type individuals. We use two parameters—market quality and market size in our analysis. Surprisingly, our theoretical analysis reveals a non-monotonic relationship between these factors and selection efficiency. Contrary to expectations, we find that as market quality improves or the number of agents increases, the winning rates of high types may decrease, resulting in reduced selection efficiency for employers. Simulation experiments inspired by Fictitious Play and evolutionary game theory are conducted to research deeper into these dynamics. Learning rules and replicator dynamics under four scenarios are designed to address the inherent volatility in agents’ strategic choices, test optimal strategies, and enable a comprehensive comparison of selection efficiency. A mechanism is proposed, derived from agents gaining experience from their usual behavior, and attempts to align outcomes more closely with Nash equilibrium, improving the optimal result. The study’s unexpected findings about single-round screening in certain conditions highlight the need for tailored selection processes in different markets. In summary, this research brings a fresh perspective to contest games. It encourages a rethink of traditional ideas and provides practical insights for decision-makers, especially in labor market.