Learning with probability of winning
Goal: The behavioural regularities of increasing effort after losing, decreasing effort after winning, dropping out after consecutive losing and over-dissipation at the aggregate level are often found in experimental data in strategic games including the ones which can be modeled as contests and all pay auctions. The environment we consider is that of limited information where agents play the game repeatedly and know their own efforts and outcomes. We assume that in such an environment agents decision making is driven by some threshold probability of winning each period. We formulate a two parameter reinforcement learning model which has features of direction learning and can make a plausible explanation for such behavioural regularities together. The model can broadly explain the experimentally observed behaviour in Tullock contests in such an environment where other predominant theories are not able to.