Mancala games refer to a large family of strategy games. The research described in this paper investigates strategies for the game. We initially identify a useful set of simple heuristics for the game and then evolve a strategy that combines these heuristics in a useful way. Our overall goal in this research is to develop robust strategies that do not require a huge look-ahead for a range of mancala games. In this paper we describe our results obtained on one particular variant, namely, Bantumi. The results are analysed to assess the effect that changing the number of seeds per bowl has on the best evolved strategy. We investigate the weakness of the best evolved strategies and discuss ways in which they can be further improved.