Prescribed burning for fuel reduction is a major strategy for reducing the risk from unplanned fire. Although there are theoretical studies suggesting that prescribed fire has a strong negative influence on the subsequent area of unplanned fire (so-called leverage), many empirical studies find a more modest influence. Here, I develop a series of simulations to explore the landscape drivers of leverage. Leverage declines with treatment level in a nonlinear, 'decay' relationship, implying diminishing effectiveness. The spatial configuration of the prescribed fire treatment has a major effect: long linear (gridded) barriers are far more effective than patch barriers, but gaps in the grid lead to large reductions in leverage. However, the extent of unplanned fires in the landscape has the largest influence such that a landscape with 3% annual extent has only one-fifth of the leverage of a landscape with 28%. Leverage decreases with the probability of spread, suggesting that treatment is less effective when fire weather is severe. For gridded designs, leverage increases with the size of individual fires, but this is not the case for patch designs. These results agree well with recent empirical studies finding that prescribed burning has only a modest effect on subsequent unplanned fire in many biomes. They also help to explain why those empirical studies report lower effectiveness than many simulation studies. In practice, leverage values >1 (replacement of unplanned with planned fire) are hard to achieve.