Elicited preference rankings for two lotteries are typically inconsistent in choice and price. We test whether pre-play learning makes preference rankings consistent. Pre-play learning denotes ex-ante lottery learning, where subjects observe playing lotteries before making decisions. We find that pre-play learning makes the average selling prices for the p-bet, of subjects who choose the p-bet, higher than their average selling prices for the $-bet. However, pre-play learning is not strong enough to equalize the rates of standard and non-standard reversals, although pre-play learning reduces the rate of standard reversals.