I like the simplicity, I like how it's easy to understand. It sounds like Hi-Lo, but in practice it's closer to Easy as 123 with more room for error.
The difficulty is in setting it up--unless you're using four prizes that are REALLY close to each other, this would probably be very easy for most contestants.
Just for the sake of argument let's pretend that the $1,000 prize is just a dollar cheaper to avoid ties:
AB ($2499) vs CD ($4500) - easy win
AC ($2999) vs BD ($4000) - win
AD ($3499) vs BC ($3500) - close win
BC ($3500) vs AD ($3499) - close loss
BD ($4000) vs AC ($2999) - loss
CD ($4500) vs AB ($2499) - loss
On paper it SEEMS like the contestant has a 50/50 chance of winning, and they would, if they were picking randomly. But they're not--to win the game all the contestant REALLY has to do in this scenario is identify one prize.
If they can identify the highest priced prize and avoid it, they win. If they can pick the lowest priced price, they win. Even if they pick both, they still win in this scenario.
The only way to lose would be to put the cheapest prize on the high end (which is hard to do--how often does the cheapest prize get picked in Most Expensive?). Even though these are made up-numbers, they illustrate that there's a LOT of winning combinations that most contestants will be able to identify.
On the other hand, if you try to lower the number of winning combinations, you have to make the prices more extreme--say. $500, $600, $1200, $1250. But then it becomes even more obvious which ones are the "high" prizes and which are the "low" ones.