It's a better than 50% chance - if the sample size is odd, you always get a real observation; if the sample size is even, you may still get a real observation (if the two median observations are equal).