To illustrate the concept of dependent trials, let's take a step back into our childhood. You're at the local grocery store, and standing before you is a gumball machine filled with exactly 10 white and 10 black gumballs. You know this because you saw the store manager refill it just five minutes ago. All the balls are thoroughly mixed together, and there's a line of kids waiting to purchase them.

Your friend, being the betting type and needing only a couple more dollars to buy that model airplane he's been eyeing, makes the following proposition. You are allowed to bet $1 (let's pretend we're rich kids) whenever you wish that the next ball to come out will be white. If a white ball comes out, you win $1. If a black ball comes out, you lose $1. Sounds like a simple enough game, you say to yourself.

The expected outcome of 0 implies that, over the long run, you are expected to neither win nor lose money if you bet on the very first ball. Of course, each individual wager will result in either a $l win or a $1 loss, but over time your expectation is to net 0 on this _50/50 proposition. Let's say, however, that one gumball has already come out, and you know that it was white.