Without double checking your math, I'm pretty sure the significant error in your conclusion here is modeling the games he's playing as coin flipping with even payouts. The specific game he's playing is actually dropping a percentage each time and is more accurately modeled as "heads I win .98, tails I lose 1". (Technically he's betting $15 for a chance to win $14.69). With that in mind, many fewer than 1 in 28 players would end up with his results.