Here's the puzzle:
Your wish has been granted, and you get to play tennis against Roger Federer in his prime in the Wimbledon final. You have only a 1 percent chance to win each point, but Roger, sporting gentleman that he is, offers to let you name any score and begin the match at that point. (So, if you’ve entertained a fantasy of storming back after being down three match points in the fifth set, now’s the time to live it.) What score can you name that gives you the best chance to win, and what is your chance of winning the title?
I think it's almost immediately obvious that you should choose the score to be something like 6-0 6-0 6-6 (6-0). ie, you're 6 points up in a tie breaker in order to win your third set. This means you have six match points, each of which is a 1/100 chance of glory. The probability that you miss all 6 is just under 6%. Specifically, they are 1-0.99^6 = 5.85%. You can add an extra 0.01% chance of you winning the match from 6-6 in the tiebreak, for a total probability of winning of almost exactly 5.86% - the chances of winning in any other way than winning 2 out of the first 8 points that you play are significantly less than 0.01%, so can safely be ignored.
Incidentally, this last point is the reason why you choose to be 6-0 6-0 6-6 rather than 6-0 6-0 5-0. In the latter case, the best you can do is 3 match points, for roughly a 3% chance of winning the title, and after Federer pulls the game back to level, the way tennis scoring works means that you will essentially never get another chance. Think of it this way: in order to win the game after the score is level in the game, you need to win at least two points in a row. But the chance of you winning any two points in a row is 0.01%, so all the extra chances you have by starting at 5-0 up are just multiples of 0.01%, rather than multiples of 1%, and even to get the stage where you win 2 points in a row is really, really difficult.
That raises the obvious follow up question. How good would you have to be to choose 6-0 6-0 5-0 rather than the tie breaker? I suspect if I were feeling adventurous I could calculate this by hand, but I'm not, and since I have to write this on a computer anyway, I might as well write some code that simulates tennis matches. You can see the graph on the right. The x axis is your probability of winning each point, and the y axis is your probability of winning the match.
Incidentally, while I've got this code running, I might as well generate a plot of how often you win each match given different probabilities of winning points (I'm not going to take into account serving, and I'm allowing the last set to be own on a tiebreaker, since presumably the difference that makes is negligible). I've started the plot at 50, since it's obviously symmetric around this. Based on this, if you have a 45% chance of winning each point, you have just a 5% chance of winning the match. This is pretty much the same as the chance you'd have of winning the most points at those skill differences: matches last about 250 points and the odds of a binomial distribution with 251 trials and p=0.45 generating more than 126 successes are about 5.5%. However, things rapidly go downhill for the underdog. If you have a 42% chance of winning each point, you win about .3% of the matches, whereas you win the most points about .6% of the time. Tennis scoring does amplify skill differences, but tennis matches consist of *a lot* of points, so the better player is almost guaranteed to win, however you structure the scoring.
You can see the (not particularly elegant) Python code that I used to generate these figure here.