Oh wow, been a while since this one was argued. In reading through it seems the debate is over the "odds" as determined by statistical analysis of a situation where one of two coins is selected and you have some toss results.
Assuming all is above board (because if it isn't we have no real means to judge an answer), then it works out this way.
Initially, we are shown there are two coins, one fair and one double-headed. One is chosen randomly using a fair process, and its identity not revealed. At this point, we only have setup information, and the odds are 50/50 for either coin.
Now the way statistical analysis works when trying to find the odds a certain assertion is either true or false is this. We assume for each case that any possibility consistent with that case could have happened. When we are told what DID happen, then we eliminate those possibilities that DIDN'T happen.
So if the coin was to come up tails, then we eliminate the 100% of the possibilities the coin is double headed, and only 50% of the time the fair coin comes up heads. What we are left looking at, statistically speaking is the 25% of the time that the fair coin would be picked and come up tails, and since that constitutes 100% of the possible remaining odds, the chance it is true is 100%.
So now, we toss the coin and it comes up heads. This happens 100% of the time with a double headed coin, but only half the time on a fair coin. We've now eliminated 25% of the possibilities, and are looking at the remaining 75%, of which 50% is the H-H coin chosen, and 25% the H-T.
Another toss and another head, and the chances of the fair coin having been chosen halves again.
Keep in mind, we don't know what was picked, we're only setting odds based on the behavior exhibited and an assumption of randomness of the experiment.
Think of it this way. The more often the coin comes up heads without coming up tails, the less likely it is that the coin and experiment are fair. You can never be totally sure, of course, but after 50 flips coming up heads, you can get pretty close to certain.
Of course, if there is ever a tails, then you are certain that the coin has a tails side.