There's also the fact that the round didn't close 10 seconds ago; things have undoubtedly changed since it did. Those changes could positively and/or negatively affect the value of the equity.
Once you're past that, your approach is basically ok. It's a bit like the Drake Equation, though, in that it's several unknowns multiplied together to achieve a very precise result that has such large accumulated error bars as to be meaningless. The best you can do with this approach is try to avoid taking a worse deal than the last investors, which doesn't mean much.
Also, no one has a 25% chance of success unless they've filed an S-1. But that's just a detail.
You can come up with a different factor but all you are doing is effectively negotiating a different price. Your example thinks it is a $2.5MM company instead.
(1/Probability of Startup Success) * (50k/10M)
So if they have a 25% chance at success, then you should be getting 4*.5 = 2%. Otherwise, you are not taking the expected value, you are taking probability of success = 1. To make matters worse, assume you wont see the pay out for atleast 5 years. One in the hand is better than two in the bush, so there is a discount there as well.
Am I missing anything here?