[Matthew Shea] If you aren't missing 50% of your sprints, you are being too conservative in your estimation process. Missing sprint goals is completely normal and part of how any agile-ish process works. This was from my training that James Coplien did. It didn't land well, as expected, but it's a truth. What James didn't include was how to solve the real problem, which is stakeholders wanting estimates and dates - something directly counter to the Scrum process as-is.
I figured out a solution to this missing piece: Track the distribution of your points completed per head for a sprint. Plot this out on a histogram to see what the distribution looks like. Once you get enough data in to form a reasonable sample size, you'll be able to determine a few things.
1) Is the distribution roughly normal? If not, you may have some serious issues in your development process that need to be addressed.
2) If the distribution is roughly normal, you can now provide not just an estimated date, but a value determining the accuracy of the date. If you have 300 points left on a deliverable, and the average delivery per sprint is 30 points, you actually have a 50% chance of delivering in 10 sprints. Use various percentiles to go from there, such as p99.