The bad news is that no... let's call it an "off-the-shelf" metric... will immediately give you good results. There are too many variables between development teams and too much variation in how different people may apply the metric.
The only way you can get good results in estimation is by taking steps to chain your estimates to reality. Estimates do not just fall in line with reality, they must be wrestled there. That means you must make estimates, compare the estimates to what really happened, and adjust your process for generating them appropriately.
For something more concrete, Joel on Software has a great article on Evidence-Based Scheduling which is worth a read. I don't necessarily think this is the "best" process but it's a starting point which is better than 95% of software teams ever achieve.