This is such a hard thing to measure, since programming isn't about how many widgets. Even how fast you complete a project. Often the team is making decisions about interface and data design that impact the project down the line. With 2 teams working, you would get 2 different products and I think it would be hard to see which is the better product without significant time. You could also have the exact same team make the same product 5 times and get 5 different products. Experiments with humans are hard