Survivorship bias and small-population variance can affect this result.
Small populations are more variant. Most of the best schools are small. But most of the worst schools are small too. Large schools tend to converge on a median. The most dramatic rates of cancer and absence of cancer are found in rural areas because they have small populations. Cities converge on a median. And so on.
We don't hear about the small, unsuccessful businesses. They don't get on panels because they no longer exist. So the small firms that did well are over-represented.
That said, that smaller projects perform better is a consistent finding in classical software engineering research.
Small populations are more variant. Most of the best schools are small. But most of the worst schools are small too. Large schools tend to converge on a median. The most dramatic rates of cancer and absence of cancer are found in rural areas because they have small populations. Cities converge on a median. And so on.
We don't hear about the small, unsuccessful businesses. They don't get on panels because they no longer exist. So the small firms that did well are over-represented.
That said, that smaller projects perform better is a consistent finding in classical software engineering research.