Machine Learning About Venture Capital Choices

European VC's missed five unicorns that are currently valued at over 2BN euro each. This study shows that VC's misallocate capital in predictable patterns by analyzing a dataset of french companies and investments. These patterns show that they are ignoring objective company metrics and are over-relying on subjective founder qualities.

Sounds like a VC firm that employs machine learning might actually out perform firms relying purely on human judgement. Of course the dataset only covered France, so it's uncertain if this research would replicate in the much bigger US market.