I believe Robin's just suggesting that to make a judgement about the effectiveness of prediction markets, you must look at a wide data set. Prediction markets clearly failed in the election, but you can't make the judgement that they're bad just based on this tiny dataset.