Bookmakers and betting exchanges were well in the "Remain" camp, signalling a 90% chance of a REMAIN victory until actual vote results started to show that "Leave" was going to win, and then all hell broke lose to reverse wrong "Remain" bets.
As P. Krugman has speculated, "prediction" betting markets do nothing other than reflect current opinion and have no special insights into the future.
I hailed this Krugman comment in 2015:
I’m not a huge believer in prediction markets, which seem more to reflect conventional wisdom than to offer profound insights....The view that "prediction" markets have a special ability to predict the future is a fundamental misunderstanding of Hayek's concept of prices as signals.
Hopefully in the future, we won't see a lot of this type stuff that appeared in WSJ in an article titled,Who to Trust on Brexit Predictions, just days ago when prediction markets were "signalling" a REMAIN win):
One reason prediction markets seem to beat the polls is that the polls capture public sentiment in a limited time frame—instead of measuring the likelihood of something actually happening, said David Rothschild, an economist at Microsoft Research in New York. Prediction markets can reflect other data such as expected voter turnout, poll discrepancies or other uncertainties that could sway the odds of an event happening, he said.
“When it comes to head-to- head competition in predicting the likelihood of an event, I’d pretty much always take the prediction market’s price,” Mr. Rothschild said.
Unlike polls, which can run into methodological issues, the prediction markets “quickly incorporate new information, are largely efficient, and are impervious to manipulation,” a 2012 National Bureau of Economic Research paper written by Erik Snowberg, Justin Wolfers, and Eric Zitzewitz found.More on "prediction" market failures here.