Michael Dent the creator of PoliticsOdds.bet, a site which tracks and graphs betting prices on political events, looks at the EU Referendum in-play betting

At 11:36pm on June 23rd, just before the first result was declared, the market was just short of 90% confident of a Remain vote. So much for markets knowing best – the market was wrong, and staggeringly confident in its wrongness.

So how did we get from there to settling the market for Leave? What follows is an attempt to analyse five hours of in-play betting. Of course, it is highly speculative, and it’s easy to create a narrative with perfect hindsight, so comments are welcome!

Inertia

When early results indicate a possible surprise event, the market is stubbornly slow to adjust. We saw it in the general election last year, when even after the shock exit poll it was possible to get excellent odds on a Conservative majority. And we saw it again here.

As results rolled in, suggesting that at the very least it was much closer than previously assumed, you could almost hear the excuses – that these must be freak results, that they did not represent a trend. But a trend of Leave over-performance against the parity model was indeed emerging, particularly by Swindon’s declaration. As a result, there was value to be had on the Leave price.

I think a bold gambler starting the night with a neutral position would have built a position on Leave here. Bets placed during this phase could have yielded a 200-500% profit.

Animal spirits

At 1:56am, a period of almost an hour of relative stability in the betting market turned to severe volatility for the next 90 minutes. The market moved fast and sharply. Within 45 minutes we had an astonishing three distinct crossover events.

During this period, I believe individual results moved the market far more than they should have done, magnified by animal spirits and confusion as gamblers were hypersensitive to any result that appeared likely to change the narrative.

Our smart gambler would have compared each result to her parity model at this point, and realising that the individual results did not signpost a significant change, would have stuck to her hypothesis, aiming to top up her position at the best prices. A more risk-averse gambler might well sit this period of volatility out and just watch the data, and no-one would blame him!

Acceptance

From 3:20am, the market began to accept a Leave outcome.

Our smart gambler had made her money by now and would have eased up on the betting. A risk-loving gambler (or one desperate to cover a loss) can still make some money here – even when the outcome looked pretty clear, profits of up to 30% were still available.

But the risk-reward pay-off was weakening fast at this point, and by the time available profits fell below 10%, any sensible gambler should have been of Leave, and perhaps building a small hedge position on Remain at >10 odds, to protect against a last minute black swan event (when the graph looks like this, anything might happen!)

A word on volumes

I was astonished by the in-play volumes. As PB reported, by 6am on polling day the EU referendum had become the first market on Betfair to surpass £50m in cumulative bets. But by the time the broadcasters called the result for Leave, the total matched on Betfair had more than doubled £113m.

So despite a four month campaign, 55% of all betting was matched in the final 24 hours. On three occasions during the in-play period covered here over £800,000 was matched in a two minute period. Most two minute periods saw at least £200,000 matched. Truly the biggest political betting event ever.

Lessons

In the chaos of in-play betting on an event like this, a calm head allows a smart gambler to make money. Or to keep their shirt – for me personally, I started the night with a significant position on Remain at 1.55 average odds. By the end of the night, I was able to finish with a small profit. I was lucky.

I shouldn’t have been able to get away with that – I should have lost my shirt. But by keeping a calm head and following Mike & others, I was able to turn a potentially very bad night for me into an OK one.

Michael Dent

Thank you to everyone who followed the betting prices at politicalodds.bet. I apologise for the server crash on 20 June and the fairly slow service during some of the most exciting moments of 23/24 June. We’ll be working on improving capacity for future events. In the next six months we’ll be closely following the Conservative leadership contest and the US Presidential Election.