Polls and betting odds

Speaking to someone the other day who is doing a lot of work around this and the difference between opinion polls and betting odds are interesting.

The betting odds are calculated on the basis of prediction combined with, of course, the betting market - what people are putting their money on. The people who put money on the outcome of the election tend to have a very high vote habit. If you bet on a result you're more likely to vote to swing the result in favour of your bet (unless you're an "emotional hedger" - voting one way and betting to recover some cash just to cheer yourself up in case it goes against you. There are people who do this). The sample is therefore self selecting and largely politically active.

By contrast the polls involve dip testing samples 5000 people spread across socio economic groups. They are polled on what they would prefer, what confidence they have in parties and individuals and what their voting intention is. However, the rate of non-voting participant is quite high as is the tendency to have a "wants/likes" opinion a month or week prior to the election which is quite different to what they do in the ballot box, if they go there at all.

Online betting has increased the field of those having a punt very significantly such that this is not really such an anomalous sample. Pollsters are now looking at the odds and factoring in betting patterns too.

Odds of up to 20-1 on (i.e. 1-20) with a modal bet of 1-16 on a Conservative majority are now being offered post the weekend's manifesto announcements. You can get 10-1/12-1 on Labour. Lib Dem 100-1.

https://www.oddschecker.com/politics/british-politics/next-uk-general-election/most-seats

odds on me giving a tug - 2000/1

I got 42 on a labour majority last night on bet exchange - don’t think they will get one but think polls will narrow significantly from here so good cash out possibilities

Polls feeding off betting odds sounds dangerously circular.

If you ever look at the raw data from polls compared to the published projections, the level that they distort the base numbers with highly judgemental assumptions and modelling techniques is astonishing. It helps to understand why they've been fairly bad in recent years as even small misestimates in things like youth turnout have a massive impact.

There’s also some analysis to be done on who bets how much, I remember following the odds on the day of the referendum, they went out to 5 or 6 to 1 against Leave on the day, and I remember reading them that the clues were still there, there were a greater number of bets on Leave, but they were for smaller amounts. More money went on remain but fewer actual bets. Various explanations were proffered, from Leave being an emotional bet vs Remain where the “smart” money was to the comparative incomes of Leavers vs Remainers.

One would assume the bookies would take account of that but frankly fvck knows if they do, and/or how this is reflected in the betting exchanges.

I would like to know more about this.

@pancake

I wasn't suggesting polls were feeding off odds. I was saying that the chap speaking about it was saying there seemed to be more rigid and reliable data in the betting odds than the polls.

Here is the current poll of polls btw

shitstix

Mutters, that’s what everyone thought until the referendum.

I can’t actually remember how the polls did vs the odds in 2017, I’ve a feeling it was much closer.

I have also been researching the question of whether there is a negative correlation between the success levels of England's Test team (not one day or t-20) and Tory election successes.

Thatcher's election years - 79, 83,87 - and Major's - 92 - saw moments of true ghastliness with occasional rescues, but largely heroic failure. The Ashes were won in 2005 under a Blair administration. We are now reverting to a culture of collapsing top order and look more like we used to.

For a small size poll, don't you need random selection for you to be able to rely on the data? The polling data pays a lot of attention to that. The betting data does not. So, whilst 5,000 people placing bets may give a good indication of how those 5,000 people will vote, that is not statistically significant so you cannot extrapolate to actual results. A randomised sample of 5,000 may enable you to make more general conclusions.

I have no idea what the size of the market is for political betting, however.

the betting markets are absolute bullshit when it comes to predicting anything, and it’s been proven so time and time again. The people who bet on political outcomes are a tiny fraction of the voting public.

quite tempted by 12/1 on a labour madge

I've taken advantage of some unusually generous betting markets in politics. As a result, I cleaned up at the 2015 election. I managed to get 20/1 on Ukip taking the Isle of Wight which, although it didn't happen, is a sign of the madness. They got a seriously chunky slice of the vote.

so what you're saying is you lost because you placed money on a bet that the bookies had thought will never happen and it didn't happen so they were right.

taken advantage is a somewhat optimistic analysis

No, I'm saying that I won big time. There were one or two small instances where I lost, but they were on hyper--inflated odds, so were worth a punt.

Mutters, do you understand betting?

umm... ok.

The example you give is a curious choice in the circumstances of your floor-cleaning bet strategy though. but carry on.

Well, it was the most extreme case of disproportionate odds that sprang to mind. Sorry, I made something like 70 individual bets that election, so forgive me if I can't remember the details without digging out my spreadsheet.

It's ok. No need to go spreadsheet nuclear on this one.

well done for winning big time, however you did it.

I haven't ever placed a bet on a political event. I did win 4/5 races on ladies day at Ascot and came away three grand richer this year, though. Happy days and thanks to Frankie D.