Investigating Live Betting Accuracy: Do bookmakers' odds match the probability of outcomes?

[nerdbettin]

Hello,

I read an article on the Pinnacle blog ( https://www.pinnacle.com/en/betting-art ... 8P6NRD6ZYK) that discusses the accuracy of closing odds. If you reduce the margin, the odds are remarkably accurate on average. Bookmakers are 99,97% accurate.

However, what about live trading odds, particularly when unexpected events occur? For example, when the favorites dominate at home, are bookmakers still accurate? During such times, the market may become volatile as people try to minimize their losses, potentially distorting reality.

To ask my question differently, has any study analyzed the accuracy of bookmakers during live betting? For instance, if a bookmaker prices the surprising Team B dominating Team A to win at 25%, do they actually win 25% of the time on average? If not, are there any websites that can help me analyze live data and compare it to real-life results?

Perhaps we can collaborate on this.

Thank you.

[ShaunWhite]

It's not possible to know if the odds on any given match at any time are correct, same with any event or horse. The accuracy can only be established as an average over very large samples. The fact bookies offer and continue to offer in-play odds suggests they find it profitable at the odds they're offering.

However, those averages are comprised of individual situations where the odds are undoubtedly incorrect, theoretically every price at any time is incorrect given that a precise probability is unknowable. But whether or not you can quantify vague situations like "dominating", apply that to a sufficiently large sample, and state that the market is on average skewed on way or the other is questionable.

Personally my first question to you would be, who gives a fck about bookies prices when Betfair prices are always better? And why use bookies when closing a position (when the current odds swing the opposite way in terms of value) means further back bets incurring the full bookies margin when to do that on Betfair means you only have to incur the <1% across the back and lay spread and lay.

But to answer your question, many people build highly sophisticated price modelling systems, fed with a huge range of metrics with the objective of identifying value at any given time, but the likelihood that any of them would be prepared to bring in someone else to collaborate would be slim unless they felt that had some insight to bring to the party. But you might find someone starting from the same level as you to work with. Or, you could consider collaborating with ChatGPT? It would certainly help you with the heavier statistical analysis stuff and any machine learning tools you might need.

You've done a good job of writing the job description of a trader though

[Anbell]

It's not possible to know if the odds on any given match at any time are correct, same with any event or horse. The accuracy can only be established as an average over very large samples. The fact bookies offer and continue to offer in-play odds suggests they find it profitable at the odds they're offering.

However, those averages are comprised of individual situations where the odds are undoubtedly incorrect, theoretically every price at any time is incorrect given that a precise probability is unknowable. But whether or not you can quantify vague situations like "dominating", apply that to a sufficiently large sample, and state that the market is on average skewed on way or the other is questionable.

Personally my first question to you would be, who gives a fck about bookies prices when Betfair prices are always better? And why use bookies when closing a position (when the current odds swing the opposite way in terms of value) means further back bets incurring the full bookies margin when to do that on Betfair means you only have to incur the <1% across the back and lay spread and lay.

But to answer your question, many people build highly sophisticated price modelling systems, fed with a huge range of metrics with the objective of identifying value at any given time, but the likelihood that any of them would be prepared to bring in someone else to collaborate would be slim unless they felt that had some insight to bring to the party. But you might find someone starting from the same level as you to work with. Or, you could consider collaborating with ChatGPT? It would certainly help you with the heavier statistical analysis stuff and any machine learning tools you might need.

You've done a good job of writing the job description of a trader though

Thank you Dallas for such a comprehensive reply , without even a hint of curmudgeonliness. The nicotine holiday seems to be treating you well recently

[Euler]

Personally, I always think the academic approach to market efficiency is flawed.

If you have two selections but both are poorly priced, when you measure the market from a top level it looks perfect as one of the two won and the other lost 100% of the time.

Markets will always be efficient in aggregate, but not in isolation.

[nerdbettin]

It's not possible to know if the odds on any given match at any time are correct, same with any event or horse. The accuracy can only be established as an average over very large samples. The fact bookies offer and continue to offer in-play odds suggests they find it profitable at the odds they're offering.

What you are saying is incorrect, but I believe you may have said it mistakenly. Of course, it is possible to evaluate the accuracy of odds. By verifying the results over a large sample size, you can determine if the odds were accurate. For example, if bookmakers priced events at odds of 2.00 (implying a 50% chance of occuring), and the events did occurred 50% of the time over a large sample, then the odds were accurate.

As mentioned in the article I referenced, the efficiency of Pinnacle without margin is particularly impressive, especially at closing odds. Why closing odds, in particular? It's because that's when bookmakers and professional punters have all the information needed, and the market has higher liquidity. So, yes, bookmakers can be highly efficient (Please read the article). However, my question is whether they remain efficient in live betting, and as far as I know, there haven't been any studies on this. Do you know any?

However, those averages are comprised of individual situations where the odds are undoubtedly incorrect, theoretically every price at any time is incorrect given that a precise probability is unknowable. But whether or not you can quantify vague situations like "dominating", apply that to a sufficiently large sample, and state that the market is on average skewed on way or the other is questionable.

You make a valid point that individual situations can lead to incorrect odds. It is true that determining precise probabilities is often impossible. However, by examining large samples and quantifying situations like "dominating" relative to the outcomes, it is possible to analyze the overall skewness of the market. Further research and data analysis are required to investigate this aspect. It's important for bettors not to solely rely on intuition but to verify it with empirical evidence.

Do you know a place where historical live odd data are already available to buy/download ?

Personally my first question to you would be, who gives a fck about bookies prices when Betfair prices are always better? And why use bookies when closing a position (when the current odds swing the opposite way in terms of value) means further back bets incurring the full bookies margin when to do that on Betfair means you only have to incur the <1% across the back and lay spread and lay.

Excuse me for the confusion earlier. My initial question relative to the accuracy of bookies indeed includes platforms like Betfair and other betting exchanges. So, using the term "the market" instead of "bookies" would be more appropriate.

But to answer your question, many people build highly sophisticated price modelling systems, fed with a huge range of metrics with the objective of identifying value at any given time, but the likelihood that any of them would be prepared to bring in someone else to collaborate would be slim unless they felt that had some insight to bring to the party. But you might find someone starting from the same level as you to work with. Or, you could consider collaborating with ChatGPT? It would certainly help you with the heavier statistical analysis stuff and any machine learning tools you might need

Yes, conducting research in this field can be a lot of work and requires a comprehensive approach..

You've done a good job of writing the job description of a trader though

I feel lot of sarcasm in your response, am I right or is it just a feeling ?

[ShaunWhite]

No sarcasm intended.