Beating the Price

[MySportsAI]

Taken from an article I wrote in 2011

For those researching betting angles to Betfair prices, Betfair SP is often the yardstick used in order to see how an approach would have performed to Betfair betting. The problem we face however is what would the BFSP actually have been if we had bet into it on our selection. The calculation of BFSP is done based on the pool of bets and lays registered or lapsed to BFSP at the off. Betting into this process especially if you are a big punter could destroy the suggestion any historical BFSP prices might have made on your research. The holy grail therefore is to bet in such a way that one gets close to BFSP or ideally beats it.

This month I have spent a little time looking at this problem and although no magic formula has been found I will report on what I have discovered so far. Some of us may remember the old adage of back a horse at the drop price in order to beat bookmaker SP. In other words if your horse opens at 5/1 and drops to 9/2 then back it at 9/2. My starting point therefore was to look at this approach with Betfair prices from 7 minutes down to the official off time at 5 second intervals. I looked at all horses that at 7 minutes to the off had a back price of < 10.0. Taking a drop price, should one have occurred since the last price, resulted in the following average price taken against the average BFSP.

Average price = 5.86 Average BFSP 6.1 Difference 0.24

Now if we contrast the above with taking a price when the price has drifted.

Average price = 5.88 Average BFSP 6.1 Difference 0.22

The drift price approach seems to have improved matters slightly which initially is something of a surprise. Returning to the drop price approach I was curious as to how often we could hope to get matched if we put in a back request at the higher price just vacated when the price has dropped.

26240 requests 22086 matched 84%

The figure of 84% would quite likely be higher in reality as I can only test what has happened at specific snapshots and not what has happened in real time between these snapshots. Clearly the percentage matched cannot be less but it‟s quite likely to be more. What about utilising weight of money, in other words we only bet when the amount of money on the lay side is greater than the amount on the back side. This strategy resulted in the following averages.

Average price 5.89 Average BFSP 6.1 Difference 0.21

A slight improvement over simple price movements but nothing to shout about. If I increase the WOM pressure ratio to > 1.5 we get the following averages

Average Price 5.9 Average BFSP 6.1 Difference 0.2

Again a small increase but we have to remember that as the ratio increase so does the likelyhood of not getting matched. Finally how do all the above compare with a simple strategy of betting at the back price close to the off time. To test this I plumped for less than 60 seconds to the official off time which produced the following averages.

Average Price 6.02 Average BFSP 6.1 Difference 0.08

The message seems to be that one should bet into the Betfair market as late as possible in order to get close to BFSP. This article has only scratched the surface as the techniques used are fairly simplistic but on their own they may help you to squeeze out those extra fractions that can make the difference between long term profit and loss.

 

[tony spencer]

Taken from an article I wrote in 2011

For those researching betting angles to Betfair prices, Betfair SP is often the yardstick used in order to see how an approach would have performed to Betfair betting. The problem we face however is what would the BFSP actually have been if we had bet into it on our selection. The calculation of BFSP is done based on the pool of bets and lays registered or lapsed to BFSP at the off. Betting into this process especially if you are a big punter could destroy the suggestion any historical BFSP prices might have made on your research. The holy grail therefore is to bet in such a way that one gets close to BFSP or ideally beats it.

This month I have spent a little time looking at this problem and although no magic formula has been found I will report on what I have discovered so far. Some of us may remember the old adage of back a horse at the drop price in order to beat bookmaker SP. In other words if your horse opens at 5/1 and drops to 9/2 then back it at 9/2. My starting point therefore was to look at this approach with Betfair prices from 7 minutes down to the official off time at 5 second intervals. I looked at all horses that at 7 minutes to the off had a back price of < 10.0. Taking a drop price, should one have occurred since the last price, resulted in the following average price taken against the average BFSP.

Average price = 5.86 Average BFSP 6.1 Difference 0.24

Now if we contrast the above with taking a price when the price has drifted.

Average price = 5.88 Average BFSP 6.1 Difference 0.22

The drift price approach seems to have improved matters slightly which initially is something of a surprise. Returning to the drop price approach I was curious as to how often we could hope to get matched if we put in a back request at the higher price just vacated when the price has dropped.

26240 requests 22086 matched 84%

The figure of 84% would quite likely be higher in reality as I can only test what has happened at specific snapshots and not what has happened in real time between these snapshots. Clearly the percentage matched cannot be less but it‟s quite likely to be more. What about utilising weight of money, in other words we only bet when the amount of money on the lay side is greater than the amount on the back side. This strategy resulted in the following averages.

Average price 5.89 Average BFSP 6.1 Difference 0.21

A slight improvement over simple price movements but nothing to shout about. If I increase the WOM pressure ratio to > 1.5 we get the following averages

Average Price 5.9 Average BFSP 6.1 Difference 0.2

Again a small increase but we have to remember that as the ratio increase so does the likelyhood of not getting matched. Finally how do all the above compare with a simple strategy of betting at the back price close to the off time. To test this I plumped for less than 60 seconds to the official off time which produced the following averages.

Average Price 6.02 Average BFSP 6.1 Difference 0.08

The message seems to be that one should bet into the Betfair market as late as possible in order to get close to BFSP. This article has only scratched the surface as the techniques used are fairly simplistic but on their own they may help you to squeeze out those extra fractions that can make the difference between long term profit and loss.

I will try it on my Matchbook data

 

[tony spencer]

I will try it on my Matchbook data

Interesting reading this again, Mark, particularly as it was written back in 2011.

I thought I'd revisit your original experiment using the Matchbook exchange data I've been collecting. I've now run the analysis across just over 3.1 million price snapshots covering 20,879 runners.

I used broadly the same methods as you, comparing the price available when a horse shortened or drifted, and then looking at the effect of weight of money.

These were my results:

Method	Average price	Final Matchbook price	Difference
Shortening	5.24	5.32	−0.08
Drifting	5.61	5.70	−0.09
WOM > 1.0	5.00	5.05	−0.05
WOM > 1.5	4.87	4.91	−0.04

What struck me was how similar the general pattern was to what you found 15 years ago. Increasing the weight-of-money requirement appeared to reduce the difference, although the improvement was fairly modest.

I also looked at your idea of requesting the previous higher price after a horse had shortened. Of 203,022 shortening observations, the previous price was subsequently quoted again on 141,814 occasions, or 69.85%.

That's lower than your original 84%, although mine is based on the price reappearing in subsequent snapshots rather than actual matched bets, so it's not a direct like-for-like comparison.

One important difference is that I've compared everything against the final observed Matchbook back price rather than Betfair SP. My snapshots also aren't at fixed five-second intervals.

The conclusion seems much the same, though. There doesn't appear to be any magic in simply following shortening prices or weight of money, but there may be small improvements in execution that add up over a large number of bets.

It's quite interesting that we're still investigating much the same questions 15 years later, just with considerably more data and processing power!

 

[markfinn]

Taken from an article I wrote in 2011

For those researching betting angles to Betfair prices, Betfair SP is often the yardstick used in order to see how an approach would have performed to Betfair betting. The problem we face however is what would the BFSP actually have been if we had bet into it on our selection. The calculation of BFSP is done based on the pool of bets and lays registered or lapsed to BFSP at the off. Betting into this process especially if you are a big punter could destroy the suggestion any historical BFSP prices might have made on your research. The holy grail therefore is to bet in such a way that one gets close to BFSP or ideally beats it.

This month I have spent a little time looking at this problem and although no magic formula has been found I will report on what I have discovered so far. Some of us may remember the old adage of back a horse at the drop price in order to beat bookmaker SP. In other words if your horse opens at 5/1 and drops to 9/2 then back it at 9/2. My starting point therefore was to look at this approach with Betfair prices from 7 minutes down to the official off time at 5 second intervals. I looked at all horses that at 7 minutes to the off had a back price of < 10.0. Taking a drop price, should one have occurred since the last price, resulted in the following average price taken against the average BFSP.

Average price = 5.86 Average BFSP 6.1 Difference 0.24

Now if we contrast the above with taking a price when the price has drifted.

Average price = 5.88 Average BFSP 6.1 Difference 0.22

The drift price approach seems to have improved matters slightly which initially is something of a surprise. Returning to the drop price approach I was curious as to how often we could hope to get matched if we put in a back request at the higher price just vacated when the price has dropped.

26240 requests 22086 matched 84%

The figure of 84% would quite likely be higher in reality as I can only test what has happened at specific snapshots and not what has happened in real time between these snapshots. Clearly the percentage matched cannot be less but it‟s quite likely to be more. What about utilising weight of money, in other words we only bet when the amount of money on the lay side is greater than the amount on the back side. This strategy resulted in the following averages.

Average price 5.89 Average BFSP 6.1 Difference 0.21

A slight improvement over simple price movements but nothing to shout about. If I increase the WOM pressure ratio to > 1.5 we get the following averages

Average Price 5.9 Average BFSP 6.1 Difference 0.2

Again a small increase but we have to remember that as the ratio increase so does the likelyhood of not getting matched. Finally how do all the above compare with a simple strategy of betting at the back price close to the off time. To test this I plumped for less than 60 seconds to the official off time which produced the following averages.

Average Price 6.02 Average BFSP 6.1 Difference 0.08

The message seems to be that one should bet into the Betfair market as late as possible in order to get close to BFSP. This article has only scratched the surface as the techniques used are fairly simplistic but on their own they may help you to squeeze out those extra fractions that can make the difference between long term profit and loss.

 

[MySportsAI]

Interesting reading this again, Mark, particularly as it was written back in 2011.

I thought I'd revisit your original experiment using the Matchbook exchange data I've been collecting. I've now run the analysis across just over 3.1 million price snapshots covering 20,879 runners.

I used broadly the same methods as you, comparing the price available when a horse shortened or drifted, and then looking at the effect of weight of money.

These were my results:

Method	Average price	Final Matchbook price	Difference
Shortening	5.24	5.32	−0.08
Drifting	5.61	5.70	−0.09
WOM > 1.0	5.00	5.05	−0.05
WOM > 1.5	4.87	4.91	−0.04

What struck me was how similar the general pattern was to what you found 15 years ago. Increasing the weight-of-money requirement appeared to reduce the difference, although the improvement was fairly modest.

I also looked at your idea of requesting the previous higher price after a horse had shortened. Of 203,022 shortening observations, the previous price was subsequently quoted again on 141,814 occasions, or 69.85%.

That's lower than your original 84%, although mine is based on the price reappearing in subsequent snapshots rather than actual matched bets, so it's not a direct like-for-like comparison.

One important difference is that I've compared everything against the final observed Matchbook back price rather than Betfair SP. My snapshots also aren't at fixed five-second intervals.

The conclusion seems much the same, though. There doesn't appear to be any magic in simply following shortening prices or weight of money, but there may be small improvements in execution that add up over a large number of bets.

It's quite interesting that we're still investigating much the same questions 15 years later, just with considerably more data and processing power!

Nice work Tony, yes very interesting that things are pretty much the same. I will see if I can find anything else on this topic. When are you next in Portugal ?

 

[tony spencer]

Nice work Tony, yes very interesting that things are pretty much the same. I will see if I can find anything else on this topic. When are you next in Portugal ?

Not until next year Mark off to the Caribbean for a month at Xmas