Efficient Markets on Betfair

[Euler]

Here is an example of one of my strategies (-100 is BSP)

Figure_1.png

I noticed that too but studied the price activity around those points and came up with the conclusion that it was the narrowing spread that caused this.

[Euler]

I've always loved a good old debate about market efficiency because the market is always efficient if you measure it in certain ways. But an efficient market is full of really bad guesses in either direction that cancel each other out to form a 'perfect' guess. But it doesn't describe efficiency at all.

There are always a ton of opportunities in every market I've looked at and having been profitable for so long you can't say the market is efficient in order that I can do that.

[Archangel]

The flip side being an inefficient market will soon be corrected as the market sees value and jumps on it. Therefore efficiency ( or close to it) is a natural conclusion

[Derek27]

You can test how efficient the market is by comparing the BSP`s Implied Probability v Actual Outcome. Using 4.0 odds as an example, if the event is meant to return a win 25% of the time it`s also meant to lose 75% of the time.

Assuming I have 539 (lets say horses) that started with a BSP of 4.0. Of those 141 won and 398 lost. My implied win is 25% but my actual win is 26.16% and actual loss is 73.84%. Pretty efficient but still a 1.16% leaning towards a Back bet rather than a Lay bet.

So if the horse has a BSP of 4.0 and we secure odds greater for our Back bet then it may be an opportunity.

My 10p worth

That's not a valid test, as I've already mentioned because it doesn't prove each horse has a 25% chance. One 4.0 might be an absolute bargain while another could be an overrated favourite who's really worth 6/1. They may average out at 25% winners but that's not market efficiency.

[weemac]

I take the perhaps simplistic view that on a micro level, say minute by minute or event by event, it's pretty inefficient, but on a macro level, say over a season or a period of years, it's remarkably efficient in its pricing. So you can argue that efficiency and inefficency can be true at exactly the same time, depending on your timescale.

Edit: My last sentence above may be a load of cretinous nonsense - I'll have to go and have a lie down and think about it.

[Atho55]

Derek,

If you want to spend your time examining the % of each horse feel free, I find it easier to look at odds as a group and do a simple test on them. I had forgotten you are 100% correct about everything so stand corrected....

[Derek27]

I take the perhaps simplistic view that on a micro level, say minute by minute or event by event, it's pretty inefficient, but on a macro level, say over a season or a period of years, it's remarkably efficient in its pricing. So you can argue that efficiency and inefficency can be true at exactly the same time, depending on your timescale.

Edit: My last sentence above may be a load of cretinous nonsense - I'll have to go and have a lie down and think about it.

I think we get what you mean.

[LinusP]

You can test how efficient the market is by comparing the BSP`s Implied Probability v Actual Outcome. Using 4.0 odds as an example, if the event is meant to return a win 25% of the time it`s also meant to lose 75% of the time.

Assuming I have 539 (lets say horses) that started with a BSP of 4.0. Of those 141 won and 398 lost. My implied win is 25% but my actual win is 26.16% and actual loss is 73.84%. Pretty efficient but still a 1.16% leaning towards a Back bet rather than a Lay bet.

So if the horse has a BSP of 4.0 and we secure odds greater for our Back bet then it may be an opportunity.

My 10p worth

That's not a valid test, as I've already mentioned because it doesn't prove each horse has a 25% chance. One 4.0 might be an absolute bargain while another could be an overrated favourite who's really worth 6/1. They may average out at 25% winners but that's not market efficiency.

This is the problem, however if you take this bucketing as the 'efficiency' you can look at how this changes on the lead up to the race, what is surprising is that the 'efficiency' at 10 minutes out is almost as good as SP, scheduled off time is the best. I will try and graph this later but its tricky.

I think the whole 'Betfair SP is so efficient, look at how close it is to EV' is a load of bollocks because as has been mentioned its just the law of large numbers and nothing more.

[spreadbetting]

If a bookmaker needs to balance anything late on by using the exchange, then someone needs firing. And will have been ages ago. Cannot see that ever being the case. They'd be burning money.

What would they have done wrong to need firing? The recent daily blog by the bookie at Ascot confirms they will usually have a horse or two that will show a loss, and therefore an uneven book, so makes sense for them to even that out on the exchange if they can

Star sports are a traditional bookie so they say they don't hedge and why would you when you're operating a 120% book?

There are some smaller time bookies that I have no doubt try and make pennies hedging from the course to the exchanges.

Star Sports may well wish to continue to take on liabilities but you'll find the majority of rails bookies and even independants/majors are more than happy to take profits when they can. Even with a 120% book they'll come a limit to when you've hit the max liability you wish to hold, a bird in the hand is better than two in a bush especially when you can simply take 6 at wolves and cover it at 10's on the exchanges, even if you've hit the max liability you wish to hold. Betting to a 120% book is no use at all if you're continually running off a lopsided book and relying on a skinner each race.
I think you're being a bit naive if you assume it's "smaller time bookies " making pennies by using exchanges, whole point of bookmaking is to balance your book not take on punters, Betdaq is generally the exchange for most bookies but that money does naturally drift across.

[Derek27]

Derek,

If you want to spend your time examining the % of each horse feel free, I find it easier to look at odds as a group and do a simple test on them. I had forgotten you are 100% correct about everything so stand corrected....

I don't want to spend time examining the % of each horse because there's nothing to examine. I was simply saying it's not a test of market efficiency and I don't believe there is such a test. That's why I was asking LP how he calculates market efficiency.

[Archangel]

In the graph you can see that the implied probability is a nice curve, this is on the basis that the implied probability is perfect. For the underlying market you would expect to see a similar but bumpy curve on the basis of small levels of volatility in the market, in pricing and because of lack of equipartition. If I combined all the data I would expect the line would likely be smoother.

Because of the poorer odds available due to the effect of commission, you can see that the inclusive of commission line is higher than the implied line. Because the market is generally efficient and priced "tightly" the commission pushes you beyond any immediate value.

[henbet22]

In sports markets there are some massive ricks at times. Usually when the macro or micro has been judged and weighted poorly.

[firlandsfarm]

In the graph you can see that the implied probability is a nice curve, this is on the basis that the implied probability is perfect. For the underlying market you would expect to see a similar but bumpy curve on the basis of small levels of volatility in the market, in pricing and because of lack of equipartition. If I combined all the data I would expect the line would likely be smoother.

Because of the poorer odds available due to the effect of commission, you can see that the inclusive of commission line is higher than the implied line. Because the market is generally efficient and priced "tightly" the commission pushes you beyond any immediate value.

I understand your post Archangel and it's a comparison I have made many times but … because of fluctuating prices up to the Off I often wonder what the graph would look like if plotted 1 minute before, at the scheduled Off or if the Off was delayed by an additional minute (we can't have prices for that). There is a certain amount of randomness about the time of the Off and so that randomness is introduced into the graph.

[Euler]

When you see something drift like this, you sense it's probably value.

Incredulous_1410_16_32_03.png

[Euler]

Go figure.

Screenshot 2019-07-13 13.57.20.png

Screenshot 2019-07-13 13.56.19.png