Efficiency of Horse Racing Markets on Betfair

[Euler]

Hmm

Voix Des Tiep_12504165_14_25_42.png

[ruthlessimon]

Not quite as clean as that one! but here's another similarish fav from today:

[Euler]

If you look at my chart it lost or gained, depending on your viewpoint 17% in two directions. Which obviously can't be efficient.

[ShaunWhite]

Surely the endpoint is the most efficient after the most 'votes' have been polled. No logical reason for any other conclusion.

But aren't these just a case of it being relatively cheap to make the price drift when the vol is low and there's little real money to resist it and then backing it back in when there's more money at each corresponding price point you layed at. (+1 obv). Before the stablising effect of significant real money it seems relatively easy to move a price in chucks and have the market follow rather than it being immediately brought back to where it was.

I don't think im seeing anything unexpected or inexplicable in these apart from early manipulation by someone who's a good judge of the likely sp. As always, the tricky part now is to converting hindsight into foresight,

[ajdal]

If you get the nod that the one paced, abundant stamina 12/1 shot that has been ridden mid-division in all his races recently is going to be ridden aggressively from the off in this race and will almost certainly still be thereabout with at least 3 fences left (bar a fall).. what kind of adjustment would you make (if any) to the price?... and at what stage pre-race would you start to attack the price ??... in fact, if few were privy to the knowledge, maybe you might even consider to push the price out first when it's cheaper ??

There really are a myriad of reasons that could alter the apparent chance of a horse even in the last few minutes...

[Trading96]

Surely the endpoint is the most efficient after the most 'votes' have been polled. No logical reason for any other conclusion.

But aren't these just a case of it being relatively cheap to make the price drift when the vol is low and there's little real money to resist it and then backing it back in when there's more money at each corresponding price point you layed at. (+1 obv). Before the stablising effect of significant real money it seems relatively easy to move a price in chucks and have the market follow rather than it being immediately brought back to where it was.

I don't think im seeing anything unexpected or inexplicable in these apart from early manipulation by someone who's a good judge of the likely sp. As always, the tricky part now is to converting hindsight into foresight,

It's also logical to suggest when matched volume per second is at it's highest is where the market is at its most efficient.

[Derek27]

Surely the endpoint is the most efficient after the most 'votes' have been polled. No logical reason for any other conclusion.

But aren't these just a case of it being relatively cheap to make the price drift when the vol is low and there's little real money to resist it and then backing it back in when there's more money at each corresponding price point you layed at. (+1 obv). Before the stablising effect of significant real money it seems relatively easy to move a price in chucks and have the market follow rather than it being immediately brought back to where it was.

I don't think im seeing anything unexpected or inexplicable in these apart from early manipulation by someone who's a good judge of the likely sp. As always, the tricky part now is to converting hindsight into foresight,

In my opinion, the only logical conclusion to draw is that the market doesn't have a clue what chance the favourite had. You can't trust a market that says a horse has a 2.1 chance when just a few minutes earlier it thought it had a 2.9 chance.

I think trading on a market corrupts the link between the Betfair market and true probabilities. The horse probably drifted largely because other horses were being backed, and I presume they went on the drift as well.

[ShaunWhite]

The horse probably drifted largely because other horses were being backed, and I presume they went on the drift as well.

Even in a two horse market I don't believe you can tell if a move is caused by the backing of one or the laying of the other. With bots soaking up the underround you'd need millisecond data to see what was driving it. In a multiselection situation you'd have to assume that the one which moved furthest was the driver as the knock on effect would be diluted across the others.

[ShaunWhite]

Surely the endpoint is the most efficient after the most 'votes' have been polled. No logical reason for any other conclusion.

It's also logical to suggest when matched volume per second is at it's highest is where the market is at its most efficient.

I've heard that said before but I personally can't see why opinion during a period of high flow would be more accurate than the final assessment after the horse has been seen going down.

But.. looking at what PW said above "I've studied the market extensively and it's typically most efficient near the off", if that 'near the off' does indeed tally with that late volume surge, then I'd be wrong to argue against either of you.

[Euler]

Iley Boy_12987891_19_39_50.png

[Derek27]

Looking at the graph above, if the true chance of the horse was at the off, that would suggest that, with the exception of the last bet matched, everybody who backed it had a value bet and everybody who laid it had a bad value bet, for the entire 24 hours or so.

In all probabilities its true chance is closer to the average price. I suspect traders were closing trades in desperation - or even during the course of a heart-attack.

[ShaunWhite]

For any given individual horse it's impossible to say what the true price was, only what the crowd thought it was. All the stats and maths might tell you that markets are most accurate 30s out or at peak volume blah blah blah but that's over 1000s of races... not for an individual race. They are animals not machines, it's an unknown known as George Bush would have said.

[Derek27]

For any given individual horse it's impossible to say what the true price was, only what the crowd thought it was. All the stats and maths might tell you that markets are most accurate 30s out or at peak volume blah blah blah but that's over 1000s of races... not for an individual race. They are animals not machines, it's an unknown known as George Bush would have said.

It was Donald Rumsfeld that said it, and unknown knowns was the one combination he left out. There are known unknowns, but it's hard to define an unknown known, because it's self-contradictory to say you don't know something that you know.

[Iron]

The evidence shows incontrovertibly that the market is extremely efficient at the off (at least the horse racing market).

For example, if you had laid all 46, 899 horses with BSPs under 10.0 in 2017, then if there was zero commission you'd have made a profit on risk of 0.12% (based on data downloaded from the relevant link in this thread - https://betangel.com/forum/viewtopic.ph ... 7&start=10).

Jeff

In my opinion, the only logical conclusion to draw is that the market doesn't have a clue what chance the favourite had. You can't trust a market that says a horse has a 2.1 chance when just a few minutes earlier it thought it had a 2.9 chance.

[Derek27]

The evidence shows incontrovertibly that the market is extremely efficient at the off (at least the horse racing market).

For example, if you had laid all 46, 899 horses with BSPs under 10.0 in 2017, then if there was zero commission you'd have made a profit on risk of 0.12% (based on data downloaded from the relevant link in this thread - https://betangel.com/forum/viewtopic.ph ... 7&start=10).

Jeff

In my opinion, the only logical conclusion to draw is that the market doesn't have a clue what chance the favourite had. You can't trust a market that says a horse has a 2.1 chance when just a few minutes earlier it thought it had a 2.9 chance.

That's what I would expect from efficient markets, but it's also what I would expect from inefficient markets, because the overround is approximately 100% so according to the laws of probability you would get that result regardless. I would expect similar results if you laid such horses 15 minutes before the off.

Genuine Approval_10623818_18_46_21.png

You're not seriously telling me that they got this one's price right just before the off ?

They didn't have a clue!