Market efficiency

[Iron]

It is widely believed that the markets are extremely efficient, i.e. that, overall, the odds you get on a horse are likely to be an extremely accurate reflection of the horse's true odds. If this is true, then it follows logically that you would expect neither to make or lose very much betting and laying at level stakes.

This morning, I put the above to the test. I downloaded my horse racing results for the last 3 months. I treated each bet or lay as if it were an outright bet in its own right, regardless of whether it was part of a trade. I also adjusted the profit and loss for hypothetical level stakes of £10 (for example, if I were betting with £2 stakes, I'd multiply the profit or loss by five).

Betting at level stakes, over just under 5,000 backs, I made about a 2% loss, which is higher than I would have expected. However, the ROI to liability on the lays (again, about 5,000) was what I would have expected, a loss of 0.2%.

I'd be interested to hear if anyone else has done this kind of analysis.

Jeff

[Iron]

PS I thought my figure for backs might have been affected my me often backing the same horse repeatedly at a particular price when trading (causing the amount of staking on particular selections to distort the overall results).

I therefore re-ran the results, but this time eliminating instances where a horse is backed more than once at a particular price.

Over just under 3,000 backs, I got a 5.4% loss (and lost 0.6% on lays, to liability).

Jeff

[LinusP]

Surely this depends completely on if you are a profitable trader, if you are in profit your backs and lays should both show an individual profit.

Peter posted this on greening up..

http://www.betangel.com/blog_wp/2011/03 ... -green-up/

[Iron]

Hi Linus

Not necessarily your backs and lays, but one of the two should (depending on your approach).

However, my point was that, if you trade at random, you should (according to theory) lose very little money at level stakes. It's therefore curious that, over thousands of bets, I did significantly worse than trading at random for backs.

Jeff

[Wyndon]

Isn't it the case that odds are extremely efficient at predicting the eventual result at the point of going into play? Although the over-round may be close to 100% some time before the off, individual odds can vary widely within this envelope. It is still interesting that you get this consistent difference between the back and lay parts of your bets.

[Iron]

It is often said that's the case, although there is a school of thought that says the odds are also very efficient overall.

Jeff

Isn't it the case that odds are extremely efficient at predicting the eventual result at the point of going into play?

[Wyndon]

It is the case. Most of my strategies over many years have been based on trying to beat Betfair SP (and industry SP before that). If you can do this consistently (without going in play!) you WILL show a profit. I don't see how any one of many possible single snapshots prior to the race will give such a close agreement with actual results.

[Iron]

How do you know?

Some people that the price at the off on Betfair used to be very accurate overall, but is no longer, due to the crazy price movements you see on Betfair all the time. How do you know they're wrong?

Jeff

It is the case.

[Wyndon]

I don't. I performed an analysis some years ago now looking at 19,273 races in which 209,222 horses took part and confined the sample to SPs between 4/9 and 200/1. I was amazed at the almost perfect straight line relationship between the probability of winning according to SP and the actual probability of winning. The correlation coefficient was 0.9925. I've seen reports that this relationship continued with Betfair SP - with the BFSP being even closer to the actual chance of success. You tell me that this has now changed. I don't discount the possibility, but would like to see some hard figures to back it up.

[Iron]

I don't.

I'm just saying that changes in market dynamics may mean that it's no longer the case.

Jeff

You tell me that this has now changed. I don't discount the possibility, but would like to see some hard figures to back it up.

[PeterLe]

Jeff
When you placed your bets did you take the odds available or did you ask for the price ?
it makes a big difference even when greening up
regards
Peter

[Iron]

When opening - both.

When closing - I usually take the available price.

With regards to opening the trade, I'm not sure your long term ROI will be worse from taking the available price vs offering to the queue if you're swing trading (although I appreciate that one tick scalping is a different beast).

On one occasion, you'll get filled at your offer price. However, on the next occasion you may find that you don't get filled before the market storms off, and you could end up a few ticks' worse off due to missing out on a nice move you'd have profited from had you taken the price.

I think it may be a case of swings and roundabouts, if you'll excuse the unintended pun!

Jeff

Jeff
When you placed your bets did you take the odds available or did you ask for the price ?
it makes a big difference even when greening up
regards
Peter

[PeterLe]

When I close, I always try to offer, it doesn't always work, but a lot of the times it does. On 800 races in the summer, it adds up.
Although not always possible when opening the trade; the same applies.

If you were to take your spreadsheet and create a couple of new columns (you can use a lookup command to an odds table) and calculate just one tick above your back price and just one tick below you lay, then check you figs again; you will find a massive difference. (Obviously that would be a perfect world, but just assume you only got it 50% of the time?)

In an analogy drawn from the car world, the lengths manufacturers go to to squeeze an extra 1 MPG out of a gallon of diesel is astounding (even special tyres and aerodynamics)

It all contributes towards the "line line" of profitability we talk about in the forum. Sometimes these tiny cages can make a world of difference.

[Iron]

When I close, I always try to offer, it doesn't always work, but a lot of the times it does. On 800 races in the summer, it adds up.
Although not always possible when opening the trade; the same applies.

Hi Peter

I have experimented with offering to close in the past. The problem is that I sometimes find myself chasing a market, and going further and further into the red.

That may be because, when a trend starts to break down, everyone heads for the door at once, which can make it hard to get matched if you're offering.

If you were to take your spreadsheet and create a couple of new columns (you can use a lookup command to an odds table) and calculate just one tick above your back price and just one tick below you lay, then check you figs again; you will find a massive difference. (Obviously that would be a perfect world, but just assume you only got it 50% of the time?)

I hear what you're saying, but when you take into account that, when you don't get matched, it's because the market is running away from your position, I'm not sure you'll be better off long term.

I'm not saying you're wrong, btw, but looking at this from first principles I can't see why I would expect to be better off from offering to the market.

As an aside, when I asked a group of financial traders whether they put in an order at the ask or take the bid when buying, they all said they took the bid, although the I appreciate that the market dynamics on Betfair may be substantially different.

In an analogy drawn from the car world, the lengths manufacturers go to to squeeze an extra 1 MPG out of a gallon of diesel is astounding (even special tyres and aerodynamics)

I'm all for maximizing efficiency - I'm just not convinced that offering to the queue is the way to do it if you're a swing trader (although I keep an open mind).

Jeff

[Iron]

PS It would be interesting to run two random trading experiments, both using stops of x and offsets of y, offering to the queue to open in one experiment and taking from it in the other.

Looking at the contrasting ROIs may give some insight into which approach is the best.

Jeff