Longshot-Favourite Bias - Rugby

[mcgin]

Datapunter,

Just started loading the BF data into a database to analyse for the bias. Is it just me or is the data for RU in the betfair files corrupted.

Take for example the Entries for the following fixture -
Group D/Celtic League/Fixtures 01 September/Cardiff v Munster
(Event id = 20021823).

Appears to have a whole pile of horses in the Selection column!

Not just that match either, I've spotted it in a few others.

[mcgin]

Selections for Saturday - 50.37pts on each:
Newcastle 1.4 (Bet365)
Sale 1.4 (Premierbet)
Worcester 1.29 (Premierbet)
Ospreys 1.13 (Stan James)

Ospreys look borderline at about 2.8% but will include them as think the average will drop with more books opening tomorrow.

I'm amazed at how many selections this is throwing up, with only a handful of games each weekend we're getting great price discrepencies. If only there was more Rugby on eh?!

[mcgin]

S. Africa are available @ 1.33 (after com) on Betfair. As theres only a few books open on this one I won't include it in the results.

[mcgin]

AS expected the average did indeed drop giving us a 3.9% difference between bet price & average for the ospreys. Not quite what I'd normally accept, but certainly not bad.

[mcgin]

All 4 come in.

Total staked 654.49
Total returned 752.32
Yield 14.95%
Winners 11.00
Loosers 2.00
Strike Rate 84.62%
Current bank 5,097.82
Profit/Loss 97.82

[mcgin]

Heineken Cup should throw up a few selections for us.

Bourgoin v Leicester @ 1.4 with betfair (1.38 after comm) 50.98 pts

[mcgin]

Ospreys @ 1.11 (1.10 after commission) Betfair. 1.06 average
Biarritz @ 1.13 Stan James. 1.08 averag
Leinster @ 1.2 (1.19 after commission) Betfair. 1.137 average

[mcgin]

Thats 51.18 on each

[mcgin]

Leicester & Ospreys both won & Biarritz looking comfortable. Fingers crossed for Leinster then for a full card for the weekend

Also from next week onwards I'll be changing the staking strategy from 1% of the bank to Kelly criterion, so expect the bets to be 4-5 times larger. You can see the way it works from the last column on the spreadsheet.

[mcgin]

Leinster obliged, bringing us up to a nice 16% yield.

Total staked 860.40
Total returned 999.91
Yield 16.21%
Winners 15.00
Loosers 2.00
Strike Rate 88.24%
Current bank 5,139.51
Profit/Loss 139.51

[davidmphill]

Nice work mcgin.
Long may it continue for you.

[mcgin]

Going to include Stade Francais today even though the best price is just under 4% better then the average. Knowing my luck now that I've upped the stakes it'll lose

Stade Francais 1.2 Betdaq (1.19 after commission). 184.55pts

[mcgin]

Cheers Dave, thought it was just me in here

[davidmphill]

Originally Posted by mcgin

Cheers Dave, thought it was just me in here

We all feel that sometimes buddy.

[mcgin]

Well I've sifted through a years worth of Betfair data and it would appear that the bias exists in Union, and to a lesser extent in League. With Union you could see a profit simply from backing the strong favourites, in League you simply loose less the shorter the price. In fact the simple strategy of backing anything less then 1.1 in both League & Union would have given you a nice 2.5% yield over 106 selections.

So, based on what I've seen in the figures I'll be modifying my selection criteria so that anything <1.1 on average will only require a difference of 3% to be included. Anything else will require a 4% margin and must be less then 1.4 on average. And from next week this will become a real money system (Watch it collapse!).

[mcgin]

Stade Francais win.

Total staked 1,044.95
Total returned 1,219.53
Yield 16.71%
Winners 16.00
Loosers 2.00
Strike Rate 88.89%
Current bank 5,174.58
Profit/Loss 174.58

[mcgin]

Perpignon 1.4 BlueSquare
Ulster 1.2 Premierbet

Stakes in spreadsheet.

[Datapunter]

Sorry mate, missed your earlier post on the Betfair data,
been working on a soccer results database where i intend to plug-in the Betfair data. But it's turning out to be a hell of a job so haven't gotten around to anything else.

Yeah, be carefull with the data as there are silly entries and inconsistencies all over, it would be fair to call it a mess, still, very usefull info to be found.

How are you approaching the timing thing. I really can't see any use in it as it just doesn't allow you to see price movements over time. Fine in case of a steamer or drifter where the price steadily moves in one direction. But if the price goes up and down somewhat you have no idea.

So how do you determine from the data what is the most likely price you would have been able to have gotten ?

The approach i am applying is this, i look at the total volume matched, then i determine the price at which 30% , 50% , 70% of the total volume is matched. Then if i'm looking at laying i take the high value, if looking at backing i take the low value. ( the 50% i currently don't use but i'm still in the process of finding a good approach ) It appears as there is not all that much spread in the prices i get this way, especially in high volume matches, but it could be the difference between profit and loss. I think this appproach will get me price info that is close that what i might expect to be able to get. ( or the price i should have been able to have gotten )

[mcgin]

Originally Posted by Datapunter

How are you approaching the timing thing. I really can't see any use in it as it just doesn't allow you to see price movements over time. Fine in case of a steamer or drifter where the price steadily moves in one direction. But if the price goes up and down somewhat you have no idea.

What I was looking for here was the 'SP' I suppose of the selection. Theory being that the prices matched most recently (assuming there's a high enough volume) are prices that would be generally available at other bookies as the early price discrepancies would get smoothed out.

So how do you determine from the data what is the most likely price you would have been able to have gotten ?

As mentioned above I took the 6 most recent prices for the selection and took the weighted average. I only considered matches where >£2000 was matched, anything less then that I felt there wasn't any liquidity whatsoever in the market (even £2000 is quite low). This I think is a pretty good indication of what price you could get at the off.

I also compared the total matched on the event with the total matched for the most recent 6 prices. Any huge discrepencies were again discounted as I couldn't reliably estimate the price that could have been taken for that match. This discounts steamers and drifters from the selections.

The approach i am applying is this, i look at the total volume matched, then i determine the price at which 30% , 50% , 70% of the total volume is matched. Then if i'm looking at laying i take the high value, if looking at backing i take the low value. ( the 50% i currently don't use but i'm still in the process of finding a good approach ) It appears as there is not all that much spread in the prices i get this way, especially in high volume matches, but it could be the difference between profit and loss. I think this appproach will get me price info that is close that what i might expect to be able to get. ( or the price i should have been able to have gotten )

I'll try this one out at some point to see what results I get.

[mcgin]

Well Ulster were making me nervous @ half time, but thankfully Humphries had a blinder of a second half to put us safe. 2 more wins. Excellent yield, but that'll get annihilated once the next loss comes along :p

Total staked 1,574.22
Total returned 1,926.07
Yield 22.35%
Winners 18.00
Loosers 2.00
Strike Rate 90.00%
Current bank 5,351.85
Profit/Loss 351.85