If you are in to reading academic papers on horse racing and betting markets like me you would have read numerous times that betfair markets do not exhibit or dramatically decrease the fav - long shot bias. But is this true ?

Well it depends on the figures used. Most academics use the average of the buy/sell or the weighted price (which is flawed in my opinion for betting).


Lets look at the $50 - $100 range to make it easier for analysis. If we just look at the raw figures and the extremes. There were 11,186 selections (most academic studies use less then a few thousand selections). The payouts would have ranged between 9621 and 15225. If we reduce this to only markets which had a buy / sell spread if 25% or less for this range you get 5208 selections returning between 4447 and 4950.

Did you notice anything ? The first one was a clear loser for the top range of 15525 and the one with the lower spread is a clear winner for the top range.

Lets assume this was dumb luck. Lets try a different range. How about $100 - $300. According tot the bias these should make even more of a loss if you were to bet them. Lets use the 25% spread again. There were 2257 selections and they returned in the range of 1360 to 1580.

From 50 - 100 the result was (in percentages) 85% - 95%. From 100 - 300 the result was 60% - 70%.

Seems to be an indication of the long shot bias here. But to be sure lets look at all races $300+. There were 396 selections and they returned 500-560 so showed a loss. There was 1 winner. Unfortunately this range didn't have enough selections to use (with the 25% spread).

So lets test below $50. Lets go to $20 - $50. According to the bias this should have a higher return then the 85% to 95% return for $50 - $100.
There were 15519 for a return of 14949 to 16442. This is 96% to 105%.

This confirms my theory that that the fav long shot bias still exists on the betting exchange markets. Academics do not seem to understand the nature of the betting exchanges in most cases and they need some real world experience. It is dimished in that on the tote you would notice it from $20+ but on the exchanges it is prevalent at only the highest levels ( $50+ or higher )

How can this help your betting .... Easy. If you are going to be betting on longshots you are better to exclude those over $50 on the exchanges as they will lose more in the long term. If you are laying you should take note. Setting a upper limit on your price band (which I do too) is limiting the best part for laying.

If you can handle the risk you could setup a simple strategy of laying all selections beyond a certain point (i'll keep that point to myself) with no form study. You are simply trustng that the universal theory of fav-longhost bias holds up as it is based on human nature.

I wrote this quickly so sorry for speling errors. Need to go but would be interested to hear other peoples thoughts on this.

as you say wes - "If you can handle the risk"

to make such a scheme worthwhile would require each selection to be layed for reasonable amount & as we all know, that inevitable day where a couple of roughies get up within the space of a few races is potentially just around the corner, so deep pockets required.

thanks for putting the ideas out there, i have a reasonable large database containing betfair staring prices now so will have a play around when time permits & report back any findings that may be of interest.

I like your thinking however i have found a major problem.

From my database which only does Saturday metro racing over the last 2.5 years.

Total selections > 50-1 = 9272
Strike Rate = 0.64%
Profit (using BF starting price) -44.29%

Looks great!

Now if we decide we want to make a living out of this
I add my filter of "total money matched on horse" > $5,000 (Not really that much money to be traded)

I get the following:
Total selections > 50-1 = 142
Strike Rate = 1.41%
Profit (using BF starting price) -22.54%

142 Horses is just over 1 bet per week and Saturdays has by far the most volume on Betfair so i would assume there wouldn't be many weekday selections either.

Anyway, I like the theory but it looks a little impractical based on my experience.

I am almost finished writing an app which coallates UK race data, perhaps your system may fair better on the UK markets as there is much more liquidity than in Aus. I will check it out when i am finished and have collected a bit of data.

Cheers
Glen

Glen,

Agreed liquidity is low and would be a problem. I am not sure on how your filter works but I will assume matched is the same as traded as per the betfair summary for each horse. In that case matched is the bet amount not the liabilitya dn instead of 5000 you would want $5 - $10. It would be rare for any 50/1 shot having more then $50 matched as that is a liability of $2500.

More importantly though, I wrote this as a response to some articles I have read on the bias ad was hoping someone could back up my claims with their own figures. I am glad though that your figures support the bias. Means it does exist and academics are blinded by what they want to see.

wesmip1,

total money matched on horse == $ traded on horse.

Agreed that you wouldn't find many bets for $50 at 50-1 however if you are only turning over $5-$10 per horse thats still pretty slim pickings for the potential risk involved of hitting a big one.

9272 Selecitons / 156 weeks = 59.4 Selections per week.

59.4 x $5 = $297 pw
or at $10
59.4 x $10 = $594 pw


Also my stats may be slightly off yours. I did a >50-1 query which menas i was looking at horses up to and including 1000-1.

Looking only at (50 - 100)
Selections= 6287
Strike Rate = 1.03%
Profit = -31.53%
(Using Starting Price)

If you were to use BF WAP then your Profit is = -24.16%

Worst case:
Take your 5% commission out, plus premium charge of 15% then you are looking at a 5% profit on BF WAP. Not bad if you can find enough bets.

Cheers
Glen

Glen,

Thanks for the post. I actually think the best range is 70-150. After that the risk is too great and before that you hit too many winners (50-70).

It isn't for the feint hearted.