Efficiency of Horse Racing Markets on Betfair

[sionascaig]

This is a distribution of BSP's (rounded) over last few years on UK horse markets. The spikes occur at crossover points. From a market efficiency point of view it wasn't quite what I expected to see - but then again!

[Atho55]

The Orange line is as I see it under and over performing as I imagine every other odd or range do within the market.

Instead of asking me to do your calculations, do them yourself using the vast array of data you have and post the answer.

The difference between the Blue Line and Grey Line is 2.59%

Deviates on individual selections, = proven. accurate over a large sample. = not proven.

[sionascaig]

The Orange line is as I see it under and over performing as I imagine every other odd or range do within the market.

Instead of asking me to do your calculations, do them yourself using the vast array of data you have and post the answer.

The difference between the Blue Line and Grey Line is 2.59%

Deviates on individual selections, = proven. accurate over a large sample. = not proven.

I wasn't asking you to do any calcs, just thought it was a interesting pic..

Anyhow, useful to include commission into calcs - even if market is efficient it doesn't mean you will make a profit.

(green line is accumulated profit in an efficient market)

[Atho55]

Apologies, I was actually replying to Shaun.

So taking Rank 1 which you have charted, are you of the opinion that selecting odds from the array that Rank 1 starts at will improve your chances?

Have looked at combinations and think it provides a small edge but never taken it any further.

[Atho55]

These are odds I think overperform over the long term I should add....

[sionascaig]

I've just started exploring the data - encouraged by some of your earlier posts - thanks for sharing... Can't find fault with any of your analysis.

I have not found any edges on the backing front although keen to explore the "reversion to mean" which I think was discussed in earlier posts / topics but the margins are so small its probably a long term (years) strategy.

On the laying front, I've found two interesting scenarios which maybe worth exploring further, again the margins are very low and require at least a years exposure ... Will send you a PM about it and would be interested in your thoughts...

Regards

[ShaunWhite]

The difference between the Blue Line and Grey Line is 2.59%

Deviates on individual selections, = proven. accurate over a large sample. = not proven.

2.59% is very little over your sample. It's never going to be spot on 0%. If you plot a linear trend on that orange line I bet the x multiplie isn't far from zero and the r-squared is well over 0.90. That says the EWR and the Implield% are correlated.

I could do it over a huge amount of data but the answer will be the same, and I don't have the inclination to spend the time proving something to myself again. When I got a decent set of data I looked for the low hanging fruit like most people do, ie to see if any particular price is a winner, or any particular time on certain ones, like favs, or any track. Nothing stood out and even if it had, without a reason for it I would have just binned it. Edges need an underlying reason otherwise they're just noise and varience, and if you try and trade an anomaly they just end up reverting and you're always on a loser.

[rik]

This is a distribution of BSP's (rounded) over last few years on UK horse markets. The spikes occur at crossover points. From a market efficiency point of view it wasn't quite what I expected to see - but then again!

maybe its to do with SP often not being an exact number?
So say SP was 10.23 it would be rounded to 10 in this graph? but 9.77 to 9.8 even though its the same „distance“ in terms of decimal odds?

[sionascaig]

This is a distribution of BSP's (rounded) over last few years on UK horse markets. The spikes occur at crossover points. From a market efficiency point of view it wasn't quite what I expected to see - but then again!

maybe its to do with SP often not being an exact number?
So say SP was 10.23 it would be rounded to 10 in this graph? but 9.77 to 9.8 even though its the same „distance“ in terms of decimal odds?

Yes, thats happening in the rounding. Was still surprised by size of the spikes given you don't know SP until its set... Was wondering if its related to the non decimal system used by some big hedgers (allowing for the fact of the change of odds increments at the crossovers)..

[Anbell]

This is a distribution of BSP's (rounded) over last few years on UK horse markets. The spikes occur at crossover points. From a market efficiency point of view it wasn't quite what I expected to see - but then again!

maybe its to do with SP often not being an exact number?
So say SP was 10.23 it would be rounded to 10 in this graph? but 9.77 to 9.8 even though its the same „distance“ in terms of decimal odds?

Yes, thats happening in the rounding. Was still surprised by size of the spikes given you don't know SP until its set... Was wondering if its related to the non decimal system used by some big hedgers (allowing for the fact of the change of odds increments at the crossovers)..

I've looked into this before, and it does seem weird at first glance, but part of the explanation is that you have to remember that far BSP (which is meaningless) is the balance between all pure SP bets, and then you have near SP which takes into account pure sp and also exchange bets.

If there's a large lay bet on the exchange available at 10 then that sort of overrides the balance on the pure SP side, so the SP bumps into that large bet and therefore the curve isnt smooth. IT's difficult to explain until you've looked at the formula behind the SP, but it kinda makes sense that the SP isnt smooth.

[Atho55]

Depending how you deal with the BSP odd provided by BF and the size of the data you hold will give differing answers. I =FLOOR(BSP,0.01) which gives less increments at the lower odds but more as you go up the odds compared to the traditional Ladder odds.

As a BSP count with 0 being losers and 1 being winners I get this at the 10 ish BSP odd range as a count.

Floored Odds Count.jpg