Place Pricing

  • I think we should have a new thread for the discussion that has emerged in the Bad EW thread.

    http://www.betfairprotrader.co.uk/2012/04/place-pricing-in-horse-racing.html

    http://www.betfairprotrader.co.uk/2012/05/study-of-betfair-place-odds.html

    This is for those that have been looking at estimating probabilities for 2nd, for placing etc. We probably all do to check if an offer is good value, if its likely enough to pick up a free bet, if the price on a place is value, if we can pick off some value on the prices in a bad ew race on the early betfair market.

    I had come up with something like the deck of cards example that tennis punter described and found it wasnt accurate. But I should have known that no work is original any more (not since the internet), so this method has already been invented by Harville in 1973 and discreditted due to a number of issues including the “Silky Sullivan” effect.

    ”
    Harville (1973) assumed that any horse that won is automatically discounted from placing and the win probabilities of the remaining horses recalibrated to sum to 1. This naive Bayesian approach suffers from various problems.
    ”

    Mentioned is Henery and Stern models too, but I haven’t found much on that. But googling for Harville seems to bring up a lot of articles/websites with exactly the discussion that I have been looking for.

  • New to matched betting?

    My Matched Betting Academy is the best place to get started. Learning the fundamentals takes 10-15 minutes and you’ll make £15 in the process.

    Learning the fundamentals takes 10-15 minutes and you’ll make £15 in the process.

    scabbadab 259

    Great idea FOG . There were some brillant posts regarding the place stuff in the bad ew thread.

    Perhaps some of that info/links/charts should be copy and pasted here for future ref.

    Nice1

    +0
    FoG_BLoG 61

    It is a bit daunting reading some of these research papers on the topic, to beat the markets its these sort of experts that we are trying to beat. Just hoping that there is some easy pickings available for those of us that can even understand 10% on this area – and take advantage of some obvious market inefficiencies (place prices in bad ew race, events where bookies have crazily enhanced terms) …

    +0
    betman 41

    Good post/links etc

    I presume the Harville approach reperforms the calculation for 3rd, i.e. assumes 1st and 2nd, then calculates probability of 3rd based on remaining horses win probabilities?
    If so then that is the place model i also recreated. I knew at the time i wouldn’t be the first and knew it would have some limitations (like the serial 2nds i referred to in the other post) but felt it would be of use.
    Interestingly the first few races i plugged the numbers in worked remarkedly well, I needed to keep the win odds updated but the place odds gravitated pretty accurately to my calculations for the off.
    Although it has since missed the mark on some races. I suspect those races were ones with identified horses that would distort my calcs, but i’m not convinced by all of the ‘discrediting theories’. That’s why i was asking about the data you had yday fog, i’d be inclined to say Harville calcs are materially accurate… but not seen numbers to prove or disprove it.

    I’d be interested to read the Henerey model, although not found it directly accessible yet only register for links. Its properly hidden too, my first google search result was an article about that stupid blonde bint who dated the UKIP guy!!

    +0
    FoG_BLoG 61

    Well perhaps I haven’t given enough time to the Harville method then. Or my race selection criteria was skewing the results from the start. I was concentrating on bad ew races and the first couple I picked were over estimating considerably the chances of some horses to place (compared with the exchange prices).

    I had come across something on Henery and Stern but even the experts are writing that it is impractical to use, it seems to involve solving two sets of integral equations.

    ”
    Henery (1981) assumed that Ti  N(θi,1) independently. This will involve solving the system of integral equations in (3) and (4) using numerical integrations and is not practical to use in real races. Similar practical difficulties apply to the gamma model proposed by Stern (1990), where an extra shape parameter is involved.

    Probability and Statistical Models for Racing (PDF Download Available). Available from: https://www.researchgate.net/publication/4748916_Probability_and_Statistical_Models_for_Racing [accessed Apr 19 2018].
    ”

    Some time when I have the time I will check how good the simple estimate on prob of places fits with the data that I have collected for race results. Maybe when the kids are grown up, so Ill have an answer for you in 12 years! (PS kids never grow up).

    I suppose even if it works okay and you can identify the “Silky Sullivan” and “Serial Seconds” and avoid using on those races then you might have something. I think with jumps too you can have a situation where either he falls or he wins. In that case his prob of 2nd/3rd wouldnt really be related to win odds. Too many variables.

    +0
    betman 41

    Or I self proved Harville method too quickly

    I also used bad each way races, but one’s where the outsiders were too ‘gappy’ and possibly inflated not exchange…. so I used best prices on oddschecker.
    On reflection that’s rubbish as we know bookies hold the win value back on such races, but maybe by fluke 2 wrongs made a right and the bookie price actually reflected the adjustment needed for place probability!!!!

    Had also considered the falling option, we’ll call that the Oscar whisky factor! (if you were familiar with that horse).

    Maybe the model works well, but you need an ‘adjustment factor’ for 2nd and 3rd which you manually apply depending on race…. although sod knows how you quantify that

    My kids are 1 and 5 so will join you a few years later!!!

    +0
Viewing 5 replies - 1 through 5 (of 5 total)

You must be logged in to reply to this topic.