- This topic has 11 replies, 9 voices, and was last updated 13 years ago by
.
- Posts
I wondered how much more efficient, if at all, the betting market is since exhanges came into being.
Presumably you would measure the market efficiency by ,measuring actual event probabibility against the market’s estimate (taking into account over-round).
Given the tight margins on betfair, in comparison to the hefty margins on and off course bookmakers enjoy, you might conclude that the BF market is likely to be more efficient (than pre-exchange SP).
The bookmakers SP market now pretty much follows betfair prices (allowing for commission) so I’d imagine it is just as efficient (as betfair) these days.
But has it improved since pre-2000? And why?
Anyone with data to provide a definitive answer?
For me, that’s one of those questions that would involve more work than knowing the definitive answer would benefit you.
Both are efficient, we know that, so the difference would likely be smallish, and the extent of that difference could easily be rendered debatable by the way you remove the overround for instance. Apart from the fav longshot bias, the takeout in a bookmaker market is not distributed proportionately as in a tote market.
Away from efficiency, exchange markets are far more transparent than old bookmaker markets (from the punters perspective that is) largely because you can see pending orders and volume traded. You can now see leakage of what I would call non-formbook based information.
Sometime last year I believe a regular TRF contributor posted a table that suggested late morning Betfair price was an accurate indicator of success.
Maybe someone with more time can find the link to the thread.
Rob
I was not around racing/betting pre exchange so find comparison difficult.
What I would say is markets are only as accurate as the general accepted process by which they are formed. In this country that means pretty much form ratings, namely Timeform.
I wondered how much more efficient, if at all, the betting market is since exhanges came into being.
Presumably you would measure the market efficiency by ,measuring actual event probabibility against the market’s estimate (taking into account over-round).
Given the tight margins on betfair, in comparison to the hefty margins on and off course bookmakers enjoy, you might conclude that the BF market is likely to be more efficient (than pre-exchange SP).
The bookmakers SP market now pretty much follows betfair prices (allowing for commission) so I’d imagine it is just as efficient (as betfair) these days.
But has it improved since pre-2000? And why?
Anyone with data to provide a definitive answer?
I have no figures to back it up David, but I believe the last 10 minutes of the betting market is more efficient since betting exchanges came in.
I used to study the form, work out my idea of the fair odds, go racing and back what I believed to be the "value" horses. Stronger morning market and many more races priced up early – means if I don’t back my value horses then some other good judge/s certainly will and they shorten up. No longer value in the market just before the off…
This also means some prices have gone the other way, too short in the morning but gone out to what seems (at least to my form study) a "value" price. However, I noticed most of my winners came from horses that shorten up. Is it because the exchange market gets to know the horses that are unlikely to run to form (for whatever reason)? ie "Form" may suggest the horse is excellent value but its actual chance may not be as good for non-form reasons. Exchanges and then on course markets have taken this in to account. Someone somewhere knows something; not fit, poor journey to the course, sweated up etc (no skulduggery suggested in the vast majority of cases). Therefore, I seldom have a main bet late/on course, even if it has gone out to (in my workings out) tremendous value.
I personally much prefer the Early Odds markets which I believe is less efficient for the "form" value student. ie More "value".Value Is EverythingI have no figures to back it up David, but I believe the last 10 minutes of the betting market is more efficient since betting exchanges came in.
I think this may be correct – I get the ‘feel’ that it is, but like Ginge I have no hard stats to prove it.
There could be other factors at play though. I did some research years ago (prior to the impact of exchanges) regarding SP and expected win ratios for about ten years worth of data (something like from 1990-2000) and found that even in such a less reactive, less sophisticated time, the win to price ratios were nevertheless staggeringly accurate and virtually exactly as expected. This may hint that the process that we see on Betfair is merely the same as the correcting prior to the off in the standard SP market of 20-odd years ago.
Many football fans insist there are far more refereeing mistakes nowadays. In reality, the difference is that there are 16 cameras at every match with endless analysis and replays. In the same fashion, we now have a rapidly-changing real-time data in front of everyone’s eyes. Maybe that’s just creating the illusion of efficiency?
I also think sometimes price contractions can be exacerbated by the herd mentality looking (quite sensibly) for positive confirmation about horses they may have been willing to back anyway. And there’s also a potential ‘wisdom of crowds’ effect, whereby punters just latch on to more obvious form selections that have simply been trading at the ‘wrong’ price.
In summary…dunno!
Mike
I did some research years ago (prior to the impact of exchanges) regarding SP and expected win ratios for about ten years worth of data (something like from 1990-2000) and found that even in such a less reactive, less sophisticated time, the win to price ratios were nevertheless staggeringly accurate and virtually exactly as expected. This may hint that the process that we see on Betfair is merely the same as the correcting prior to the off in the standard SP market of 20-odd years ago.
That would suggest the markets are no more efficient now. Perceptions that it is may be because of the transparency of information – as mentioned in the post above.
Away from efficiency, exchange markets are far more transparent than old bookmaker markets (from the punters perspective that is) largely because you can see pending orders and volume traded. You can now see leakage of what I would call non-formbook based information.
I’m sure I saw some stats on ‘late drifters’ once but can’t accurately remember the gist of it.
Rob – it is accurate/efficient Rob, I’m just wondering if it is more accurate than pre-exchange markets.
I did some research years ago (prior to the impact of exchanges) regarding SP and expected win ratios for about ten years worth of data (something like from 1990-2000) and found that even in such a less reactive, less sophisticated time, the win to price ratios were nevertheless staggeringly accurate and virtually exactly as expected. This may hint that the process that we see on Betfair is merely the same as the correcting prior to the off in the standard SP market of 20-odd years ago.
That would suggest the markets are no more efficient now. Perceptions that it is may be because of the transparency of information – as mentioned in the post above.
On the face of it, it’s remarkable and not a little counter-intuitive that SP – be that bookmakers’ SP or BFSP – converge to expected theoretical true chance given a sizeable sample, but it’s largely explained by The Law of Large Numbers (the convergent sample) and its constituent Bernoulli Trials (the win-lose ratio at each price)
http://en.wikipedia.org/wiki/Law_of_large_numbers
http://en.wikipedia.org/wiki/Bernoulli_trialsTherefore one would not expect a difference between the win-to-price ratios of SP and BFSP, other than the former being slightly more divergent from the expected due to the in-built bookmakers’ over-round
I’m sure such as Glenn and Cavelino Rampante could expand further on market efficiency specific to horseracing prior to the necessarily wise-after-the-event SP data, but I for one haven’t a clue really, not for want of trying
In fact Efficient-Market Hypothesis (and it’s various sub-hypotheses) in the broader sense is a subject shrouded in theory, conjecture, mystery and scholarly debate
http://en.wikipedia.org/wiki/Market_efficiency
I think, IMVVHO
I price the Premier League up every year at the start of the season. Every team is given a 19/1 chance.
How would some of the methods suggested above rate the efficiency of my tissue?
Some would rate it as remarkably efficient, which suggests that they aren’t very good methods.
Handicappers need to get to a weak market situation like poker where we play ‘The Game’ against only 5 heads round a real or virtual table for high stakes whilst being filmed and envied for our multi-millionaire lifestyles.
Thinking a bit more about efficiency of the market and the methodology used to measure this.
If we look at a group of 3/1 shots, the market has estimated the probability of each individual horse winning at 0.25. This is just an estimate of its winning probability however, and will be inaccurate to some degree of other.
So, the method I used initially to evaluate market efficiency to prompt the discussion (ie. Calculating the actual probability based on the actual result vs the estimated probability) will only tell us if that price is accurate on the whole. The market can be inefficient at a micro level (per horse or race) but efficient at a macro level, over the long term.
The opportunity for backers lies in backing horses where the probability of winning is greater than the market has priced it.
So, consider the two scenarios represented graphically below.
http://i840.photobucket.com/albums/zz321/cormack15a/dc_zps8f5c9693.jpg
Both lines on the graphs represent a data set of the probability of 500 individual horses winning a race they contested. The graph shows a representation of a notional ‘true’ probability that they will win (The ‘true’ probabilities in a horse race do exist but, as we are all acutely aware, because of the number of variables at play in a horse race it is almost impossible to assess wholly accurately, unlike a dice throw or coin toss where, assuming a fair throw or toss, the probabilities can be accurately assessed prior to the event. I’ve assumed for the demonstration/discussion here that these ‘true’ probabilities were known for the horses concerned and that, statistically, they formed a normal distribution).
In scenario A (the red line) the mean of the true probabilities is 0.25 (3/1) but the range of true probabilities is relatively narrow (a standard deviation of 0.02).
In scenario B (the dark blue line) the mean is the same (0.25) but there is a wider range of ‘true’ probabilities (a standard deviation of 0.05)
In both scenarios each individual horse was priced by the market at 3/1 (in other words the betting market had correctly estimated the mean probability of each data set as 0.25)
In the actual event the horses in each data set won 125 times out of the 500. (This is wholly to be expected because the ‘true’ probability of the horses in each data set had a mean of 0.25).
Using the ‘efficiency’ measure described in my earlier posts (average win probability vs market price) both scenario A and B turn out to be equally ‘efficient’. The probability estimated by the market (3/1 or 0.25) matched the mean of the actual probabilities (0.25) exactly.But do both scenarios represent the same opportunity for the backer? Which of the two scenarios offers most hope for the backer who considers himself to have an edge in terms of assessing the true probabilities when compared to the overall market? Surely the one where the variation is higher. Horses whose price is ‘wrong’ should, in theory, be easier to spot by the discerning backer.
- You must be logged in to reply to this topic.