Using the Closing Betting Odds to test for a Tipster's Skill
Posted 20th December 2016
In August I wrote an article showing how the Pinnacle closing line (or closing odds) can be used to estimate the advantage we might hold. Specifically, with the bookmaker's margin removed, my data analysis revealed that the ratio of the odds you bet to Pinnacle's odds at closing is a good predictor of your expectation. For example, if you bet odds of 1.8 and they closed at 1.65 (with a margin of 3%), your predicted profit expectation would be given by 1.8 divided by 1.65 divided by 1.03, or about 5.9%. I've also previously looked at whether a set of 48 tennis tipsters was beating the market in aggregate by outperforming the closing odds. This time I want to look at 4 individual tipping records separately to give you an idea of how we can use the closing line of an efficient bookmaker (i.e. a bookamker who doesn't ban winners) to test for whether a tipster is skilled, lucky or faking it.
The first of these records comes from the football tipster Jeremy Price. Now verifying to Betrush, he was for many years the top football tipster working at the tipster supermarket Betadvisor. I've previously written about how Betadvisor has largely been selling coin tossers. If any skilled tipsters did exist there, they were largely hidden amongst the noise of random chance. Since then, Betadvisor has assumed new ownership, a huge increase in marketed tipsters and an unwillingness to publicly share its data for analysis. Talking to Jeremy Price it appears he left Betadvisor on the grounds that his tips were allegedly being resold (by whom is not made clear) and that his tips were experiencing very fast price movements once advised. Customers were not happy and Betadvisor, it was alleged, did not do enough to manage this (presumably in the form of controlling the number of subscriptions). Anyway, all that is by the by, what we really want to know here is whether Jeremy Price really is any good.
One way to determine this is to statistically analyse his performance and compare it what might happen by chance. If his performance was highly unlikely to have happened by chance, we might reasonably conclude that he had something more (i.e. prediction skill). This was the approach I adopted for analysing tipsters I had verified in my second book How to Find a Black Cat in a Coal Cellar: the Truth about Sports Tipsters. One drawback with this method is that it can take a long time to be confident that a tipster is not just randomly tossing coins. Instead, let's now look at Jeremy's advised odds and compare them to their closing prices.
Since beginning to submit tips to Betrush in November 2016 Jeremy has advised a total of 26 tips (as of 17th December). [You can download these tips here, along with those for the other tipsters I have analysed for this article.] Of these, 23 saw a shortening in price from his advised figure to the one published by closing, with only 3 lengthening. [All but one was with Pinnacle, the other being with SBOBet.]. Assuming we would expect 50% to shorten and lengthen where the tipster was just guessing, a chi-square test comparing expectation and observation shows his performance to be highly statistically significant (p-value about 1 in 100,000). Similarly a t-test comparing the two sets of odds is also highly significant (p-value about 1 in 60,000). On average Jeremy out performs the closing market by about 10%. With a typical margin of about 3% in his tips we should expect his long term yield to around 6 to 7%. We have managed to predict this from just 26 tips and a record of only 2 months. With nearly all his tipped prices shortening by closing time, the implication is that Jeremy Price is a marked man at Pinnacle.com.
In fact Jeremy's long term record at Betadvisor was almost 13% from well over 1,000 tips, although it should be noted that this was up to April 2016, the only entry for his page in the Wayback machine. Betadvisor, sadly, are in the habit of removing past tipsters records from their website so they are not available for scrutiny. Since most of these will be losing records, you can see clearly what the intention is by doing it. Explanations for why my prediction for his long term yield is lower than his previous Betadvisor performance might include good luck (in his earlier record) and the betting marketplace becoming a tougher environment to secure an edge. As more and more punters and tipsters take up the challenge of betting, the betting market arguably becomes more and more efficient through a process known as the paradox of skill. Although Jeremy has operated in very niche betting markets which possibly explains how he's come to find his advantage in the first place, over time we should expect it to become harder and harder to maintain it. Nevertheless, a 7% yield from even-money betting is not something to be scoffed at once you fully understand and appreciate the difficulties involved in finding any kind of edge at all.
A couple of other clues exist to reveal Jeremy's genuine skill in forecasting. Firstly, for his most recent 4 tips for which full odds movement records are available from Oddsportal, we can see that when he tips his price, the bookmaker cuts it dramatically in one move. One tip, for example, was cut from 2.00 to 1.78 in one go. Part of that will be due to the small liquidity in the markets Jeremy targets and part because he's obviously marked as a highly knowledgeable tipster. Secondly, the timing of the odds shortening is typically within the first minute after the advice is issued. Any wonder, then, that Betadvisor customers subscribing to his service were complaining of odds unavailability. Conceivably, the bookmaker was slashing odds after just one bet. If that was Jeremy himself then it's difficult to see how anyone else could secure their advantage for themselves. Given the size of first odds movement, Jeremy's tips are arguably unprofitable except for the first person to bet them. If that isn't consistently the same person then everyone is wasting their time and money here, and even if it is, then all but one is wasting them.
The second record I've looked at comes from another Betrush tipster called Eredivisie2000. As the name might suggest, this tipster bets only on Dutch football teams. Prior to proofing to Betrush, he's also been posting on Blogabet, showing a pretty healthy 6% yield from over 700 picks. Sadly, my analysis of 23 of his most recent tips since he joined Betrush in November leave a question mark as to whether a lot of that 6% is just chance. Although 17 of the 23 picks witnessed a price shortening by closing, his average advantage over the closing odds was only 2%. After Pinnacle's margin is removed, I would predict that he'll barely break even over the long term. Eredivisie2000 may very well disagree with this point of view, as would any tipster I might challenge using this methodology. However, all I'm really doing here is describing what Pinnacle thinks of his tipping abilities, not what I think. Time, as always with these things, will tell who is right.
The third record I've analysed comes from Betadvisor. Aris Papadopoulos is a football tipster showing a mighty 25% yield from nearly 300 picks and a very solid looking growth trend. According to my p-value calculator, we might expect such a record to happen by chance just once in about 40,000 times. By any stretch this would be very strong evidence of a tipster who knows what he's doing. Sadly, the analysis of closing prices paints a different picture. Neither the chi-square test nor t-test are remotely significant for his 21 picks in December, with 8 lengthening, 13 shortening, and an average advantage over the closing odds of just 1% (unprofitable once the margin is removed). Pinnacle clearly isn't particularly bothered by his appearance in their betting markets.
Judging by the recent comments posted about his tipping ability that coincide with a December downturn in performance, it's entirely possible that Aris has inexplicably changed strategy, focusing much more now on Over/Under markets rather than Asian handicap as he had previously. To get a clearer picture of his ability as perceived by Pinnacle, we would need to look at the closing odds for his full record. Nevertheless, the omens are not good, with the suggestion that long term his record will simply regress towards the mean.
The final record I've chosen to analyse comes from Pinnacle Tipster, a football tipster specialising in Asian handicap betting. Despite the name, it's not clear whether the odds the tipster quotes come from Pinnacle. Nevertheless, other bookmakers typically follow the 'Pinnacle Lean' so it's reasonable to use their closing odds to assess the tipster's advantage over the closing line. Pinnacle Tipster claims a truly phenomenal performance of 78% win rate from of 1,000 picks averaging even-money prices. This would translate into a yield of something like 55%. The more savvy readers amongst you will know where I'm going with this, but let's do the analysis nonetheless.
As of writing, the 18 picks issued in December saw 5 prices shorten with 13 lengthen. The average ratio of tip odds to closing Pinnacle odds was 0.976. In other words, Pinnacle Tipster appears not to have the slightest idea how to spot a mistake in the betting market which will steam after he advises it. Such a finding would be completely at odds with a tipster claiming a long term advantage of 55%. Should such a tipster actually exist, we would reasonably expect to see closing odds in the region of 1.3 for advised prices of 2.00. Plainly this is nonsense. Bookmakers don't make such large mistakes consistently enough to allow this sort of thing to happen when pricing up matches. The conclusion, therefore, is obvious: Pinnacle Tipster is cheating.
Seemingly, there are two ways this could be done. 1) Wait for the match to finish and then quote the betting odds. Such an approach, however, would be detectable by all closing prices matching the tipped prices. This does not appear to be the case. 2) Many of the losing tips are removed from the record before it is published. Given my experience of a handful of tipsters who have tried this approach even when verifying their tips with me, this would seem the most likely explanation. Of course, it won't come as a surprise to learn that the tipster monitor who verifies Pinnacle Tipster, Tipster Connection, it itself a serial scammer whom I have previous shown to be associated with most of their services it monitors, by showing them to have the same Google Analytics account IDs. Equally unsurprising are the performances of some of Tipster Connection's other marketed tipsters. Pinnacle Tipster in fact is only 14th out of 16. Most of the tipsters have strike rates in the high 80%s, with one above 90%. Evidently, only the most stupid of punters will be suckered, but given that Tipster Connection has been going for so long, plainly there are still enough out there for the owner to make a living from committing such fraud.
In summary, the closing line, and more specifically that of Pinnacle.com, can be used with great effectiveness as another tool to assess the quality of a tipster. Before you consider buying their tips, always check to see if they are beating the closing market, no matter how good their profits record may appear to be. Of course, once you've established that they are, determining whether you'll actually be able to bet their advised prices before they shorten is another matter altogether, as the analysis of Jeremy Price's tipping record has shown.