Friday, 12 April 2019

A Priori, Patterns and Sceptics

Not for the first time, and probably not the last time either, Joseph Buchdahl generated some responses and interest with this Tweet yesterday:

I'm not sure Joseph has got his ROI calculation quite right, because an ROI** of 104.8% from 3,511 bets would certainly be impressive. I think from looking at his chart, theoretical profits are actually 168.53 points, which is a far more modest, but still decent, ROI of 4.8% and a P-value of 2.341 assuming average odds of 3.0, a far cry from the 0.0006 tweeted. Or am I missing something here? 

Joseph shares another profitable "system" although again hugely inflates the actual ROI% and P-value, and then a losing "system" although most readers would be very happy with an ROI of 92.8%! 

His conclusion was: 
don't be a schmuck and believe everything you see has a reason for being so.

Learn what regression to the mean is and don't be fooled by randomness.
Joseph later retweeted his initial Tweet and directed it to someone called Massimiliano “Maxi” Imparato with the message:
It appears that this Massimiliano, who started looking at Draws late last year, is now selling something called the "Golden6 system: new betting strategies to take advantage of draws" using slogans such as:
The best time to bet on draws was 7 years ago. The second best time is now.
While I agree with some of his spiel, his overviews of his six systems are heard to understand, and some appear to be complete nonsense, with the word 'pattern' appearing several times.
Golden6.4 Missing Draws: The home team comes from a small streak without drawing, and a game of uncertain outcome is played. This run without draws can cause value in their odds.
Aren't ALL games of "uncertain outcome"? And what does a 'small streak without drawing' mean? Basically nothing, except that a pattern has been found that has no predictive value at all. I guess a small fortune was lost this year on Tottenham Hotspur, chasing that overdue Draw!

However, I do know that backing the Draw in certain types of matches, has been, and continues to be, profitable. 

Presumably Maxi is offering a money-back guarantee for any seasonal losses, a step first taken by this ground-breaking service, even if it was never needed?

As I have written before, the idea of the Draw sometimes being value was first suggested, at least to me, by Derek McGovern in his 1999 book "On Sports Betting...And How To Make It Pay". Page 87. 

His ideas were reinforced ten years later by an article by Kevin Pullein, a name many readers are probably familiar with. Pullein stated that one of the major factors influencing the result of a football match is "the difference in ability between the two teams". Not rocket science, but the article went on to explain why the draw can be value in certain matches.

The work of David Sumpter more recently has confirmed the idea that in some matches, the Draw offers value. Sumpter has written about this using the Big 6 for his system, although Joseph dismisses the results writing:
I don't believe his [Sumpter's] "inefficiency" is real. The sample size is too small. I've given you samples in the thousands with big profits to closing Pinny prices, never mind junk bookies, that are totally worthless.
I don't like using all the Big 6 matches myself, as teams have off seasons when they are not so Big, and it's more important to select matches where the win probabilities are close, whether they be Big teams or not.

And agreed, it is a small sample, since we only have Pinnacle's closing data going back to the 2012-13 season, but we have prices from other books going back to the 2000-01 season - all thanks to Joseph I might add.  

These pre-Pinnacle seasons had over-rounds in excess of 112%, but if we adjust the odds to reflect an over-round of today's 102% for the nearly 19 seasons in the EPL, looking at Big 6 (or earlier Big 4) matches where the two teams had win probabilities within 10% of each other, the ROI is 29.14%

Yes, it's a small sample, but it's all we have. There is only one Big 6 and one EPL so we have to make the best of it.

Despite the roots of the 'Draw can be value' idea going back twenty years, and presumably satisfying Joseph's requirement for some logic behind the idea:

...but you would be a fool to look for anything and believing there was any point to it unless you had previously developed a methodology that could explain why this pattern occurred.. That is the point of this thread.
To my way of thinking, it's not a pattern, and it's not random, if you call it out ahead of time. 

Joseph remains unconvinced, because he is looking for statistical proof:
No one has convinced me that draws systems don't simply mean regress because they are based on nothing. 
They're not based on 'nothing' if they are based on ideas first aired in public twenty years ago, and as I have previously written, "betting markets are not ideally suited to the traditional standards of proof, mainly because they are constantly evolving and adapting and if any edge is present, it should soon be discovered and the market will adapt accordingly."

If we forget the Big 6, and use all the data we have from those (almost) 19 seasons, and just look at matches where the two teams had win probabilities within 10% of each other, the results are still good, with an ROI of 7.79% from over 500 matches. At 20%, the ROI is 9.65%, from 1,107 bets. 

It's interesting to look at the Draw price in those 1,107 'close' games. You might expect it to be generally around 3.4, which is indeed the average, but the price ranges from 2.88 to 4.08. If you back only in matches where the Draw is priced at the higher end of scale, profits are greater. E.g. at 3.6 and above, the ROI is 30.9%.

I guess the choice is that of the individual - wait until we have statistical proof of a system, (with 30 Big 6 matches a season, I'll be 137 by the time we have a big enough sample), by which time paradoxically, the secret is out and the edge has long gone, or we make hay while the sun shines, and if the system stops being profitable, look elsewhere. You're certainly not going to make any money waiting for solid proof from the sidelines.

Edges should not last for long, but they do tend to persist far longer than it would seem reasonable to expect and that is because betting markets are not made up of rational people. 

Does it make sense that backing small underdogs on the road (College Football) would be profitable 17 of the last 18 seasons? That small road 'dogs in the NFL are profitable 11 of the last 13 seasons? That NBA road favourites after a loss are profitable 11 of the last 13 seasons? 

I could go on, but there's a difference between identifying inefficient markets (because the public likes favourites, or doesn't like favourites if they lost their previous game) and finding random patterns in past data as Maxi seems to have done, at least in some systems. Apparently he has something like 150 which is a sure sign that he's a little confused and selling trash based on patterns and data mining. When Golden6.147 starts to show a loss, it'll quietly be dropped while the virtues of Platinum12.833 are extolled. 

Doctor C tweeted:
I think the example [from Joseph] is to highlight the pitfalls of blindly believing some of the false patterns that can be found when data mining and the importance of splitting data into training and validation sets. I think....
Joseph wrote:
IMO the only way to profit from draws is to back draw prices that are deliberately offered as loss leaders.
That would only make sense if a bookmaker were saying "you can have 3.8 on the Draw if you also bet on a horse in the 3:30 at Chepstow". I don't think it works that way. Why would a sportsbook offer a "loss leader"?

A Draw, like any other outcome, is offered at a price because the book wants to attract money on that outcome, and balance their ledger. If money continues to pour in on the Home and Away, the draw should lengthen. 

Messrs. McGovern, Pullein and Sumpter have all explained why the Draw price is sometimes higher than it should be, and my experience with the Draw tells me they are on to something. Anecdotally, how many times has someone, a neutral, watching a game on TV, told you they put a fiver on the Draw? 

A Lucky A Day tweeted:
I wouldn't read into this that they risk a loss in just about every market. I would interpret this as meaning that their books are seldom evenly balanced, so that one outcome will always be more favourable than another to them. I'm sure on occasion they may even take a loss, and perhaps are happy with a push on one result but a bigger win on others, but if they are intentionally taking risks on markets, then they are not the well run business that they appear to be, and we probably should withdraw our money now. There's also likely a PR element to that statement from their traders. A "we're all in this together" kind of thing.

If Pinnacle's lines and prices never moved, then I could accept the statement above, but the fact is that they do move, often significantly, and the reason is simple. They want to attract bettors to one outcome and deter them from another, and they are not shy about tweeting out meaningless 'betshare' numbers to help accomplish this.

For the record:    

** The return on investment formula is:
ROI = (Net Profit / Cost of Investment) x 100 

1 comment:

christoff-ka said...

nice. btw, his 104% is a total profit return of 4%.