Thursday 30 May 2019

Rationale Thinking

Most of my systems have sound logical reasoning behind them, including the Totals system mentioned yesterday for Major League Baseball. 

The system takes advantage of the fact that most punters have preconceived, and fortunately for us, incorrect ideas about what total might represent a good bet.

To use extremes to make the point, a total set at 14 would appear to the average punter as a high number, triggering the instinctive and quite natural response that "the value must be on Unders", while a total set at 5 would prompt the opposite reaction.

My thought process, as was the case with NBA totals, was that in some scenarios, a high total is value for an Overs bet, and likewise a low total offers value on the Unders. 

Note the "some scenarios". A blind backing of Unders below, or Overs above, any number won't work. Those scenarios included various factors such as the rules for each league, and the number of at-bats per game. 

This was the a priori (as philosophers and betting skeptics refer to it) behind this idea.

The important thing is that there was a logical reason why this system should work, and as it has turned out, as least so far, it does. 

Will the market correct? As with other ideas I've shared, it should, but it often takes its time in doing so.

Look at my College Football (Small Road Dogs) 50/50 system (left) which has had a winning record for 18 seasons straight.

It could be luck of course, as skeptics might have it.

Or.... Or it might just be that favourites are over-rated by the average college football bettor. Surely not. Everyone knows home advantage is worth three points. Maybe not always.  

But at what point does even the biggest skeptic become a convert? If 18 years doesn't do it, then probably 25 years wouldn't. I'm skeptical by nature myself, but I'm also open to a logical idea supported by years of evidence.

A Lucky A Day raises the possibility that the Totals System might be the result of data mining. To me, data mining is an exercise that produces systems such as:
Back the Home Favourite in an American League game when the pitcher lost last time out, the visiting team won their last game but gave up two runs or more in the third, the temperature was 73F or hotter, and the game is being played on a Thursday afternoon. 
(Incidentally, the day of the week can be significant, not because the result is impacted, but because the markets change, as has often been noted on betting forums regarding weekend horse racing).

Here's an actual example of a nonsense 'finding':
99 pitches or more, or the price was 162-plus, and they are 0-7 SU. Maybe. 

To me, this kind of thing (I hesitate to call them systems), is the result of data mining, but I'd argue that it's not data mining when there is rationale involved, rationale defined as:
a set of reasons or a logical basis for a course of action or a particular belief.
In this case, I have a theory about where and why totals markets might be inefficient and then see if the results show that to be the case. So far, they tend to do just that, especially if A Lucky A Day's 1 in 10,000 estimate is accurate!

Remember too that we're not looking for wins, but profits. We are looking for inefficiencies in the market. Find value, and the profits will come.   

Wednesday 29 May 2019

MLB Totals Simplified

The @PinnacleSports Twitter account highlighted a 2016 article on totals betting in Baseball, with the following quote highlighted:

"At most bookmakers it is effectively impossible to beat the Over because unsophisticated bettors tend to go Over not realizing that the numbers don’t add up the natural way."
By extension, if the market is so efficient that you can't win on Overs, then you can't win on Unders either, but it's not that clear cut. The article itself includes this comment:
Even at Pinnacle we can’t be sure that we have the best of it, especially when the person betting is sharp, so often we have to move the line and offer what we think is likely to be value. While the under is more likely to offer you value on a game where you don’t particularly like your side – especially if you’re playing at another bookmaker than Pinnacle – the over is where you can have vastly the best of it in the right spot. It’s never a sure thing that a pitcher will get it done.
As is often the case, the article complicates what is really quite a simple problem to solve. Let the market do its stuff, and take advantage of its inefficiencies.

I apply a similar strategy in baseball to that for the the NBA. When the total is (relatively) high, my expectation is that the public will be deterred from backing Overs, and Overs can thus offer value. 

Similarly, when the total is (relatively) low, the Under doesn't appeal to the unsophisticated public, and this selection can offer value. Remember:
Most punters lose. They are ill-informed, intrinsically lazy, psychologically flawed, impulsive, ill-disciplined, incapable of appreciating the importance of affect to the decision making process and prone to imitative and repetitive behaviour.
MLB has two conferences, and two sets of rules, with significant differences which need to be taken into account, and games hosted by the Colorado Rockies always have a higher total, something else that needs to be treated as an exception. 

The average total of runs per game over the past five seasons (including the current 2019 season) is 8.3 in the National League, 8.6 in the American League, and 11.0 for games in Denver.

The Totals markets are, at first sight, remarkably accurate.
For the National League, Overs and Unders are split as shown above, and the American League is close to 50/50 at 50.8% in favour of Unders.

First sight can be misleading however. 

By adding a few conditions to my Totals bets, for example with Overs, I want the favourite to be the away team, (since that increases the chances of both teams batting all nine innings and 54 at-bats is much more favourable than 51 when you're wanting runs) it's not hard to see where the market weaknesses are. 

I reverse the rules for the Unders, and as you can see below, this is also a profitable strategy.

Results for the latest five seasons:

It's a simple, and extremely time effective, strategy. There may be sophisticated models out there, requiring a lot more effort, and requiring huge databases for evaluating pitchers and batting line-ups, but how many of them can match or exceed the double digit ROI percentage on Overs since 2015? 

Likely not many. Simple is often best - the KISS principle.

Sunday 19 May 2019

Cassini's Law

There's a law known as "Cassini's Law", which states that the moment you start talking about a long winning sequence, along comes a loser. Or in this case two.

The T-Bone System ended its winning run at 13, a number not renowned for being associated with good fortune, with not one, but two losses on Friday, which takes the profit on the basic systems down to 9.25 and 8.60 points on the Straight Up and Run Line respectively.

Results for May to level stakes so far:

Meanwhile in the NBA Playoffs, the strategy of backing Unders is profitable so far. 

The logic I use here is that in the playoffs, the total in Eastern Conference games over the last ten seasons is six points fewer per game than in the regular seas, while in the Western Conference, the drop in points is 2.51. 

However, weighting so that recent seasons are more impactful, and the averages are 4.61 and 1.58 points per game.

This season, the regular season average total for Eastern Conference games was 219.7 points, so backing the Unders in games where the line is around 215 points or higher should be unpopular with the public, and thus offer value to the sharper bettor. Likewise in the West, the total where the squares balk should be around 222 points.

So far, we have a 10-7 record in the East, while in the Western Conference, we have a 5-2 record. A 15-9 record is a 22% ROI, but with so few matches in the playoffs, be wary of some variance with this idea.     

And with that, I leave you for a week. I'm off to Las Vegas shortly at work's expense for a different type of conference, but hoping to find a few hours to play some Craps and watch some games. Craps may have a -EV, (1.36% or less if playing optimally and taking the odds), but it's a lot of fun, a very social game, and after a big win a few years ago, I'm still in lifetime profit. Not to mention the free drinks, meals and rooms they will offer you after a long session.

A hot roll at the Craps table is like a winning sequence on a sports system. You want to keep pressing, but know at some point the 7 will show up. 

Friday 17 May 2019

Repetition Compulsion

A thread appeared on my Twitter timeline on the topic of whether someone can be so bad at betting that money can be made by fading (opposing) their selections.

It's an idea I remember reading about years ago. The argument was that in betting, even negative information is useful.

Just a couple of months ago, I wrote that:

Given that my grandmother would be expected to pick winners at a rate of around 50%, and she's been dead for over 30 years, this didn't seem to be a hugely impossible task.
The fact is that when you are betting on coin-toss events, picking your selections at random means you should break even over the long term, excluding commission of course, which is what ultimately beats most bettors.

Losing intentionally is as hard as winning, as anyone who has tried arbitrage betting with the intent to lose on Betfair and win elsewhere to avoid Premium Charges, will be able to confirm. Well, anyone except Tony Hargraves, for whom the laws of probability apparently conveniently suspend themselves in this regard.

While many people, some successfully, put in a great deal of effort to find value, enough to beat the odds and commission, and become profitable over the long term, it's not clear why anyone would make such an effort to intentionally lose.

As the always interesting @RufusPeabody wrote:
The idea that there are people so bad at betting that fading them — betting the opposite of what they’re on — provides value is preposterous. Yes, the average bettor is worse than a coin flipper (I’d guess like 49.7%) because of biases that they fall victim to, but not by much.
Think about it logically — what talent could someone (ostensibly trying to win) have for picking losers? Where is that skill coming from? And what does that say about the efficiency of the market if that’s possible?
@JFCPicks contributed to the debate that:
Overreacting to recent results is the only thing I can think of that's liable to result in long-term losing. Theoretically, it should be just as hard (if not harder) to lose consistently betting -110 than it is to win consistently betting -110.
Indeed. While this bias should be self-correcting, in my experience there are several markets where it is not. The success of the T-Bone System is based on this bias. 

Profitable, so-called 'revenge', systems aren't successful because the team tries that little bit harder next time out. Their success is because the market is inefficient, the inefficiency due to the aforementioned overreaction by the betting public, and thus the market, to recent results.  

Speaking of which, hopefully some of you are following the T-Bone System again this season, and enjoying one of the longer winning streaks currently (see left). 

The 'official' ROI for straight up bets is 19.4% this season, and for Run Line bets it is 23.9% while to a level stake, the ROI is 20.9%.

There may be one more winner to add to that list, as the Athletics just qualified yesterday at -140 and beat the Tigers by 14 runs, but sometimes these borderline qualifiers drift and don't get counted officially. 14 runs would be the largest margin of victory for this system in over a year - when the Athletics were on the receiving end of a beating in Houston. 

Of course, this inefficiency should disappear, but these biases are persistent little devils, which is good news for those of us who enjoy making hay.

Hot favourites in baseball have performed well in recent seasons as readers will know, and after a losing March / April have resumed their winning ways this month.
Again, this isn't because favourites have a life of their own, and are pulling their socks up and trying harder this month. The success of this strategy is down to the public's resistance to betting on hot favourites in baseball, and the resulting inefficiency that is there to be taken advantage of.

The probability of these returns being by chance? Using a calculator I downloaded some time ago, I believe from Joseph Buchdahl's web site, I get the following.

Run Line:
Straight up:
Interesting how much difference there is between the straight up bets at short odds and higher stakes compared to the run line and its more even odds and stakes.

The numbers actually look too good to be true, and I should probably remove this post and all references to baseball betting, but even if they are correct, not everyone will be convinced by them, and that's a good thing. 

A one-in-a-million probability wouldn't sway some people. 

Matthew Trenhaile woke up after a long nap, more accurately after a long  hibernation, and has resumed activity on Twitter, which is good news. He is well worth following and recommended the BettingMarket web site.

Somewhat relevant to this post are these opening lines from an article there titled: On the psychology of the betting market:
Most punters lose. They are ill-informed, intrinsically lazy, psychologically flawed, impulsive, ill-disciplined, incapable of appreciating the importance of affect to the decision making process and prone to imitative and repetitive behaviour. When punters lose they typically come back for more (repetition compulsion); when they win, they become overconfident and prone to risk-seeking.
Long may this continue.  

Monday 13 May 2019

Shameless

While many of us are used to some rather strange, and often unpleasant, content from the Daily Mail, the below Tweet from their North-West football reporter Dominic King was a late contender for the "Most Ridiculous Tweet of the Season" award:

I personally don't see anything wrong with this practice at all. The league season is over, but even if it wasn't, I don't see why offering prices for next season, or any future seasons, should be considered 'shameless'.

Is there supposed to be an official mourning / celebration period at the end of each season during which respect is shown and the vulgar business of betting suspended? First I've heard of it.   

Possibly some projection at work here, as working for the Daily Mail would certainly fall into the 'shameless' category. 
Daily Mail, 15 January 1934

And if you want a more recent example of shameless from the Daily mail, here's a 'story' from today:

https://twitter.com/sportsdrzaf/status/1127650172710858757

In the financial markets, prices are quoted for several years into the future 
Weather forecasts also make long range projections, and The Old Farmer's Almanac has done so since 1792. Shameless!

I'm not sure why football markets, or any markets for that matter, should be any different, and the use of the word 'shameless' in this context, is absolutely ridiculous. 

For those shameless enough to be looking to back Manchester City to win the league in 2019-20, Pinnacle have them priced at 1.714.  

Sunday 12 May 2019

EPL 2018-19 Records

With 36 goals in the final round, the record for goals scored in the Premier league was broken with an average per game of 2.821, mostly due to the increase in Away goals which also hit a new high this season: 

Not surprisingly, Away wins ended the season above one in three for the first time in English top flight history: 
The Draw finished at a Premier League record low, the lowest percentage for Draws since 1931-32.
At the individual club level, no club had previously drawn fewer than three games in a Premier League season, but this season two teams (Manchester City and Tottenham Hotspur) managed that feat, with Tottenham a mere 15 minutes away from just one Draw all season. 

Two Draws in a season is a top tier record fewest for a 20 or more club season - Sunderland set the record in 1908-09, so for two clubs to match this in the same season (103 seasons in total) is quite remarkable. 

At the other end of the Draws table, only Southampton made double figures this season with 12, three clear of the next highest total. Like Tottenham, Southampton also drew their final game of the season. 

Twelve is the joint lowest 'winning' total in the Premier league era, matching Aston Villa's 12 in 2005-06, although in that season three other clubs also made double figures. 

This Premier League season is also only the second where the title would have gone elsewhere if we were still in the pre-1981 era of two points for a win. Manchester United would have won the 1994-95 title on goal difference from Blackburn Rovers, and Liverpool would be champions today. 

Look forward to more fascinating insights in the coming months. 

Tuesday 7 May 2019

End of Season Goals

With just one round to go in the English Premier League, only a somewhat improbable clean sweep of ten draws next Sunday will prevent this season ending with the lowest Draw percentage in Premier League history.

It may come as a shock to some, but there was actually top level football prior to 1992 and here (left) is how the current season ranks overall for all seasons where fewer than one in five matches ended as a Draw.

Draws on the final day of the season don't occur any more or less frequently than in other rounds - some seasons zero (2017-18), others as many as six (2006-07), with an overall average of frequency of 21.67% since 2000-01. 

Backing the Draw in matches where the true difference between the teams is less than 25% and the ROI (using prices adjusted to an over-round of 102.5%) is 8.6% on the last day, not much different from the 9.13% over the 19 (almost) full seasons. 

No surprise that, once again, there are no Big 4 or Big 6 contests on the final day of the season. The last time this happened was in 2003 when Chelsea beat Liverpool 2-1 to clinch fourth place. It's almost as if this is by design. 

For those wondering about Aways, which is quite possibly no-one at all, the percentage currently stands at 33.24%. This season is already guaranteed to be the joint highest (tied with 2013-14) for Aways in top level English football ever, and just one more win will set a new high, with four more wins pushing the percentage above 33.33%. 
The total number of goals will also be interesting to watch. 31 more, and we will have an all-time Premier League record, while a low scoring final day with less than 16 goals (highly unlikely) will see the season finish eighth all-time. 

The final day sees an average goals per game of 3.12, with a high of 37 and a low of 23, so a new record is well within reach.

Saturday 4 May 2019

Treasure For Sale

The infamous Gadoon Kyrallos, aka Spanky tweeted:

I have a copy of that book, and actually re-read it after reading Spanky's recommendation thinking I may have overlooked something, but aside from some interesting background as to how money line and point-spread betting came about, and the role of a bookmaker, and the importance of finding the best prices (soft lines), but surely anyone who is reading this knows about the principle of only betting when you have value. There's little else of any value in it as you might expect from a book written before the Internet was in common use. I replied to the tweet:
I re-read this book a few weeks ago. It’s outdated and not worth 80 cents, never mind $800. Got some interesting historical references admittedly but won’t give you an edge in today’s markets.
Spanky's response: 
My success in sports betting is based on the underlying principles in that book. To each his own brother. One mans junk is another mans treasure.
The conversation continued:
Seriously, if anyone wants to offer me $750 for the book, it is yours, and I'll even pay the postage and insurance.

** Update 5th May 7am: I don't make a habit of editing posts, but I just now came across a 2007 review of the book in question on the Las Vegas Revealed web site, and thought I would post it in full:

Reviewed by Nick Christenson
July 18, 2007

I only rarely write reviews of out-of-print books, and then usually only for those books that have some sort of timeless significance, such as Mike Lee's Betting the Bases, or David Hayano's Poker Faces. Certainly, I can't say that Jack Moore's, The Complete Book of Sports Betting belongs in this elite company. In fact, I debated with myself as to whether I should write this review at all. I do think that Moore's book is interesting, though, and it deserves more attention than it has gotten. The reader will ultimately need to judge whether it is worth seeking out or not.

Moore offers a new, at least new at the time of its publication, strategy for winning at sports betting. As such, the principle target of the book are those who have tried other methods of winning at sports betting, and presumably failed. It is these folks who are in a position to appreciate the novelty of Moore's approach. The author does not take for granted prior knowledge of sports betting and its terminology, though. He provides a great deal of background aimed at equipping the novice with enough information to understand the the complex jargon of the bookmaker. Even just a decade after publication, some of the methods for quoting odds that Moore describes are no longer in common use, but the introductory material is good and devoid of the silliness that pervades many other books on sports betting.

About 60% of the way into the book, we finally learn Moore's secret to winning at sports betting. This method is simple enough to be described here, and since the book is now out of print, I don't feel bad about revealing its secret. The method Moore espouses is to assume that the consensus betting line is an accurate measure of the proper game line. Then look for shops with betting lines that differ significantly from that consensus, and then bet them. With this method there is no handicapping, and Moore calls it the "blindfold method". The author is not the only person who espouses this method as a component of a winning sports betting strategy. The well-known sports bettor who goes by the name of "Fezzik" uses a similar method as the cornerstone of the way he bets.

While I believe that Moore's strategy is fundamentally sound, I'm not as impressed with some of the details he provides. He states his opinion on how many points of deviation from the consensus line one needs in order to make bets of a given size, but these seem to be based on the author's intuition and experience rather than a careful examination of years of data from each sport. Also, in some sports, especially American football, not all points are created equal. Moore explains the value of the "3" and "7" in the NFL and college football earlier in his book, but in his betting method a line move between 2.5 and 3.5 should be considered the same as a line move between 4.5 and 5.5. Anyone who has studied football results will understand that this is absurd.

Consequently, I would say that Moore's "blindfold method" represents a reasonable beginning for a strategy that can be used to win at sports betting, but is incomplete as it stands. Some more number crunching, especially for sports such as football, will be necessary in order to turn this into a reliable winning method. Also, the author's recommendation that having three different "outs" at which one may bet is probably not sufficient to generate much betting volume. This is due to the efficiency with which information propagates these days, largely due to the Internet. On the other hand, the Internet makes it much easier to obtain access to multiple lines and efficiently compare them. The upshot of this is that one probably wants many more than three outs to successfully apply this method.

Moore's book is interesting, and if it were still in print (and accounted for the emergence of the information age), I would be happy to recommend it to those interested in sports betting. Because its methods aren't "sexy" and don't allow bettors to put money on their favorite teams more often than others, Moore may not have ever gained a large following. However, since his book is devoid of the sort of misinformation that fills so many other volumes, he deserves significant credit for accomplishing at least that much. I don't think it rises to the level of "must read", and in my opinion those who have made it this far into this review now know over half of what the book has to offer. Still, the book deserves some consideration by those who are interested in the subject of sports betting.

Capsule:

In the out of print book, The Complete Book of Sports Betting, Moore advocates a market-based approach to winning at sports betting. I believe the basis for his method is sound, although I don't believe the book contains everything the bettor needs in order to succeed with it. I wouldn't rate this book as a "must read", but since it is significantly better that the average book of this genre, and because it approaches sports betting from a novel vantage point, I certainly wouldn't discourage those interested in sports betting from seeking out a copy.

Again, if you would like my copy, please send offers, while bearing in mind the ebay price is currently just under $500. 

Friday 3 May 2019

Draw Collusion

When a draw suits both teams in Italy, the game will end in a draw. It's all to do with the mentality of the Italian people. They see nothing wrong in such an arrangement. - Liam Brady
I have written several times before on the Draw in Serie B, and it is that time of year again with Future Greyhound calling me to task for my observation that the Draw can never be the Favourite in a football match: 
Enjoy your stuff Cassini. But as a pedant I must point out the Draw definitely can be Fav in football. Draw Collusion has been a studied topic (USA v Germany 14 WC & Uruguay v Mexico 10 WC) and think Draw was Fav in one of those from memory. Don’t get me started on Seria B!
Nothing wrong with being a pedant, and of course, the Draw can, and sometimes is, the market favourite, but my statement was intended to refer to matches expected to be played competitively.

Future Greyhound mentions Serie B (actually he references Seria B, but only a pedant would point out the mistake) and since Pinnacle's Closing Prices were added, i.e. the 2012-13 season, the Draw has been the favourite in 37 matches, including six this season already. 

Incidentally there were no prices available on another 25 matches, all of which were May games, 10 of which resulted in Home wins, while the other 15 were all Draws. Possibly other bookmakers were offering prices on these games, but I doubt that you would have been able to bet too much on them.

For those 37 matches where we have Pinnacle's prices, backing the Draw would have been profitable to the tune of 23.4%

In 489 other matches where the Draw was priced at 3.0 or shorter, backing the Draw would have lost you money, something that is not easy to do in this league.

Blindly backing the Draw in the Pinnacle era, and you'd be up 160.35 points from 3,070 matches, a decent ROI of 5.22%. My strategy of calculating the 'true' difference between the teams increases this to 7.59% across almost seven seasons.