Tuesday, 16 July 2019

The Siren of ROI

It was quite an exciting weekend for sport, with the Tampa Bay Rays beating the Baltimore Orioles on Saturday night another win for the T-Bone System...

OK, so maybe that wasn't most people's idea of the sporting highlight of the weekend, which for me was my son making his debut in the boxing ring, and opening his career with a first round TKO after about 90 seconds. 

For most people, I suspect the highlight was either the Formula 1 British Grand Prix, Wimbledon tennis or a cricket match.

The last two were certainly dramatic, with presumably vast fortunes won and lost on the exchanges for those who like donating their money to court / pitch-siders, but for me, those sports are to be enjoyed for themselves rather than as markets to trade. I did see that Federer had traded at 1.01 but didn't even look at the cricket. An edge in cricket, tennis or motor racing, I do not have. 

With football still a few weeks away from getting interesting again, sports investing at this time of year is all about baseball, with plenty of time for research for the new American Football seasons. 

I mentioned the Aestivation System I play in a recent post, and someone suggested another improved system for this time of year, with the qualifiers that the game be an intra-league game, that the home team has more wins this season than their opponent, and are favourites playing at home.

This is certainly impressive, all-time the records are:

But one can get caught up in the race to find a glamorous ROI. Remember my 2011 post:
Those who’ve spent a lifetime maximising ROI, I guess you’ve now realised that in a punting context, those who are able to grow their bank balance more significantly are, by most people’s definition, the more successful.
So, in summary...Return on Investment for show, Rate of Bank Growth for dough. £, not %.
The key to this strategy is simply that the market undervalues favourites out of the All-Star Break, which is all I concern myself with.

The only point in eliminating perfectly solid bets for the sake of an improved ROI is if you don't have the time to place the bets required by the system and need to focus your time and energy on the big dividend generators.


Value is so hard to find though, that it'd be a waste to leave money on the table. 

One strategy I use is variable staking, so the basic system from the above example would be to back all favourites, but back the higher ROI bets with larger stakes. 

Why ignore those Away Favourites with the miserable ROI of "only" 16.7% just because the Home ones are at 21.1%?

Regarding the intra-league qualifier, very few first matches after the break are non-conference affairs, to be precise just five, so while they do have a losing record of 2-3, (1-3 at home, 1-0 away) that's a meaningless sample size.

It's not hard to find high ROIs, but typically they come at the expense of only offering a few bets. 

If you back Chicago teams playing night games against Divisional rivals as dogs when Joe West is the home plate umpire, you're looking at an ROI of 130%, but 6 games from 2007 isn't too much evidence! 

Sunday, 14 July 2019

Changing Numbers

Either it's a remarkable coincidence, or this blog is more influential than you might think.


Following on from my recent Centering The Pill post, Axios jumps on to the story with this piece:

MLB accused of juicing baseballs following historic home run surge

Yesterday's Home Run Derby news cycle began with All-Star Justin Verlander emphatically saying that the league is juicing the baseballs, adding more fuel to an already raging fire.

By the numbers: MLB teams are projected to hit at least 6,463 home runs this season, which would break the all-time record set in 2017 (6,105) by almost 400, per WashPost.

"It's a f---ing joke. Major League Baseball's turning this game into a joke. They own Rawlings, and you've got [commissioner Rob] Manfred up here saying it might be the way they center the pill. They own the f---ing company. If any other $40 billion company bought out a $400 million company and the product changed dramatically, it's not a guess as to what happened. We all know what happened.
Manfred, the first time he came in, what'd he say? He said we want more offense. All of a sudden he comes in, the balls are juiced? It's not coincidence. We're not idiots. ... They've been using juiced balls in the Home Run Derby forever. They know how to do it."— All-Star Justin Verlander, via ESPN
Context: MLB does in fact own Rawlings, but they purchased the company last summer — more than two years into this ongoing home run surge. Obviously, that hurts the argument that the balls were suddenly altered once MLB took control.

On the other hand, multiple independent studies have shown that, beginning in 2015, the balls changed. So perhaps MLB was influencing the design before the purchase?


Hitters are being coached to elevate the ball like never before, and that is certainly contributing to the record numbers, but something is going on with the balls, and the players know it.

The bottom line: I'm not sure we've ever seen anything quite like this juiced ball saga, and MLB's "nothing to see here!" stance grows more disingenuous by the day.


And here's the crazy part: If they'd just admit that something significant is going on with the balls — instead of gingerly suggesting that maybe something is afoot — would anybody be mad? Fans love homers, and even pitchers would likely appreciate the transparency.

Instead, Manfred continues to play defense, seemingly worried that anything he says will make the league look bad when, in reality, the only thing making the league look bad is that they refuse to fully acknowledge what every fan and player is thinking.

Well some people might be mad, yes. Anyone who uses stats to come up with an informed opinion might feel that the playing field has been tilted a little, but this is one of those times when someone spotting such a trend can do very well before the market catches up.

Which is why it is always worth looking at the numbers and at off-season rule changes. 

Not to give too much away, but the NFL is a great sport for this strategy. 

For example - look at how the missed extra point attempts jumped from 8 in the 2014 season to 73 in 2015, or the jump for two point conversions from 29 in 2014 to 84 last season.

There are also changes in strategy, many triggered my data analysis, for example fourth downs are attempted more frequently today than they were several years ago. With change comes opportunity.    

Saturday, 13 July 2019

Back From Aestivation

After a slight setback last season, the strategy of blindly backing favourites in the first game after the All-Star Break returned a profit, which at 11.35 points was the highest since 2010.  

I mentioned this in 2016, so hopefully some of you made some money, although I must say that my inbox hasn't been overflowing with thank yous or donations. From 2016: 

One trend that did hold true again this season was that of favourites doing well in their first games after the break. The rationale for this is that although the better teams are likely to be better represented at the All-Star game, many of those will only put in a less than exhausting cameo performance, while the majority of their players enjoy the break and get some rest which apparently benefits better teams more than worse teams.
While I'm not typically a big fan of once a year systems, (a losing day lingers in the memory for a year), this one does appear to have some merit.  

Going back to the dawn of the MLB database, season 2004, (2007 for the Run Line), this idea has double digit ROIs of 13.6% on the Run Line, and 14.9% on the Money Line, and a total profit of 74.65 points.
Running Joseph Buchdahl's p-value calculator, the probability that the Money Line returns are by luck alone comes out at 0.24%, or 1 in 417. 

Tuesday, 9 July 2019

Sesquicentennial Season

It's hard to believe that the 2019 American Football season is now less than three and a half weeks away with the first pre-season NFL game kicking off on August 1st.

The College version of the game is on August 24th, and this season will celebrate its 150th anniversary on November 6th this season, with the generally recognised first game having been played in 1869 between Princeton and Rutgers, although the rules were more like Football (Association) than Football (American). Rutgers won 6-4, but a week later Princeton won 8-0 and the National Championship was shared, albeit retroactively! 

College Football can be complicated with 130 schools (teams) playing in the NCAA Division I Football Bowl Subdivision (FBS) with most (124) playing in 10 Conferences. Six teams are Independent. 


The ten conferences are themselves considered in two groups, the "Power Five" which are the strongest, and the "Group of Five" which are the weaker ones.  
 
Overall, as readers of this blog will know, the strategy of backing small road 'dogs is a long-term winning strategy with a winning record every season this millennium, although in three seasons the hold (at -105) meant a small loss.

Some more observant readers may notice that the table of results on the left is slightly different to those previously published.

The reason is that I am restricting selections to the Division 1 games. The small number of other games included in the database don't add a lot of value to the results if it is almost impossible to get a bet on.

As a result the 8.1% ROI from 1,740 selections is now down to 7.9% from 1,657 selections, which established over 18 years should still be worth getting up for, although of course, some will say it's just chance and that there is no such thing as a public bias favouring home teams in this sport. Long may that continue.

Most college games are conference games, and the ROI from 1,120 games is comparable at 7.2%, but not all conferences are the same. 

In recent years, defined as the last five seasons, four of the "Group of Five" conferences each show a double digit ROI, with the Mountain West showing a small loss. 
In the elite, and modestly monikered, "Power Five" conferences, recent overall returns are steady, if not spectacular. 
With each conference having its own personality, there are of course differences between individual conferences, and with some conferences sub-divided into divisions, there are differences there also. 

My favourite conference, for personal reasons, is the currently top ranked Southeastern (SEC) conference, which is split into two divisions, East and West. 

Don't ask me why Missouri play in the East, when only two teams in the SEC are located further West. (Actually you can ask me why, as I do know the answer, but it's an odd alignment on the face of it). 

The West is the stronger of the two, with Alabama, Auburn and LSU winning 8 championships since the Bowl Championship Series was established in 1998, 
with the East winning 3 - Florida (2) and Tennessee. Alabama has also won 5 of the last 10 championships, and lost in another two finals.

As there is so much interest in Alabama, the markets do offer some opportunities. Back Overs when they play on a neutral ground (last ten seasons 76.9%), fade them at home and back them away (the public again overrates the home team).  

In conference games between East Division teams, there is no edge on small road 'dogs, either recently (10-10-0) or historically (33-34-1) while in the West all-time, they cover 54.2% of the time, and 66.7% in recent seasons. Look for Auburn to cover the 3.5 points on 21st September when they play Texas A&M.

When West teams play in the East, small 'dogs cover 55.6% of the time, (71.4% when the host is coming off a loss), while when East goes West, only back the small 'dog if they won their previous game as these teams cover 60.5% of the time. 

Monday, 8 July 2019

Centering The Pill

As I have mentioned before, sports change - rules change, tactics change, and sometimes equipment changes.

I have written many times about the NBA and the increase in three-point shooting, and the consequent increase in scoring, and a similar change, though perhaps less noticeable, is being seen in the MLB.

The MLB, or at least professional baseball in the United States, has been around for a while, starting in 1871. 

In that season, the average number of Home Runs per game was 0.19, and between that inaugural season and 1920, the average ranged between 0.06 and 0.39.

In 1929, there was half a Home Run per game, and in 1950 the 0.75 per game mark was reached. 

And then came the 'steroid era' which resulted in an average of one Home Run a game being reached in 1987, and again in 1994 where it stayed until 2010. 

Since then it has averaged 1.06, but very recent seasons have seen the average at 1.26 (2017) and 1.15 (2018).

All this is a long-winded build up to pointing out that this season, the average is at all-time high of 1.37, helping push the Runs per game average to 4.81, a number not seen since 1950 other than in the 'steroid era'.  

The reason given by MLB for this huge increase in Home Runs is that the balls have less drag, i.e they fly further, due to an improvement in the manufacturing process.

More drag means that the ball doesn’t travel as far. Now, according to Manfred, one of the things that may be happening is “they’re getting better at centering the pill. It creates less drag.” That helps the ball travel farther to create more opportunities for home runs.
It's also been observed that players have changed their swings in recent years to take advantage of this, and that pitchers are throwing more off-speed pitches, which is not as easy as throwing a fastball. This leads to more mistakes, and while a pitcher can often get away with a mistake heading to the batter at 95+ mph, a slower pitch can end up over the fence.
Indeed, Statcast reveals that only four of the 450-plus homers this season were on fastballs 95-mph or harder
More Home Runs means other changes to the way the game is played, for example the numbers for Stolen Bases and Caught Stealing are at lows not seen since 1971 and 1940 respectively. Clearly there's less incentive to make the effort to steal a base, if the hitter is going to belt a Home Run.

All good for anyone playing an Overs System, and the one I've mentioned is making record profits this season as we go to the All-Star Break.  

Sunday, 7 July 2019

Going To The Dogs After The Break

In the article I referenced yesterday by Brad Allen, he noted that:

Sports Insights has found that blanket-backing favourites had been profitable before and after the break in recent years, although those studies only look at the one game either side.
While 'recent years' isn't specific, looking at the last five seasons, I didn't find this observation to be true.

Certainly blindly backing all favourites in the final game before the All-Star Break is a profitable strategy, +6.7% Money Line, +12.1% Run Line, while the first game after the break sees a profit on the Money Line of 6.7% and on the Run Line of 12.1%*.

* Corrected on 13.Jul.19 

Brad then wrote that:  
It would be interesting to see the results from a whole week before and after if anyone fancies running that…
Well who doesn't like to look at these things? 

Again looking at the last five seasons, the answer is that in the week before the All-Star Break, backing all favourites is profitable, but only slightly with the Money Line +3.48 points from 576 games, while the Run Line is +15.63 points from 575 games.  

Not an ROI to set the pulse racing, but a lot better than the week after the break, when the Money Line is down 22.45 points and the Run Line is down 33.61 points.

So is there a system here backing 'dogs in the week after the All-Star game? 

Blindly backing 'dogs will lose you a little, but there does seem to be an edge in Divisional games, where the Money Line has an ROI from 191 matches of 10.7% and the Run Line is 6.0%.

A final teaser: With a little further refinement, the ROIs from 48 games are 38.2% and 32.6% respectively, and profitable on both the ML and RL every season since 2011.

Saturday, 6 July 2019

Breaking Down The Break

I mentioned Major League Baseball's All-Star Break this week, and an interesting article on the topic was published by @BradAllenNFL posting the question:

Is there any truth to the idea that betting baseball in the second half of the season is a losing proposition?
The All-Star Break (ASB) in baseball does seem to be more impactful than those in the NHL or NBA. I actually have one system that is active for the first games back after the break, but this, like my early season systems, is more for fun than any serious money given the small sample size. 

Brad's article was triggered by his observation that:
The last few years my baseball betting seasons have followed a familiar pattern; get nicely ahead by the all-star break then slowly drip profits back in the second half.
Conventional wisdom would have it that early in the season, off-season changes mean that the strengths and weaknesses are not yet fully known, and that as the season progresses, outcomes should be more predictable.

As readers of this blog will know, the evidence suggests that this simplistic view is incorrect. I talk about favourites a lot, mostly because in recent years, its been hard to lose money backing hot favourites. 

While the term 'hot favourite' is not a technical term, for baseball I use the -200 line (1.50) or shorter to define a hot favourite, and the before and after All-Star results from backing all these favourites blindly is:
These results combine the Money Line (ML) with the Run Line (RL), and so 52% of the profits (excluding 2019) come before the ASB, which rather like the Leave margin for Brexit, is insignificant. 

For 'white hot favourites', i.e. those at -300 (1.33) or shorter, the numbers since 2014 show a more significant split.

Excluding the current season, before the ASB, Money Line Favourites are down 1.95 points while after the break they are +19.60 points

For the Run Line, before the break has an ROI of 5.7%, while after the break it's a loss of 4.3%. The sample size is just 79, so don't read too much into these numbers. 

When it comes to my basic T-Bone System, the big difference is in games played in National League ballparks, where prior to the break, the ML and RL ROIs are 15.3% and 18.6% respectively, while after the break they lose 4.6% and 10.4%. The American League hosted matches are profitable throughout the season.

Regarding the Totals markets, my Overs System looking to benefit from 18 innings rather than 17 is profitable before the break +9%, versus 6.2% after, while the less lucrative Unders System, which looks for 17 inning games, is up 2.3% before, and 2.1% after, the break, although again, here games in National League parks are loss makers after the break. 


Basically, I agree with Brad's conclusion that edges do evolve as the season progresses, but that...
if you’re sharp enough to turn a profit on baseball in the first half the year there’s no reason you can’t back it up in the second half. Just watch out for some funkiness around the all-star break.
Month by month, backing hot favourites does show July to be a problematic month, something I've mentioned before:
Contrast the profits in July before the break of 16.80 points with the losses of 11.80 after the break, although all the losses (and more) come in the first game of a series. Avoid these, and July is profitable throughout.

One final thought from Brad was this:
Finally there’s the human element of the bettor involved. “MLB is a grind,” Andrews says. “Fifteen games a day nearly every day takes a toll on a bettor. This is the point where you just get so tired of the grind that you need to take some time off. By mid-July the focus of the nation turns to the NFL training camps beginning to spin up. Getting ready for NFL season is tempting.”
The beauty of my strategies is that they take about five minutes a day to process so burnout isn't an issue, and while I have started to prepare for the upcoming (American) Football season, it doesn't impact my MLB at all.

Thursday, 4 July 2019

From Russia, With Overs

Fascinating, or at least mildly interesting, that this blog has become so popular in Russia of late. All-time, as might be expected, the UK and USA are miles ahead:

...but in the last week, hits from Russia are almost three times those from the UK.
The most popular post of the last month is A Tale of Two Podcasts in which the wisdom of Rufus Peabody is contrasted with the naiveté of Mel.

Rufus' comments led me to the book The Logic of Sports Betting by Ed Miller and Matthew Davidow, and an excellent read it is too, especially if you are interested in US Sports and markets. Plenty of thoughts on American Football, basketball and baseball, the latter of particular interest at the moment.

My post from May titled MLB Totals Simplified explained my thought process regarding the correlation between home teams winning and the effect on totals, and this idea is mentioned on page 164, although the impact of the different leagues, and thus rules, isn't discussed. 

For the record, the strategy for Overs has a 72.6% strike rate this season, and 68.4% since the post, so those of you who followed along are now at least a little closer to retirement.  

Mel, on the other hand, continues to have memory problems: 
Yes, "a full time trader" forgets he has another account he can use. That certainly seems credible and professional. 

Monday, 1 July 2019

MLB - June Summery

Hopefully some of you cashed in on the advice in yesterday's "London High" post, with the Colorado Rockies v Los Angeles Dodgers game completing the double for Overs after the Boston Red Sox v New York Yankees game. An added bonus was that the Dodgers won the game, as they were a T-Bone selection.  

June is now in the books, as they say, and as usual it was a good month for the 'hotties'.

For the super hot teams, those at -300 or shorter, the season finally moved into the green after a shaky start:
The Basic T-Bone System took a small loss of 0.4 points on the Money Line, 3.4 points on the Run Line, but still up for the season by 8.3 and 2.95 points respectively. 
If you followed my May advice regarding Totals betting, you are probably patting yourself on the back for a wise decision, with Overs winning 62.3% of the bets in June for an ROI of 20.3% from 65 matches. 

A slightly higher ROI for the Unders, with 13 winners from the 21 selections. Numbers for the season so far, with a little historical perspective:
We're actually a little more than halfway through the regular season already , and it's only a week or so before July sees the MLB season interrupted by the All-Star break, which is being played in Cleveland this year. 

Sunday, 30 June 2019

London High

The MLB data covers 16 seasons (including the current one), which is 38,171 games, and never has there been a Total higher than 14.5 runs.

Until today. 

Unless there's an adjustment, the London Series game between the 'home' Boston Red Sox and visiting New York Yankees had the Total set at 15, not high enough to prevent Overs from another comfortable win.

The previous high of 14.5 runs was set in April 2010, for a day game at Wrigley Field, and again the outcome was Overs as the Chicago Cubs beat the Arizona Diamondbacks 13-5.   

Since 2007, there have only been 12 matches with a Total higher than 12.5, and 9 have gone Over. Those public biases helping the informed bettor perhaps, although it's a small sample.

Incidentally, today is the first time ever that two such games have been played on the same day.

The higher totals tend to be set in National League ballparks of course, and all-time the record in these games (Total 13+) is 36-27-7, which at Pinnacle's -105 works out to an ROI of 10.4%.  

In matches where the two teams scored 13 or more between them in the previous game, the ROI is 26.2% suggesting that when the batters are hot, they stay hot.

Saturday, 29 June 2019

Football Unders and London Overs


Although the correlation between the Draw and the Under is well known, Joe posted the above question a week ago. 

The prices on the Under / Over markets are not sharp, with the overround on the average prices significantly higher than those for the closing match odds as provided by Pinnacle, courtesy of Joseph Buchdahl's Football Data site.

Nevertheless, I did take a look at the past four EPL seasons, which is 1520 matches and a decent sample size. 

The average overround was 105.4%, and Over 2.5 was the result in 805 matches (53%). 

In matches where the teams 'true' win probabilities were within 25% of each other, the average goals per game was 2.56, while in more one-sided games, the average was 2.79.

Backing the Under 2.5 goals in these 279 matches would have resulted in a loss of 16.01 points (ROI -5.74%), which is pretty much the overround, and where the threshold is 10%, the 109 matches resulted in a small profit of 2.95 points. These games averaged just 2.51 goals per game.

Switching sports for a moment, and I'll update June's numbers after the weekend, but the visit to London for MLB offered a great value bet on Overs.
London Stadium is smaller than your typical MLB ballpark. It’s only 385 to dead center field and 330 down the line to either foul pole. The distance to “deep” center is shorter than any current MLB ballpark and well below the league average of 402.6 feet.
The Overs showed up as a selection anyway, with a total of 11.5, but a little homework suggested 2.08 was huge value. I'm pretty sure I've never won a Totals bet after one inning before!

Monday, 24 June 2019

17 Leagues, One Season, One Result

As promised, here are the results of backing the Draw in the 2018-19 season. 

As was the case from 2012-2018, overall we again see the expected improvement as the matches become more competitive. 

Coming in to this season, the evidence indicated staying away from the smaller leagues of Belgium, Netherlands, Portugal and Scotland, as well as England's League One.

How did these leagues fare last season? More of the same essentially. 

Prior to this season, the profitable leagues were the top two tiers of the Big Five countries, and again backing the Draw in all competitive matches in these 10 leagues was profitable:

As most readers will know, this season saw a record low strike rate for Draws in the English Premier League, the fewest since 1931-32 which most readers won't remember.

Even with the EPL's 14.43 point loss at the 'Difference less than .25' level, the overall return of 3.2% for the 'Top 10' was the same as for 2012-18.

For serious bettors, the biggest concern should be the increase in over-round last season. It's a topic covered previously in the blog, but every one of the 17 leagues saw an increase this season, with the EPL at an average of 103.5%. That may not seem like a lot, but it makes a huge difference over time and it's a worrying trend.

The full seven season summary looks like this:
In 2017, I referenced David Sumpter's conclusion, based on five EPL seasons 2011-2016, that:
It turns out that when two well-matched teams meet (i.e. the probability of a home win is only slightly bigger than the probability of away win) then draws are under-priced.
The seven seasons for which we have reliable data show that this continues to be true in the big leagues, but not in the fringe leagues.  

Just as I thought I was done, and could enjoy the summer, Joe threw another idea out there:
Looking at the average prices for the Over / Under for the EPL from last season, we're looking at an over-round of 105.2% which might be a problem. That there is a correlation between Unders and Draws will not be a surprise to readers of this blog but it might be an interesting exercise to look at some more data.   

Saturday, 22 June 2019

17 Leagues, 6 Seasons, One Result

A few days ago, Joe tweeted the above message. 

The 17 leagues are comprised of the top two tiers of the Big Five leagues (England, France, Germany, Italy and Spain), plus Leagues One, Two and National in England, and the top leagues in Belgium, Netherlands, Portugal and Scotland. 

Not surprisingly, each league has its idiosyncrasies, not to mention varying over-rounds, and Joe followed up with: 

Before I get to last season, here are the numbers for the seasons for which we have Pinnacle's Closing Prices which are available at Joseph Buchdahl's Football-Data.co.uk site.

Some additional clarifications are that:

  • The seasons covered are the six seasons from 2012-13 to 2017-18
  • All Profit and Loss calculations use Pinnacle's Closing Prices
  • Any matches where Closing prices are missing have been excluded
  • 'Competitive' is defined as matches where no team has a 'true' win probability greater than 0.5
  • 'True' means after the over-round has been removed, i.e. the sum of the probabilities equals one. They are not truly true, but they are close.
First, the overall story, and I hope the picture is self-explanatory.

From the top, we have 40,524 matches and backing the Draw in each would have lost you 2.3%, which is basically the bookmaker's vig. 

There are faster ways of losing money than backing the Draw, but we can do better.

The next block shows the outcome of backing the Draw only in competitive matches, and the loss in this category is 1.4%.

Then I looked at matches where the two teams are within 25% of each other in terms of win probability, and finally at the relatively exclusive category where teams are within 10% of each other.

To keep Joe somewhat happy, 2018-19 is still to come, here are the numbers for Belgium and Netherlands:
As Joe suspected, these leagues do not follow the overall pattern, in fact the Netherlands results are the exact opposite of the overall results. Just back the Draw when a team is odds-on, and count the money, except that the market seems to have corrected since 2015, so this strategy is not recommended.

The only other league which is upside down like this is League One, although here we go from bad to worst with poor results across the board.
The other English leagues all follow the expected pattern, with the Premier League noticeably strong, something that will come as no surprise to readers of this blog.

For the Big Five leagues, the results for the top two levels are:

I'll update these results with 2018-19 in the next few days. 

Meanwhile, if anyone is interested in detailed results from any of the other leagues, let me know.

Tuesday, 18 June 2019

A Tale of Two Podcasts

It was the best of podcasts, it was the worst of podcasts.

Well worth 50 minutes of your time is the Pinnacle Podcast with Rufus Peabody. Rufus focuses on three sports, two of which align with my sports markets of interest, and his comments confirm my belief that profitability is all about exploiting inefficiencies in the betting markets, that the reason these inefficiencies exist is because of biases, and that you need to think creatively and logically as markets sharpen and the number of people who are "good with numbers" continues to climb.

Also interesting to me was that Rufus, like myself, finds the challenge of beating the markets, i.e. the competition, more of an incentive than the money itself. As I wrote more than ten years ago:
I was thinking in the shower this morning, (yes, once a week, whether I need one or not), Betfair really is the ultimate video game. I've never been one for games, (friends at work spend hours playing Call of Duty - why? What's the point?), but in many ways the exchanges are one big on-line game. It's me versus an unknown opponent. My opinion versus yours, except in this game the points are real money.
Or as Warren Buffett describes beating the markets:
"In a sense, the game that I'm in gets more interesting all the time. It's a competitive game, it's a big game, and I enjoy the game a lot"
It's not often that I can compare myself with Warren Buffett.

There's plenty of other content that is worth listening to, if only to understand the level of competition you are up against if you are trying to create a model, and how the markets have changed over the years, although Rufus is a relative youngster!  

He also claims at the end to not be a speaker, and hopes people will make it through the podcast, but for me it was easy. 

This was a very good listen.

On the other hand, not worth 50 seconds of your time, are the same old nonsensical ramblings from Mel "Scientia Trader" who is back spouting his rubbish after the "huge" set back of losing £472.35 in March.

Mel is most definitely not a speaker, but appears to be blissfully unaware of this, offering viewers / listeners a 100 minute mono-tonal lecture, littered throughout with "you knows", or at least the parts I selected at random were, and remarkably absolutely no content of interest. I take my hat off to anyone who can sit through this, and if there is anything of interest, please let me know along with the time. 

One unintentionally amusing moment I did stumble upon is when Mel explains how he had a problem registering with bet brokers who don't accept UK customers, but then suddenly realised he had dual citizenship. Problem solved! 

Amazing stuff. I mean, who among us hasn't suddenly remembered that we have dual citizenship when it comes to betting? While it is true that this is a useful advantage, it's not likely true that its value only dawns on a bettor several years into his betting adventure. As a plot twist, this one is a little bit of a stretch.

I shall leave to the reader to draw their own conclusions as to why the loss of such a small amount, a miniscule draw-down from the profits already (supposedly) reaped, would trigger a ten week disappearance, and the need to "Reflect, Recover and Restart", but to be clear, the idea that anyone can consistently have a win rate of 55% to 60% for an outcome with a probability of .357 (i.e. 2.8 in decimal odds) is complete and utter nonsense.
Some markets might be inefficient, but none are quite that inefficient, or would remain so for long, and as I have written before:
This risk to reward ratio implies an average price of 2.8, i.e. a win probability of 35.7% for his bets, so anyone achieving a 50-60% win rate at that price would clearly, and rapidly, be on the way to a fortune and keeping very quiet about it.
If, by some miracle, Mel really does reside in a world where the laws of probability actually suspend themselves [say Hi to Tony while you're there], the idea that the amount risked should be 1.5% is similarly, literally incredible. 

Using Kelly, the suggested stake with this kind of an edge is over 28%, which should be a clue that something doesn't quite add up! 

Even a more modest quarter-Kelly suggests a stake in excess of 7%, so either Mel is terribly confused about the edge he has or he's clueless about how to make the most from it. 

The evidence suggests that both are true.

I did attempt to listen to a few minutes to see if Mel had learned anything, but once the name of fantasist Adam Heathcote was mentioned as a source of inspiration, it confirmed that thinking logically is still not Mel's strong suit. 

Not a good listen - unless you have trouble sleeping.

"You can't speak butterfly language to a caterpillar." - Unknown   

Sunday, 16 June 2019

Flipping Variables

I mentioned recently the importance of having a logical reason why a system should work, rather than using 'data mining' to find a back-fitted system which has no predictive value at all.

For example, yesterday's post shared an idea for a system which takes advantage of the idea that the issue of time zones in the NBA may not be fully understood by the public, and as hopefully everyone understands by now, these situations offer an opportunity, at least in the short-term.

Contrast the solid logic and rationale of this idea with my tongue in cheek example of:

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.
One of the forums I frequent regularly has such 'systems' recommended by contributors, but sadly most have them have multiple conditions applied. For most of these, there is simply no logic or rationale as to why this condition might lead to a market inefficiency. Fortunately, there is the occasional nugget that does makes sense, but these tend to be rare.

There was a voice of reason who jumped in on one 'idea' as it spiraled out of control, and explained the problem quite neatly, and while I have taken the liberty of correcting errors of spelling, punctuation and grammar to improve readability, the gist of the comment remains valid.

I need to speak out in terms of experience here before someone gets hurt.
The system included here has little likelihood of future success, and here is why. My experience, which is plenty, has shown me that when you back-fit a situation with so many conditions (at least 12 here), you're no longer predictive in terms of the future, but creating the most perfect flow chart of the past.
Most of these systems are built on finding something with a modest ROI, and then experimenting with variables, meaningful, or meaningless, adding only those that make the situation appear better.
If you experiment with a lot of variables, then you will "stumble" into those that make a certain situation look better, but in the process, you "contaminate" future predictability.
What you are left with is thinking and believing you have found the holy grail of sports betting, only to be fooled by a false premise of profitability. I guarantee you over the next 100 games, the win rate will be closer to a 0% ROI than to the back-fitted ROI.
We are putting faith in a system that is built upon 12-15 conditions, many of which are used, not out of known predictive advantages, but out of anything that improves the win rate. By definition you will find something that is better than the one condition variable.
Think of this. You flipped a coin 200 times, and without adding any filters your results were 100 heads and 100 tails, no predictability of future flips.
You video taped every flip. Now you go back and analyse what happened. You notice that when you flipped with your left hand, heads came up 55 times out of 100, while when you flipped with your right hand, heads came up just 45.
So you add the variable, if flipped with left hand, 55% of the time you get heads!
Further looks see that if you placed the coin in your left hand from your right hand you got 30 heads and 20 tails, but when you picked the coin up instead without any use of your right hand, you got 25 heads and 25 tails. So you're now up to a way to get 60% heads!
Next, you noticed if you paused for more than 10 seconds after placing the coin in your left hand from your right, you got 17 heads and just 8 tails. You now have a situation that generates 68% heads, just by flipping the coin from your left hand after placing it there from your right hand, and waiting at least 10 seconds before you flipped it.
So my question is:
If you then flipped the coin 300 more times, doing everything the way you did to get 68%, what percentage of heads is expected from the 300 flips?
The answer is 50%!
The variables used above were selected not based on anything that is predictive, but based on anything that made the system look better!
THAT IS THE PROBLEM!
The heads scenario here was built the same way, none of the added conditions are predictive in any way!
The moral of the story is this:
Build a concept on known +EV variables, not a system built to make +EV concepts.
What do I mean by meaningful variables? When you do a search in a given sport with one variable, and it shows an advantage, then you have found a meaningful variable.
Build a pile of meaningful variables for a given sport, then stack the meaningful variables, to get meaningful situations.
Hope that helps everyone here.
I should add that it is quite possible to identify a variable that appears meaningless but for which it later turns out there was a valid reason all along. A Tweet from A Lucky A Day referenced the 15th a few weeks ago.
Some readers may be familiar with the apparent illogical pattern in the 15th round of Sumo Wrestling competitions mentioned in Freakonomics where, in certain contests, the supposedly 'weaker' wrestler was winning 80% of the contents.
In a sumo tournament, all wrestlers in the top division compete in 15 matches and face demotion if they do not win at least eight of them.
The sumo community is very close-knit, and the wrestlers at the top levels tend to know each other well. The authors looked at the final match, and considered the case of a wrestler with seven wins, seven losses, and one fight to go, fighting against an 8–6 wrestler.
Statistically, the 7–7 wrestler should have a slightly below even chance, since the 8–6 wrestler is slightly better. However, the 7–7 wrestler actually wins around 80% of the time. Levitt uses this statistic and other data gleaned from sumo wrestling matches, along with the effect that allegations of corruption have on match results, to conclude that those who already have 8 wins collude with those who are 7–7 and let them win, since they have already secured their position for the following tournament.
There was an underlying reason for this apparently illogical finding, even if the reason wasn't understood until later, so it's important not to dismiss every impacting variable as meaningless. 

Saturday, 15 June 2019

Fast and Three

So in the space of two days, both the NHL and NBA seasons are over, and it's all about baseball for the summer. 

At the start of the 2017-18 NBA season, I shared some thoughts regarding the totals markets that even the most critical of readers would struggle to find fault with.  

Actually, the only fault was that I underestimated how rapidly the totals would increase, and the suggested entry point of 215.5 soon became unmanageable, although had you been on all of them, you'd have been rewarded, but 431 bets over six months is a lot of work.

For the 2018-19 season, the initial band I targeted turned out to be even more ridiculously low, resulting in a whopping 701 bets! 
Whatever your entry point, the key to this idea is that the public have been slow to adjust to changes in the game, and the resulting higher totals.

In the five seasons between 2010-11 and 2014-15, the total was never set at 230 or higher.

In 2015-16, there were five such occurrences, in 2016-17 there were 30, and last season there were 233. That's fairly conclusive evidence, although sceptics might want to wait a few more years to confirm. 

Even the traditionally lower scoring Eastern Conference has joined the party. Having never previously had a total set higher than 230 points, 2018-19 saw no fewer than 42 such totals set, and as you might have guessed, the public balked at such a novelty, biases kicked in, and consequently Overs went 26-16.

The number of three-point baskets is the biggest factor in this increase in points. Looking at the regular season, in 2005-06, the average number of 3-pointers scored was 5.7 from 16 attempts; numbers which exactly doubled in 2018-19 to 11.4 and 32. These numbers have climbed steadily over the past seven seasons, and show no sign of slowing down.

One thing that is perhaps a little surprising is that the strike rate from 3-point attempts is pretty much the same as it was 25 years ago. Last season's 35.5% is actually slightly lower than the 35.9% of 1994-95.

The 3-point line was introduced in 1979-90, and it took teams a while to get up to that percentage, but since then it has remained in the 33.9% to 36.7% range.

The reason there are more attempts these past few seasons, is that the pace of the game (as measured by possessions per 48 minutes) is increasing. Last season, for the first time in 30 years, the number of possessions hit 100. This was to be expected, as the pre-season rule change reducing the shot clock from 24 seconds to 14 seconds after an offensive rebound was implemented for this purpose. 

While offensive rebounds stayed about the same, defensive rebounds hit an all-time high, the result of more missed shots.   

While there is a limit as to how high the pace can go, we do know that it has been as high as 107.8 and as that was in 1973-74, the first season the statistic was recorded, it's quite likely to have been higher at some point prior to that. 

For the second consecutive season, and the only two seasons since the introduction of the 3 point line, the ratio of free throws to field goals was less than 1:5. 

Bottom line is that I don't see the totals as peaking just yet. There's room for the pace to increase, and for improvements in shooting accuracy. Kyle Korver's 53.6% in 2009-10 is the benchmark.

The only question is when will the markets catch up.  

Following Eastern teams when playing in the West as underdogs had a winning record, but a small loss, while the 'tired' dogs eked out a small profit. 
The idea of 'Eastern' teams being value when playing 'Western' teams received a somewhat wider audience than this blog when the Guardian published an article on the topic at the start of last season (no wonder edges disappear!) including this comment that readers of this blog were already aware of:
It’s just one more home court advantage for West Coast teams, hosting sleepy teams from the East.
One problem with defining teams as Eastern or Western is that a match between teams from the two conferences isn't necessarily a match between teams from different time zones, with the Central Time Zone having both 'Eastern' and 'Western' teams. The Eastern Conference spans two time zones, while the Western Conference spans three.

So an Eastern (Central Time Zone) team playing at a Western (Central Time Zone) is at far less a disadvantage than an Eastern (Eastern Time Zone) team playing at a Western (Pacific Time Zone) team.

After reading that article, I made some adjustments to the "Tired NBA Eastern Road 'Dogs in the West" system, excluding matches between teams in the same time zones for example, but the research uncovered a fact that was even more interesting. Games involving teams traveling west don't have as many points scored as the market expects. In a league where the talk is all about the increase in scoring, finding an edge on Unders was promising to say the least. 

Here are the results from backing the Under when Eastern Time Zone teams headed west for a game in a different time zone for the past ten seasons:

As the prior would suggest, the edge is stronger in the Pacific time zone than in the Central, with the Mountain zone hosting relatively few games. 

Using Joseph Buchdahl's spreadsheet, the chance that the results in the Pacific time zone are luck is 1 in 106. I share this idea because I'm a generous chap, and by the time the 2019-20 season rolls around, you'll all have forgotten about it! 

And for what it's worth, I use the previous season's average points total (adjusted) as my entry point because not all Unders are value.