Sunday, 5 January 2020

Change Is Constant, Mean Not So Much

"The only constant in life is change" - Heraclitus
By some distance, although admittedly with no competition, yesterday's post is already the most popular of the year, prompting comments in abundance on Twitter.

Mel (Scientia Trader) does indeed appear to have vanished from the face of the earth, with Dionysios confirming that:
Indeed, hopefully the latter. Unfortunately for Mel, the laws of probability do not suspend themselves just because you want them to, and the claim of a 50% to 60% strike rate being sustainable at a price of 3.8 (implied probability 0.26) was always quite ludicrous. Hopefully he is alive and well, and gainfully employed and didn't lose anything more than his pride. Of course, it is always possible that his huge edge was genuine, and that he is now retired and living on a private island somewhere. Nah. 

A Lucky A Day sent me some links to data for the Draw and Under prices, with the free version being the less useful (but nevertheless interesting) opening prices from both bet365 and Pinnacle. There's about 210,000 rows of data so I may be gone a while!

The closing prices come at rather a high cost, and I'm not sure senior management here, still seething after seeing my 2019 alcohol expenses, will be approving it.
In addition to being very helpful, Mr Lucky did comment that:
Personally I do think draws get overlooked and inefficiencies could last a long time!
Agreed, and the Draw has served me well over the years with thanks due to Derek McGovern for planting the seed in my brain back in 1999.

They also served many readers well when I shared the XX Draw selections for three seasons in the last decade:

Not a huge return for the effort, but For comparison, backing the Draw blindly in every game in those leagues for the three seasons would have lost you 5.8%

In some ways it's a shame I had to discontinue the spreadsheet, as it was based on the idea of expected goals before xG was even a term, but as I've mentioned before, it took several hours a week to make all the updates and a promotion at work meant that was no longer a viable proposition. For most of us, time is a precious commodity and as the cost of spending hours watching a sport in-play waiting for an entry point, so there is a cost to spending several hours a week updating the numbers for ~100 clubs.

Graeme Dand back in 2014 had this to say on the Draw:
I really buy into the idea that few punters back draws and they are more interested in backing a team to win (same idea as why no one backs 0-0 when watching a game as they want to see goals!)
Another expert who saw the potential for value in the Draw (and Under for that matter) was Daily25, who had this to say back in 2012:
I love the "in the right circumstances" qualifier, (aren't all bets profitable "in the right circumstances?") but a quick look at the free data would suggest the Under / Draw relationship I mentioned yesterday might have some merit. 

The Premier League data isn't quite current, with the latest matches from early December, but they go back to the 2010-11 season which is a fair way.

Stripping out the matches where one club is odds-on or the Draw is priced at 4.0 or greater, backing the Draw in matches where the Under is odds-on produce the following results:
I may need to change the wording from "odds-on" to "favourite" but more to come on this, and thank you to A Lucky A Day for the pointer.

Then there was this comment from Joseph, a little passive-aggressive in tone perhaps, but Joseph is more into the theory side of betting than the practical side, so I'll assume positive intent:
The problem with talking about a regression to the mean when it comes to sports and betting, or at least my technical analysis based form of betting, is that there is no specific and invariable mean, because both the sports themselves, and their related betting markets are constantly changing. 

In sports for example, from 1888-1892, the average number of goals per game in the top flight of English football was 4,37, and the percentage of matches ending in a Draw was 14.5%. Is this the mean to which the number of goals is expected to return to?

As mentioned many times previously, sports change over time. Rules are updated and scoring becomes easier or harder. Look at the impact of the 1925 changes to the offside rule in football. Points for events are changed (e.g. Rugby and NFL). Strategies change, e.g, the higher scoring seen in the NBA since 2012. Divisions, conferences and schedules are rearranged. The list is long.

The net effect of this is that data from one season often cannot be usefully compared with another season and the idea of numbers regressing to a mean is nonsense. Values are not constants.

Similarly with the betting markets that accompany these sports. For one example, one need only look at the football markets from Joseph's own website from 2000-01 where the over-round (William Hill) averaged 112.54%. This season the average (Pinnacle) is 102.77%, which marks a huge difference and you can't simply compare the results of a strategy without recognising this difference.

Markets also have to change because the underlying events change. I've written before that if you are able to foresee how a change in the sport might impact results, you should make hay while the sun shines because at some point the market will catch up to you. It has to, and only a fool would think that it won't. 

Of course winning strategies should eventually go "tits up" as Joseph puts it, and there should be no need to "save face" when this happens, unless you are claiming your golden goose will live for ever, and yes, there are probably gullible fools out there who would believe that line. 

As Alfred Lord Tennyson put it so eloquently:
'Tis better to have loved and lost than never to have loved at all.
Assuming you didn't lose all your profits of course.

That some biases appear to be so strong that systems like the Small Road 'Dogs have had more winners than losers in effectively a coin-toss for 19 seasons and counting is the real puzzle:

All my strategies, and those of anyone, should have a limited shelf-life. The key is to identify current market inefficiencies, use them to your advantage until the market catches up and it is no longer profitable, and move on to the next idea.

Remember that the ideas offered on this blog are all based on technical analysis and are all offered for free, although donation to the retirement fund are always gratefully received. 

The main idea for this blog is to show the process required to identify market inefficiencies rather than offer selections themselves. I hope it also shows that making a profit from betting is also possible with the right approach.

Mean regression might be meaningful (pun intended) if you are measuring the performance of a tipster who is using fundamental analysis to identify selections, but for a technical system where the market is always changing, it is not. 

The idea that markets evolve shouldn't be news to anyone who is actively betting. I don't believe Joseph actually bets on a regular basis, and thus may not be aware of these changes, but as others on various forums have pointed out over the years, successful strategies can stop working literally overnight. 

You're basically buying a property on top of an eroding cliff. At some point, your house is going to fall into the sea, but you should be able to enjoy the view for a while before that happens.

Build up your bank while you have the edge, move a stop-loss up as your profits increase, and stop betting when (not if) your stop-loss gets hit. The idea that one should wait until the system has been 'proven' over 10,000 events, possibly going back years when the markets were quite different, is frankly nonsense. We're not testing a coin-toss or a roulette system - we're playing in constantly changing markets. 

The idea that betting markets evolve shouldn't be a surprise - the financial markets do too, and a punter, like an investor, needs to stay flexible:
The important thing to remember is that there is not a set rule you can apply. You must factor in what is going on in the world. For example, if the economy is in trouble, corporate earnings can be worse than expected. This lowers investor expectations, and stock prices will go down. Even if the market seems fairly valued at a P/E ratio of 14, bad times could cause the market returns to continue on a downward spiral with the P/E ratio going much lower.
On the other hand, during booming economies, corporate earnings can continue to rise, and stock prices can continue to rise for many years in a row. A P/E ratio of 16, or even 20, does not automatically mean the market is overpriced. In the early ’90s, many thought the market was overvalued based on P/E ratios, and thus they missed years of great returns from 1994 to 1999.
The point is that the betting landscape is constantly in a state of flux. New players with new ideas and sometimes sophisticated analytical tools, come in to the markets, sportsbooks with revolutionary business models appear or disappear: 
When Pinnacle began trading in 1998, most gaming companies used bonuses to lure potential customers. Rather than offer sign up or reload bonuses, Pinnacle became the first sportsbook to introduce a reduced margin pricing model or reduced juice, thus deriving profit from lower margins but a far higher turnover. Pinnacle began offering -105/-104 ($1.95-$1.96) prices on head-to-head match odds and spreads, significantly undercutting the standard -110 ($1.91) pricing model that was used by its competitors.
These changes are why the challenge of betting profitably is so much fun. 

What happens when NBA teams play faster and increase the number of possessions per game? There are more points scored. Do the 'traditional' sportsbooks sit and wait for mean regression? No, of course not. There is no 'mean'. There are 'means' over certain periods of time, but they are generally pretty meaningless (sorry, did it again). 

Finally, Mark had this to say on the topic of bookie restrictions:
Hopefully by mentioning him, his interest will continue for at least another day or two, but my point on this topic was really that anyone with the ability to become a long-term winner at betting is unlikely to be someone who would be surprised at having their business declined. Or as I put it last year:
Are you suggesting that someone with the skill to win at betting long-term is someone who would be unaware that they might not be able to extract money from a business indefinitely? This seems highly unlikely to me.
Anyone starting out today on Betfair is certainly well aware that in the unlikely event they are successful to the extent of the lifetime limit, there will be a Premium Charge applied on future profits. The complaint from the relatively few customers impacted at the time, (in reality, far more people were complaining than were, or would ever be, affected by the charges) was that the lifetime limit was being retroactively applied, and if you were already there or almost there, it was unavoidable, whereas if you were just starting out, it was fairly easily avoidable.

But life goes on, for me it's only a hobby, and we adapt.

Hopefully a few of you were with me on the Unders last night in the NFL Wild Card games. With two winners already, the worst case scenario for the weekend is a small loss after commission, but while my 'official' results will record the wins at 1.952, the actual bets were made at 2.08 (Tennessee Titans @ New England Patriots) and 2.02 (Buffalo Bills @ Houston Texans) so it will be a profitable weekend no matter what. Fly Eagles Fly. 

1 comment:

Mark said...

I hope your readers do not think that when you hit a profit threshold on Betfair you automatically hit Premium Charge. This is not the case, there are other conditions that have to be met and certainly if you are a high turnover low ROI winner then you probably have nothing to worry about