Wednesday 30 May 2012

Diversification

Thanks, and apologies to Maurice, for it was he who came up with the idea of looking at the Half-Time 0-0. He wrote:

Let's hope it will be another good year with the XX draws. The 0-0 Half-time was my idea. It's not a big leap from unders to 0-0, to 0-0 HT, so maybe others have thought about that too. I also want to test that coming season with low stakes. I will use the classic XX draws and not the extended.
Matt very kindly came back with some numbers for previous seasons, after my last post which showed that the draw had performed very strongly. As I suspected, based on the fact that the edge on the unders wasn't huge, it does appear that this season is a statistical blip for him, although when selections are subjective, it's hard to tell whether the selection criteria from season to season remains the same, or is subject to subconscious biases - "I picked Norwich four times, and they didn't win once" for example, is likely to influence your opinion of Norwich next time they show up on the radar. I'll be watching these selections next season for signs that the draw bias might be continuing, but a strike rate of 34.3% on draws is historically high.

Here's what Matt wrote:
This was something I looked into when I was doing the review, primarily because as you say I noticed there was a fair bit of crossover with the FE bets and your XX bets and they are typically the same "type" of games.
Unfortunately I think it was just a random case of the recommended bets happening to hit a lot of draws this season (34.3%). I wouldn't expect that long term.
The shortlist bets this season only had 28 draws from 117 bets (23.9%)

2010/11 season there were 22 draws from 97 bets (22.6%) for the recommended bets and 39 draws from 152 bets (25.6%) for the shortlist bets.

2009/10 season before that it was 31 draws from 121 bets (25.6%) for the recommended bets and 50 draws from 129 bets for the shortlist (38.7%).

2008/09 season the rec bets and shortlist bets were all together as one and the draw rate was around the 25% mark from memory.
So I dont think there's much mileage in backing the draw blindly although there certainly could be an improvement with manual intervention.

I think it is interesting that we both pretty much target the same kind of games but bet on different outcomes, yet both make a long term profit. Before the start of the season we will have to compare results and have a look at what happens when there is a crossover of FE bets and XX bets. Could be an interesting angle there.
Well, it's before the start of the season now, and I did just that, giving up several hours of my life that I will never get back, a sacrifice for readers and subscribers everywhere. I took a look back over the season just gone to see what those 'crossover' matches were, and how they fared, and the findings are that:

There were 13 matches, 6 were draws and 5 were home wins. We both lost on two of the matches.


Returns from the strategies of backing the draw, Under 2.5 goals, and Home wins in these 13 matches gives the following:

Draw: 8.6 pts. / Under 2.5: 2.04 pts. / Home: 3.83 pts. / Total: 14.47 pts.

Had Borussia Dortmund's two matches not seen very late 'draw-breaking' goals, the results would have been even more interesting, but even so, 6 draws from 13 selections is a very impressive strike rate. Worth monitoring next season perhaps. 

Incidentally, if you subscribe to Football Elite on the basis of what you read here, please use the exclusive Promotion Code "ILOVECASSINI". It won't benefit you in any way whatsoever, but it'll maybe make Matt feel like he needs to pay for my next holiday. Clacton, here I come.  

ROIs are meaningless from such a small sample, but available upon request. Six matches were goalless at half-time, and no less than four matches stayed that way. Backing 0-0 would thus have been profitable too, which leads me to a comment from fairfranco, who wrote:
An excellent set of results.
Just wondering, while your draws coming in may be more satisfying, and was the initial push behind your selections, the under 2.5 goals clearly made a lot more money (LSP) last season.
With the higher strike rate and much lower LLS and overall season profit surely unless the correlation between the 2 changes then the ROBG will be much higher on the Unders than on the draw.
Does this not make backing the Draw pretty redundant?
Not redundant, no. It remains a useful way of taking advantage of an edge. I reserve the right to change my mind on this, but I see spreading the bets across different markets somewhat similar to diversifying your investments. Statistical variance is going to mean that even over an entire season, one bet type will perform better than another. 

Most financial advisers recommend a blend of investments - for shares, a certain percentage large-cap, some in mid-cap, some small-cap, some International, etc. It would be rather silly to put all your money into one stock. If, in an up-year, the stock market climbs by say 10%, the FTSE100 will show a big difference between the performance of companies, even those in the same sector, but by spreading the investment (diversifying) you smooth out the returns, and limit the volatility. Draws at an average price of 3.55 may not seem hugely volatile to some, but we're not all made the same when it comes to risk tolerance.    

For example, my XX Draws were profitable over 156 matches, to the tune of 16.85 points, outdone by the Under 2.5 goals on the same selections, which returned profits of 30.28 points.

But then Football Elite's selections were also profitable over 96 matches, with the return from backing the draw 19.64 points However, these selections outperformed the Under 2.5 profits which were 5.76 points.

Both selection sets were profitable, but the most profitable way to play them varied. In fact, it's highly likely that backing Under 1.5, or Under 3.5, or the Half-Time 0-0, or Full-Time 0-0 or Time Of First Goal might well have been even more profitable for either, or both, selection sets. All the above should live or die together in the long run, but short-term, no. If there is an edge on the Under 2.5 goals market, it is because the market expects more goals than my spreadsheet does. That difference will also reflect an edge in prices across the markets above and some others too.

Rather than put 10 units (for example) on the draw for each selection, I can see some merit in spreading that money across several markets, and diversifying the risk. 

So yes, backing the draw could be considered redundant, but it's profitable, and as seen from Football Elite's results, the Unders isn't necessarily going to be a better return than the draws. Losing runs will be shorter, but the losing run on the Under 3.5 or Under 4.5 goals markets will be even shorter still. 

Last season, I backed the draw and the Under 2.5 with the same stake, and next season will probably keep the stakes the same, but add more bets to at least some of the markets mentioned. The 0-0 really will be a perfect draw next season! 

Just for fits and giggles, I summed the profits from backing the XX and FE selections as draws, and the total was 33.79 points. Backing the Under 2.5 in the same matches, and the total is 36.04 points - pretty close. 

Finally, no post is complete without a comment from BigAl, and he kindly stopped by on his way home from hospital, after having his sides stapled back together. They apparently split from laughing too hard! He had this to say:
It makes me chuckle that someone who has spent a not insignificant amount of time picking holes in the systems / tipping services of others, is marketing his own subscription service using........
ESTIMATED ROI AND ASSUMED PRICES!!
Yes, I can see the funny part there myself, but not wanting to go to Casualty any time soon, I shall try to control my merriment. 

BigAl is quite right. My choice of words is certainly misleading, and could be considered rather amusing I suppose. 

I should, of course, have made it absolutely clear that the 'estimated ROI' and prices would have been more accurately described as 'UNDER-estimated' ROI and that the 'assumed' prices will, in practice, almost certainly exceed those available for the 'Classic' selections. Apologies, and thanks to BigAl for pointing that out. This will be the last time I listen to my lawyers incidentally. 

Seriously, while the Classic selections are at the heart of the service, and are what they are judged on to this point, the inclusion the bonus 'Extended' selections next season, comes at the request of every subscriber (bar one) who responded to my request for input. No one has to use them, but they will be included so that subscribers have that choice. I don't have much supporting data other than their long-term strike rate, as I have made quite clear, but draw prices are robust little devils, and I would be very surprised if the average draw price on 'Extended' draw selections, turns out to be less than that of the 'Classic' selections. The strike rate is the key statistic here, the draw price is 'almost' secondary, simply because it is not going to be significantly lower, unless there's something fishy going on. 

Speaking of which, I may have an explanation for the under-performing results from Serie A - games are being fixed! Not unheard of in Italy of course, but the latest scandal might show that a lot more matches than most of us thought, are not being decided on the pitch.   
As disappointing as those results from Italy are for the 'Classic' selections, the strike rate in Serie A for the 'Extended' selections is the best of all leagues - 64 winners from 189 selections, an implied price of 2.95. 

There's that benefit of diversifying again, this time across leagues, and selection range. 

I've created a simple web site for the XX Draws with results from last season, and the prospectus. 

1 comment:

BigAl said...

If I can raise a point.

You often justify your ideas with "average prices". I hope you can see that this is a dangerous approach. I suppose in simple terms, its only the price of the winners which count when you're analysing the final returns of this type of stuff. The returns of the winners in the data for which you don't have prices could conceivably be low enough to produce a negative ROI even if the average price were to be what you think.

To be fair, your xx draw prices are fairly constant due to the similar type of matches you pick but you have indirectly picked up on the above issue with your comments on the Italian results.

Also, a word of caution. I suspect you've fallen into data mining traps with some of your analysis. It's a common mistake and can easily cause long-term problems