There were a couple of interesting comments posted by 'Your New Blog Reader' on a post from a few days ago. He was discussing about whether it may be better to play different strategies at different times of the year and it's actually something I mentioned in the TFA forum during February's bad spell as this was when I first realised there is potentially an issue after Christmas with my ratings. Here's both comments.
That was definitely a good read. I think it answered all of my questions about using variable stakes for your system. One important thing that I noticed (which you also mentioned in this post) is the fact that most losing months (with the exception of October 2008) came in the beginning of the year (or around second half of the season). I wonder what factors could be affecting this (probably some teams opting for a comfortable draw when they are safely in the top or may even player fatigue). But i think the important thing is that the first part of the season has historically been more profitable and this may very well be something to look out for next season. Anyway, thanks for answering my earlier queries. I look forward to reading more of your blog!
By the way, I've been looking at the performance of your system when using DNB and DC for the month of Jan - May. Please note that all results were taken from your April 11 post "Backing the Draw?".
From Jan - May 2007:
The H/A would have won around: 595.6 units
The DNB would have won around: 369 units
The DC would have gotten: 88.8 units
From Jan - May 2008:
The H/A: 208 units
The DNB: 165 units
The DC: -1.7 units
For the first 2 years of back testing your system. Backing the H/A would have produced around 50% more than simply backing the DNB (803.6 units vs. 534 units). The results for DC would not even be close. But the results for the DNB do get interesting once we look at the last 3 years.
For Jan - May 2009:
The H/A: 104.3 units
The DNB: 133.6 units
For Jan - May 2010:
The H/A: 266 units
The DNB: 229.9 units
For Jan - April 2011:
The H/A: -39.3 units
The DNB: 73.2 units
For the last 3 years, the DNB has outperformed the H/A by around 31% percent. Although I think this might be more of a reflection of what has been happening for the last 4 months (i.e. a ridiculous increase in draws), since 2009 and 2010 seem to cancel each other out. In addition, you said on "future analysis" that you may have overstated the profit of the DNB, so this might not be as promising as it looks.
Well I just wanted to point out that it might be an interesting thought to use variable stakes (i.e. using bigger values for months leading up to December) and then consider backing DNB instead (if this trend continues for the next season) to get a smoother ride since their recent ROI might not be as far apart anymore.
The picture below shows the returns for all the different types of betting (traditional H/A, DNB, DC, Draws) for each month going back to September 2006. Draws means backing the draw in each game selected by the system. I've also summarised the data by calendar month at the bottom of the monthly table. (Please be aware that for the purposes of continuity when reading the blog, my historical results for DNB/DC were wrong until recently as I was overstating DNB and understating DC. It's corrected now but it does mean that looking back when I've discussed these results before, they may not entirely tie up with the more recent results.)
If you look at the bottom of the table, you can see that there is a summary that splits the monthly data into months from Aug-Dec and Jan-May. The results clearly point to the fact that the first half of the season is more profitable than the second half of the season historically and we've certainly seen that this season also.
The same is true for DNB and DC (less so than DNB admittedly) and it is definitely the case for backing the draw. A substantial loss is shown in the first part of the season but a smaller loss in the second part of the season.
Now, looking at this data, all you can say is that you're more likely to make money in the first half of the season than the second half of the season but there is nothing too alarming that would want to make you think seriously about adopting a different approach from traditional H/A betting throughout the season.
However, and this is quite important and was overlooked by me earlier this year when I looked at this in the TFA forum, these results do include the seasons 2006/07 and 2007/08 which include some backfitted results. Although the results are not fully backfitted in these years, there is an element of backfitting in these years and therefore, they may be skewing the performance of the systems slightly as the backfitted results from 2002-2007 were much better than the backtested results in 2008 and 2009.
Stripping out the years with any backfitted results, how do the results look now?
Clearly, we can now see that there is a potential issue with the systems after Christmas. The ROI for the systems before Christmas is a healthy looking 24.7% but after Christmas, this falls to a lowly 7%. This includes this season's results which was a very high 30%+ before Christmas and a -ve ROI after Christmas.
Looking at DNB, it is 14.3% before Christmas and 8.5% after Christmas. Not quite the same drop off as traditional H/A betting but still a drop in return after Christmas.
Interestingly, looking at double chance, it looks much smoother. A return of 11.2% before Christmas and a return of 11.9% after Christmas.
You can see why this phenomenon occurs when you look at the returns from backing the draw. Before Christmas, this makes a near 600pt loss and after Christmas, this makes a profit of 377pts!
Obviously, this year is included in the results but even looking at this without this year included, backing the draw after Christmas was break-even before this season which is why double chance looks so good.
Although I don't want to draw too many conclusions at the moment and definitely now before I've spent an enormous amount of time on the ratings this summer and trying to understand how I can improve their profitability, I think there is plenty of evidence to suggest that I should be looking to get the draw on our side in some way or another. Whether this is done using DNB, DC or even a combination of DNB,DC and H/A betting throughout the season, I think there is enough evidence to involve me exploring this option for next season.
Thanks to 'A New Blog Reader' for the comments. I hope the above lets you see that you are actually onto something here although how we use the above info will always run the risk of backfitting something to fit the results which is not a route I want to go down!
In terms of variable staking, I probably agree that variable staking may actually work well with something like double chance betting, especially since the strike rate is very high and the average odds are low. Hence, it could be a steady way to increase your bank and your staking. Once I've decided on how I'm playing the bets next season, I'll then look at how to maximise the returns using different staking plans. This is definitely the last step in the process though for me and considering the work I need to do this summer, I won't be looking at this for a while!