I’m slowly gearing up for the end of season analysis that I need to do to look back on the results this season. I actually enjoy analysing the results and building new systems more than following the bets and tracking the results at times as once you’re betting live, there isn’t much you can do to change it. However, when you are analysing and assessing the edge you have (or lack of it in some cases), it leads me to think I can do things to improve the performance in future.
Anyway, I started spending some time last night building a proforma that will be used to analyse the results of each system. It’s much in line with what I did last summer if I’m honest. The biggest difference this season is I’ll be giving out the spreadsheet to all members of the service, so they can spend time looking at it themselves, rather than screwing their eyes up, trying to read it on the blog! Another advantage of being a member next season and joining this Summer! ;)
One thing that made me chuckle when I looked at this last night was the overall performance of the established systems if I look at all proofed results to date. Amazingly, after 5,444 bets, the ROI is as close to 10% as you can get (well, a profit of 547pts). When I started out, one of the aims was for all system bets on these 11 systems to generate a return of 10%. If I’m honest (as I usually am on here!), I thought that I would settle for 5% as long as the 6 combined systems gave me a 10%+ return. As it happens, the 11 systems are at 10% and the combined systems are at 15%. Ahead of expectations so to speak!
Aside from noticing this (which to be fair, is only interesting to me!), I had a quick glance at the performance by league for all systems combined. When I analyse the results by system soon, you’ll see the impact a bit clearer but sometimes, you can get too immersed in data and you can’t see the wood for the trees. I suspect I’ve been a victim of that this season although as always, it’s nice to know what’s going on but you can’t do a lot about it!
Here’s the profit and loss for Homes/Aways, split by league and season at the top. I’ve then ignored season as a filter and finally, I’ve taken out the Home and Away split. Hence, 3 tables, all showing the same overall results but just different splits.
I’ll be completely honest again and say if someone had asked me to rank the leagues in terms of performance after two seasons, I would have went for:
1st – SPL
2nd – League One
3rd – Premiership (although this season has been brilliant which I’d looked at a few weeks back)
4th – Championship
5th – League Two
6th – Bsq Prem
As you can see, looking at the bottom table and the results to date, I’m not really close. Well, the Premiership is the best league (still can’t believe that!), then it’s the SPL following closely behind. Then we have League One in 3rd. I guess you could say I got the top 3 in the wrong order. :)
In 4th place is the Championship which I got correct but the biggest shock for me is the performance in Bsq Prem and League Two. I thought I was about break-even in both leagues if I’m honest but jeez, a 5.9% profit in the Bsq Prem (that must have all come after Christmas as in a piece of analysis at the turn of the year I think it was, this league was losing badly!) and wait for it…..a 9.5% loss in League Two over the last two seasons.
WOW. How did I miss this? I don’t know if I’m honest.
So, a loss of 120pts from 1,268 bets in League Two. Looking at it by Home and Away split, it’s actually a profit on Homes of around 42pts, so it’s a great big loss of 162pts for Aways in this league. Gulp.
Split by season, it’s a loss of 111pts last season and 51pts this season. I can’t actually recall worrying about League Two last season but if I look back at my notes last Summer, I did notice the losses on some systems last Summer. However, the backtested results were so strong in League Two, I concentrated on tweaking the systems for Premiership bets and sort of ignored League Two’s performance and put it down to too many draws and a bit of bad luck.
I think one powerful thing with having the returns for AH0.25 and AH0.5 in the results now is that it helps to understand what has gone on in some cases. I didn’t have this last Summer!
In this example, I can see quite clearly that the draw is playing a massive impact in the performance of these League Two bets. I think the best way to do this is to compare League One with League Two as they have had a similar number of Away bets.
Here’s a table showing a comparison:
So, a 152pt profit in one league, a 162pt loss in another league over circa. 975 bets. Surely this is evidence that the ratings work in one league and don’t work in another isn’t it? Well, I wouldn’t bet on that!
If you start working along the table (bearing in mind, the more right you go, the more draw coverage you are using), you can see that the returns on the systems really start to converge.
By the time you get to AH0.5 (which is basically Double Chance which means dutching the selection to win and draw to win the same amount, regardless of the selection winning or drawing), you can see that the returns are almost similar. One league has turned a loss of 162pts into a profit of 64pts (6.6% ROI) and the other league has turned a profit of 152pts into a profit of 85pts (8.6% ROI). On league has made an additional profit of 226pts by including the draw coverage and one league has made an additional loss of 67pts by including the draw coverage.
Clearly, League Two in this case has suffered horrendously with draws in this sample of bets which has resulted in this freak set of results but does it actually mean that League Two will continue to be unprofitable going forward?
Well, there are many ways to judge if a system has an edge over the bookies. Many people will use the simple test of whether it can make a profit backing every team outright. If this is a true test, then the conclusion has to be that the systems have no edge in League Two and therefore, there is an argument for dropping League Two bets completely.
On the other hand, we need to bear in mind the sample size here is less than 1,000 bets. If the systems turned a small % of these draws into wins, the systems performance would look much, much different. For a start, the average odds in League Two on these Aways is quite high at 4.50 (7/2) and therefore, every game turning from a draw into a winner is actually a swing of 4.5pts or thereabouts. Hence, although a deficit such as this looks quite large, it is only a small swing in a small number of games (bear in mind there is correlation amongst systems to factor in) and therefore, if the results of 10 games had turned from draws into victories, then the system could be in profit over the two seasons. Of course, any more swings against the systems and we’d be looking at a larger loss!
Overall then, what does any of the above tell us? As with a lot of analysis I post on the blog, it’s more thought provoking for us, rather than leading us to concrete conclusions. I personally don’t see an issue with League Two’s performance based on the fact that it can be explained by the number of draws on the systems in that particular league. I have another 2,000+ backtested results that tell me the ROI in League Two should be in line with other leagues, so do I suddenly believe that the backtesting is incorrect and there is some inherent flaw in ratings in League Two?
I think it’s easy for me to say to everyone to ignore this but bear in mind, I know how the ratings are built. There is nothing within the ratings that can bias the results in League Two (I don’t use league as any sort of filter or rating factor) and therefore, I would confidently predict that we’re seeing a lower than average number of away winners in League Two over the past two seasons for high priced Away teams. Not only that, they aren’t actually losing when they aren’t winning, there is clear evidence to say that they are drawing, rather than winning.
Is it a trend or a blip? Well, I’ll tell you in 5 years if I should have stopped betting on League Two matches but until then, the sample size is too small to know for sure.
However, if I look at the backtested results in League Two, then you can see clearly that the trend started in the 2009/10 season and therefore, the systems have actually lost for 3 seasons straight. This makes it much more interesting now as we have another set of data to add to the mix. Therefore, over the last 1,311 games, a loss of 183.5pts has been created in this league.
As you’ll probably tell from reading the above, I wrote the whole post assuming that I didn’t know the backtested results and the conclusions you would reach without these results being known but now that you know the systems lost in 2009/10 before the systems went live, you have to (no matter how much I blabbed above!) come to the conclusion that this needs further investigation this Summer.
So, add that to the list of things I need to do this Summer. What is going on with League Two? I’ll get back to you in August I suspect but if I forget, someone remind me to look at this please.
No comments:
Post a Comment