Tuesday, 28 May 2013

2012/13 Season Review Part Three - Systems 31,32,33,41,42

I think this review (and the New combined systems review next) are going to be the most difficult to write since if I base it on live results, we have systems 31-33 with two seasons of live results and 41-42 with one season. If I base it on two season’s results, we have a mixture of live and backtested results and as we know, last season wasn’t indicative of the performance this season as it has been better for the New Systems this season than last season. If I go back too far, results get clouded with too much backtesting and therefore, there is no right or wrong answer here.  Assuming that people want to play some of these New systems after this season, then I think they have to take into account more than just one season to get anything meaningful for analysis purposes when we look at results by league etc.

After a bit of consideration, I am going to look at the last 3 season’s results for the detailed analysis at the end of the review. This gives a big enough data sample to draw meaningful conclusions, it includes last season which was very poor for the combined systems and it includes one season of full backtesting which will gives the results a bit of a boost but it’s the trends were are interested in, rather than the actual figures in the detailed analysis section.

Any previous backtested results prior to season 2010/11 can be found on the historical results section on the blog.

Season 2012/13 Performance


These systems started the season very well and have probably ended the season in a bit of a slump if I’m honest but similar to the first two reviews, when you look at the overall season results, you can’t have too many complaints.

System 31 was the star this season I think and to achieve a 10.2% ROI from 733 bets is very good over a season.  It is very disappointing that system 32 couldn’t improve upon this return and ended up with a slightly disappointing 5.8% return. Hard to know where it went wrong for system 32 this season as it was picking from a very good set of bets but ultimately, didn’t manage to do that well at all. 

The biggest shock though (and maybe the biggest shock of all the TFA systems this season) was system 33. It was the only system that made a loss last season from all my live systems and going into this season, I didn’t give this system much hope at all. However, from the very start this season, it has performed very well and to end the season with a 34.2% ROI from 104 bets is stunning.

System 41 was in its first season live and a return of 4.3% from 942 bets is disappointing considering it was above the 5% return for most of the season.  It actually started the season very well and at one stage, 10% ROI wasn’t outside the realms of probability but the second half of the season was tougher for this algorithm. 

System 42 did really well though and managed to increase the ROI to 9.8% from the 4.3% on system 41.

Overall, 214.6pts profit from 2,545 bets at an ROI of 8.4%. A very good season and definitely lots of hope for the future here I think.

All Live Results (last 2 seasons for 31-33 and last season for 41-42)


I think the ROI trend is the key again here and as you move up the systems, the ROI increases. Wind the clock back 12 months and it wasn’t like this for systems 31-33 but I think as I said then, it was most likely down to variance. 

System 31 suddenly looks like a system that could be worth following now with an ROI of 7.2% from 1,404 bets. Even after a disappointing season this season, system 32 looks OK with an ROI of 8.1% from 503 bets. Finally, we can see that system 33 has recovered somewhat from a losing first season and now has an ROI of 13.8% from 230 bets. Still early days of these systems with two seasons behind them but so far, so good I think and the fact they have come through a poor first season actually gives me hope that they may be OK systems long-term.

Systems 41 and 42 have the same comments as above since this is their first season.

In terms of ROC, system 31 led the way this season and is the first system I’ve reviewed so far who has managed to double the betting bank in a season. An ROI of 107% is amazing.  I should also pick out system 33 who somehow achieved an ROC of 93.5% this season after losing money last season! Would have taken a massive leap of faith to follow system 33 this season I think but shows that you shouldn’t judge a system on one poor season if you think it is still working as you hoped when it was built.

Interesting that system 42 managed a better ROC than system 41 this season but looking at the bet number, I would say this wouldn’t be the normal thing to expect. 

Overall, the average ROC’s per season are quite decent for these systems and this is something that should be taken into account when people are building their portfolios going forward, whether it be these systems or the Combined systems.

Statistical Significance


I think the key here is how much more statistically significant system 31 is compared to the other 4 systems. Again, something that you should take into account. Some of the least statistically significant systems I’ve reviewed so far appear in this group of New Systems, so I think that’s another point to remember when setting portfolios for next season. These systems have started well but it is very early days for them I think. I’ve been saying the same all season if I’m honest and nothing has changed now we’ve reached the end of the season.

Targets for 2013/14 Season


I think that when looking at targets, we need to remember the number of live bets and statistical significance when deciding upon how accurate these targets may turn out to be. It is early days for these systems and therefore, the targets we set probably come with a bit more risk than the systems where we have 3 seasons of live results. Of course, it may be the case I understate the targets and they are easily beatable but more likely is the case that we undershoot the targets as we don’t really know the true underlying performance yet.

I didn’t mean to be cautious but a quick look at the 2012/13 ROCs and then the targets and it does look like the targets may look low. However, the targets are based on the long-term view and more importantly, I think system 31 and 33 probably outperformed last season, so you don’t want to be banking on this every season to make your return.

Systems 31 and 42 lead the way with a target ROC of 50% next season. Again, based on this season it looks low but based on the first two seasons for 31 in particular, it’s not too far off I think as a target.
The other 3 systems have a target ROC of 29%-40%.

I think the interesting system is system 33. Based on the returns from last season, this target ROC for next season looks very low. During the first season when the system had a losing season, it didn’t have any massive drawdowns, it was just generally hovering around break-even all season and then finished with a small loss. Hence, the betting bank isn’t huge and if it creates anywhere near the profit it created last season, the potential is massive for this system. However, with the very low turnover, variance may play a part (probably played a part last season in creating an amazing return) and therefore, it’s a risky proposition but might well be worth following in a portfolio of systems.

Detailed Analysis (Not going to discuss this too much given the fact it is a mixture of backtested and live results)

Performance by Season


I think the interesting thing to point out here is the fact that the overall returns from last season and this season are so similar. An 8.3% return two seasons ago and an 8.4% return this season. If you go back another season, you can see how much better the backtested results were before the systems went live.

It’s hard to draw too many conclusions from looking at the results from the first season as I mention in the introduction but I think any trends from the first season shown can hopefully be used along with the last two seasons to allow us to draw some meaningful conclusions.

Performance by League


You can see that the two poorest performing leagues are the SPL and League One. A slightly different pattern from what we saw when we looked at the Established Systems in the previous reviews where it was clear that League Two was an issue.

Performance by League and Season


The most interesting thing here is the fact that League Two made a substantial profit last season! This goes against everything we saw with the Est Systems.  Interestingly, the Premiership and SPL were the worst performing leagues last season.

Performance by Home and Away


Aways are more than twice as profitable as Homes and the trend is pretty consistent across all 5 systems.

Performance by Home and Away by League


Some strange figures thrown up here compared to the Est Systems. Biggest difference is probably the fact League Two Aways are profitable here and League One Homes are loss making. A substantial loss is created on SPL Homes and the SPL is proving a tough nut to crack across all the algorithms I think as the Est Systems struggled in the SPL too. 

No comments:

Post a Comment