Wednesday, 23 May 2012

System 6 Review

As mentioned on the last post, there will be a series of system reviews appearing on the blog in the next few weeks.  They’ll all be collated at the end and published in a single page.

To understand the reviews fully, you will need to have obtained a copy of the spreadsheet that was available to download on the last post.

System 6

In a way, system 6 will always hold a place in my heart as it was the very first real system I developed. Well, that’s not exactly true. Although it is the first algorithm in the UK leagues I developed, when I started out playing with data back in Jan-10, I wasn’t planning on looking at UK leagues. Hence, the very first thing I built was a rating algorithm for the European leagues but after a few weeks of playing with data, I decided against starting with the European leagues as quite simply, I couldn’t work out how the Away bets worked in European leagues. I could basically back Homes blindly and make a profit in the top European leagues with a couple of odds filters applied and although this appealed to me at the time, I wasn’t keen on just becoming a one trick pony, so the Euro leagues got shelved and I started again with the data for the UK leagues.  Systems 1-5 would have been systems on the Euro leagues, so that explains why system 6 is the first system I have!

Anyway, as usually happens, I digress. So, how has system 6 performed this season? Well, if I’m honest, I think it’s done brilliantly.  A profit of 41.9pts across 712 bets is an ROI of 5.9% and that isn’t bad for a base algorithm with very few filters applied. If you can make 6% ROI from backing every game, then it isn’t too difficult to narrow down the bets and make good profits.  I’d go as far to say that it may well be the 2nd best rating algorithm I know of for the 2011/12 season. Only beaten by its younger brother, system 21, the 2nd rating algorithm I built!

Importantly for me, the algorithm has outperformed its first season, over a larger number of bets.  I applied a few additional filters to algorithm one last Summer and you can see the impact from the strike rate dropping to 33.3% this season when it was 41.7% last season. By reducing the number of shorter priced homes and cutting out shorter priced away bets completely, it has clearly improved the performance of the ratings. In addition, the inclusion of more bets at higher average odds hasn’t been a disaster either. Hence, we get more bets now, at higher average odds but importantly, the profitability is better all round.  I can’t complain about that.

You can see the system has struggled if using any form of draw coverage this season.  Last season, there were too many draws on the system and this is evident by the fact that you increased your profitability last season on this system by covering the draw! I won’t apologise for the fact that AH betting hasn’t worked as well overall this season as like I said last Summer, the systems aren’t judged on AH returns, they are judged on outright profitability and therefore, any tweaks I apply are meant to improve profitability of backing outright and in some cases, this may be to the detriment of AH returns.

There were 7 winning months from 9 this season, compared to 6 from 9 last season.  Not bad consistency over the two seasons if I’m honest.

League One and League Two were both loss making this season but a profit was made on both if dutching the draw. At the opposite end of the scale, the Bsq Prem made a good profit outright but a terrible loss if dutching the draw too.

It’s a funny one as people will look at this and say League One and Two were poor and Bsq Prem was good but I’m not so sure myself. At times, there isn’t a big difference between winning and losing when your teams are drawing and therefore, I don’t particularly like to see a system which works when backing outright but loses badly when using some draw coverage! Makes me think that it could have been much worse if a couple of goals had gone the other way whereas a system showing a profit if dutching the draw makes me think it could have been much better if a couple of goals had gone my way!  All psychological though if I’m honest but interesting nonetheless.

Similar to last season, Aways did much better on algorithm one. An ROI last season of 5.7% was followed up with an ROI of 8% this season.  The Homes had an ROI of 2.2% last season and 1.7% this season!

It’s a tricky one to explain if I’m honest but I do think it has been a tougher year for Homes this season compared to last season, so maybe the returns don’t tell the full story. I remember picking up on this last summer and saying I wish I could get the Homes on system 6 to be as good as the Homes on system 21 but at the end of the day, that’s why the system 6-21 works better than 6. I know 21 picks better homes, so no one should really ever back a system 6 Home unless it is a pick on system 21 too.

Not surprisingly, Aways were to blame for the deficits in League One and Two this season.  Interestingly, a quick look at the Bsq Prem shows that a 25% loss was made on the home bets in that league. That explains why AH returns were so low in that league this season. The homes are much lower odds, so if you aren’t making a profit outright, they’d all need to be drawing nearly to make any sort of return covering the draw too!

Aways did lose in League two last season too and therefore, that is definitely something to keep an eye on next season I suspect. 3 seasons losing in a row and I’d be worried that the league needs some attention.

The success story of the season has to be Premiership Aways.  Going into this season, they were losing in 4 of the 5 seasons and I took some action last Summer to try to correct this. Interestingly, it hasn’t reduced the number of Aways appearing as bets but it has improved the profitability! An ROI this season of 43.8%. I’d take that every season! :)

Looking at the P&L graph since the system went live, it is definitely showing a strong upward trend and I’m pretty confident that the first rating algorithm has a significant edge over the bookies and other rating algorithms in the market. A new maximum drawdown was reached this season of 29pts but given that the average odds are much higher now than ever before, we do need to be careful about putting too much weight on the historical drawdowns as they are going to be higher now than they were.

There will be no changes to this algorithm for next season.  I tweaked it plenty last Summer and it appears to have done the trick, so no changes this Summer.

As for betting banks, I suggested 55pts last season on the blog and having seen a 29pt drawdown this season and knowing the average odds are higher than before, I’d have to suggest this should be increased.  A 60pt-70pt bank would be the call I suspect going forward and based on this season’s performance being repeated, this would be an ROC next season of around 60%-70%.  Based on the betting bank suggested last Summer, the ROC this season was 76%.

My target for next season would be an ROI of between 4%+ across something in the region of 700 bets.  Would achieve a ROC of 40% in this scenario (4% ROI, 70pt bank).  Would be a decent target for a basic system as I think the system outperformed this season a little if I’m honest.

No comments:

Post a Comment