Thursday 30 June 2011

Similar Games Model

I'm not sure if I'm suffering from some sort of addiction to Excel but you won't be too surprised to hear that I'm looking at potential new football systems for next season!

I think my one worry with everything I've done on the football modelling so far is that it is all based around a ratings approach to football modelling. I won't try to pretend to anyone that I believe there are better ways to do this as this approach is one that has stood the test of time and ratings are used by many people to correctly predict football games. I'm also using factors that are widely believed to be the most successful factors when building football ratings, so I feel like I have most angles covered.

However, what happens if my approach suddenly stops working as well as it did during backtesting and during last season? This could be caused by inherent flaws in my models that haven't come to the fore as yet. It could also be caused by others cottoning on to what I'm doing and building similar or better models than I'm using. Lastly, but the least likely, it could be caused by bookmakers changing their odds in such a way, that my edge is eroded. I don't think my 9% return last season will have bookmakers giving up just yet!

There are clear advantages and disadvantages to using ratings for football modelling but over the last week or so, I've started to think that I could use my data to look at another ways to build systems.

Now, before I start on this road, I don't want anyone to think I'm going to end up with another 10+ systems and I don't want people to get too lost in what these new systems say.

To put it in perspective, if my new systems (31-33) carry huge risk next season, then any other systems I'm tracking for next season are going to be off the scale when it comes to how risky they are. In addition, they won't be based on my good old trusty method of ratings analysis, so they might turn out to be rubbish. Hence, these really are paper tracking systems.

Interestingly, I was exchanging emails with a blog reader yesterday and I happened to mention that I was thinking of doing something outside of my current ratings methodology and once I explained what I was thinking of doing, he wrote back to me and said there is actually a name that is used to describe this......

'Similar Games Model' - I learn something new everyday!

Basically, my idea was to trawl through my data and try to build an algorithm that identifies historical games that featured statistically similar teams and try to see if I could find an edge with these games.

For example, I find a team who's coming off 6 defeats, playing a team who's won their last 6 games. The form team is a heavy favourite and the other team a big outsider. Is there an edge here of backing the underdog or even the favourite?

Now, that's an extreme example but you get the idea. I write something that can pull out games where the teams involved are statistically linked in some way. I then see if there is a way or determining whether or not there is an edge there. As a last check, I'll apply my ratings from the 3 algorithms I've built and see if any edges appear that I can maybe exploit.

Well, going by the fact I'm writing this post, you'll know that I've got something already but unfortunately, it was in an Excel sheet and I should have put more thought into the output format! I thought putting the output into different Excel tabs would be a good idea but I've used the maximum memory I think my PC has and ended up with a lot of tabs!

After a bit of manipulation (thank God for the Indirect function!), I've got some data out that looks sensible.

As you can imagine, there are lots and lots and lots of ideas here but they fall over due to lack of volume of bets. I'm not in the game of adding up different ideas to make a system and therefore, you either find something that works for a fairly large sample or you don't find something at all.

Obviously, you can relax the criteria a little and widen the rules and that creates enough of a sample although it dilutes the returns a little possibly as if you have a criteria that works, expanding it will lead you to include more rubbish!

Anyway, the crux of it is that I have a selection of systems that throw up bets. I'm at the stage now of going through each system and checking what sort of system it is and whether I could use it next season.

I think I'd ideally be looking for 3 systems max to follow as it would be interesting to test whether this methodology works.

For me personally, it would be really interesting to see how these systems would compare against my other systems.

Watch this space......

Sunday 26 June 2011

I'm still alive.....

Just a quick update to show that I’m still alive. I’m struggling for content for the blog at the moment, so instead of writing for the sake of writing, I’ll just keep quiet until I’ve got something worth reading to post.

If anyone wants me to take a look at something or has any questions about the systems, feel free to post a comment and at the very least, it will give me something to write about hopefully!

Just to show that I’m not resting on my laurels, I’m working away on my own strategy for my betting portfolio next season and I’m also working on improving the spreadsheets that I use to find the bets each time there is a set of fixtures. This sort of thing can feel like hard work at times but if I can save myself 30 mins each time I do the analysis of the fixtures, it will help me in the long-term.

I'm counting down the days to the start of the season but similar to last season, I will wait for 6 games to be played before I have any bets. I thought about reducing this to 4 games last season but my ratings actually need 6 games in order for the formulas to work properly. If I used 4 games, I'd need to use some adjustments and it's not worth the hassle I think.

Monday 20 June 2011

Brilliant Observation......

Apologies for the lack of updates on here in the past few days. I'm not sure what happened but it was like someone flicked a switch across at the SBC forum and the questions were coming at me from all angles which was a nice change! I spent a good few hours last night answering all of these questions, so I didn't have time to get around to looking at a question Grant had posted at the end of last week.

I don't want to paste his question but you can see it on the last post. In summary, Grant was suggesting that the betting banks you'd need for home bets wouldn't be the same as you'd need for away bets and therefore, the betting banks I suggested could maybe be improved up on. Grant even went as far to say that he thinks that you could maybe play twice the stakes on home teams and it wouldn't affect the overall drawdown of the systems. Hence, double the stakes but play the same betting bank I suggested before!

I could have looked at this for any system but what I've done is look at this for the 6 systems combined from 6-21 through to 8-22.

There are a few ways to look at what Grant has suggested. Firstly, I can analyse the drawdowns of the overall systems compared to the Home bets and the Away bets. Secondly, I can also look at the monthly returns of the overall bets versus the Home bets and the Away bets. Lastly, I can look at exactly what happens if I played twice the stake on home bets as compared to away bets to see if Grant was right.

Drawdowns



As you can see, the drawdowns for the overall systems is 90pts for H/A betting and 45pts for DNB. This is what I showed in my last post and I suggested 225pts as a bank for H/A and 112pts for DNB.

However, if we split this down as Grant suggests between Homes and Aways, we can see the drawdowns on both.

For Homes, the drawdown is 19pts for H/A and only 9pts for DNB. So, based on the same theory (2.5 times worst losing run), we’re looking at a betting bank of 48pts for H/A and 24pts for DNB or thereabouts.

For Aways, the drawdown is 96pts for H/A and 46pts for DNB. That would amount to 240pts for H/A and 115pts for DNB. Hence, only slightly higher than what I suggested as a betting bank for the overall system.

P&Ls



I could say lots but the key thing worth pointing out here I think is the fact that the Homes graph is much smoother than the Aways graph. Nothing too surprising considering the average odds for each type of bet.

Monthly Results



Again, I could write tons here but the key point here is that the overall system has less losing months than either backing just the Homes or the Aways. A little bit surprising possibly but then again, combining different types of risk and having more bets in a portfolio should help smooth your monthly returns.

And finally……….Is Grant right? Well, if you have looked at the above drawdowns and put 2+2 together, you’ll have a fair idea what the answer to this question is.

2pts on Homes and 1pt on Aways – The Results



Well, I take my hat off to Grant. An exceptional observation I think and something that takes portfolio betting to a new level with these systems if you are into this sort of thing.

The drawdown for H/A is only 82pts and the drawdown for DNB is 42pts. Hence, by increasing the staking on your overall portfolio (thus increasing the profits), you can actually decrease the risk of your portfolio.

Now, reducing your betting bank from 225pts to 205pts may not seem like an amazing thing but it’s a near reduction of 10% in your betting bank. In addition, by staking double on Homes as compared to Aways, you are increasing your ROI and your profits. Being able to increase your profits, increase your ROI and reduce your betting bank at the same time is something that everyone should strive to do! Increasing profits and reducing risk is what this game is all about and Grant has come up with a way to do this.

This post will either make perfect sense or it will fly so high above some people’s heads, they’ll get confused by it. Simply, I am NOT saying that anyone should stake more on homes or decrease betting banks etc. All my suggested betting banks are made clear on the blog and I suggest that 99% of people who follow these systems next season play to these betting banks and can hopefully make a good profit. The other 1% can have great fun trying to maximise profits whilst reducing risk at the same time as it’s the sort of thing that geeks do. ;)

Tuesday 14 June 2011

Reply to Peter

There is an excellent comment in the comments section on the last post. Over the past 6 months or so, I have had some good questions asked on here but this one probably ranks above most others. One thing I struggle to know at times is whether or not readers fully understand the way the systems work and the results I'm posting. Questions like that from Peter help me in the sense it gives me something to discuss but importantly, it also challenges my thinking which is a great thing for me personally.

I know from the comments posted on here over the past few months that there are a lot of intelligent people reading this blog and they probably get a lot from it. I can't help but feel this blog would be a better read if I got more comments like this which then allows me to expand on the work I've done and the thinking behind the systems etc. I think what I post on here is just the tip of the iceberg but unless people ask questions and get me thinking, I think I'll struggle to improve things going forward.

I hoped that by posting on the SBC forum, I'd get asked lots of questions and people would challenge my thinking on there but so far, it hasn't happened. I can't even get a comment on the thread in all honesty which is confusing for me personally and a little bit frustrating as I'm now struggling for what to write next. I could park the blog and the SBC thread until September but I hoped I'd spend all Summer explaining the systems and models and how they work and looking at ways to maximise profits next season but I really need people to ask me stuff to give me something to respond to.

Anyway, Peter's comment is a really good one and it does challenge me and asks some very pertinent questions. I'll deal with one question at a time.

Looking at last season,the systems 6,7,8,21,22 all performed poorly compared to previous seasons.Do you think this was a one-off or would you say last season's results were indicative of what may happen next season?I dont understand why anyone would follow these systems next season after last season.

I think that's probably a very fair comment Peter. I've spent a lot of the last few weeks looking at these systems from last season and the results and wondering whether or not these systems are actually worth following in their own right. After all, I have shown that it is much better to be backing teams that appear on two rating algorithms rather than just appearing on one algorithm.

However, that's not to say that the individual systems aren't worth following. The 5 individual systems made a combined return of 5.5% from 1,760 bets last season in a live environment. I've applied a couple of additional filters on away bets after last season and tweaked the base ratings by a tiny bit which should increase the 5.5% return hopefully.

Looking at it subjectively though, I can see why a 5.5% return over so many bets might actually appeal to some people. If you are playing lower stakes and want a very high turnover of bets, these systems allow you to have enough turnover to employ this strategy.

If you are driven by higher ROI and a reduced number of bets, then I do think that these systems may not be fit for purpose. However, last season was a very poor season for these individual systems and there is no guarantee that next season will be as bad. These systems all looked very good until Christmas but suffered badly after Christmas with the high number of draws they were hitting.

Overall then, I can see why most people wouldn't choose to follow these individual systems next season but I can also see why they could appeal to the right sort of person.

Like many I assume,I will be concentrating on systems 6-21,6-22 and the like this season.Have you looked at how these 6 systems perform in a portfolio?If not, can you take a look.Can you also take a look at how systems 7-21,7-22,8-21 and 8-22 perform in a portfolio too please?I like these systems but I would prefer over 500 bets a season for a football system.

I agree with you Peter that most people will concentrate on these double systems. After all, the results from the first live season were very good and therefore, I personally would be more confident about these systems next season if I was forced to choose.

I haven't looked at how these 6 systems had performed in a portfolio. Here are the results:



The results look good I guess and it is appealing to have an ROI of 16.3% from last season and a sample of over 1,000 bets.

If I look at the top 4 systems, then the results are:



This looks even better with an ROI of 21.1% from last season and a sample of 541 bets.

You suggested you wanted a portfolio of systems that produce 500 bets a season and this might well be the portfolio of systems for you Peter. Obviously, there is more analysis you'd need me to do before betting your life on that portfolio next season........

You mentioned the idea of only backing homes on one of these systems as a possible way to achieve a high strike rate, 10%+ return. As you may already know,Football Investor runs a strike zone system featuring these types of bets (including bets from top European Leagues).Last season,it provided a very nice return for my portfolio.Can you look at devising a portfolio of home bets from the systems please that will generate around 200-300 bets at an ROI of 10%?Backing 50 homes in a season isn't a system I'd follow.

I mentioned this before on the blog but it really is up to the individual to go away and look at some of the stuff I've posted on the blog and if they come up with anything, I can confirm whether or not their calculations are correct. However, since you seem like a nice guy and have clearly done some work already Peter, here's a quick look at the homes on the 6 double systems. e.g. 6-21.6-22 etc.

The returns are as follows:



A remarkable 21% ROI last season with a strike rate of 64.4%. This is over 239 bets. At the end of the day, you can do this yourself if you look at the returns by system and look at home/away bets and so on.

I disagreed with a comment on your last post regarding leverage and the fact you don't get much leverage with these systems in a portfolio.Surely you would look at the performance of the portfolio and then calculate the betting bank as a result of the longest losing run the portfolio has encountered.You wouldn't do as you suggest and sum up the betting banks for each system.

This is an interesting one and I think you may have caught me out here. When I wrote this in my last post, I wasn't really thinking about this but I'll admit you are correct (I won't admit this too often that I'm wrong!)

If I look quickly at the P&Ls for the portfolios of systems 6-21,6-22 etc. the P&L and Drawdown graphs look like this.





If I look at the drawdown graph, the max drawdown was seen last season and was a substantial loss of 90pts for H/A betting and 45pts for DNB.

Based on these drawdowns, I would have suggested a bank of 225pts for H/A betting and 112pts for DNB.

If I sum up the banks I suggested for each system, it comes to 255pts for H/A betting and 140pts for DNB. Hence, if you are playing this particular portfolio of systems, you could use a lower bank than summing up each of the individual banks. I stand corrected.

I think my point was more around the amount of leverage you get from running a portfolio like this is nowhere near the leverage you'd get from 6 independent systems. The issue is the correlation of the bets that appear on each system.

Anyway, I hope this answers all of your questions Peter. Just get in touch if there is anything else you want to know.

Sunday 12 June 2011

Lessons Learnt From Last Season

One advantage for anyone following these systems from next season is that this is the second season of football modelling and betting for me personally. Along with a group of others, I followed the original set of systems last season and have gained an enormous amount of info that I hope to learn from for the upcoming this season.

Some of it is very basic and common sense but we also learnt a few gems of info that will hopefully stop others from making the same mistakes as we made last season. This post isn’t going to be an exhaustive post of everything we learnt and I’m sure I’ll think of other things later on which I will mention but hopefully I’ll capture most of the important things here.

Be Consistent – This relates to your staking throughout the season and the systems you follow. The two main errors that people made last season (myself included!) was to take a short-term view of the systems and even shorter term view of how you are staking the systems.

Due to the fact that last season was the first season, I think all of us following started with low stakes and quite quickly realised that the systems were capable of getting near the historical returns. Hence, we increased stakes after a few months and this proved to be a poor decision.

Secondly, due to the fact that there were so many systems that we could follow at the start of the season, it was very easy to pick a system, see it start slowly or make a small loss early on and then see other systems making decent profits. Without a second thought, we’d jump onto the other systems and then watch as they hit a bad spell and the original system we were following would recover!

Both of these things seem like common sense now but at the start of last season, after making very little in the first month or so of the season, it was easy to move systems and then after another couple of months, increasing stakes seemed like the most obvious thing in the world!

In summary, if you are following a system or a portfolio of systems, stick with it throughout the season and more importantly, play level stakes throughout the season. This is probably quite important as there is a clear seasonality pattern in the historical results and live results where returns dip after Christmas. Hence, increasing staking isn’t the best idea for these systems.

No Leverage – If you are playing a portfolio of systems, please remember that the system bets are correlated and in the case of the systems from the same rating algorithm (e.g. 31,32,33), the systems have a very high correlation. Hence, if you choose to follow system 31,32 and 33 for example, then every team that appears on 33 must also appear on 31 and 32. Hence, all system 33 bets will have 3 times your average stake on it.

What you can’t do, is look at the 3 system banks I suggest, sum them up and then decide to play 50% of the overall bank due to the fact you will get some leverage by playing 3 systems!

I struggled last season with the whole leverage thing as I was following a lot of bets and trying to squeeze my bank as far as it could go to allow me to maximise profits with as small a bank as possible. The only thing you do though is put your bank under pressure when things aren’t going well and that’s never a good thing. Stick to the betting banks suggested and assume very little leverage from playing more than one system.

Don’t make up your own way to play the bets! – There was one really good example that sticks in my head from last season. A few guys chose to follow 2 systems (one from each rating algorithm) but only back a team if it appears on either system. This has a fundamental flaw in the logic which they found to their cost quite quickly!

If you are only placing 1pt on a team that appears on either system, then you are likely to be hitting a lot of very weak bets. If a team only appears on one rating algorithm, it is actually a very weak bet. As we can see from the results last season, when a team appears on both systems (e.g. 6-21 or 8-22 etc.), then this is a much stronger bet than a team that only appears on system 22 for example.

The outcome for these guys was that they were following 2 systems that were both profitable but by playing the bets in this way, they were actually making a loss even though both systems were independently profitable! Thankfully, as soon as they told me they were following this strategy, I showed them how bad a strategy it was and they switched to another strategy!

If anyone does come up with an idea of a way to play the bets, it won’t take me too long to pull together the historical results for their strategy and show them the results, so they can ask me for help.

Worst Systems You’ve Ever Seen! – I can guarantee that there will come a time next season when this thought will go through your head! On any given day or any given weekend or even any given month, the systems will go through a bad patch that will be so bad, you will wonder if the edge is gone. If it happens at the start of the season, you will wonder if there was ever an edge there in the first place! (Hopefully that doesn’t happen this season….)

Last season, I think everyone following the systems had this thought at some points and at various times, it did seem like these systems were so bad, they’d never pick another winner.

I think on one weekend last season, I had value bets in 21 different games and didn’t hit a winner in any game! Considering the average odds were about 2/1, it was nearly impossible to do this but the systems managed it.

Like most things in this gambling game, you can’t judge the systems on any day or any weekend and as last season showed, you can’t even do it on any month as the returns can vary so much on a monthly basis.

When these thoughts do go through your head, it’s important to remember why you are following the systems and look at the historical drawdowns and the historical P&L and try to remember that these systems have been there before and come out the other end!

Best Systems You’ve Ever Seen!
– I can (hopefully!) guarantee that there will come a time next season when this thought will go through your head!

It’s been well documented on here and on the blog that the results for the first few months last season were probably as good as anything most of us have seen in the gambling game. The ROI for all systems being tracked was at 25%+ and 7 of the 13 systems had ROI’s above 35% at Christmas time last season! The sample size was only 1,195 bets at that point in time but clearly, based on what the results were then, there were thoughts that I may have built something that could make us all a lot of money! I made plenty of money in the first half of last season myself and I’m not really a big gambler.

I don’t know what will happen this season but I’m sure there will be the odd day, weekend or even month when we look at the returns and think we’re one step to the Holy Grail at this game! The truth is though, if you are thinking like this, a correction is probably due…..

Don’t overstake – This one really is common sense but simply, there is no point in overstaking these bets.

For those wanting to follow systems 31-33, 6-32, 21-31 or 6-21-31 next season, the bets will be posted up free of charge in the SBC forum but will also be made public on the blog and also to any subscribers to my other systems. Hence, there is no reason to overstake these bets. I don’t quite get why people overstake bets which are given out free of charge as surely the staking should take into account the subscription fees.

Secondly, even if there is a fee for the other system bets that were proofed last season, it will be cheaper than most other subscription services. Again, this should be factored into the staking and there is no need to look to maximise profits in a season. Think of this as a long-term strategy to make money!

The New Systems Are Unproven – I don’t mind repeating what I said the other day to the guys who followed my bets last season but simply, anyone who is taking the TFA bets seriously this season should remember that the 6 new systems are completely unproven.

If given a choice of following the systems from last season or the new systems for this season, you have to really follow the ones with live results under their belts. That isn’t the same as me saying that no one in here should follow these new systems this season but at the end of the day, I have 11 systems with a season’s proofed bets and 6 systems with no proofed bets. If given a choice, you should always go with the systems which have proofed bets.

However, for all I know, the new systems might have an amazing season next season and the systems from last season might bomb completely. I don’t know what will happen but I know that I won’t be playing the same stakes on the new systems as I’ll be playing on the systems from last season. Others choosing to follow some of my systems next season should have the same idea.

For those only interested in my new systems (after all, these are free and being posted on here next season), then I suggest they play very small stakes or even paper trade next season. There is no rush for these new systems and after the time and development that has gone into this project, there is no rush to make substantial profits during a trial season.

And lastly……

Don’t blame me! – Anyone choosing to follow the system bets are in the very same boat as me when it comes to trying to make a profit from the footie. All I can do is share all the historical results, try my best to make sure the modelling is sound and then work very hard to find the bets, track the results, tweak the systems and so on.

If my ratings suddenly go down the tubes and no longer work (seems to happen with footie services every so often!), then I won’t be held responsible for the results. I have 9 years now worth of results that show that my ratings and systems have an edge but one day, that edge might evaporate and I’ll be left with a bunch of useless and outdated systems. I’d like to think that would never happen as I track everything so closely and as soon as I thought this was a possibility, I’d pull the plug on the project. However, I can never be sure that this methodology will work forever and therefore, there will always be a chance that the ratings stop working as well as before.

Friday 10 June 2011

A few final answers.....

As discussed last night, I was conscious that there were some outstanding questions from a post I wrote before to try to answer any questions people had asked me about the systems for next season. The questions are here http://the-football-analyst.blogspot.com/2011/05/question-and-answer-session.html and the answers to some of the questions are here http://the-football-analyst.blogspot.com/2011/05/answers-to-questions.html
From what I can make out, the questions I have left to answer are:

Q). What betting bank does each system need for next season? What are the drawdowns for each system? Can I graph the drawdowns for DNB? Can I compare the volatility of DNB with backing the teams to win?

You can find the answer to this either here http://the-football-analyst.blogspot.com/p/system-reviews-201011.html or here for the new systems next season http://the-football-analyst.blogspot.com/p/new-systems-for-201112-season.html
Obviously, when I was writing my review of the systems from last season or writing a preview of the new systems for this season, I had these questions somewhere in my head! Hence, the answers are already on here.

There was also a question that appeared a little bit later on.....

Q). Can I create a system which is for Home Bets only and has a high strike rate and a 10% return historically?

Well, I'm not going to create a system for this specifically as if someone is only wanting to follow home bets on a system, all the details are included on the blog as I always show the results split between home and away bets.

I can see that backing home bets on system 6-21 has produced a return of 28.9% and a strike rate of 67% over 5 seasons. Last season when the system was live and proofed, it made 11.7pts from 69 bets with a strike rate of 62.3%.

System 7-22 has produced a profit of 41% across 5 seasons with a strike rate of 70%. Last season when the system was live and proofed, it made a profit of 8.9pts and an ROI of 36.9% with a strike rate of 70.8%.

I could go on and show all the other systems and records for home bets only but simply, all the results are on the blog and are easily accessible. If you click on a picture and then click on it to enlarge it, you can see all this easily.

If people have any questions or would like to ask about betting banks for a particular system (i.e. home bets only), I'll happily pull together the historical results and check the drawdowns etc.

Overall then, I think that means I've answered all the outstanding questions I had on the blog (excluding questions about plans for next season!). If anyone does want me to answer any other questions or look at something for them, then please leave a comment and I'll get back to you.

Thursday 9 June 2011

Finished My System Building....

I’ve finished pulling together the data for the final 3 systems that will be live from this season.

In case anyone spots a small discrepancy between the returns I quoted on my last post and the ‘true’ results for systems 6-32,21-31 and 6-21-31, it is simply down to a small change in the odds I used.

When I was building systems 31,32 and 33, I used a list of 9 bookmakers and took the best odds from these at the time of kick-off. However, I noticed last night when I was pulling together system 6-32 that I had small differences on system 32 compared to system 6 for the same team! I then read the blog (yeah, it’s written on here!) and saw that I used a list of 7 bookmakers when building my other systems last Summer. Not a disaster but cutting out 2 bookmakers has meant a loss of around 0.5% return on each new system. If anything, it shows I’m honest with my analysis!

Anyway, the results are all updated. http://the-football-analyst.blogspot.com/p/new-systems-for-201112-season.html

I’m now finished with my system building this summer. I WON’T be building another system. You need to draw a line somewhere and I could build systems until the cows come home but at the end of the day, if I have an edge, I don’t need to prove it on 100 systems. I proved it last season with profits on 12 of the 13 systems that were live.

I’ll let the dust settle and then I’ll get into thinking more about next season. I know some people are waiting on hearing my plans for a possible service next season as they’d prefer access to my bets before they appear on the blog. I’m leaning towards doing something down this road but I need to decide whether I want the commitment and possible aggravation that comes with running a service. Having been there before and not really enjoyed it, I’d hate to think it would spoil my enjoyment of the football modelling. It spoiled my enjoyment of horse-racing for a long time!

I also remember that there were some questions that were unanswered from my Q&A post and I’ll need to go back and answer these.

Here’s a few tables with the historical results of these new systems side by side. I know it sometimes helps to look at systems side by side if anyone is thinking of following one of these next season.


Monday 6 June 2011

The Final Systems for 2011/12 (I hope!)

This will be my last post before I take a wee break. Had a few hours unexpectedly tonight, so thought I’d use the time wisely!

As discussed on my recent post, my last task this summer is to look at the systems and try to create another couple of systems that may be worth following going forward.

People might wait their whole life for a winning system but at the moment, I seem to have them coming out of my ears…..

Anyway, I’ve looked at the possible combinations of systems and simply analysed the number of bets and profit generated across the last 5 seasons. The results are below.



I’m trying to work this through in my head but I’ve got 11 systems that have 2 season’s fully backtested results and one live season of results from last season.

I’ve also got 3 new systems (31-33) that will be trialled from next season.

That means I’ve got 14 systems in total. As well as recording all the results, I need to find all the bets each week and record odds etc. as well as place all my own bets.

I’m trying to work out just how many systems I will be able to track as I would hate to think I’m taking on too much here.

I think I can manage another 3 systems and ideally, I’d want one system to be with rating algorithm 1 and 3, one system to be with rating algorithm 2 and 3 and one system with rating algorithms 1,2 & 3.

Looking at the table of results, I have settled on 3 final systems to meet my requirements. At the end of the day, I can’t follow every combination of bets!

The final 3 systems will be 6-32, 21-31 and 6-21-31.

I’ll need to go away and pull together all the results and stats for all 3 systems and add them to the other 3 systems we are tracking this season.

Another advantage of using these 3 systems is that they include my weakest systems from last season 6 and 21, so anyone who is following my main systems this season shouldn’t be affected too much if I post these bets in the SBC forum next season.

Once these 3 systems are built and I have all the results, I’ll share them on here and that means we are ready to go to war for next season.

After that, I’ll go over some ground rules about how these systems work and any advice I can give anyone who fancies following these systems next season. I learnt a lot last season along with some others, so I have plenty of advice to give!

Sunday 5 June 2011

More stats for Systems 31-33

I've decided to create another section on the blog where I've posted up all the important stuff anyone would want to know about systems 31-33.

http://the-football-analyst.blogspot.com/p/new-systems-for-201112-season.html

I've also worked out the betting banks for each system and they are shown on there too.

I think that brings me to the end of my rating analysis and backfitting work this summer. It's not quite the end of the system building as I want to explore the correlation between the 3rd rating algorithm and the other 2 rating algorithms but that is just Excel work really. Nothing too difficult from here on I hope.

I want to look at combining the 3rd algorithm with either of the first algorithms and look at systems such as 6-31, 6-32 and so on. I will probably look to add another 2 or 3 systems if I can find anything that works particularly well when I backtest the results.

Lastly, I will also look at whether we can use the 3 algorithms together to find a set of bets. No idea if this is possible as the 3rd algorithm seems to throw up totally different bets to the other 2 algorithms but I'll take a look at it and see what I can find.

I'll obviously give regular reports on here how the next bit of work goes but I'm going to take a bit of a break now as I'm shattered after doing all the work for systems 31-33. I need to give my brain a break!

I'll be back later in the week I guess.

Saturday 4 June 2011

System 33

Last but not least, system 33 is done.

One thing I’ve learnt over the last 9 months or so is that trying to narrow down the basic system bets from the ratings I’ve built to find only the strongest bets is fraught with danger. Yeah, it may be the case that you can pinpoint the very strongest bets but in any season, due to the fact there are less bets, it can cause havoc with your P&L due to the smaller number of games.

For anyone who followed my bets last season, they will have seen this phenomenon occur on system 9 and system 23.

System 9 made a profit of 0.4pts from 48 bets last season and system 23 made a loss of 3.5pts from 74 bets.

Systems 8 and 22 included far more weaker bets than systems 9 or 23 and yet, they still created much higher profits than these more selective systems.

Simply, it is down to short-term variance. I don’t think I have any doubts that system 23 was picking better bets than system 22 but at the end of the day, system 22 had 245 bets and system 23 had the 74 bets last season. Hence, over a sample such as 74 bets, even the best system might hit a bad spell and not make a profit over a small sample of bets. This led me to drop system 9 and 23 going forward as they don’t have enough bets.

Clearly, you don’t want to dilute your sample of bets too much as it can create too much extra volatility.

Why am I mentioning this……

Well, system 33 might possibly fall into that same trap this season if we look at the results for this system.



There are a few issues I have with this system now that I look at the results.

Firstly, I don’t like the fact that there are nearly twice the number of bets in season 06/07 compared to season 10/11. The bets per season seem quite variable. What is causing this change of trend?

Secondly, I don’t like the fact the returns are so variable by season. Season 08/09 has a return for DNB of 7.9% and yet, the previous season it had an ROI of 22.8%!

Thirdly, due to the lower volume of bets, the periods of stability in the P&L seem to go on for months! I can find spells of 4-6 months quite regularly where no profit is made or very little profit is made. This isn’t ideal for me as next season, I could make no money until after Christmas or likewise, I might not make any money from November till the season end!

Lastly, the system only seems to have about 100 bets a season now. As discussed above, this is pretty small for a season worth of bets and it wouldn’t take too much bad luck to turn this system from a winning system to a losing system I suspect.

On the flip side though, the returns from system 33 are better than systems 31 and 32 and therefore, I can’t deny that system 33 does seem to contain the highest value bets these new ratings can find. Any system which produces a return of 25% for traditional H/A betting and 15% for DNB returns has the potential to be a goldmine!

Anyway, I’ll leave that there for now.

Thanks for the comments to my last post. Looking back, it sounded like I was a bit desperate to know someone was reading the blog but when you’re as involved in this as I am at the moment, it’s hard to know if anyone is getting this or not. I’m in danger of getting to the stage like some of the mad scientists I see on Dragon’s Den. They create something that could be the best thing that’s ever been invented but no one else understands it and therefore, it is only meaningful to them! I don’t want to get to this stage with my footie modelling………

Friday 3 June 2011

And system 32........

I’m starting to feel like I’m getting a bit drained with the footie as doing this, on top of working my ass off all day at work isn’t easy. I’m spending most of my lunch hour doing footie stuff and then another few hours in the evening. Part of the issue is that I’m like a dog with a bone at times and I should be pacing myself as it’s a long summer. At this rate, I'll be all set to go in a few weeks and I can't have my first bets until September! :(

It’s slightly more difficult at the moment as it feels like I’m updating 3 places at once. As well as keeping the blog ticking over, I’m keeping the TFA forum guys abreast of any developments with these new systems and plans for next season and I’ve also got a thread on the go in the SBC forum which is where these system 31-33 bets will be proofed first next season (as well as then appearing on the blog!).

I said in my last post that I was slightly disappointed with the results thrown up by system 31 but I also said it was the first step in the process. I can narrow down the number bets by stripping out the lowest value bets that appear on system 31 and this will give us a system which I've called system 32. This contains all the medium and high value bets thrown up by the ratings algorithm.

Well, the results for system 32 are much better and thankfully, I can start to see some light at the end of the tunnel.

Here are the results:



The same observations apply to system 32 as well as system 31 but in summary, the system has achieved a return of around 20% for backing the teams to win and a return of around 14% for backing DNB during backfitting (the first 3 seasons is the best proxy I think). This is a good deal better than system 31 which is good news.

I don’t want to analyse the results too closely before finishing system 33 and posting that but the first really interesting observation I've spotted is the monthly results pattern in 2011. Under all my other systems, January, March and May were profitable last season but February and April were big losses. Feb and April were actually the two worst my systems have ever had due to the high number of draws I seemed to hit, so I was curious how these systems would cope with these months since they are fitted on DNB.

System 31 experienced a profit in Feb and a loss in April I noticed for DNB and traditional H/A betting.

Looking at System 32 now, it has a loss in Feb and a profit in April.

The key point is though, I still can't get away from the fact that February and April aren't good months. In one way, this is disappointing as I was hoping backfitting on DNB would turn a profit in both months but as I've said on the blog a few times now, I really believe February and April are freak months for football betting on these leagues. Hence, seeing these results on system 31 and 32 just reinforce this fact I think.

Anyway, that's another quick update and system 32 is now built. I'll finish system 33 tonight or tomorrow and post that too.

After that, I'll spend time pulling together all the statistics you'd ever want to know about each system and getting the results tracking into the same format I use for my other systems. Once this piece of work is done, I'll share more info on these 3 new systems.

Once I have a wee break, I'll then look at systems such as 8-33 etc. and I'll also think about whether or not it is worth looking at a system such as 6-21-31 which would mean backing teams only when they appear on 3 different algorithms! Having seen the results from backing teams that appear on 2 algorithms, I would think that backing a team that appears on 3 independent ratings algorithms would be very good but then again, there may not be enough bets!

I’m noticing that the readership of the blog is dwindling. I’m guessing this is due to the fact it’s summer and the season is over and most people were reading the blog for the bets each week but feel free to comment or ask questions etc. to try to give me something to feed off. It’s a lonely game writing to myself each time! It’s my 150th post on here too, so I deserve a comment for that if nothing else. :)

Thursday 2 June 2011

Introduction to System 31

I'm in a position now to share the first cut of results from the new rating algorithm. I have taken the very early decision to just ignore Premiership Aways going forward for this set of ratings. Anyone who reads the blog will know how difficult it has been to make money backing Premiership Aways over the past few seasons and quite simply, it is much more difficult for away teams to win away from home than it is at home in the Premiership.

You'll get a team away from home who is playing much better than the team at home they face and the bookies will price up the home team at 5/4 and the away team at 2/1 and no matter what way you look at it, the away team will represent value on nearly every set of ratings you can formulate. I paid close attention to Fink Tank in the second half of the season last season and they'd usually agree with my ratings and have the away team as value but ultimately, it is very difficult to get the away team to win.

Although these new ratings are looking at maximising the returns from Draw No Bet, I isolated the Premiership Aways for DNB using these new ratings and it was a similar trend to my other ratings. The very high value bets will break even if you are lucky but the weak value bets thrown up week in, week out are heavy loss makers. Teams like Stoke, Blackburn and newly promoted sides will always be taken on with my ratings as they are so bad away from home and usually get outplayed at home too by classier teams but at the end of the day, this is a results business and these teams somehow manage to win games at home! Keeps it simple if I just exclude these games from the outset.

You can see it simply by looking at the Premier League table last season and teams like Birmingham, Blackburn, Stoke, Bolton, Newcastle, West Brom, Fulham, Everton all lost 5 games or less at home. When you consider they have to play the top 6 at home, then you can see that they didn't lose too many to other teams of the same calibre as them. Away from home, they lost 10,12,13,12,9,10,7,7 respectively. Simply, if given the choice between two equal ability sides in the Premiership, even though the odds are 5/4 for the home team and 2/1 for the away team, you don't want to be backing the away side!

Over time, this trend may change and the odds for away sides may drift as people look to lay them and then it will look foolish to filter the bets this way.

Just so everyone is clear, for these ratings, value is defined where the odds I calculate for DNB is greater than the odds being offered for DNB. Seems like it's simple and very obvious but it's not! I first started doing it by looking at the odds for a win or draw against the odds I calculated for a win or draw but of course, that would be defining value for Double Chance and not DNB! Hence, it took a wee bit of time to get my head around defining value in the case of DNB and even now, it's still not crystal clear as it depends on the home/away probability and the draw probability but the home/away probability carries much more weight than the draw probability.

Common sense really as the stake you place on the draw is much smaller than the stake you place on the team to win with DNB! Hence, my first two sets of ratings can look at a game and define there is no value to back either team but my new ratings tell me that there is value in backing the home or away team and the draw! Simply, the win price may not be value on its own but by bringing in the draw, this creates additional value to turn it into a value bet for DNB.

Anyway, leave me to worry about trying to interpret my ratings. Blows my head thinking about it and it's me who's building them!

Below is the first set of results from the ratings.



A few observations before we move on:

1). Don't place too much emphasis on season 09/10 as 50% of the results from this season are included in the backfitting process
2). Seasons 06/07,07/08,08/09 and 10/11 give the best indication of the likely profitability of the ratings going forward
3). Season 10/11 has a higher number of draws than I would like to see in any season and therefore, 10/11 isn't the best indicator of future performance of this system either I suspect
4). Therefore, the first 3 seasons actually represent the likely profitability of these ratings
5). In summary, expected returns may be circa. 15% for traditional H/A betting and 9% for DNB betting

Overall, I'm slightly disappointed with the results during backtesting but then again, it isn't too surprising. Unlike my previous systems, the ratings here have been fitted on DNB and not traditional H/A betting in the earlier seasons and therefore, the 15% for traditional H/A betting was never the aim! The 9% is the aim.

How good is 9% for the ratings for DNB? Well, if you compare it to my systems last season, these systems had an ROI of around 9% for DNB (and traditional H/A betting!). However, last season was an exceptionally good season for DNB and therefore, the 9% last season was probably above expectations. The 9% I'm looking at here is looking at the 3 worst seasons for DNB.

The key thing to remember though is that 9% is only from backing all the teams that appear as value. The next stage of the project is to look at the bets and how much value is apparent within each bet. Clearly, the higher the value, the higher the return and therefore, filtering the bets by value should enable us to increase the returns.

All the value bets will appear on a system called system 31. What I'm going to do is develop systems 32 and 33 now and then spend time getting the results into the same format as my other systems. These results will always be kept separate though so as people don't forget these are untested in a live environment and don't mix them up with my other systems which have now have one season of results.

Once I have the new set of systems 31-33, we can then look at whether we can do something really cool and look at teams that appear on the 3 rating algorithms! We could also mix up the other systems with the systems 31-33 but all of this is for later this summer. Lots more work to do before we get to that stage!

I'll be back next time with an update of system 32 hopefully.