Tuesday, 4 February 2014

January 2014 Monthly Review

As a one-off this month, I’m going to spend a bit of time discussing the trends within the underlying results we have seen this season. By underlying results, I simply mean the results of betting 1pt win on every Home selection, every Away selection and every Draw in all UK games I’ve rated so far this season.

I personally have always kept OK records of the underlying results for all UK leagues as I download the data from footballdata.co.uk like most playing this ratings game but I was aware that my results analysis was probably not as detailed as some others in this game. Therefore, I’ve spent a little bit of time pulling together the underlying results in a little more depth than I was doing before. The purpose of this is so I can look at the trends in a little more detail and therefore, I’m going to share some of my findings in this post. I also have the tables set up so I can refresh them easily in future. 

I’m sure some reading this will be thinking that I’ve went down this route to try to find excuses for my results this season and in a way, that is partly true. I do believe my ratings still have an edge and therefore, I have to believe that there are other factors at play this season. When I look at the underlying results, I can see where the issues are and therefore, I think it helps me to understand what is going on this season. Of course, it doesn’t change my results and at the end of the day, I’m judged on the results the systems achieve and so far this season, they haven’t been good enough.

OK, I think the first point to make is that I always simply use the best price from a host of bookmakers in the data from footballdata.co.uk. Given I don’t use best prices for my own system bets, I could never match the returns quoted in the underlying results but that’s immaterial really. The purpose of this is to look at the trends rather than the specific numbers. I have concentrated my analysis on the last 3.5 seasons as this is when I started betting on football but I have the data going back further. The issue with going back further is the football odds aren’t quite as competitive and the results from 10 years ago are much worse overall than the last few seasons, so let’s keep this analysis to when I started rating games myself which was the 2010/11 season.  All data is to the end of Jan-14.

Let’s firstly look at the results by season, split between Homes, Aways and Draws.


OK, there is a lot to take in here, so let me explain what this table shows. It shows the games rated, the % wins for Homes, Aways and Draws and then the corresponding ROI for backing all Homes, Aways and Draws. It also shows an overall ROI which would be the result if you had placed 1pt on the Home, Away and Draw for every game I’ve rated.  At the right-hand side, it shows the ROI for each of my 4 algorithms for each season. I have only looked at live results, so alg 3 missed the first season and alg 4 only started last season.

I think the first thing that stands out is the fact that all 4 of my algorithms are underperforming this season compared to previous seasons. Two of the algorithms are actually loss making this season and so far, I’ve never had an algorithm make a loss over the season. None of the algorithms are having a disaster though and losing 1% and 0.6% across a high number of bets is no disgrace. The other two algs are showing a small profit.

Looking at the underlying results then, you can see that for Homes, this is the first season where following all Home bets hasn’t made between a 3%-5% loss. So far, backing all homes has made a 0.2% profit. In terms of Aways, the first two seasons lost just under 2% before last season’s freak season where backing all Aways made a profit of 7.4%. This season though, something of a freak again where backing all Aways has made a loss of 6.7%.

In terms of Draws, these have performed slightly worse this season than the last 3 seasons and have lost 4% so far which is the worst season of the 4.  In terms of overall ROI, the loss this season is the worst of the last 4 seasons (last season showed a profit which was an amazing stat!) and if backing 1pt on every Home, Away and Draw this season, you would have lost 10.6% of your stakes.

What does all of this mean in the context of the TFA ratings and TFA systems?  Well, around 65% of all TFA bets are Away bets. It is slightly more confusing this season as I have Draw bets too but before this season, it was quite steady at a 65% split for Aways and 35% for Homes.  Therefore, if there is a bias against Aways (in the sense, it is more profitable to be backing Homes), then this means it will impact my ratings and systems.

However, it isn’t as easy as this to understand the impact on my ratings this season. If we look at the returns by odds range, it gets even more difficult for my ratings.

Here’s the returns by odds range for Homes:

As this split shows, backing all odds on homes has shown a profit of 7.3%. Unfortunately, my ratings and systems don’t tend to find much value in odds on home bets (traditionally, more difficult to recognise value for this odds range) and therefore, a very small percentage of my home bets are odds on. However, my ratings do find plenty of home bets priced 2.00+ and therefore, even though backing all Homes has been a marginally profitable strategy this season, my ratings and systems are not really playing too much in this odds range and therefore, the home bias isn’t picked up by the systems as much as I would have hoped this season. Many of my home bets are priced 3.00+ too which compounds the issue. 

Here’s the returns by odds range for Aways:

Now, this is where my ratings and systems have a second issue. Historically, finding value in away bets priced 2.75 or lower is difficult according to my rating algorithms and therefore, I basically have a filter on all systems which kicks out all away bets at 2.50 or below even if my ratings find value in the bets. Not only that, after the poor performance of these Aways in the first season of live bets, I decided 2.75 was a better cut-off and therefore, I very rarely give out Aways at less than 2.75. Hence, I have missed out on the best performing Away bets this season too!

90% of my Away bets are priced within the odds range of 2.75-6.00 and as you can see, trying to make a profit in this price range this season has been very difficult. An overall loss of 9.1% for all of these bets.

OK, summing together the findings for the Home and Away bets by odds range, it starts to paint the picture of what has happened to my ratings and systems this season! Homes may be profitable overall but I am missing out most of the profitable bets and Aways losses would have been minimised if I had been backing short-priced Aways in addition to my usual bets.

What does all this mean in the longer-term?  Well, it doesn’t mean anything I think! Nothing I’ve seen this season has changed my findings longer term that it is very difficult to make money backing odds on Homes and short priced Aways. I do track the system bets (loosely) which don’t make the cut due to the odds being too short and over the first 3 seasons, both odds on Homes and short priced Aways were loss making.  This season, both sets of bets are showing a good profit but that’s simply due to the underlying conditions being easier than normal.

I think the other thing I’ve learnt a lot about over the first few years betting on football is that you need to go against the crowd to make money long-term and maintain a long-term edge. Backing odds on Home bets may be doing great this season but long-term, there can’t really be an edge backing the favourites all the time and they just don’t represent value. Likewise, backing favourites Away from home isn’t a strategy to make money and it’s not the type of bets that my ratings work well on.

Below is a table I’ve pulled together. Admittedly, if I had more time, I would have split the data between Homes and Aways but I decided to stick with the overall results. 


This table shows the same data I showed earlier by season but splits it by month. I’m not expecting anyone to be able to do any statistical analysis to show that there is perfect correlation between my algorithm results and the underlying results but I think this table allows us to make some observations.

Clearly, when there is a large loss for the underlying Away results, generally, it is more difficult for my rating algorithms to make a profit. We can see that in Sep-13 this season. Aways lost 17.4% and my 4 algorithms lost 9%-30% each! The same is true for Jan-14 where Aways lost 16.9% and 2 algorithms made a loss. March-13 saw another 2 algorithms make a loss and then Feb-12 saw all 3 algorithms suffer large losses as Aways lost 16.2%. Going further back, we can see the same in Jan-11 and Apr-11 where Aways suffered large losses and the algorithms suffered.

Of course, the degree of loss on each algorithm is dependent on the split of Homes and Aways on the algorithm (varies by algorithm) and then how the loss is made up by odds range. If the Aways suffered a large loss because all short-priced Aways lost that month, my algorithm will have avoided the damage. On the other hand, if the opposite is true (as in this season), then my algorithms will struggle.

I could go on and on here and start looking at the split by League but the data samples get very small. I will mention League Two Aways though. This season, backing all League Two Aways blindly has made 10% profit and therefore, my ratings have really capitalised on this factor this season. As many of you will know, going into this season, most of my algorithms had made a loss on League Two Aways over the last 3 seasons and therefore, anyone who took a decision to drop League Two Aways have been hit hard by the underlying results being so favourable in this league compared to other leagues. That’s just bad luck but it shows what can happen in the short-term at this game.

Before I move onto the monthly review, I’ll try to sum up my thoughts on what the above means for my own betting and for the service.

Prior to this season, I honestly believed that a season was a long-time in football betting. I rate around 2,425 games a season on each algorithm and therefore, surely that’s enough games to allow any trends to level out and surely each season will give my ratings the opportunity to show the edge they have? Now, I’m not so sure. I’ve never had a losing season (or year) since I started betting on football but at the moment, it’s looking more and more likely that this could be the first season where I don’t make a profit on the football.

However, if I look at the last 4 seasons in total, there isn’t much to worry about. If I had been offered 3 fantastic seasons and a break-even season at the outset from the first 4 seasons, I’d have snapped it up.  Now I’ve had the first 3 seasons, having a break-even season would be very disappointing but as I’ve come to realise, I’m in this for the long haul. A break-even season hurts and given how much I will have staked this season as well as the time and effort, to make nothing will be painful but if anything, it makes me more determined for the future.

Right, onto the monthly review.

Est Systems


The table above shows the results by value band and split by Home/Away bets for algorithm 1.  The obvious thing to point out is the split by Homes and Aways. Homes had a good month and made a profit of 7.1pts from 20 bets. 11 winners from 20 bets.  Aways had a tough month with 8 winners from 32 bets and a loss of 2.5pts.

In terms of the value bands, <5% Aways had a nightmare and lost 6.7pts from 16 bets. All the Home bands made a profit this month. Again, looking at a month in isolation isn’t much use given the very small bet number but I say that every month!

Overall, a profit of 8.8% from 52 bets is a decent month for this algorithm.


Algorithm 2 didn’t have such a good month and overall, made a profit of 0.4pts from 42 bets. Homes and Aways both made 0.2pts this month, so a bit different from algorithm 1 and the general underlying results this month. 

The <5% Away bets was the best performing band for Aways whereas the <5% band for Homes wasn’t great. The two highest bands for Homes both had 100% records this month although there were no bets >25% and no bets between 15%-20%.

This algorithm was the 2nd best algorithm coming into the season but so far, it has found this season really tough. Hopefully it can push on over the remainder of this season.

Here are the results for the Est systems this month:


Overall, a profit of 6.4pts from 185 bets. An ROI of 3.5%. Similar to previous months this season, the more selective the systems were, the poorer the results became.

Systems 7 and 8 both had losing months which is disappointing given system 6 had a winning month. System 22 did well though and managed to improve on the returns on system 21.

Systems 6-21 and 6-22 both had good months (6-22 hit 4 winners from 8 for a 56% ROI) and therefore, going by previous seasons, you would look for 7-21 to 8-22 to build on this. Unfortunately, as has happened all season, the returns fall down a hole from system 7-21. 7-21 thru to 8-22 all had losing months.

Fed up saying this every month but this season is so different to previous seasons. It’s a shame on systems 7-21 thru to 8-22 as before this season, the ROI’s were exceptional on these systems after 3 live seasons but there has definitely been a correction this season.

Overall, although the ratings had an OK month, the systems haven’t capitalised on the ratings again. Systems 7-21 thru to 8-22 continue to disappoint as they have done all season and it will take a minor miracle for these 4 systems to not suffer a loss this season. All 4 systems are heavily in the red and have been all season.

New Systems

Here’s the results for algorithm 3:


I guess the first comment to make is that whilst the other algorithms have been struggling all season, this algorithm had somehow managed to achieve a profit in the first 4 months of the season in difficult circumstances. However, it has taken a bit of a hit this month and has made a loss of 8.1pts from 73 bets. Interestingly, it is the Homes causing the issue and they made a loss of 6.9pts. Aways lost 1.2pts.

Drilling down into the Homes, we can see that the <5% band did the damage with only 3 winners from 19 bets and a loss of 11.6pts. I keep taking some comfort (not too much mind you given I’m ignoring these ratings this season!) that when the lower value bets are losing, we can easily escape the damage by just missing these bets out of our portfolio of bets.

The high value home bets were also 0 from 3 but it these bets that have done exceptionally well so far this season and therefore, I’ll give them a poor month now and again. Overall, a poor month for Homes but the type of Homes on this algorithm are a little bit different (average odds are 2.99) and therefore, I think a bit of volatility is to be expected.

Here’s the results for algorithm 4


A really poor month for this algorithm although the low value Aways bets were to blame. An overall loss of 13.7pts from 94 bets. However, a loss of 18.7pts for <5% Away bets probably tells you all you need to know about this month on this algorithm.

4 winners from 33 bets is shocking but strip these bets out and this algorithm didn’t have a bad month. Actually, the <5% Home bets lost 3.3pts too, so the low value bets performing poorly sum up the month on this algorithm.

It’s disappointing that these bets have had a poor month but as I said above, long-term, it’s no bad thing. If I can find a better way to filter the bets to remove the worst bets (aside from using the system filtering), then this can only benefit my betting and the service generally.

Homes made a profit of 3.5pts from 36 bets. An OK month for Homes but no chance of undoing the damage done by the Aways.

Here’s the results for all the New Systems this month:


Given both rating algorithms lost 21.8pts in total, to be posting an overall loss of 17.4pts from 351 bets is very pleasing. Not quite as good as posting a profit of course but when the ratings have a nightmare month, the systems can only do their best to minimise this loss and they did that well this month.

Systems 32 and 33 had a storming month even though system 31 had a losing month. System 42 wasn’t quite as fortunate and seemed to make things worse than system 41!

However, when you get onto the combined systems, you can see 31-41 only suffered a small loss even though both individual systems had a nightmare month and the top 4 combined systems all managed a decent profit. (That’s how the Est Systems are supposed to work if they are reading this!)

If you look at the results before this month, it does appear that we have seen something of a correction this month. The lower systems (31 and 41) have been performing well all season but they struggled this month whereas the higher combined systems have had a good month after struggling all season!

Overall, a losing month and a disappointing month for the New Systems but longer term and even this season, the results still stack up well. These aren’t the worst systems this season, far from it!

Misc Systems

Overall, a profit of 18.8pts from 154 bets for a 12.2% ROI. A really good month and I’m about to get very  picky I suspect with my forthcoming comments but I’m hard to please at times!

Looking at the split by system, 5 of the 6 systems achieved a profit of 13%+ this month and therefore, it’s hard to be too critical. However, given the way this season has gone, I don’t think I should be surprised to see that STOZ is the system letting down this set of systems this season and has been all season. Coming into the season, STOZ was the best of these 6 systems but like the rest of my systems, this season doesn’t seem to mirror previous seasons! An 8pt swing between STOZ and TOX/STOY and it’s very annoying! All 3 systems hit 9 winners each, STOZ seemed to have 13 more bets though!

Historically, it’s these additional bets that STOZ has that separates this system from the other two (the two are 95% correlated) but this season, it is these additional bets that are causing STOZ to lag behind the other two systems.

My other moan this month about these systems is the split by Home and Away bets. Historically, the Home bets on these systems have been much better than the Away bets but this month, it’s been a horrible month for home bets and a great month for Aways. A 6.7pt loss for Home bets and a profit of 25.6pts for Away bets. Not only that, all 3 SG models made a loss on Homes and these systems have the best homes from any of my system bets!


It really is a bit of a head scratcher if I’m honest and for someone like me who has really enjoyed seeing the same trends repeated each season, this season is proving a nightmare to understand. I’m sure the same is true for others following and those who base staking plans and portfolios on the live results, they’d be backing the wrong systems with the wrong staking plan I suspect!

Overall, I can’t knock the returns this month and it’s another step forward for 5 of the systems as they recover from the bad losses in September. STOZ needs a boot up the backside!

Draw Systems


Another very solid month for these systems and my confidence in these systems continue to grow and grow. I’ve been on a journey with these systems whereby at the start of the season, I had more faith in these systems than the Euro Systems (hard to not have after the mess the Euro Systems made last year!) but after a very slow start and a lack of confidence, as results have grown, my confidence has grown.

This month is the 5th live month and that’s 4 monthly profits in a row. Overall, a profit of 39.7pts from 239 bets for an ROI of 16.6%.

The first draw algorithm (D1) made a loss of 6pts this month but the second draw algorithm made 14pts profit. Pleasingly, the 6 combined systems all made profits again and these systems continue to impress.

I’ve said this before but I only follow 3 of the draw systems in my overall portfolio but they have really saved my bacon this season. Of course, I wish I followed more but it’s easy to be brave with hindsight at this game and as I’ve said a few times this season, you make your own luck at this game. If these systems had crashed and burned this season, I would have regretted following any unproven systems in my portfolio but that’s the chance I took. The fact I’ve been paid handsomely so far (still early days this season!) is a little fortunate but I’m not complaining!

The next few months are big months for these systems and if they can finish the season in line with current returns, I expect these will become a much bigger part of the service next season.

As an aside, one thing I am considering for next season is a set of draw systems for the Euro leagues. It would be a lot of work and not sure whether I can fit it in without dropping other systems but I’ll have some big decisions to make at the end of the season I think regarding the future for my own betting and the service. I’m always looking for ways to diversify further and I expect Euro Draw systems are probably an obvious place to go. 

In terms of other ideas for next season since I’m on that topic, I think one issue this season has been the lack of Home bets in the UK leagues and although they are traditionally not the most profitable bets my ratings find, one option I could have would be to have some bespoke Home systems. One issue with the way my algorithms work is that they really are suited to Away bets generally and therefore, having some Home bets (including some short prices) would maybe help smooth the returns and even if I could make 2%-3% ROI from Home bets at short odds, I’d be more than happy with that given the ease of staking and the ease of odds availability.

Anyway, back to the review and another good month for the draw systems.

Euro Systems


After a 100pt loss last month and me tearing the systems apart in the monthly review and the blog results updates, they’ve bounced back with a 104.1pts profit from 219 bets. Not only that, the filtering by system worked like an absolute dream this month.

Before I get on to this month’s results, let’s recap where we have been with the Euro systems before now. Last season, I built two Euro rating algorithms across the 4 leagues (Italy, France, Spain and Germany). Both algorithms were a disaster with system E1 losing 4.5% and system E6 losing an amazing 9.3%. However, the filtering by systems seemed to work OK and it seemed like a bit of a miracle that the top 3 combined systems made a profit.

I binned both these algorithms in the Summer, started again and with the very same data, had a second go at building two algorithms. So far this season, it was a similar story to last season whereby the ratings weren’t doing well but by cross referring the algorithms, the higher combined systems were again in profit somehow.

Roll on another month and boom, the Euro systems are now back in the game and maybe (it’s a mighty big maybe!), these Euro ratings are not as bad as I first feared. Those who remember the Euro systems from last season will remember a number of false dawns with the Euro systems and being honest, I didn’t have any confidence in the ratings and systems all last season. I remember reading comments from people saying these Euro systems looked OK when they were in profit but from what I could make out, they were being kept afloat by some big priced away winners and overall, I didn’t care too much for the systems.

I had that in mind when I built these new Euro ratings last Summer and therefore, one of the things I tried to do was reduce the average odds and switch the emphasis from Away bets to Home bets. The average odds last season were much higher than this season as a result. However, with the apparent Home bias in these Euro leagues (backing all Homes blindly traditionally doesn’t lose too much), I figured my best chance of succeeding in the Euro leagues was to have more home bets.

These new Euro systems have now been live for 4 months and we have 3 months of profits and a really bad losing month. Thankfully, this month wipes out last month completely and we’re back in profit for the season.

In terms of my thoughts on these systems, it’s still early days. I have more confidence in these Euro systems than the same systems last season (that’s not saying too much!) but compared to the other UK systems, these Euro systems would be quite a long-way off I think.

I chose to follow 2 Euro systems this season in my portfolio to give me some interest and combined with the Draw systems, these systems have saved my bacon. Again, I wish I had followed more but I got burnt a little last season from the Euro systems (would have been worse but I jumped ship before season end!) and I wasn’t about to get burnt twice!

One thing I should mention about these Euro systems is the number of draws they find. If you look at the AH returns, I could be wrong but I don’t think I’ve had any set of systems produce such high AH returns in a month. What this tells us is that with a few less draws and a few more wins, this month would have been absolutely astronomical and therefore, the month was maybe one of the best months any of my sets of systems have ever had! To create 100pts profit in a month is good but to also create 75pts for AH0 on all bets is amazing. Last month, the systems lost 100pts but only lost 34pts for AH0 betting. 

Overall, a great month for these systems and a solid start to their season but a long way to go. These systems are a little behind the UK systems and they haven’t reached the halfway point of their season yet.

Under/Over Systems


Keeping this short. These systems have been teetering on the edge of the cliff for the past year and a half and they toppled over the cliff this month! A loss of 10.5pts from 44 bets.

These systems are already looking forward to their retirement this Summer. I’m looking forward to not having to track the results of the systems or spend time finding the bets each time either!

Overall

At a high level, a profit of 141.1pts from 1,192 bets. An ROI of 11.8%.

In reality, as I said a few times on the results updates, the actual results most people would have been achieving were nowhere near these results and therefore, it was a tough month. To be fair, when the Euro systems lost 100pts last month, that brought down the returns for that month, so they have boosted the returns this month!

It’s difficult to see how you can make a profit this month, even with a well diversified portfolio. To make a profit, you needed to be following as many Draw and Euro bets as possible and as I said on a recent results post, if that was your strategy this season, you were following the wrong systems!

The season is going to be remembered (if that’s the right word?) for what has happened to the UK combined systems I suspect. We’re now past the halfway mark and systems such as 21, 7-21 thru to 8-22, 33-41, 33-42 and STOZ are all loss making for the season. As many reading this will be aware, before this season, these systems had 1,2 or 3 solid seasons behind them and had achieved very good returns, so it’s disappointing they’ve all struggled so far this season.

In reality, some of it will be due to the fact the conditions have been tougher this season as outlined in the beginning of this post, some of it will be due to a little bit of luck not going our way (never had so many big bets lose 2 goal leads) and the rest of it will be down to the systems just not having as good as a season as the last 3 seasons. Poor seasons happen and my systems are no different to any other systems with an edge but if you get a bad mix of these 3 things as some of the systems have had this season, it can lead to a really poor set of results.

Of course, some following the service this season are possibly sitting with a portfolio including a number of the systems that have struggled the most and therefore, they must be wondering what they have done wrong this season.  Have they done anything wrong?


I’ll leave you with that last thought and for my next blog post, I’m going to touch on a few issues people have had this season following the service and try to respond to some fair criticisms thrown my way recently. 

No comments:

Post a Comment