Wednesday, 10 December 2014

Season review do far

I don’t tend to write too many season review posts apart from season end as quite simply, I don’t think any time period shorter than a season is too meaningful. Yes, I do write monthly reviews and yes, I even go as far as detailing all results on this blog for every day my ratings and systems have bets but anyone with half a brain knows that what happens on a day to day basis when it comes to gambling is fairly random. If it wasn’t random and I knew when I was more likely to win or lose on any day, I’d build this into my models and my staking plan and increase my profitability. :)

I think any time I have done season review posts it’s always off the back of poor runs of results and most times, it’s to have a period of reflection for myself more than anyone else. Anyone who reads my blog posts will know this blog isn’t an advert for the service and is simply a record of all the system results over the season. I suspect some people only read the monthly reviews and if you only read these, I’m not sure you get the full flavour of what we are seeing in a season. I think this season maybe falls into this category and therefore, after 3 poor months, I think it’s better to stand back and take stock of the situation.

There’s not going to be a lot of depth in this post and it’s going to be a high level overview. I think I sometimes get too lost looking at details of how each rating algorithm and each system is doing but at an overall level, people are generally following the ratings or following a combination of systems or a mixture of the two. Therefore, I’ll try to look at the results in this way in this post.

Here’s the results so far this season for each algorithm,:

Coming into the season, algorithm 3 was the best algorithm, algorithm 4 was the worst algorithm and algorithm 2 just shaded algorithm 1. Therefore, I think these results aren’t too far out from what I would expect in terms of how the systems are doing relative to one another.

Algorithm 3 has made a 4.5% profit from 298 bets so far and although it’s not a vintage season, it’s hard to expect the algorithm to do that well when Aways have been performing so poorly. Algorithm 4 has made a loss of 7% from 252 bets and this is disappointing, even taking into account the difficulties this season. Algorithm 1 has a profit of 1.3% and algorithm 2 has a loss of 0.1%.

I think in a more normal and probably an ideal season, algorithm 3 would be sitting at c. 8%, algorithm 1 and 2 at 5% and algorithm 4 at 3%. Hence, all 4 algorithms are substantially down on where they should be I think based on historical experience.

I think this season, algorithm 3 is the most followed algorithm, followed by 2 and 1. Not too many will be following algorithm 4 but of course, quite a few are following unique bets and therefore, a unique bet on algorithm 4 may be pulling down the returns a little based on these results but I could be wrong.  I know Steve (Daily 25) is following algorithm 4 this season as I figured that it was a potential way to follow TFA bets with less odds movements but I didn’t expect the algorithm to do as badly as this! Even though it was the worst algorithm, it has been profitable up until now, so it’s disappointing to see it making a loss so far this season.

Overall, not the best results over the first 3 months for the algorithms but it’s not a disaster. All 4 algorithms favour Away bets heavily and therefore, given the carnage on Away bets, I don’t the results are too bad apart from algorithm 4.

This season, quite a number of people are using the value ratings to filter the bets to try to improve the profitability of the ratings and systems. I have always been wary of this myself but after seeing fairly conclusive results last season, I decided to base my staking plan in part of the value ratings so far this season. Unfortunately, it has been a disaster so far and the following tables will help me (and you!) understand why.

Rather than looking at the results by algorithm and value band, let’s look at the total results for all 4 algorithms together and by value band:

Straight away, we can see the issue the algorithms are having this season. Historically, the low value bets were marginally profitable and on some algorithms, the low value bets were loss making. So far this season, we are seeing the complete opposite picture. Bets with less than 5% value have created all the profits and those of you filtering these out or staking according to value are basically missing all the profits so far this season.

A profit of 34pts for all bets <5% value, a loss of 34.8pts on all bets >5%. As you would expect, the average odds increase as you go up the value bands and with the issues we are having on Aways this season, it makes sense that the highest value bets are struggling.

I think one part of the value ratings that I haven’t investigated in enough detail is the Home v Away bias when it comes to % value. Clearly, with the way value is defined within my ratings, Aways are always going to have more value on average since it is based on odds and Aways have a much higher average odds than Homes. This means that those of us filtering by value or staking by value ultimately end up with more riding on Away bets. Not really ideal and in a season like this, missing Home bets for being so called ‘low value’ is a killer for the P&L.

This is definitely something that needs more thought this Summer and rather than having a value % based on odds, I maybe need to relook at how I define value for Homes and Aways.

Here’s the same results split by algorithm:

Those following each of the algorithms and filtering by value can probably relate to the results shown here. Although algorithm 4 is the disappointing one this season, the carnage could have been avoided by applying a value filter as suggested during the Summer. Unfortunately, playing the same way on any of the other algorithms (which was also a suggestion!) has resulted in losses as the wrong bets have been winning.

What I would say though is that we are only 3 months into the seasons and there are a helluva lot more historical results that show that filtering by value isn’t the worst idea in the world. Given the results this season aren’t in line with the previous seasons, I’m not totally surprised the value ratings aren’t proving a good indicator but I’m hopeful this will change over the next few months and hopefully by season end, the season may look something like previous seasons!

As I outlined in the introduction, using the algorithms and/or value ratings are one way to play the bets I provide each week. The other way is to simply select a portfolio of systems to follow and follow these systems in a portfolio. It has been discussed to death on the blog before about the pros and cons of a portfolio approach but I feel that as long as you aren’t following correlated systems within a portfolio, the portfolio approach has its benefits. This season, I invested 75% of my betting bank in a portfolio approach and the other 25% in an approach using the algorithms and value ratings. So far, the portfolio approach has been much more successful but both approaches have been disappointing so far to varying degrees.

Going into this season, the record for profitable systems was:

2010/11 – 11 live systems, 11 profitable systems (100%)
2011/12 – 22 live systems, 21 profitable systems (95%)
2012/13 – 41 live systems, 33 profitable systems (80%)
2013/14 – 52 live systems, 35 profitable systems (67%)

I remember doing the work on this in the Summer and writing about it on the blog but if you followed a random selection of 4 systems each season in a portfolio, the chances of making a profit were over 90% in each season.

How does things look this season?

Here’s the results for all systems to the end of November:

There are 61 live systems this season. At the moment, only 22 of the systems are in profit which is only 36% of the systems. As you can see, compared to the previous seasons, this is a much lower % and shows the difficulties the systems are having this season.

Doing the same analysis I did in the Summer, what are the chances of making a profit if selecting 4 random systems in a portfolio? Here’s the same sort of table I showed in the Summer:

As you can see, if following 4 systems this season in a portfolio, the chances of being in profit at the moment are only 39%. Considering this number was 90%+ for the last 4 seasons, you can see how poor this season has been to date if using this portfolio method.

To finish this post off, I thought it was worth having a quick look at the 10 SBC portfolios I tracked last season and posted about during the Summer. Admittedly, some people may have used these portfolios and tweaked systems based on what I wrote in the review of these portfolios in the Summer but at the very least, this gives us an indication of how these systems are performing compared to last season. Here’s the results of the 10 portfolios for the season so far:

7 of the 10 portfolios are in a marginal profit situation and 3 of the 10 portfolios are in a marginal loss situation. Simply, break-even for the 10 portfolios isn’t too far away from the truth.  If anyone has added Euro Draws to their portfolio (I did!), then it may be slightly worse than break-even but not a disaster.

Again, similar to what I said above, it highlights the fact that it hasn’t been the easiest strategy this season but then again, it’s not been the worst strategy in the world.

Overall, I think this post provides a quick overview of a few strategies this season and how they are doing. I think break-even is a common theme that jumps out to me and if you are break-even this season, I think you are in line with where I expect. Anyone doing better than this is outperforming and anyone doing worse is under-performing. I’m sure there are outliers at both ends of the scale.

For results to improve, we need to see more Away winners in the UK, more Euro Draws than we are seeing and we need the higher value bets to start performing a bit better than so far this season. A combination of all 3 would turn this season around quickly but I suspect we’ll see some improvements in some cases and some deterioration in others!

Let’s see what the next 6 months bring……


  1. So Graeme, same old question, what is the P/L for each unique selection to level stakes?

    Also with your value calculations if your system makes something 2.0 (Evs) and best price is 2.5 (6/4) does your value calculation make that 10% value (50% - 40%) or does it make it 25% (10/40 * 100)?



  2. Hi Matthew.

    Placing 1pt on each bet for every UK algorithm is basically at the start of the post. It shows the P&L to placing 1pt on systems 6,21,31 or 41. Depends what algorithm you want to follow?

    Following all 4 algorithms together and just placing 1pt on any bet appearing is not a way I suggest anyone follows (puts the same weight on a unique bet on algorithm 4 as algorithm 3 which can’t be right!) but the P&L so far for this is 4.77pts profit from 439 unique bets. An ROI of 1%.

    Being honest, no matter what way you cut the data for the 3 months so far, I think break-even is about right. I personally think break-even isn’t a bad performance when Aways are losing so badly. Using your unique strategy, Aways have lost 30pts from 240 bets. Homes have made 34.77pts from 199 bets. Quite an amazing return for Home bets but as I keep saying, if you can’t make a profit backing Homes over the first 3 months in the UK, you should really find a new hobby!

    In terms of the value calcs, as I alluded to in the post, I’m maybe coming around to your way of thinking. I know you define value by the first method whereas I have defined value by the second method. My method means that I end up with potentially bigger value on bigger prices. I rarely find 10% value on Home bets but on Away bets, my prices are often way out of line and I’m seeing 20%+ value on big price Premiership Aways for example.

    I think when we get runs like this where Homes are doing much better than Aways, anyone following my value calcs must be struggling. I can’t get near break-even for Aways, so seeing teams at 20%+ value just makes me laugh. If I’m break-even at Pinny odds, I’m losing fortunes at minimum odds but I need my ratings to be achieving 10% ROI for the ratings to break-even at minimum odds. When my ratings are break-even, value becomes a bit pointless.

    Cheers for the comment. Stop backing the same Euro teams as me please Matthew. :) Every time I see a tweet saying you’ve backed a team I’m on, they never win!!!!!


  3. I have just looked and I think (sure someone can pick me up if I'm wrong!) if you didn't bet on any unique bet that appeared system 41, you would have an ROI of 26.4% this season under your idea. :) I knew there was a use for the 4th algorithm....just a shame it's a way to filter out the worst bets! :(