Tuesday, 4 June 2013

2012/13 Season Review Part Seven - Euro Systems E1, E2, E3, E6, E7

Similar to the last review, this review is going to be quite short. I’m not going to spend a long time reviewing a set of systems that have clearly failed over the course of the season and given the fact I can scrap these and start again, there is no point in going over old ground.

For what it’s worth, I don’t even think what happened this season will help me any more next season as from looking at the systems during the season, the underlying ratings were never getting anywhere near the profitability they achieved during backtesting which is as good a sign as any that the ratings aren’t much good.

I’ve mentioned a few things in emails to various subscribers over the past few months about why I think these leagues are more difficult to make a profit in than the UK leagues. End of the day, if I knew what was causing the systems to underperform, I’d have corrected it last season but it’s not easy.

This was my first season of really paying attention to these leagues and from what I could see, the majority of bets in the leagues were for the bigger priced teams away from home who are in the bottom half of the league. No real difference to the UK bets but I think one thing I noticed this season is that the Home bets seem to have a bit of a bias towards them and the favourites appear to perform much better than in the UK leagues.

I may take this into account next season and adjust the criteria around the odds. Playing on bigger priced aways seems to create a real edge in the UK but in the Euro leagues, I’m not sure the edge is as easy to find. I’ve been playing about with the data most of the season in the background trying to find an edge or two to explore but it’s not easy to spot anything obvious. I actually think I’d maybe have a better chance with the draws in these leagues given the draw rate and basically, if you are backing something blindly that doesn’t make a large loss, you have a much better chance of succeeding. In the Euro leagues, I’m backing big priced Aways and they make a large loss in general, so trying to find an edge in a place where a large loss is created if backing all Aways is pretty difficult!

I’m guessing some reading this are familiar with the favourite-longshot bias but I suspect a large part of the issues I’m having with the Euro leagues is linked to this. There is a short article written by Joe Buchdahl on this very topic which I think covers the subject in enough depth for most of you reading this.  http://www.football-data.co.uk/blog/favourite_longshot_bias_football.php

Of course, one way around this is to adjust my thinking and rating building techniques and concentrate on Home bets at lower average odds in the Euro leagues but as anyone will know who’s built rating algorithms, trying to find an edge at lower odds is much more tricky than finding an edge at longer odds. However, as the Euro bets have shown last season, finding an edge during backfitting and backtesting and then seeing it work in a live environment is a different thing altogether.  It makes me appreciate my edge in the UK leagues more than it did before!

Anyway, leave that for later this Summer when I’ll have another go at trying to get something to work in these Euro leagues.

Season 2012/13 Performance & All Live Results

Not sure how you describe these results apart to say they are terrible! I guess the only positive you can take from the results is the fact that yet again, the filtering appears to work with my systems but when you are working with ratings as bad as this, it would take a miracle to get any system to make a profit.  I wouldn’t quite call it a miracle but seeing E3 with a profit is actually some achievement. I already know this from the UK systems but in terms of system building, my skills are pretty good. My issue with the Euro leagues is my ratings are rubbish to start off with, so I’ve next to no chance of making anything work with the systems I’m building.

One thing that does stand out is the average odds are pretty high. I realised that during the season but my ratings were definitely getting sucked in by the high odds for some Away teams but as I said on a few occasions, the market was taking my ratings, chewing them and spitting them out!

The impact of the draw can be seen on some of the AH returns. At various points in time, the draw was playing a bigger part than it appears here at the season end but a lot of this is down to the run the ratings experienced in the last few months. Again, nothing in the historical data pointed out an issue with the end of season games but clearly, my ratings were useless and again, they got sucked into overpriced away bets against home teams who needed to win. Long-term, the strategy may be profitable but I wouldn’t like to bet on it.

Targets for 2013/14 Season

Given what happened last season, I think my target remains the same as last Summer. I want to try to find an edge in these Euro leagues with a couple of different rating algorithms. If I can do this, I can then cross refer the algorithms and build systems that can hopefully create better return than just following the profitable ratings is the key to making a success of these Euro leagues going forward. It’s not easy and as I’ve commented before on the blog, one attraction to these leagues is the fact that there isn’t many ratings analysts doing what I’m doing on the Euro leagues. The UK leagues are saturated by people using rating algorithms and this is reflected in the competition for the prices at times. The fact that not many are able to do this on the Euro leagues (and those I am aware of using ratings in these leagues aren’t doing great either) suggests it’s maybe impossible to do but I’ll give it another go next season.

Detailed Analysis

Performance by League

I can only laugh at these results. There are two leagues profitable outright (France and Spain) and the other two leagues have massive losses. However, if you look at the AH returns, the two leagues with massive losses appear to have been hindered by a lot of draws (an 18% outright loss in Italy is turned into a profit if using AH0.5!). Therefore, I would suggest that it’s not as straightforward to look at this table and say that France and Spain were the strongest leagues and Italy and Germany the weakest leagues.

Yes, based on outright betting, this would be true. Based on AH0.5 returns, Italy goes from the poorest performing league to the best performing league.

Overall, the results aren’t great but backing the draw in Italy for all the system bets was a massive profit this season, so I suspect the systems weren’t that far away from a good return here given the average odds. When the average odds are nearly 3/1, there is a massive swing between a draw and a win. A 4pt swing and with a crazy amount of draws in Italy, in another season, it could have been a much better season there.

As for the other leagues, the results aren’t great and the best performing leagues were still poor overall.

Performance by Home and Away

Homes were the big underachiever here with a substantial loss of 9% whereas Aways only lost 3.2%. However, whilst I said above that there appeared a home bias and a favourite bias, if you look at the average odds here, I think you can see the issues.

The Aways thrown up are on average 5.12 and even the Homes on average are 2.55. Hence, in the majority of games, we’re backing unfancied home teams against better fancied away teams and of course, we’re backing big priced aways against odds on homes. In a season where favourites seemed to do well and Homes especially, it’s not hard to see why the systems struggled so much!

I’ll try to pay particular attention to this aspect during the ratings building later this Summer but it’s natural for my ratings to get attracted to some juicy prices as it’s easier to find value in these as the favs tend to be underpriced (according to my ratings) but we then get into the favourite-longshot bias again and we’re trying to find an edge where it is much more difficult to do so!

Performance by Home and Away by League

Spanish Aways stand out enormously here with some amazing profits across the board. However, aside from French Homes, the rest of the datasets are all losing money. Likely to just be a lot of variance at play here with the sample sizes and therefore, it’s hard to draw any conclusions apart from to say that the ratings were rubbish and these systems were rubbish!

Performance by Odds

I stared at this for a while and I couldn't really find anything too interesting! 


  1. I believe each league has it's own intricacies and you would need to come up with separate models for each league, a lot more work, but I don't think one model to fit them all will work.

  2. Hi Steve.

    I think you may well be right mate but doing this brings with it many issues I think. Sample size would be the key issue for me. Building ratings on a few seasons of data from a single league is next to nothing in terms of bet numbers. How robust the ratings would be is anyone’s guess.

    In addition, given my other ratings work from the Premiership down to the Bsq Prem, it’s hard to believe the performance could vary so much between the 4 Euro leagues to effectively nullify any edge my ratings could find.

    I could maybe try to build systems from the ratings for unique leagues (build a system for Germany, Italy, France, Spain from one algorithm) but this would take me away from the idea of cross referring algorithms etc.

    I personally would have no confidence in building a system for a league and just following it level stakes. I know most others do this but I’ve really bought into the whole multiple algorithms, system build process I go through and therefore, this would be a good bit away from what I do now with no guarantee of success and no record of it working for me.

    I might try to do something outside the box this season and you have given me an idea now with maybe a system for each league alongside my usual systems I am looking to build from the ratings but the idea of having unique rating algorithms by league is unworkable I think.