Monday, 8 August 2011

A factor called X

It felt a little strange watching the games last weekend and not having any financial interest in the games at all. As I've said a few times now, I'm really disciplined when it comes to the punting side of things and I won't be having any bets until the first 6 games have been played.

However, I'll admit to being a little envious when I read Cassini's blog at the weekend and saw some picks posted for the first set of European games. Reading through Cassini's blog this morning, I missed the comment Cassini had posted when he said he rolls over the form from the end of last season but he won't have much confidence in it until the first 6 games are played.

Well, there is nothing to stop me from doing the same on here. We're all adults on this blog (well, if you read some of the anonymous comments I delete, that would be debatable!) and therefore, there is nothing to stop me from trying to do something similar with the leagues I play in.

I'll stress again that I won't be backing in these games but I like having some sort of interest in the games and for someone like me, trying to find a way to roll on form between leagues and from one season to another is quite intellectually challenging but also very interesting!

So, I'll be posting up the first set of bets this weekend for a bit of fun. However, I don't want anyone to think these are anything other than a bit of fun, so I won't be putting them into the systems that we're so familiar with on here. What I'll do instead is just post the teams, the % value I think they represent and we can track these bets over the next few weeks. Of course, if things went really well, then I would think about starting next season on day one which would be pretty revolutionary when it comes to ratings and supplying bets to others (aside from Sir Cassini who is doing this already!)

Of course, I'm deliberately putting myself in a no lose situation here because if the bets all bomb over the next few weeks, I'll be taking great pleasure from the fact I'm not betting on them and not suggesting anyone else does. On the other hand, if they do amazingly well, I'll be slapping myself on the back for being able to run a ratings algorithm effectively using no live data from this season.

One thing I need to give a bit of thought to is how I deal with teams who are relegated and promoted. At the moment, I'd be having the mortgage on Swansea away to Man City I suspect this weekend looking at my ratings from the end of last season. I need to look at adjusting the teams who are promoted and relegated by a factor called X. I'll use the last 5 seasons and adjust for how promoted and relegated teams do in their first few games before their rating adjusts from the previous season's finishing position.

For the Bsq Prem games, I'll just stay clear of the newly promoted teams for a bit as I'm not about to try to interpret the lower league results by season. Not worth the hassle for a bit of fun!

I'll also try to say which teams are value for DNB too just in case DNB does really well with these bets early on in the season.

So, I'll be back in a couple of days with these bets. Just need to work out how much to reduce Swansea's value by! ;)


  1. A football manager dictates the way his team plays. If a club retains it's manager from last season then perhaps the stats will still apply. EG Spurs v Everton might be a very similar game to last year. I always apply weightings from the way certain managers set up there teams for League home and away games and European home and away games.

  2. Hi Andrew.

    Thanks for the comment.

    One thing that's very important to me throughout this footie project is that there is no human intervention here when it comes to the ratings. Everything has to be done in Excel using numbers and therefore, that's what my dilemma is. It's easy for me to say that team A has a new manager and may play differently this season but that's not how my ratings work. It has to be fully automated.

    Some people would see a fully automated ratings system as being a weakness but having been there in the past and trying to interpret ratings and form for horse-racing study, this fully automated system has much more advantages. Looking back at past season's to understand how new teams have performed as they've moved up from one league to the next is in my opinion a much better indicator of whether a team is likely to be value or not based on my pure ratings from last season.

    Of course, this doesn't mean my ratings will pick any winners or make any profits during the first 6 games of the season. I tried it last season in private and they made an overall loss and I tried it on Saturday and they made a small loss too.

    I'm seeing these first few games as a bit of fun before the real business starts in a few weeks time.



  3. Hi Graeme

    Good blog by the way.

    I just thought a good analytical exercise would be to compare the results of the matches between two teams who have retained their managers. This might just give you the edge during the first 6 games of the season.

    I used give lectures regarding Flex to engineering students. This allows you to introduce a decision engine which takes into account stats from other systems like prometrics and make knowledge based decisions.

    I have always considered why certain teams historically do well against the same teams even if the teams are placed at either end of the league table.

    After 3 years I am close but not quite there:-)

    Take a look at Flex Knowledge based systems.

    Kind regards


  4. Hi Andrew.

    Thanks for the prompt re. Flex Knowledge. Had a quick surf and read a little. Seems like something that would take a lot of time to understand fully given my starting position! :(

    In terms of your point about managers, you could well be right but it's not something that it's in my data. I don't run a database with every factor in football like some of the stats guys do. I'm very much results orientated and shots on goal is my key factor I use.

    Historical results is a factor I have in the data which is something you're getting at. However, after thinking I was onto something with this, when I started testing it live during the end of the 2009/10 season (when I was still developing the ratings), I couldn't find a statistical link in the English leagues. It worked really well in the European leagues (especially Italy!) but I couldn't see the statistical significance in the leagues I ultimately chose to settle on.

    It was actually so strong in the Italian league, you could use this as a filter to simply select the bets and you'd make a profit each season. The bookies clearly don't use this factor at all (lots of arguments why it is meaningless all over the net incidentally which shows people haven't analysed the data well enough in some leagues!) and therefore, it is very easy to find value with this factor. I can't remember the stats off by heart but I remember a couple of games where the away had only beaten the home team once in 20+ games and yet, they were favourites to win the game! That's the sort of lay that I'd like to make all day long, regardless of the relative strength of the teams.

    Good to hear like you're close to something. I feel the same here but this game has a habit of biting you just when you think you're near to something I find!



  5. Hi Graeme

    Just arrived at your blog and must say I'm very impressed especially the fact you are taking a ratings approach to football...something I am just beginning to tinker with having done so successfully with racing and golf over the last 10 years.

    I wanted a ratings approach with football to cover the time during the off-season in golf and during the jumps season in horse racing. Am looking forward to reading your blog and analysis throughout the football season and hope to learn a fair bit along the way arriving relatively late on the scene with regards to football betting so I have some back-reading to do on your blog.

    Keep up the excellent work