Sunday 29 August 2010

Specialisation


After some basically reserve teams were put out by some of the Premier League clubs in midweek’s League Cup matches, I mentioned that I had made the decision not to include these results for Premier League teams in the Elo ratings.

I received a great comment, Anonymously, which said:

It seems reasonably fair to cast aside many of those Carling Cup games. But there are still some worthy of inclusion e.g. Wigan and Sunderland. You have information on these teams - use it.

Ignoring teams with weakened teams raises another question. What do you do when teams are forced to play League games with teams weakened to various degrees by injuries and suspensions? Adjustments need to be made. You can't rely on ratings which steadfastly ignore what strength of team has been fielded.

Say, for example, Chelsea had to play their next 10 games without Drogba. As far as I can tell from what you've said about your ratings, this wouldn't show in your ratings.

It's not easily done (accurately) but there are ways and means. When he is absent you should be tweaking your workings so Chelsea's results without him are achieved with a lower Elo rating.
These are all valid ideas, but the problem is that there just isn’t the time available, and more importantly, they require adding a lot of subjectivity to the ratings. The suggestion to exclude say Blackpool’s result because they made 10 changes to their team, but include Wigan’s and Sunderland’s results means that a line needs to be drawn somewhere. What number of changes invalidates a team’s result? If it is reasonable to exclude a team after they make 10 changes, should I exclude a team making 6 changes or 4? It seems better to me to treat these matches as nothing more than friendlies.

As the poster says, I do not adjust the ratings if a star player is absent from a team for some reason. Again, it’s too subjective. No one would deny that Drogba is a key member of the Chelsea team, but if he doesn’t play, someone else replaces him so evaluating his absence becomes fraught with problems. If he is replaced by a youth team player, that’s one thing, but if Chelsea loan a £30 million player to fill in, that’s quite another.

The comment continues:
It's similar to the type of change you need to make if a team has a numerical advantage for part of a game. I presume you already do this. Playing the best part of a game with 10 men makes an enormous difference.

Likewise, what should you be doing if a team signs a player who clearly, and immediately, improves their team. You need to up their rating appropriately. Your weighting of results alone will not be enough to capture the effect (there are players who fit this bill and won't need bedding in time)
I do not adjust the ratings to account for red cards. Again, the problem is that it is subjective. One of the most misleading headlines in football is “10 Man Team Beats Other Team” and then you read the report only to find that the sending-off was after 83 minutes or something. A sending off in the 90th minute is clearly less influential than a first minute dismissal.

The approach I take to “special situations” such as if a team makes a significant change, managerial of players, or undergoes another significant change such as relegation or administration, is to continue to update the ratings objectively per the results based formula, but flag the team as “on probation” meaning no bets, or maybe small ‘interest-only’ bets.

As the writer says,
To try and do those things for all the Leagues you maintain your ratings for is a near impossible task. You would need immense knowledge of European football and there wouldn't be enough time. You would need to know every player in every League and take time analysing every match and line-up.

Which is why it's important to concentrate on Leagues you know well and can follow properly.

There will be experts in every League who help shape the markets and will take in all these factors and many more besides. Ratings purely based on results are not enough to beat the markets long-term and anyone who thinks they are is simply kidding themselves. Specialisation into certain markets and expertise and knowledge in these markets is key.
It’s hard to disagree with the statement that one needs to specialize, and anyone who chooses to bet on football markets is up against a lot of competition. There is so much press coverage of the top leagues, and the markets are so efficient, that the implied probabilities are usually pretty accurate, but it is by no means impossible to find an edge as evidenced by the success consistently achieved by Football Elite.

Here is someone who specializes in the top leagues of England, France, Germany, Italy and Spain and season after season makes a profit. I am not familiar with his selection process, but he starts with certain filters, and after further analysis the end result is that he finds selections that are value. Very few selections in fact, and perhaps that is the key right there.

For me, my ratings are the first step, with the ‘on probation’ flag the next filter, with my ‘subjective’ research the next step. Football is not my main interest on the exchanges, (specialising on less popular sports works better for me when it comes to real money), but it is a fascinating challenge. At first sight, it seems like finding value should be a simple matter. Two teams, three possible outcomes. How hard can it be?

Well, five years on, and while the losses have been staunched and last season was profitable, there’s plenty of room for improvement. Matches are not decided on paper or in the computer as results like Tottenham Hotspur 0 Wigan Athletic 1 remind us almost every week, but winning at sports betting is a marathon and not a sprint. Some days it seems that nothing goes right, yet on other days, everything follows the script. September may be close, but most leagues still need a few more rounds of matches before I have the same confidence in the ratings that I had last season.

2 comments:

Scott said...

this is an area that can be quite lucrative if you specialise in it. As you say, you're either a raw ratings man or you get heavily into team selections. Bettorlogic (.com) have a section called PlayerLogic, which measures the impact of a player by results. Not just Drogba and Torres, but also the unsung players - you can set the query yourself. Basing it on selections is even more effective at a lower level where few punters know any more than the top goalscorer.

Anonymous said...

To beat the market consistently it's imperative to take into account the other factors aside from ratings themselves.

Pure ratings can be misleading for reasons discussed. If you can integrate information into ratings as you go along then you have taken a big step forward and the ratings will be more accurate. If you don't, then the raw ratings become more and more meaningless as you ignore more and more factors.

To simply discard all information, such as red cards is folly. Of course there's a difference between a first minute sending off and one in the 90th minute. Research and analysis is required to help estimate the impact.

The comments on Drogba, with respect, are a little naive. Of course you're right in saying there's a big difference between replacing him with a youth player or a £30m superstar. But you have that information and should therefore use it. Again, research and analysis is required to estimate the impact. It can be done.

The small things make a (relatively) small difference. However, given any edge in these markets can also generally be considered to be small, then these things take on high significance and are often the difference between value and non-value prices. Yes, it can be time consuming to keep on top of them - yet another reason for specialisation.

There is also one crucial factor evident in (virtually) every game. You've kind of touched on it in a recent post. It's key to success and has been the subject of a huge amount of research over the years although I am yet to see academics quite get to the bottom of it.