Tips Dec 17 2011

Back Sporting Gijon *4
Lay Hearts *9
Back Hartlepool-Colchester Draw *4
Back Freiburg *3
Back Augsburg *2
Back Hertha Berlin *2
Back Wolfsburg *1
Lay Schalke *6

[Edit] All but one of these bets lost. -26

Tips 13 Dec 2011

Genoa *4

[Edit] Lost; -4. These Serie A bets are going badly.

Tips 12 Dec 2011

Roma *4

[Edit] Lost: -4

Tips 11 Dec 2011

Espanyol * 5
Zaragoza * 2
Bologna * 6 (?!)
Cagliari * 3

[Edit] Espanyol won, the rest lost. -3

Tips 10 December 2011

Back Siena *4

[Edit] Lost. -4

Tip, 27 Nov 2011

Serie A: Cesena v Genoa - back Cesena *3.

(First tip under the new system, it's quite exciting)

[Edit] Result - win! +5

Filtering

Now I'm filtering aggressively, on both team statistics (e.g. only consider games where the away team has a very low average number of corners won per game) and on odds (e.g. only consider games where the odds for the home team are greater than 5.0).

This means I will be placing far fewer bets. My tip for Saturday is: place no bets.

September 2011 Analysis

My total for September 2011 is -22. The Championship was profitable for me, but Prem, League Two and Bundesliga all made a loss.

There were at least three bugs in my program. It looks like one of these skewed the results of my analysis of previous seasons, making some leagues look more profitable than they really were. In particular League Two seemed to be profitable in previous seasons, but actually made me a large loss this season. The second was an inaccuracy in estimating lay prices from back prices, again causing me to overestimate profit predictions. The third bug was a difference between the analysis of previous seasons and predictions for the current season, meaning that I made a few bets that I shouldn't have made according to my own model, which lost.

Having fixed these bugs, I need to rerun all my analyses of previous seasons. At the same time I'm going to test some more subtle strategies for picking games and for betting.

Predictions 27 Sept

Posting after the fact, but these were the bets I made:

Lay Brighton *3 (backer's stake)
Lay Reading *2.5
Lay Blackpool *3
Lay Derby *2.5
Lay Hull *1.5

The Reading lay lost (right at the death), but the rest won. +5 today.

Predictions 24 Sept 2011

Lay Arsenal *5 (backer's stake)
Lay Chelsea *14
Lay Liverpool *5
Lay Man City *6
Lay Man Utd *8
Lay Tottenham *4

[Edit] Results: only the Man Utd bet won. Total -7: not good.

Predictions 18 Sept 2011

Value bets today:

Fulham v Man City - Fulham or draw [Edit: bet wins]
Man Utd v Chelsea - Chelsea or draw [Edit: bet loses]
Southampton v Birmingham - Birmingham or draw [Edit: bet loses]
Hannover v Dortmund - Hannover or draw [Edit: bet wins]
Schalke v Bayern Munich - Schalke or draw [Edit: bet loses]

[Edit] Only two out of five correct, but because of differing stake sizes, I made a small profit.

Predictions for 13 September 2011

Some League Two predictions:

Macclesfield v Morecambe: Home or draw [Edit: Result - draw]
Hereford v Aldershot: Home or draw [Edit: Result - away win]
Crawley v Swindon: Away or draw [Edit: Result - away win]
Barnet v Plymouth: Away or draw [Edit: Result - home win]

[Edit] Two bets won and two lost. For me, a small loss.

Predictions 11 September 2011

Koln v Nurnburg: Lay Koln.
Wolfsburg v Schalke: No bets.

[Edit] Result: Nurnburg; bet wins. For me this recouped about half of the losses from yesterday. Still cash negative overall.

Predictions 10 September 2011

Today I am laying the home team in each of the six Bundesliga matches.

[11 September] Four home wins out of six. Cash negative. I suspect the unusual run of away wins in August 2010 has skewed the training data. Again, I find myself wondering how much training data is the "right" amount to use.

2011-2012 predictions

2011-2012 predictions start when teams have played two home games and two away games each. So a few weeks away yet.

Scibet

I hadn't previously noticed that http://www.scibet.com has an archive of their predictions for previous seasons. At least I think that's what the "Results" pages are. They're not labelled very well. Anyway, I investigated a bit, and found that my own method makes more money, except in the last quarter of the season, where Scibet's average and variance are both higher. Maybe English leagues are a weakness for them.

Kelly shortcomings

The Kelly bet isn't perfectly suited to football betting. It gives the optimum stake size when making repeated bets at the same odds and probability, but this is not the situation in football betting. And it can't help when placing bets on several markets at the same time either.

Kelly bug

There was a bug in my Kelly calculations! Effectively I was using the estimated probability as the stake size instead of the proper Kelly calculation. Hopefully when I retest with the correct calculation the results should be better.

Unfortunately I think this is a symptom of being forced to work on my system in very short bursts and thus lacking time to refactor and fix bugs. I wonder how many bugs remain...

Divide and conquer

I've been experimenting with splitting up the season into sections and treating them separately, with some success. The first half of each season seems pretty intractable though.

Resetting every season

I think resetting the model every season was perhaps a mistake. It makes it too sensitive to periods of unusual results. Now that I have data from several seasons, it's more feasible to experiment with using the last few seasons for prediction.

But I still think using too many past seasons should be avoided. My feeling is that the game changes pretty quickly, so the factors which will predict results will also change pretty quickly.

Past seasons

After tweaking my model for the 2009-2010 and 2010-2011 seasons, I tried it out on previous seasons. It showed mixed results:

2003-2004 reasonable profits
2004-2005 good profits
2005-2006 big losses
2006-2007 reasonable profits
2007-2008 small losses
2008-2009 good profits
2009-2010 huge profits
2010-2011 big profits

I need to decide how much these results matter to me. Of course I would have been happier if every season had showed a profit, but the 2005-2006 season was five years ago, which is a long time.

What data is valid?

How should I decide what data to include in my analysis?

Is there most similarity between games in a particular season? A particular league? A particular team?

Is there more similarity between a 2009-2010 Premier League match and a 2010-2011 Premier League match, or between a 2010-2011 Ligue 1 match and a 2010-2011 Premier League match, or between a 2010-2011 Premier League Chelsea match and a 2010-2011 FA Cup Chelsea match?

So far I have been analyzing data by season. I reset my analysis at the beginning of each season. But within each season, all teams contribute to the model. Is it really reasonable for a Blackburn-Bolton match to affect the model I use to predict a Spurs-Fulham match in the same season?

Positive profit

I've found a method that shows profit for both the 2009-2010 and the 2010-2011 seasons. Unfortunately the variance is still pretty high - at one point in the 2010-2011 season, the account shows a large net loss, before recovering to show profit. So now I want to reduce the variance.

It's hard to know whether two seasons of data is sufficient. On one hand, it's not a lot of games, making it hard to separate real trends from random noise. On the other hand, it seems likely that the sport of football changes quickly, and that older seasons are less similar to the upcoming season than recent seasons are. So perhaps it's not wise to analyze too many past seasons.

Predicting football

I'm a software engineer with interests in statistics and betting. I've been working on a personal project to predict English Premier League football results using AI. I've been working on data for the last two seasons and developing algorithms to decide when and how to bet. If I can develop the system to the point where it shows a profit for the last two seasons, I'll run it live on the upcoming 2011-2012 season, posting the predictions and results here.