Predicting The English Premier League-using stats

[Matthew]

Some of you will remember me from last year.....

Last years thread (Predicting The English Premier League-using stats-Strike Rate 57.2% - Yield 16.9%)

I have calculated the following raft of stats for every match since 2000-2001 season:

Variable Name	Description	Variable Level
Date	Date On Which The Outcome Match Took Place
Home Team	Name Of The Home Team	Outcome Info
Home Score	Goals Scored By Home Team	Outcome Info
Away Score	Goals Scored By Away Team	Outcome Info
Away Team	Name of the Away Team	Outcome Info
Result	Result Of The Match	Outcome Info
LeaguePos_H	League Position	Home Team
TotWon_H	%age Of Games Won To Date	Home Team
TotDrawn_H	%age Of Games Drawn To Date	Home Team
TotLost_H	%age Of Games Lost To Date	Home Team
TotGFPG_H	Goals For Per Game, Season To Date	Home Team
TotGAPG_H	Goals Against Per Game, Season To Date	Home Team
TotPPG_H	Points Per Game, Season To Date	Home Team
HomeWon	%age Of Home Games Won To Date	Home Team
HomeDrawn	%age Of Home Games Drawn To Date	Home Team
HomeLost	%age Of Home Games Lost To Date	Home Team
HomeGFPG	Goals For Per Game, At Home	Home Team
HomeGAPG	Goals Against Per Game, At Home	Home Team
HomePPG	Points Per Game, At Home	Home Team
WL8_H	%age Last 8 Games Won	Home Team
DL8_H	%age Last 8 Games Drawn	Home Team
LL8_H	%age Last 8 Games Lost	Home Team
WL5H_H	%age Last 5 Games at Home - Won	Home Team
DL5H_H	%age Last 5 Games at Home - Drawn	Home Team
LL5H_H	%age Last 5 Games at Home - Lost	Home Team
GIN2W_H	Number Of Games Played In The Next Two Weeks	Home Team
GIP2W_H	Number Of Games Played In The Past Two Weeks	Home Team
S-1_POS_H	Finishing Position In The League Last Season	Home Team
S-2_POS_H	Finishing Position In The League Season Before Last	Home Team
NSH_Home	%age Of Home Games When Team Not Scored	Home Team
NSH_Tot	%age Of Games, Season To Date, When Team Not Scored	Home Team
CSH_Home	%age Of Home Games When Team Not Kept A Clean Sheet	Home Team
CSH_Tot	%age Of Games, Season To Date, When Team Not Kept A Clean Sheet	Home Team
Promo_H1	Was This Team Promoted To This Division Last Season	Home Team
Promo_H2	Was This Team Promoted To This Division Season Before Last	Home Team
LeaguePos_A	League Position	Away Team
TotWon_A	%age Of Games Won To Date	Away Team
TotDrawn_A	%age Of Games Drawn To Date	Away Team
TotLost_A	%age Of Games Lost To Date	Away Team
TotGFPG_A	Goals For Per Game, Season To Date	Away Team
TotGAPG_A	Goals Against Per Game, Season To Date	Away Team
TotPPG_A	Points Per Game, Season To Date	Away Team
AwayWon	%age Of Away Games Won To Date	Away Team
AwayDrawn	%age Of Away Games Drawn To Date	Away Team
AwayLost	%age Of Away Games Lost To Date	Away Team
AwayGFPG	Goals For Per Game, At Away	Away Team
AwayGAPG	Goals Against Per Game, At Away	Away Team
AwayPPG	Points Per Game, At Away	Away Team
WL8_A	%age Last 8 Games Won	Away Team
DL8_A	%age Last 8 Games Drawn	Away Team
LL8_A	%age Last 8 Games Lost	Away Team
WL5H_A	%age Last 5 Games at Away - Won	Away Team
DL5H_A	%age Last 5 Games at Away - Drawn	Away Team
LL5H_A	%age Last 5 Games at Away - Lost	Away Team
GIN2W_A	Number Of Games Played In The Next Two Weeks	Away Team
GIP2W_A	Number Of Games Played In The Past Two Weeks	Away Team
S-1_POS_A	Finishing Position In The League Last Season	Away Team
S-2_POS_A	Finishing Position In The League Season Before Last	Away Team
NSA_Away	%age Of Away Games When Team Not Scored	Away Team
NSA_Tot	%age Of Games, Season To Date, When Team Not Scored	Away Team
CSA_Away	%age Of Away Games When Team Not Kept A Clean Sheet	Away Team
CSA_Tot	%age Of Games, Season To Date, When Team Not Kept A Clean Sheet	Away Team
Promo_A1	Was This Team Promoted To This Division Last Season	Away Team
Promo_A2	Was This Team Promoted To This Division Season Before Last	Away Team
ResLS	Result Of The Match Last Season	Matchup Stats
ResSBL	Result Of The Match Season Before Last	Matchup Stats
RevResLS	Result Of The Reverse Fixture Last Season	Matchup Stats
RevResSBL	Result Of The Reverse Fixture Season Before Last	Matchup Stats
RevResTS	Result Of The Reverse Fixture This Season	Matchup Stats
System 13 Rating (A)	This Is A System Taken From A Book. It Works By Figuring Out What The Result Should Have Been Based On The Numbers Of Shots (On And Off Target) And The No. Of Corners. This Then Builds A Table Based On The "True" Results And Compares Teams Points Per Game Measure. The Five Different Measures Reflect Slightly Different Points Strategies. Eg. 3Pts For A Win At Home - 5Pts For A Win Away From Home. This System Was The Best One Highlighted In The Book For Predicting Home And Away Wins.	Comparison
System 13 Rating (B)		Comparison
System 13 Rating (C)		Comparison
System 13 Rating (D)		Comparison
System 13 Rating (E)		Comparison
System 9 Rating (A)	From the very same book, this system was the best at predicting draws. Its in 5 separate variables for the same reason as above. This system merely compares league positions based on leagues built on the actual results but with differing points strategies.	Comparison
System 9 Rating (B)		Comparison
System 9 Rating (C)		Comparison
System 9 Rating (D)		Comparison
System 9 Rating (E)		Comparison
h_win_climb_places	How Many Places The Home Team Will Climb If They Win	Home Team
h_win_into_champs	Will A Win Get The Home Team In To A Champions League Position	Home Team
h_win_into_uefa	Will A Win Get The Home Team In To A Uefa Cup Position	Home Team
h_win_out_of_rel	Will A Win Get The Home Team Out Of The Relegation Positions	Home Team
a_win_climb_places	How Many Places The Away Team Will Climb If They Win	Away Team
a_win_into_champs	Will A Win Get The Away Team In To A Champions League Position	Away Team
a_win_into_uefa	Will A Win Get The Away Team In To A Uefa Cup Position	Away Team
a_win_out_of_rel	Will A Win Get The Away Team Out Of The Relegation Positions	Away Team
Gamebookers Home Win Odds		Bookmakers
Gamebookers Draw Odds		Bookmakers
Gamebookers Away Win Odds		Bookmakers
Interwetten Home Win Odds		Bookmakers
Interwetten Draw Odds		Bookmakers
Interwetten Away Win Odds		Bookmakers
Ladbrokes Home Win Odds		Bookmakers
Ladbrokes Draw Odds		Bookmakers
Ladbrokes Away Win Odds		Bookmakers
SportingBet Home Win Odds		Bookmakers
SportingBet Draw Odds		Bookmakers
SportingBet Away Win Odds		Bookmakers
William Hill Home Win Odds		Bookmakers
William Hill Draw Odds		Bookmakers
William Hill Away Win Odds		Bookmakers

I then put this data through a statistical modelling package called KXEN (www.kxen.com).

The 4 seasons 2000-2004 are used to build the model, 2004-2006 is used to validate the model and last season (2006-07) is used as a blind test to see how the model would have worked had I used it.

Once the best model to describe the data is built, Kxen "to's and fro's" between the best model and the validation sample to come about a model that best describes the data yet is robust enough that it can be used in the real world, a tradeoff.

The output from this model is a %age chance of HW / D / AW.

In preparation for this upcoming season I have played around with the cutoffs which qualify a bet (the %age diff between my assessment and that of the bookie) to retrospectively fit the best possible result for last season. This worked well last year when I did it midway through the season and then continued with these cutoffs.

So, had I used this model last year and used the cutoffs I found (obviously self-fulfilling) I would have observed something in the order of this:

	TOTAL	HW	D	AW
Stake	£281.00	£74.00	£105.00	£102.00
Return	£377.95	£88.87	£160.95	£128.13
Profit	£96.95	£14.87	£55.95	£26.13
SR	53.4%	77.0%	43.8%	46.1%
Yield (%)	34.5%	20.1%	53.3%	25.6%
HW Odds:	1.62
D Odds:	3.45
AW Odds:	2.97

This is based on level stakes £1 per selection.


With profit arriving like this:


This is the model I am using going fwd against the old model from last year, again retrospectively fit to achieve the best possible result.

As you can see the new model drives greater profit from fewer bets which is an obvious increase in yield(%).

I'm intending to use %age staking to the equivalent of 4% of my bank.

My bank currently stands at £1,050 (me and 20 mates, £50 each). I am expecting 2 more to send me the cash so by the start of the season my bank should be £1,150.

Here's to as good a season as last season .

Any thoughts? Questions?

I'd welcome your comments as I am always looking to improve things.

Matt

[Matthew]

Date Home Away Prediction Odds Bookie Stake Correct Return 11/08/2007 Everton Wigan D 3.7 Betfair £42 0 11/08/2007 Sunderland Tottenham D 3.55 Betfair £42 0

[muppet77]

good luck plop. this looks really exciting!

[muppet77]

quick question mate - do you have it fully automated now
ie. you just fill in the date, teams, score, shots etc
and it churns out a 1x2 %?
do you do this in excel?

[barrhead73]

Glad to see you back for another season Plop.

Good luck.

[LaRoz]

Good Luck

Looking forward to seeing plenty of these

[Matthew]

Thanks all.

It's all about the !!!!!


Muppet77,

"quick question mate - do you have it fully automated now
ie. you just fill in the date, teams, score, shots etc
and it churns out a 1x2 %?
do you do this in excel?"

It's fully automated in the sense that I update scores and shots etc (via a download of the latest EPL file from football-data), then its just a few macros