Not a system or strategy exactly but rather more help requested in developing one

[happygooner]

Re: Not a system or strategy exactly but rather more help requested in developing one

sure no worries' date=' good luck with the cusp :ok[/quote'] Cheers mate, my head'll explode before I get there. Not sure if there's point in taking h2h into account at all anymore. I'm all at sea haha!!

[servecold]

Re: Not a system or strategy exactly but rather more help requested in developing one im also looking into elo ratings an ive only managed to test the spurs chelsea game so far these were the elo ratings i came up with

ELO%

64.57

82.90

52.54

i still need to find away to weight these thought about /2 as a possibility

ELO%

32.28

41.45

26.27

whats the best way to back check these result once i have a few more i will post them up as i have only kept stats from xmas on each game, as im having to punch in the data game by game

[Datapunter]

Re: Not a system or strategy exactly but rather more help requested in developing one

Doesn't seem like something that would be too quick running through the hundreds of games on Saturdays card though

Don't worry about it at all. Once you have your "strategy" worked out it usually isn't a big deal to write a supporting program, in whatever form, that can run through all the matches on a Saturday (sometimes over 600 btw :lol) in a matter of seconds. A database like from footballdata is fairly readily available. ( I happen to know of a site to be launced soon that will have all footballdata has, and loads loads more, won't be entirely free though )

[happygooner]

Re: Not a system or strategy exactly but rather more help requested in developing one

Don't worry about it at all. Once you have your "strategy" worked out it usually isn't a big deal to write a supporting program, in whatever form, that can run through all the matches on a Saturday (sometimes over 600 btw :lol) in a matter of seconds. A database like from footballdata is fairly readily available. ( I happen to know of a site to be launced soon that will have all footballdata has, and loads loads more, won't be entirely free though )

Excellent. Well I'm getting there on my criteria so after that it's just the work involved. Have laid Concordia Basel tonight.

[Datapunter]

Re: Not a system or strategy exactly but rather more help requested in developing one Just a thought here, let me ask you a question: The result of a match, what % is the result of talent, training, practise, strategy, etc... aspects that can be reasonably measured, and what % is simply down to "luck of the day", an aspect that cannot be measures or in any way predicted ? For example would you say the result of any match is 60% down to ability and 40% down to "luck". Or some other % . When you have the answer for yourself see if that changes your perception of the importance of the factors you mention. Recent home form, particulary H2H. If you assume the % "luck" is simply 0 then all stats can be considered important. If you take the position the % "luck" is actually very high then for example a H2h from years ago becomes totally meaningless.

[happygooner]

Re: Not a system or strategy exactly but rather more help requested in developing one

Just a thought here, let me ask you a question: The result of a match, what % is the result of talent, training, practise, strategy, etc... aspects that can be reasonably measured, and what % is simply down to "luck of the day", an aspect that cannot be measures or in any way predicted ? For example would you say the result of any match is 60% down to ability and 40% down to "luck". Or some other % . When you have the answer for yourself see if that changes your perception of the importance of the factors you mention. Recent home form, particulary H2H. If you assume the % "luck" is simply 0 then all stats can be considered important. If you take the position the % "luck" is actually very high then for example a H2h from years ago becomes totally meaningless.

Luck certainly plays a part otherwise you wouldn't get many surprises at all but in the selections I have been trying to narrow down to I've found a s/r of just over 90% so far from over 90 selections. This is uaing a complicated mixture of things. Home form measured against away form is the basis, followed by h2h form. However this h2h form is again only based on the home fixture for current home team. I've no idea how it's going to work in the long term but it is logical. Rad Beograd would still have been laid @ 11 and would still have won as to my mind that was a huge shock that I still can't go back and think well if I ruled this out it wouldn't have been a selection. In truth you could go back and take every loser out if you were prepared to have a sample size of 10 in 6 months but it wouldn't help much. There have been another 11ish selections at those odds which have all been successful.

[Datapunter]

Re: Not a system or strategy exactly but rather more help requested in developing one At the odds you are talking about even a couple of losers can potentially be devastating. So you need a large sample size to get any meaningfull kind of testing. Luckily for you i think it should be possible to create such a research set. If you have past results, match results, goals scored, including opening odds would that be sufficient data for a backtest ? Or is other data also involved ?

[happygooner]

Re: Not a system or strategy exactly but rather more help requested in developing one

At the odds you are talking about even a couple of losers can potentially be devastating. So you need a large sample size to get any meaningfull kind of testing. Luckily for you i think it should be possible to create such a research set. If you have past results, match results, goals scored, including opening odds would that be sufficient data for a backtest ? Or is other data also involved ?

Odds go all the way from 3 though so not many that high. Goals scored only counts as in a f/a difference in the h2h history. Everything else is based on the actual results ie h, d or a. I only have results from 23rd Feb and wouldn't know how to backtest beyond that without spending hours/days/weeks doing so.

[Kanga]

[a1ehouse]

Re: Not a system or strategy exactly but rather more help requested in developing one Kanga, was it from here? http://www.uk-neural.net/football2.html - looks a bit much even from my nerdesque perspective!! Datapunter - I like the idea of a % luck factor (would also need to include the MikeRileyQuotient!!) given its a qualitative judgment; so you could include an additional team weigthing and include in the calculation. In reality though I'd have thought yin-yang/Karma an all dat jazz would balance the figures out overall and personally would stick to just stats when developing the core system. Anything else applied on top is also down to your own luck in finding the winning formula.

[Kanga]

Re: Not a system or strategy exactly but rather more help requested in developing one

Kanga' date= was it from here? http://www.uk-neural.net/football2.html - looks a bit much even from my nerdesque perspective!!

Yep that's the page .... I tried putting the formula into excel and it works ok but I have no knowledge of automating this to run against data ... manually it would take ages.

[a1ehouse]

[Kanga]

Re: Not a system or strategy exactly but rather more help requested in developing one Nice work A1 :ok - I don't suppose you have past odds data handy to calculate profitability ? < -0.33 for aways and > 0.42 for homes ?

[Smokin Joe]

Re: Not a system or strategy exactly but rather more help requested in developing one

Don't worry about it at all. ( I happen to know of a site to be launced soon that will have all footballdata has, and loads loads more, won't be entirely free though )

DP, can you give any more details on this new site, such as when it will be available and what it is likely to contain?

[a1ehouse]

[Kanga]

Re: Not a system or strategy exactly but rather more help requested in developing one

Mid range of 42 to 92% (in bold) is actually quite profitable at 20.02 points (with average odds of 2.02). I am quite impressed, we may be on to something here!!

For Homes I make that a ROI of 22.75% :loon Looking at the Aways < -33% = 16.7% (not so many) I don't know how many bets were = -33% so couldn't calculate this from the above but it's looking good. Overall 126 bets = ROI of 21.52% :ok

[Smokin Joe]

Re: Not a system or strategy exactly but rather more help requested in developing one I ran A1ehouse and Kanga's numbers for all of the Football-data files. The results are poor. However, as a check I thought I would compare my 08/09 EPL numbers with what is being quoted here. I used the best odds from approx 10 books according to football-data. I back the home team if Genex > 0.42 and back the away team if Genex <-0.33 I had 108 home bets, losing £7.19 (£1 level stakes bet), and had 31 away bets winning £4.77. Those results seem quite different to a1ehouse's, so to check my figures a1ehouse can you confirm that Wigan v Hull on 22nd March had a Genex rating of 0.576, with Man City v Sunderland on the same time having a prediction of 0.567?

[Datapunter]

Re: Not a system or strategy exactly but rather more help requested in developing one

DP, can you give any more details on this new site, such as when it will be available and what it is likely to contain?

Similar datasets as you can find on footballdata but, datasets updated at least daily, with much more details, in different file formats like csv, xml, mysql, and multiple sports. Should be launched sometime in April with Tennis, followed by Horse Racing then sometime during the summer the football section ready for next season. Won't be free but pricing will be affordable based on the principle of the more you buy the less you pay. Something like ATP results, or results for all GB based football competitions will be around 30-40pounds "once only" with an update for a full season at 5 pounds, somewhere in that price region anyhow. Can't promise anything just yet but PL members will most likely be able to get a discount right away. That's all i can say for now, to be continued... :lol

[Smokin Joe]

Re: Not a system or strategy exactly but rather more help requested in developing one So I guess you know the person who will be running it then :)

[Datapunter]

Re: Not a system or strategy exactly but rather more help requested in developing one I should hope so, i've been working on it for months :lol

[a1ehouse]

Re: Not a system or strategy exactly but rather more help requested in developing one

I ran A1ehouse and Kanga's numbers for all of the Football-data files. The results are poor. However, as a check I thought I would compare my 08/09 EPL numbers with what is being quoted here. I used the best odds from approx 10 books according to football-data. I back the home team if Genex > 0.42 and back the away team if Genex <-0.33 I had 108 home bets, losing £7.19 (£1 level stakes bet), and had 31 away bets winning £4.77. Those results seem quite different to a1ehouse's, so to check my figures a1ehouse can you confirm that Wigan v Hull on 22nd March had a Genex rating of 0.576, with Man City v Sunderland on the same time having a prediction of 0.567?

My figures are for the EPL 0809 season, so I will double check later today - in the middle of DIY'in at the mo, but this is what my db pulls out:

GeneXPro

SeasonID	Div	Division	Date	HomeTeam	AwayTeam	FTHG	FTAG	FTR	GeneX
0809	E0	Premier League	22/03/2009	Man City	Sunderland	1	0	H	64.29%
0809	E0	Premier League	22/03/2009	Wigan	Hull	1	0	H	63.23%

Points per game: Wigan: 1.37 Hull: 1.10 1.37-1.10 = 0.27 Man City: 1.27 Sunderland: 1.07 1.27-1.07 = 0.20 Goals per game: Wigan: 1.00 Hull: 1.17 Man City: 1.53 Sunderland: 0.97

[Smokin Joe]

Re: Not a system or strategy exactly but rather more help requested in developing one

Points per game: Wigan: 1.37 Hull: 1.10 1.37-1.10 = 0.27 Man City: 1.27 Sunderland: 1.07 1.27-1.07 = 0.20 Goals per game: Wigan: 1.00 Hull: 1.17 Man City: 1.53 Sunderland: 0.97

Mine are a bit different: Points per game Man City - 1.21 Sunderland - 1.10 Wigan - 1.31 Hull - 1.14 I wonder why we are different? I have realised that my goals per game calc is incorrect (I just took the average goals of the home team at home), so I will need to look at that again.

[a1ehouse]

[muppet77]

Re: Not a system or strategy exactly but rather more help requested in developing one just for anyone trying to follow this, my version of Excel has the arctangent function as ATAN and not ATN as in the equation used earlier. swapping all the ATNs for ATAN works fine. :ok

[a1ehouse]

[muppet77]

Re: Not a system or strategy exactly but rather more help requested in developing one don't backtest too far buddy as surely the original equation was based on the data that you will be backtesting with? just a thought. (i don't know too much about the eqn and its history so you may be fine)

[a1ehouse]

Re: Not a system or strategy exactly but rather more help requested in developing one

don't backtest too far buddy as surely the original equation was based on the data that you will be backtesting with? just a thought. (i don't know too much about the eqn and its history so you may be fine)

True, think you're right mate, no point in reinventing the wheel, but would also like to compare it across divisions/countries/my own ratings model. Next job is to get the odds appended into my db. Anyone done it already or got a function to do it!? The footydata files are all over the place which makes automation a nightmare and why I've put it on the back burner...

[Kanga]

Re: Not a system or strategy exactly but rather more help requested in developing one Wow, you guys are great - I remember the ATN / ATAN problem I had at the time now you mention it Muppet :loon.

[Smokin Joe]

Re: Not a system or strategy exactly but rather more help requested in developing one A1ehouse, can you confirm what the league points were for the 2 examples I gave earlier, I want to see if I have my formula correct?

[a1ehouse]

Re: Not a system or strategy exactly but rather more help requested in developing one

A1ehouse' date=' can you confirm what the league points were for the 2 examples I gave earlier, I want to see if I have my formula correct?[/quote']

	Games	Gls	Pts	Pts/Game	Gls/Game
Man City	30	46	38	1.27	1.53
Sunderland	30	29	32	1.07	0.97
Wigan	30	30	41	1.37	1.00
Hull	30	35	33	1.10	1.17