Pullein Power

[Vilamoura04]

Re: Match Ratings -snappy title coming soon I know, grex. I know :$ The St Mirren Aberdeen game is a league match though :loon But it is a midweek league match :\

[Grex]

Re: Match Ratings -snappy title coming soon From my back-testing, I've found nothing unusual about mid-week games.... but I have to admit I've never analysed Scottish leagues (a bit strange, coming from a Scot ;)). Anyway, I've decided to take the plunge and put a few quid on Aberdeen at AH(0)..... kiss of death, I know. BTW, we're still waiting for the "snappy title".... any thoughts ?

[Vilamoura04]

Re: Match Ratings -snappy title coming soon How about "Pullein it off" :eek I'll never get a job on the tabloids will I

[Grex]

Re: Pullein Power OK, I'll be the first ignoramus to ask the question......"Pullein Power" ???? :unsure:unsure:unsure

[Vilamoura04]

Re: Pullein Power *shrugs* Kevin Pulleins system :$

[lunatism]

Re: Match Ratings -snappy title coming soon

A friendly word of advice....... for the reasons you allude to in your post' date= steer clear of cup games ;)

seconded - cup games bring a whole new set of variables to the fore, and unless you've superbacktested (hmmm, superbacktest? erm....let's say years and years worth of data), then it's something worth sidestepping. That actually brings something else to mind - have you thought about the effects of derbies and rivalries on your matches? Because from personal analysis (i'll find the exact research I did one day and start a thread on it), i've found that in the UK, there is roundabout a 30% decrease in reliabilty when there is an element of rivalry in the match. By 30% decrease in reliabilty (using bookies odds as a base), I mean that where the favourite odds are less than or equal to 60% (1.66), the favourite wins 67% of the time if there is no element of rivalry - include fixtures that DO contain elements of rivalry, and the favourite only wins 38% of the time. I'd go into more detail, but that's just from some basic notes i'd jotted down to keep in mind when planning out new system ideas... If you like, just send me all your backtested fixtures in the simple format HOME ¦ AWAY ¦ RESULT and then i'll run tests on your data to see how it is affected (and if you used the data I sent you before - all the better, since all the team names will already be in my database) :ok

[Grex]

Re: Pullein Power

Kevin Pulleins system :$

Thanks for filling me in..... excuse my ignorance :$

[Grex]

Re: Match Ratings -snappy title coming soon

Anyway' date= I've decided to take the plunge and put a few quid on Aberdeen at AH(0)..... kiss of death, I know.

I did warn you :sad

[Vilamoura04]

Re: Match Ratings -snappy title coming soon

seconded - cup games bring a whole new set of variables to the fore' date= and unless you've superbacktested (hmmm, superbacktest? erm....let's say years and years worth of data), then it's something worth sidestepping. That actually brings something else to mind - have you thought about the effects of derbies and rivalries on your matches? Because from personal analysis (i'll find the exact research I did one day and start a thread on it), i've found that in the UK, there is roundabout a 30% decrease in reliabilty when there is an element of rivalry in the match. By 30% decrease in reliabilty (using bookies odds as a base), I mean that where the favourite odds are less than or equal to 60% (1.66), the favourite wins 67% of the time if there is no element of rivalry - include fixtures that DO contain elements of rivalry, and the favourite only wins 38% of the time. I'd go into more detail, but that's just from some basic notes i'd jotted down to keep in mind when planning out new system ideas... If you like, just send me all your backtested fixtures in the simple format HOME ¦ AWAY ¦ RESULT and then i'll run tests on your data to see how it is affected (and if you used the data I sent you before - all the better, since all the team names will already be in my database) :ok

Hi lunatism. There isnt anything in the system per se for derby games, I had intended to add some interpretation of the selections of the system as we go. I agree that the derby 'factor' is important and I will defintely consider this when running the model, thanks for the prompt. The research I have read on derby games came to the same conclusion as you, that statistically the favourites perform much worse under derby conditions. But also that didnt mean the underdog won more often, it was that draws came up more often. In fact I think they found that blanket betting draws in derby games to level stakes was a viable medium to long term strategy.

[lunatism]

Re: Match Ratings -snappy title coming soon

Hi lunatism. There isnt anything in the system per se for derby games, I had intended to add some interpretation of the selections of the system as we go. I agree that the derby 'factor' is important and I will defintely consider this when running the model, thanks for the prompt. The research I have read on derby games came to the same conclusion as you, that statistically the favourites perform much worse under derby conditions. But also that didnt mean the underdog won more often, it was that draws came up more often. In fact I think they found that blanket betting draws in derby games to level stakes was a viable medium to long term strategy.

Shame that i can't seem to locate the spreadsheets i'd worked on, because i'd be able to go into detail about actual results, rather than simply being able to say that the favourites underperformed. Off the top of my head though, i'm sure your point is valid - underperforming favourites usually would result in a draw rather than the away win.... But with that in mind, perhaps increasing the draw% a couple of notches if the match was a rivalry match would reap dividends, and save lost money on backing the favourite? Although it might not seem immediately relevant (I assure you, it is), by chance are you a Football Manager owner/player?

[Rushian]

Re: Match Ratings -snappy title coming soon Vilamoura04, first of all good luck with your system! I see your point regarding your decision to choose the most likely outcome assuming the odds represent value according to your system. Betting solely on value may prove (even more) profitable in the long-run than your strategy but one also has to consider money management issues and the risk of bankruptcy. If the odds are too long, a long losing run is possible and the probability of ruin (if your bank is not large enough to accommodate these potential consequent losses) is very real. By choosing the most likely outcome, you may reduce your long-run yield but you also reduce the probability of going bust.... I think! :unsure

seconded - cup games bring a whole new set of variables to the fore, and unless you've superbacktested (hmmm, superbacktest? erm....let's say years and years worth of data), then it's something worth sidestepping. That actually brings something else to mind - have you thought about the effects of derbies and rivalries on your matches? Because from personal analysis (i'll find the exact research I did one day and start a thread on it), i've found that in the UK, there is roundabout a 30% decrease in reliabilty when there is an element of rivalry in the match. By 30% decrease in reliabilty (using bookies odds as a base), I mean that where the favourite odds are less than or equal to 60% (1.66), the favourite wins 67% of the time if there is no element of rivalry - include fixtures that DO contain elements of rivalry, and the favourite only wins 38% of the time. I'd go into more detail, but that's just from some basic notes i'd jotted down to keep in mind when planning out new system ideas... If you like, just send me all your backtested fixtures in the simple format HOME ¦ AWAY ¦ RESULT and then i'll run tests on your data to see how it is affected (and if you used the data I sent you before - all the better, since all the team names will already be in my database) :ok

lunatism, I'd be interested to learn a bit more about your findings. First of all did you consider only local derbies or did you also include more general types of rivalries? And if it was the latter, how did you decide on what constituted a rivalry and what not? Would you be willing to share the lst of derbies that you have considered? The reason I'm asking is that I also had a go at this by introducing a "derby variable" in addition to my ratings in my probability model. Unfortunately, this variable did not prove to be statistically significant. Whether this was down to only a small share of matches being local derbies in my total sample size (don't have the figures now, but I think I had only around 3-4% as derby matches) or whether the effect was already explained the other explanatory variables (i.e. my ratings), I'm not too sure. Intuitively, I would also have expected the derby factor to be a "leveller" so that favourites are pulled towards evens and the odds of longer-priced teams to come down a few ticks. Unfortunately I have not proved anything like this for now.

[Vilamoura04]

[Vilamoura04]

Re: Match Ratings -snappy title coming soon

Shame that i can't seem to locate the spreadsheets i'd worked on, because i'd be able to go into detail about actual results, rather than simply being able to say that the favourites underperformed. Off the top of my head though, i'm sure your point is valid - underperforming favourites usually would result in a draw rather than the away win.... But with that in mind, perhaps increasing the draw% a couple of notches if the match was a rivalry match would reap dividends, and save lost money on backing the favourite? Although it might not seem immediately relevant (I assure you, it is), by chance are you a Football Manager owner/player?

Ive found one of the articles I read on derbies that back up your findings http://www.guardian.co.uk/football/2006/mar/24/newsstory.sport11 Im sure theres more though I will hunt them out I hope this will be of interest to you also rushian Yes Ive played football manager for decades now :$ Go on why do you ask?

[Vilamoura04]

Re: Pullein Power Another one http://www.thefreelibrary.com/Football+Trends:+Kevin+Pullein:+Honours+often+shared+when+pride+at...-a066402579 There are some non Kevin Pullein ones too though I must remember where I found them

[lunatism]

Re: Match Ratings -snappy title coming soon

First of all did you consider only local derbies or did you also include more general types of rivalries? And if it was the latter, how did you decide on what constituted a rivalry and what not? Would you be willing to share the lst of derbies that you have considered? The reason I'm asking is that I also had a go at this by introducing a "derby variable" in addition to my ratings in my probability model. Unfortunately, this variable did not prove to be statistically significant. Whether this was down to only a small share of matches being local derbies in my total sample size (don't have the figures now, but I think I had only around 3-4% as derby matches) or whether the effect was already explained the other explanatory variables (i.e. my ratings), I'm not too sure. Intuitively, I would also have expected the derby factor to be a "leveller" so that favourites are pulled towards evens and the odds of longer-priced teams to come down a few ticks. Unfortunately I have not proved anything like this for now.

Hey Rushian, it's been a while, huh? Anyways....you might have noticed that I was deliberately not using the word derbies, and rather going along with rivalries in my last few posts? The reason being that many a night i've sat and pondered what matter created an "issue" between two teams, and realised it was simply something I could not deal with:

• I thought about distance (so now I have the longitude and latitude of every team in my database so that I can calculate distance between the two teams), but then you would get the whole issue with countries with spread apart teams (like Scotland), where most teams are really far apart, yet there are still rivalries
• I thought about past players/managers (now now i have past/present complete player lists for all the teams in my database), but then what happens when a player leaves on good terms (Silvestre to Arsenal) or when that player was jsut not important to care about enough? Or what if the fans were glad to be rid of the player, or simply hate the player rather than his team (perhaps Adebayor to Man City)
• I thought about past clashes (so i maintain a record of all high yielding red/yellow card matches), but sometimes the refs are just dashing them out for fun....and sometimes teams are just plain dirty, or the surfaces aren't decent
• I considered ground sharing (so i have a record of that too), but that doesn't always work out as an almighty rivalvry too

And with all this data, I still was unable to assign any kind of factor/percentage to the apparant rivalry...and exactly how long does it take for a rivalry to dissipate? (don't ask me, haven't a clue)....So i was ready to give up the idea, because even if i manually did all the calculating, there would be some teams that I couldn't find any info about that absolutely hated each others guts (eg, look at the Russian Premier league, they have at least 5 teams from Moscow, do they all hate each other? nope, just a select few with animosity against each other), so it would take someone with expertise in each league (or at least country) to do this properly.... (long story huh fellas?...) then one day, as i was playing Football Manager, i had some problem with the game and went onto their forum, where i saw an advert for league scouts - they wanted folks to gather as much info as possible on teams in their own countries (finances, players, backroom staff, etc - i'm sure you all know the stuff), and i noticed RIVALRIES was something else they were keen on. Then it hit me, i've SEEN rivalry percentages before - in the FOOTBALL MANAGER EDITOR :cow:cow:cow Mammoth task ahead of me - definately. Was I deterred? Nope... For several weeks i manually wrote out all the rivalry rating for every team in every league that I have in my database, and then all the teams in the leagues below my leagues just in case somebody got promoted :) And so I basically have percentage ratings for all the "rivalry" fixtures for the leagues I cover....cool, huh? :loon So now when I compile my daily fixture info, i always have a "rivalry percentage" tagged on too - it helps a LOT, because obviously one can't be expected to know every single rivalry, and it means that the whole thing can be automated in terms of eliminating games that have even a sniff of hate about them. Now, i'm sure that after reading this, folks would understand my hestitance at handing out this info - since it took forever to put together. But hopefully, by making people aware that it can be done - a kinder soul than I would post up the info (but it's probably something that shouldn't be passed around too much, otherwise maybe the FM people will decide to make rivalry details unavailable in future editions (they probably wouldn't...but i'm honestly worried that they could :()).

[Rushian]

Re: Pullein Power lunatism, your dedication never seizes to impress me! WOW :loon I don't know how accurate that info would be, but I guess it won't be a million miles away from truth, as FM is very well researched if I understand correctly. I'm also positive that having a percentage value rather than a binary 0/1 (like I had) would provide more information to the model. I'm off now, but I'll give this some more thought. P.S. Indeed it's been a while! I'm not on here as often as before as I'm involved in some additional things too, but it's always nice to occasionally pop in. :ok

[lunatism]

Re: Pullein Power

I don't know how accurate that info would be' date=' but I guess it won't be a million miles away from truth, as FM is very well researched[/quote'] I had the same doubts, but I bypassed them with the same logic - if FM is so realistic, their rivalry ratings must be too. But I wasn't happy just to take my word for it, when i was in Madrid, I made a lot of international friends - in fact, most of the leagues that i'm currently working with, so I went over almost every rivalry fixture with my pals, just to check (luckily they all had a smattering of an understanding of football) and happily found only a handful of inconsistencies, which were corrected. So i'd say it's at least 95% accurate (on my mates say so)

[lunatism]

Re: Match Ratings -snappy title coming soon

how much can you adjust the ratings/selections to account for team news?

first off - my theory is that quality will always show through in the long term, so short term, sure, you'll make your losses down to team selections, but more often than not, that extra "quality" will help your favourite to win, but if you're the proactive sort (as I am), you'll not sit back and hope.... But it's not really an easy thing to deal with is it? I'm currently in-between a long term solution to this problem. I've written some stuff in excel to extract player info for each team, and then i will maintain a player database of appearances, goals, cards, etc and then these will be manipulated to create a "player worth" grade for each player in the team. And then in turn a % for the team in terms of players of high value being available etc (so that I could say that Man UTD are currently at 50%). But maybe now you would say that Man UTD are missing Ferdinand, who's hardly contributed at all this season, yet we all know they are missing him dearly - so i'd also factor in the actual value of a player (cost), and perhaps wage demands too (because a club would obviously pay their most dear performers a lot more than the dross keeping the bench warm).... Of course, this is all in the ideas stage at the moment, but I really WANT to do it. Because I hate getting stung by so called big clubs not putting out their best and then getting beat by the lesser fellows. Obviously, this wouldn't eliminate all the nonsense results, but i'm quite sure it would make a HUGE difference - especially since it would be an objective rather than subjective decision ....the alternative is to manually check the team news for each fixture for both home and away teams, and with the numbers of fixtures i'm dealing with, i simply can't. Plus, think of being able to look back over the course of the season with a query like: "home team that is less than 50% strength" and being able to churn out the results...most likely this could turn up some very "unexpected" high priced winners (well, that's what i'll be hoping for anyways) :lol

[Vilamoura04]

Re: Pullein Power :eek incredible stuff lunatism, really look forward to your results There were sites around that could tell you the percentage of games team won when each player was and wasnt playing, like footballanorak, but they all seem to have fallen by the wayside now. I am impressed with the lateral thinking using the football manager database :lol Obviously it is never going to be 100% accurate but it must be as close as you'll ever get.

[Vilamoura04]

Re: Pullein Power Update following last nights results Likely Winners only 6.54 profit from 49 staked Value only 8.36 profit from 50 staked Likely winner AND Value 4.64 profit from 23 staked Looking like staines passive may have been onto something when he suggested following the value picks only. The sample size is still a bit small to draw firm conclusions however. To confirm the bets that you would take if you followed each system tonight: Likely winners: Newcastle and Dundee U Value: Newcastle and Hamilton Hybrid (Likely winners AND Value): Newcastle only

[relf]

Re: Pullein Power Seems that a combination of the likely winner and value has the highest yield, given your sample.

[KThom]

Re: Pullein Power Is the previous result analysis league specific, or are the historical 14 seasons not seperated by league?

[Vilamoura04]

Re: Pullein Power I dont have the book in front of me but if memory serves there is not league specific data for previous seasons, and the 14 season backdata is for the english professional leagues. However the author does state that the data applies across all major leagues, suggesting that specific analysis has been carried out and similar (or similar enough to make it not worth the the hassle) results have arisen.

[relf]

Re: Pullein Power

Update following last nights results Likely Winners only 6.54 profit from 49 staked Value only 8.36 profit from 50 staked Likely winner AND Value 4.64 profit from 23 staked Looking like staines passive may have been onto something when he suggested following the value picks only. The sample size is still a bit small to draw firm conclusions however. To confirm the bets that you would take if you followed each system tonight: Likely winners: Newcastle and Dundee U Value: Newcastle and Hamilton Hybrid (Likely winners AND Value): Newcastle only

Just happened to do some back-testing on EPL for the previous few seasons. Here's the results if you were back the likely winners for all 380 matches (1pt each). The odds are from Bet365. 2009/10: 194.38/200 (-2.8%) 2008/09: 370.56/380 (-2.5%) 2007/08: 394.41/380 (+3.7%) 2006/07: 355.76/380 (-6.3%) 2005/06: 397.63/380 (+4.6%) Not as bad as I initially thought.

[Vilamoura04]

Re: Pullein Power Hmm thanks relf. Maybe a system that uses likely winners and allows 'bad' value but limits this to a certain percentage would work. So, say, qualifying selections would be likeliest winners where value is better than minus 10%. There seems to be a lot you can do with this, difficult to know where to start.

[KThom]

Re: Pullein Power Also one other question if I may. What is the correct way to gauge the "value". I mean if true odds were 4.0 (25%) but you calculate the winning chance at 50% (2.0) what is the correct sum to do here to measure the value? Is it (50%) - (25%) / (50%) = 50% or (50%) / (25%) - 100% = 100% or (50%) - (25%) = 25%

[relf]

Re: Pullein Power

Hmm thanks relf. Maybe a system that uses likely winners and allows 'bad' value but limits this to a certain percentage would work. So, say, qualifying selections would be likeliest winners where value is better than minus 10%. There seems to be a lot you can do with this, difficult to know where to start.

Initially I thought blindly betting on likely winners would definitely put one in the red, but it turns out that you can get profits on some seasons! That's pretty amazing, considered the likely winners tend to have low odds, which might not be able to cover the upsets. As for the issue of value, different people have different systems and different ratings. So judging from the back-testing, if you can find good value for likely winners, it's not really difficult to have a good and consistent yield for EPL (haven't done back-testing on other leagues). I'm pretty new to this as well (you can see my join date). Keep exploring your options and try out new things!

[Vilamoura04]

Re: Pullein Power

Also one other question if I may. What is the correct way to gauge the "value". I mean if true odds were 4.0 (25%) but you calculate the winning chance at 50% (2.0) what is the correct sum to do here to measure the value? Is it (50%) - (25%) / (50%) = 50% or (50%) / (25%) - 100% = 100% or (50%) - (25%) = 25%

My maths isnt very good but what I have tried to acheive is expressing value as the difference between actual and fair odds as a percentage of the actual odds. I have done this using the calculation actual odds divided by fair odds minus 1. So for your sum it would be: actual 4 / fair 2 = 2 2 minus 1 = 1 or 100% Another example would be fair 2 actual 1.75 actual 1.75/fair2 = 0.875 0.875 -1 = minus 0.125 or minus 12.5% Is that the right way to do it?

[Vilamoura04]

Re: Pullein Power

Initially I thought blindly betting on likely winners would definitely put one in the red, but it turns out that you can get profits on some seasons! That's pretty amazing, considered the likely winners tend to have low odds, which might not be able to cover the upsets. As for the issue of value, different people have different systems and different ratings. So judging from the back-testing, if you can find good value for likely winners, it's not really difficult to have a good and consistent yield for EPL (haven't done back-testing on other leagues). I'm pretty new to this as well (you can see my join date). Keep exploring your options and try out new things!

I guess the thing is that blanket backing the most likely winner is fine on a night like tonight when there are bugger all matches on. Its a bit more difficult on a saturday when theres hundreds of games to choose from, you can't bet them all.

[lunatism]

Re: Pullein Power

:eek incredible stuff lunatism' date=' really look forward to your results[/quote'] Ye - I wish I could click my fingers and have it ready to go, but unfortunately it will take some time to get it out of the ideas stage, and it's not the #1 priority for me at the minute....but i'd probably start a thread and daily post my strength ratings for each team.