Importance (or lack of) of recent form

[Smokin Joe]

Thought I would start a thread that hopefully might get a bit of discussion going. I have anlysed the results of over 16,000 football matches. I looked at the previous 6 games of the teams playing and awarded points as per the English league system (ie 3 for a win and 1 for a draw). I then deducted the away team's points from the home team's points to give a "Form Rating". Example; Home team has 10 league points from last 6 games and the Away team has 5. The Form Rating would be 5. The results for all of my data seen Home team lose 7.3% of stakes, and away team lose 8.6% of stakes - so no huge difference there. If I then look at games where the Form Rating > 7 (994 matches) the home teams lost 8.1% of stakes and the away teams only lost 1.4% of stakes. On the other hand, in games where the Form Rating

[lunatism]

Re: Importance (or lack of) of recent form interesting analysis there SJ. Did you look at the home and away teams location form or overall form? Because otherwise it might have a detrimental affect on the results you're getting eg. if the home side played just 1 home game in their last 6 (but they won their last 5 aways), then up against an away side who has won 5/6 of their last away games, you might favour the home side, unless looking at the location...because then you might favour the away side

[Smokin Joe]

Re: Importance (or lack of) of recent form

interesting analysis there SJ. Did you look at the home and away teams location form or overall form? Because otherwise it might have a detrimental affect on the results you're getting eg. if the home side played just 1 home game in their last 6 (but they won their last 5 aways), then up against an away side who has won 5/6 of their last away games, you might favour the home side, unless looking at the location...because then you might favour the away side

When I said last 6, I actually looked at the last 3 home games and last 3 away games for each time.

[lunatism]

Re: Importance (or lack of) of recent form

When I said last 6' date=' I actually looked at the last 3 home games and last 3 away games for each time.[/quote'] That's what i assumed - I don't know how easy it would be for you to do (depends on the data you have at hand), but would you be able to make the same tests looking at last 6 homes games (by the home team) and last 6 away games (by the away team)? Otherwise you're just looking at overall form (and that's split between 3 home 3 away, which might not be considered fair since there could be a BIG time difference) between matches....not even taking into account winter breaks (although their effect is negligible), rather than the form of a home team (at home) against an away team (away).... Which (in my opinion) is much more likely to produce conflicting information Hope this helps ;)

[Smokin Joe]

Re: Importance (or lack of) of recent form

That's what i assumed - I don't know how easy it would be for you to do (depends on the data you have at hand)' date= but would you be able to make the same tests looking at last 6 homes games (by the home team) and last 6 away games (by the away team)? Otherwise you're just looking at overall form (and that's split between 3 home 3 away, which might not be considered fair since there could be a BIG time difference) between matches....not even taking into account winter breaks (although their effect is negligible), rather than the form of a home team (at home) against an away team (away).... Which (in my opinion) is much more likely to produce conflicting information Hope this helps ;)

Personally, I agree with what you say, but the premise I started out on was that the normal pool of "mug punters" would consider just the recent form (ie last 6 home games), hence why I wanted to see how inefficient using the last 6 games was. I think it can be shown that placing emphasis on recent form is incorrect, and I wonder if there are any avenues to exploit this phenomenon.

[lunatism]

Re: Importance (or lack of) of recent form

Personally, I agree with what you say, but the premise I started out on was that the normal pool of "mug punters" would consider just the recent form (ie last 6 home games), hence why I wanted to see how inefficient using the last 6 games was. I think it can be shown that placing emphasis on recent form is incorrect, and I wonder if there are any avenues to exploit this phenomenon.

You're 100% right - and i just hope the mug punters have the sense to see that. I'm sure that it's not just the mug punters who believe that a simple look at the form table would suffice to tell who's going to win what game. But then there are ways of interpreting the 6 game form table that can still mean very useful data can be extracted, which, combined with other data can produce results at the other end of the spectrum. Perhaps you have at hand (don't do it unless you think it would be beneficial to this discussion), the same tests run on league position, you would be able to see if there are any patterns between the two. Perhaps you could exploit the "higher in the table so must win" fallacy that grips mug punters (and even bookies sometimes) - either with league table or form, or perhaps combined? Us lot (PL folks with *that* much time on our hands) should really come up with a better name than mug punters (even though that aptly describes the majority of them), after all - that would be suggesting we're the divine bunch, right? :rollin

[Smokin Joe]

Re: Importance (or lack of) of recent form

You're 100% right - and i just hope the mug punters have the sense to see that. I'm sure that it's not just the mug punters who believe that a simple look at the form table would suffice to tell who's going to win what game. But then there are ways of interpreting the 6 game form table that can still mean very useful data can be extracted' date= which, combined with other data can produce results at the other end of the spectrum. Perhaps you have at hand (don't do it unless you think it would be beneficial to this discussion), the same tests run on league position, you would be able to see if there are any patterns between the two. Perhaps you could exploit the "higher in the table so must win" fallacy that grips mug punters (and even bookies sometimes) - either with league table or form, or perhaps combined? Us lot (PL folks with *that* much time on our hands) should really come up with a better name than mug punters (even though that aptly describes the majority of them), after all - that would be suggesting we're the divine bunch, right? :rollin

I use the football-data files to generate my information and it doesn't have a league position. It would be difficult to generate the league position based on the format of the data. I would have no problem inworking out how many league points each team had, but it would be difficult to turn that into league positions (as you would then need access to all the other team's league points at that same date). It would probably be doable one data file at a time but difficult to build into my automation routine that can run through all the FD files one after the other automatically. Any ideas?

[Zoroaster]

Re: Importance (or lack of) of recent form Very interesting research. I've argued for quite a long time that "recent form" is a complete myth and has little or no information for us as value seekers. I remember reading an essay years ago (it may have been on here, although I can't remember) about a study that was done concerning increased likelyhood of a team winning, had they won their previous game. It concluded there was zero increased probability, which went along with my debunking of "form" as a tool we should value highly.

[Smokin Joe]

Re: Importance (or lack of) of recent form .

[Smokin Joe]

Re: Importance (or lack of) of recent form

Very interesting research.

I thought so, but I was hoping for a bit more enthusiasm from other PL systemites. I had done work previously which showed that including recent form as a selection criteria actually resulted in worse performances than ignoring that particular criteria. Hence why my findings this morning were what I expected. The question is, how to exploit it?

[lunatism]

Re: Importance (or lack of) of recent form

Very interesting research. I've argued for quite a long time that "recent form" is a complete myth and has little or no information for us as value seekers. I remember reading an essay years ago (it may have been on here, although I can't remember) about a study that was done concerning increased likelyhood of a team winning, had they won their previous game. It concluded there was zero increased probability, which went along with my debunking of "form" as a tool we should value highly.

Zoroaster - the jump from not taking the form league position too seriously to discrediting all form in general is a bit steep. Don't you think number of goals scored in previous 6 matches (home or away ...or both - since you can usually see both) is at all handy? Don't you think that perhaps a team with only one win in their last 6 games may mean something? Perhaps it hints team not gelling so well (stats speak louder than words IMO....tables say the teams not scoring - doesnt matter what the quality of their attack is - the team's not scoring, etc). Like i said before - SJ has done this on overall form too, and to completely wipe out form on the basis that overall form league position hasn't shown a positive correlation to number of wins is not really a wise strategy. Past 6 home games (for home side) and past 6 away games (for away side), I think that those might show a better correlation, but so what if they didn't? Like SJ said - it's only the uninformed who'll just look at the league position and make judgement who will win. I've had plenty of success using form in the past (not really positions, but rather the other stats), so I could never agree with what you just said. I guess it depends how much work you want to put into reading the data in front of you

[Smokin Joe]

Re: Importance (or lack of) of recent form

Don't you think number of goals scored in previous 6 matches (home or away ...or both - since you can usually see both) is at all handy? Don't you think that perhaps a team with only one win in their last 6 games may mean something?

Loon, the thing is though, that you could reasonably expect the market to have priced those factors in, indeed they may even have over priced them - which is the point of view I was coming from. Of course you would rather be on a team in good form and whom scores goals regularly, but if the market has accurately estimated the chances of that team winning then it is no good to us. So you want something that shows a team is in "good form" but that the market underprices............

[Rushian]

Re: Importance (or lack of) of recent form Interesting discussion guys! I'll try to add my bit: It may be the case that a team's "form" (however one defines it) is not reflected correctly using only points gained or even goals scored. My general feeling is that goals or points gained are not necessarily good predictors of how likely a team is to win a match. In sports like football were the goal instances are quite rare on average, a "lucky" goal (i.e. a goal against the team that you would expect to score it) changes significantly the retrospective image of how the match went. The same applies to points gained: a "lucky" draw/win gives a really distorted view of a team's recent form. My gut feeling suggests that other indicators, which have a higher number of instances should (statistically) behave more nicely. Hence a from indicator based on such measures e.g. shots may lead in fact to a conclusion that form is indeed important in prediction. Just a thought.

[Zoroaster]

Re: Importance (or lack of) of recent form

Don't you think number of goals scored in previous 6 matches (home or away ...or both - since you can usually see both) is at all handy?

With goals in particular, I'd say that recent form was almost totally useless. As football is such a low scoring game, the variance of total goals over such a short period will contain no useful information whatsoever - in my humble opinion.

[lunatism]

Re: Importance (or lack of) of recent form

With goals in particular' date=' I'd say that recent form was almost totally useless. As football is such a low scoring game, the variance of total goals over such a short period will contain no useful information whatsoever - in my humble opinion.[/quote'] i have to strongly disagree - but just my opinion based on the work that i've done looking a LOT at league and form tables ;)

[happygooner]

Re: Importance (or lack of) of recent form

i have to strongly disagree - but just my opinion based on the work that i've done looking a LOT at league and form tables ;)

You still at the library loon:lol?

[Cogge]

Re: Importance (or lack of) of recent form Where did you get your odds from? Bookmakers generally adds a favorite bias margin, meaning that if the overround is 6%, it's not evenly distributed (odds(1)*1.06 : odds(x)*1.06 : odds(2)*1.06). Favorites are generally more heavily betted than underdogs, which means they would place something like 9% overround on the favorite and a 3-4% overround on the underdog. "The results for all of my data seen Home team lose 7.3% of stakes, and away team lose 8.6% of stakes - so no huge difference there. If I then look at games where the Form Rating > 7 (994 matches) the home teams lost 8.1% of stakes and the away teams only lost 1.4% of stakes. On the other hand, in games where the Form Rating < -7, placing £1 on all home teams seen just a loss of 3.4% compared to a massive 14.4% loss on the away teams." Analysing your results from this angle you get: If the teams are even/equally good, betting blindly the loss on home teams should be the same as the loss on away teams (overround is even). Your result: Home 7.3%, Away 8.6%. If the teams are uneven you will lose a lot more in the long run when betting blindly on the favorite: Your result: +7 Form - Home: 8.6%, Away: 1.4% -7 Form - Home: 3.4%, Away: 14.4% Now it's of course not the entire truth behind these numbers. But it does offer a slight explanation.

[Smokin Joe]

Re: Importance (or lack of) of recent form

Where did you get your odds from? Bookmakers generally adds a favorite bias margin, meaning that if the overround is 6%, it's not evenly distributed (odds(1)*1.06 : odds(x)*1.06 : odds(2)*1.06). Favorites are generally more heavily betted than underdogs, which means they would place something like 9% overround on the favorite and a 3-4% overround on the underdog. "The results for all of my data seen Home team lose 7.3% of stakes, and away team lose 8.6% of stakes - so no huge difference there. If I then look at games where the Form Rating > 7 (994 matches) the home teams lost 8.1% of stakes and the away teams only lost 1.4% of stakes. On the other hand, in games where the Form Rating < -7, placing £1 on all home teams seen just a loss of 3.4% compared to a massive 14.4% loss on the away teams." Analysing your results from this angle you get: If the teams are even/equally good, betting blindly the loss on home teams should be the same as the loss on away teams (overround is even). Your result: Home 7.3%, Away 8.6%. If the teams are uneven you will lose a lot more in the long run when betting blindly on the favorite: Your result: +7 Form - Home: 8.6%, Away: 1.4% -7 Form - Home: 3.4%, Away: 14.4% Now it's of course not the entire truth behind these numbers. But it does offer a slight explanation.

Cogge, are you sure that you are correct? I thought it was a given that you lose proportionally more by backing at larger prices.