new idea on prices. why do we bother?

[muppet77]

i have been thinking about generating prices for the purposes of football betting. many of us try to look at elo type rating systems to convert to actual % chances of H, D, A, and then imply odds from these.

i was wondering - what if you just CAN'T find the true chance of events ocurring? we all search for the Holy Grail, but suppose it is just not possible?

i think (or suggest) this because i was thinking about bookies odds, and asked myself a few roundabout questions:

do they really know the exact price of a match?
well, not exactly.

do they have a fair idea of the price of the match?
well, generally.

do they offer that price?
no, i believe they have a good idea, then they apply their over round and create a price for the punters. as long as this buffer zone or margin of safety is large enough, then they will win in the long term.

so is it worthwhile trying to determine if a team has a 56% of a win or tweaking the sytem / rating to see if we can make it more accurate and the chance is actually 58%? what's the difference?

what i think is worth looking at is something along the lines of "games like"......that is, rather than saying,
"right, ManU are playing Liverpool, and my rating sytem gives 12.4 points to ManU at home and 9.6 to Liverpool which converts to HDA of 57%, 25%, 18%"....

instead saying
"ok, ManU are a top team who are playing at home against a second level team"

or categorising 5 or 6 TYPES of games that can be played.
obviously this needs discussion, but it could be something like

type 1: an even steven match - very tight
type 2: a potential league winner against a demotion favourite
type 3: a team that is better than another but may draw
etc

obviously this part is the crucial part and can be done a number of ways.

what factors to look at?
home/away goal diff? league positon? point per game?

once the measure has been chosen then we can almost work backwards.

group each TYPE of match and then work out the average odds offered by a bookie. hopefully they will be similar for each type of game. for example a 'top vs bottom' type has average odds of 1.30 and a 'class one against a class two' type game may average odds of 1.75.

thus by doing this, tissue prices are almost being found.
(as a tester i found the average odds for games that finish as a draw, +1, +2, -1, -2 there is a really strong pattern in the average odds, not that this is immediately useful, but suggests that there is a pattern)

once these averages have been found (they will be inaccurate for INDIVIDUAL games) they can be applied to betting situations.

so if a game is identified as "type 1" and the tissue price suggests home should be 1.30 and betfair is offering 1.65, then you can back that team.

i just believe that many of us are trying to overcomplicate the process.....we may calculate Bolton to have a 35.765% chance of winning. well, on the day they have injuries, players' motivation, weather etc which all will affect the odds. the bookies can't get all of this spot on, and so they use the over round as protection.

wow, that's the longest post i've made.
does anyone:
...understand what i'm trying to say?
...have any advice or contribution?
...think i'm talking bobbins?

[duffman]

muppet77

interestingly enough, i've had a similar idea as yourself for years and was actually looking to share in others opinions but havent had much luck. I've tried a few postings but I dont think people have a clue what i'm on about.

I also believe that the solution to finding value is to compare current odds in a market with the same odds in a previous market under similar circumstances. The only problem comes with defining those circumstances..and i think when you talk about 'types' that you have the same basic idea as myself.

I think, for the FA Cup final a bookie would say something like..ok, both are top class teams who are well rested and no particular run of form. Man U will be in high confidence after winning the league, but I dont think that will matter..so I think we will call it even and go 2.7 on both teams...

don't see an elo ratings or percentages there...

So what if an online system (obviously this is a very basic example) could give you the previous odds under the same circustances i.e. when a team:

team type - title contenders
rest period - well rested
form - neutral
confidence - high

played against a team:

team type - title contenders
rest period - well rested
form - neutral
confidence - normal

You could instantly compare odds and find:

ManU Draw Chelsea
-2.7--- 3--- 2.7---- Current Odds
-2.2-- 2.9 ---3----- Average of Previous Odds (under similar circumstances)

Which finds that the 2.7 on ManU may be value. You can then check other markets as well under those circumstances.

Let me know what you think. Not sure if youre a programmer, but i've an idea on how to make this interesting and make such a system work..

[Sgt.Sunshine]

I thought about this a while ago too, and came up with so many "Types" that there was no difference between a conventional model and the "types" model .

I seem to remember trying it when all the "conventional" models I had tried failed to get a reasonable and consistent edge. IMO if you have a model that spits out probabilites and you get an edge, you get an edge. I can't see that the outcome of splitting games into groups this way will be any different, as you still get a probability at the end of it.

Duffman's idea of looking back to previous games played under similar circumstances is interesting, but the question still arises of which factors are important. If you get this right, then IMO any model incorporating these factors will work.

[muppet77]

duffman, i am not a programmer as such, but i am quite handy on the old excel and have tried many ratings systems and stats analysis.

i quite agree with your comment, duffman and sgt.

there seems to be two massive questions

(a) how many 'types' should there be?
(b) what are they?

duffman what did you have in mind (PMail me if you prefer)

[OTIUM]

Spot on, why waste time and energy trying to beat 40 full-time compilers when a bit of rain before kick off changes everything?
Go to the lower leagues, arb, trade, pinch a few pounds or wait, wait and wait until you see a mistake or a 10% edge.

[Sgt.Sunshine]

Duffman/Muppet - I do some programming and am always interested in new ideas. If you need some help .....

[merlin]

.... must admit that I like a decent system as well as the next man, but

I'm afraid I see very little mileage in this type of project and agree with

Otium that the compilers are often not far away when displaying their

odds.Give me something any day that I feel that I can get my teeth into!

Good luck in your search for the Holy Grail!

[duffman]

Yeah but it that 'not far away' where punters try to get their value. You cant say compilers are right all the time... For example..In OTIUM's example, we know its raining but the compilers dont. We could then use the system to show previous market odds in the exact same circumstances but when it was raining. We then compare them and find if we can get any value. Surely its not a million miles from our holy grail

[muppet77]

this is exactly what i have been saying guys - they can't be EXACTLY correct - there is always some sort of randomness or noise - be it rainfall, snow, injuries, crossbars, referees, mentality of players, motivation etc that is difficult to quantify.

all the compilers do is their best, then hide behind a profit margin. this is where BF can be used.

what was your idea duff man - post here or pmail - doesn't fuss me.

[duffman]

muppet,
if you (or anyone else who ha similar ideas) wants to mail me at an account i've just set up oddsidea@gmail.com, i can show them how I plan to do this and we can disuss similar ideas or spot flaws etc. Cheers.

[muppet77]

you got mail

[merlin]

.... you got mail again!

[si1965]

I am sure the compilers use a considerable number of factors when it comes to setting the prices. Even assuming that there are some factors more significant than others e.g. current form, any system would need to identify and quantify these.
The danger with a system such as this lies in what you don’t measure – for example, you could be in possession of a piece of information e.g. a key striker has not made it through a late fitness test. The prices set by the bookies are in the ‘normal’ range according to your new system, so it appears as though there is some value to be had because of the missing striker. What you don’t know is that there could be a number of other minor factors that actually went in the teams favour and that these along with the absence of the striker has already been factored in – hence the ‘normal’ odds.
Not sure how sophisticated the bookies are when it comes to the lower League and minor Leagues stuff – certainly there is not as much information as there is for the top European and UK Leagues. I think that in these cases you are much more likely to find a trend and bookies posting ‘generic’ prices for certain types of matches (as outlined by Duffman). Having said that, the impact of key players, injuries and suspensions in these Leagues (because of the small squads) are much more significant and easier to track.

[duffman]

Originally Posted by si1965

I am sure the compilers use a considerable number of factors when it comes to setting the prices. Even assuming that there are some factors more significant than others e.g. current form, any system would need to identify and quantify these.
The danger with a system such as this lies in what you don’t measure – for example, you could be in possession of a piece of information e.g. a key striker has not made it through a late fitness test. The prices set by the bookies are in the ‘normal’ range according to your new system, so it appears as though there is some value to be had because of the missing striker. What you don’t know is that there could be a number of other minor factors that actually went in the teams favour and that these along with the absence of the striker has already been factored in – hence the ‘normal’ odds.

Okay, very good point and I really appreciate you taking the theory seriously into account. I agree about a 'normal' range being set, but only on two things..the general strength of the two sides and who has home advantage.

However, my thinking (using your example) on anything else that can be considered a factor (beit a late injury, good form, bad weather, internal problems etc. - however major or minor)....ends up in an interpretation on how it will further affect the game, and its the odds compilers interpretation that gets represented in the bookies odds. Value punters will try to differ in interpretation and find some value.

For example...a late striker in injury doesnt actually mean 'a late striker injury' in betting terms...it means the interpretation that - 'this team might have problems scoring' so the bookie may adjust the +/-2.5 goals market slightly. A value punter might know that the guy taking the strikers place is the next maradona....so he interprets that the will score more goals.. and tries to find some value..?

So....I think if we come up with a definitive list of these interpretations in relation to markets we can start to build systems (not necessarily my system) that kicks ass! - thats my thinking anyway, anyone else agree?

[merlin]

.... I have read the foregoing with interest fellas but I'm afraid I keep

reaching the same conclusion.No matter what traits and factors you

consider as likely variables which may or may not affect the true result

of the game/games in question, one would still need a crystal ball to be

able to do what you want to do.To be able to factor in just some of the

reasons why a chosen set of facts could affect circumstances,would need

to be effectively carried out pretty much as each match is kicking off, to

be sure that most factors are accounted for.Definitely a no no as far

as I am concerned.I have read your link on the subject you so kindly

provided Steve and chose two reasons why I differ:-

"A value bettor will be generally unconcerned about backing the underdog, or perhaps more relevantly backing a team which he thinks will not win (underdog or not), provided there is value in the bookmaker's odds. A value bettor estimating the likelihood for the Liverpool win to be 70%, with a 15% chance of a Sunderland win would back the away team, despite believing that Liverpool should win."

Goes against the grain with me! Secondly :-

" Successful betting, then, is really all about understanding and managing probabilities. Know the true chances of a sporting win, and there may be profitable opportunities waiting at the bookmakers. As Geoff Harvey says in his book Successful Football Betting, "Find the value, [and] the winners will take care of themselves."

..... and I find it hard to believe this quote above.

Food for thought anyway!

[muppet77]

hi merlin, mate. good to speak.

i think that the general thrust of the thread is to address the fact (well documented) that the betting public is usually correct when operating as a single unit. the best example is betfair - where a very solid market with lots of money tends towards the true odds of an event.

"A value bettor will be generally unconcerned about backing the underdog, or perhaps more relevantly backing a team which he thinks will not win (underdog or not), provided there is value in the bookmaker's odds. A value bettor estimating the likelihood for the Liverpool win to be 70%, with a 15% chance of a Sunderland win would back the away team, despite believing that Liverpool should win."

Goes against the grain with me! Secondly :-

really?! without value, then you cannot profit in the long run?
strike rate is not everything - it's the edge over the bookie that earns your dough over time. surely this is the fundamental first law of betting?!


anyway, this thread is going to try to address questions that steer away from a statistical nature - the bookies will be able to price most matches up better than you or i - and hide behind the overround.

betfair has changed all this, with the millions of bettors pushing and pulling prices to their true value.

we aim to try something out that uses this - the great knowledge of the combined masses into one.

[merlin]

.... again all theory Martin, I'm sure there are punters out there making

money each weekend by careful selections and perhaps varying their

betting technique and strategies each week.I ask how many punters

would truly back against Liverpool in the foreging example! Not me!

[Sgt.Sunshine]

merlin I am surprised like Muppet about your comments regarding value. Maybe I misunderstood (do you really think those quotes are wrong??).

I do agree though about the rest of your post. I think taking a lot of subjective variables into account just before kick-off, or even leading up to it, is very difficult. To quantify these variables and remove the subjectivity would require an enormous amount of work and a large sample size. Even then I don't think you would get much of an edge. IMO the trick is to keep things simple .....

[Sgt.Sunshine]

Duffman's talking about circumstances got me thinking about tennis. I don't know how much mileage there is in this, but last month in Monte Carlo Federer was 1.12 to beat Ferrero, and today he was 1.26 to beat Ferrero in Hamburg. Both Masters series, both on clay. Federer has a little wobble in between and his odds shift quite a lot. I wonder how general an occurrence this is, or how often it happens. It depends on your view of what "circumstances" are I guess. Also noticed that Gonzalez was 1.8 tb Mathieu today, due I guess mainly to Mathieu's recent good run. I am a big believer in betting against form, and it might be that tennis odds fluctuate in favour of form maybe a bit too much??

[merlin]

.... your final sentence in post 18 Sgt answers it all!