Fair or not fair ?

[ChrisB]

Re: Fair or not fair ?

language should never be apologized for :ok so how do you arrive at the conclusion a team's fair odds are 2.50 ? and how do you know that value to be correct ?

i take initial average odds of few bookmakers + margin security = 2.50 so, if it's not the good methodology, i would like to show mathematically that i'm wrong. thanks :)

[Mulkis]

Re: Fair or not fair ?

i take initial average odds of few bookmakers + margin security = 2.50 so, if it's not the good methodology, i would like to show mathematically that i'm wrong. thanks :)

Wait a minute, please. If this your conclusion comes from any of my post in this tread, I am telling you, it is wrong! We had a splendid discussion here, but we have no final resolution regarding "fairness" of the bookmakers starting odds!

[Datapunter]

Re: Fair or not fair ? Ok, ChrisB, here is a scenario: I present you with 1000 bets, each bet is offered at odds of 2.00, you bet on each one and end up break even, with a strike rate of 50% . Perfectly acceptable scenario, right ? Now please pick a number between 1 and 1000.

[ChrisB]

Re: Fair or not fair ? Mulkis, this is my personal conclusion. you use R2 and i don't know if it's related with average of bookmaker. why this approach is wrong, perhaps i misunderstand... I totally agree with you for the the final resolution here the parameters - numbers of bookmakers - a combination of bookmaker who use different type of customers - a combination of "shop" bookmaker and "internet" bookmakers. - Perhaps bookmakers who use betradar. brainstorming is open. :)

[ChrisB]

Re: Fair or not fair ?

Ok, ChrisB, here is a scenario: I present you with 1000 bets, each bet is offered at odds of 2.00, you bet on each one and end up break even, with a strike rate of 50% . Perfectly acceptable scenario, right ? Now please pick a number between 1 and 1000.

400. i don't see what you mean...

[Datapunter]

Re: Fair or not fair ? Ok, now we are looking at bet nr. 400 and i present this bet to you at odds of 2.00, if i follow your reasoning and conclusion then this bet has "fair odds" of 2.00, right ? Since 2.15 is bigger than 2.00 your conclusion is you have found positive value. So if i was to offer 2.15 you would bet on it, right ?

[ChrisB]

Re: Fair or not fair ? with my margin security, 2.15 is definitively not a value. i'll don't bet on it. But if you offer 2.25 , i take :)

[Datapunter]

Re: Fair or not fair ? Excellent, i'll give you 2.25 on bet 400 no problem. Now then, none of these 1000 bets actually have an expectation of 50% and "fair odds" of 2.00 , not a single one. What i did was i took 500 bets with an expectation of 40% translated in "fair odds" of 2.50 Then i took 500 bets with an expectation of 60% translates in "fair odds" of 1.66 And then i mixed them all up completely randomly. Given this scenario you will agree that: a) the strike rate on all 1000 bets will be 50% ( simple average of 40% and 60% ) b) at odds of 2.00 you will break even c) each bet has "fair odds" of either 1.66 or 2.50 but never 2.00 ( because that's how i set it up ) d) if bet 400 is one of the 1.66 ones you have huge value e) if bet 400 is one of the 2.50 ones then even at odds of 2.25 you have negative value Can you see that you cannot determine the "fair odds" of an individual selection based of the average of a group of selections ? There is more to this conversation but before i can continue you must understand this single point first.

[Datapunter]

Re: Fair or not fair ? Either way, i'm off to bed, to be continued....

[ChrisB]

Re: Fair or not fair ? ok, your scenario, is one possibility among many. if we have scenario with 42% strike rate and average WINING odds at 2.6, datas are differents. i also disagree with concept 50% strike rate and fair winning odds at 2.00. if you have fair winning odds at 2.00, you will have more than 50% strike rate. datapunter, if it's not possible to determine value bets with opening odds, in this case, how do you proceed ? i may be stubborn, but you have not demonstrated why bookmakers have wrong opening odds. Any opinion is apprecied :)

[Mulkis]

Re: Fair or not fair ?

i may be stubborn, but you have not demonstrated why bookmakers have wrong opening odds. Any opinion is apprecied :)

Yes, please someone, demonstrate this. So far we hear only "common sense" arguments and no real data or tests. In contrary, as already shown, opening odds do correlate significantly with final outcomes.

[Datapunter]

Re: Fair or not fair ?

i also disagree with concept 50% strike rate and fair winning odds at 2.00. if you have fair winning odds at 2.00, you will have more than 50% strike rate

WOW, this is so unbelievably basic i'm actually amazed. If all the odds are 2.00 and you accept these to be the "fair odds" then exactly what strike rate do you expect ? and how do you arrive at that rate ?

[Datapunter]

Re: Fair or not fair ? I will get to the conclusion at the end, but we need clarity on each step in order to get there.

Can you see that you cannot determine the "fair odds" of an individual selection based of the average of a group of selections ?

ChrisB, you haven't actually answered this question.

[ChrisB]

Re: Fair or not fair ? another value angle : in horse racing, if you back all rising odds and lay all droping odds you will make profit. it means very well that opening odd determinate the probability of horse to win. we can think that it can be the same thing for soccer...

[Datapunter]

Re: Fair or not fair ? I will get to that later on, this needs to go step by step otherwise it doesn't make sense.

[ChrisB]

Re: Fair or not fair ?

I will get to the conclusion at the end, but we need clarity on each step in order to get there. ChrisB, you haven't actually answered this question.

i feel that you love me... :) the answer is simple. backtesting only , able to tell us if we can determinate fair odds with group of bookmakers. and mulki has provided a solution: what do you think about horse racing study ?

[Datapunter]

Re: Fair or not fair ?

Now then, none of these 1000 bets actually have an expectation of 50% and "fair odds" of 2.00 , not a single one. What i did was i took 500 bets with an expectation of 40% translated in "fair odds" of 2.50 Then i took 500 bets with an expectation of 60% translates in "fair odds" of 1.66 And then i mixed them all up completely randomly. Given this scenario you will agree that: a) the strike rate on all 1000 bets will be 50% ( simple average of 40% and 60% ) b) at odds of 2.00 you will break even c) each bet has "fair odds" of either 1.66 or 2.50 but never 2.00 ( because that's how i set it up ) d) if bet 400 is one of the 1.66 ones you have huge value e) if bet 400 is one of the 2.50 ones then even at odds of 2.25 you have negative value Can you see that you cannot determine the "fair odds" of an individual selection based of the average of a group of selections ?

I know there's a bit lost in translation but please follow this conversation step by step. The answer is either "Yes i see it" or "No i don't".

[ChrisB]

Re: Fair or not fair ? yes i see it, but i don't see in what direction you want to go exaxctely. please give me a concrete example (not with 400 theorical bets but 5 or 10 concret bets) and What Alternative do you propose ?

[Mulkis]

[ChrisB]

Re: Fair or not fair ? thanks Mulkis for the explanation. :) interesting table. - wich bookmakers do you select ? - we see on the table that subgroup 1.31-1.40 have a high deviation. do you think this deviation will be reduced if we add margin security (ie 5%) on opening odds ?

[Datapunter]

Re: Fair or not fair ?

he argues that those odds are "fair"

Mulkis, you need to have patience. His argument / conclusion is wrong, that's the point i'm trying to bring across. But i cannot talk about why the conclusion is wrong until the mechanism is understood.

yes i see it, but i don't see in what direction you want to go exaxctely.

Ok, we have established (i hope) that you cannot determine an individual selections "fair odds" from the average of a group of odds. So to conclude that bet 400 at odds of 2.00 has "fair odds" of 2.00 because that happens to be the average is a wrong conclusion. As clearly given in the scenario here it can only be 1.66 or 2.50. Now we come to the correctness of the average of 2.00.

[Datapunter]

Re: Fair or not fair ? The scenario is still the same:

What i did was i took 500 bets with an expectation of 40% translated in "fair odds" of 2.50 Then i took 500 bets with an expectation of 60% translates in "fair odds" of 1.66 And then i mixed them all up completely randomly.

If you bet on all 1000 bets at odds of 2.00 you will break even at the end. Therefore the odds of 2.00 are COMPLETELY ACCURATE..... on average. But here's the key: they are only accurate if you actually bet on all 1000 bets. As soon as you bet on a sub-group the accuracy is no longer valid. When you bet at odds of 2.00 when the odds should actually be 2.50 you will end up making a loss. But as you also bet at 2.00 on each bet where the odds should be 1.66 that profit compensates the loss made on the 2.50 ones. So on average the numbers add up perfectly. I'm beginning to really like ChrisB and because of that every 5th bet i'm going to offer him odds of 2.25 . So out of 1000 bets 200 are offered at 2.25. Great, ChrisB bets on all of them and makes a very tidy profit. After all 200 of them have odds above average and therefore result in profit. ChrisB, you agree you make a profit ?

[Mulkis]

Re: Fair or not fair ?

thanks Mulkis for the explanation. :) interesting table. - wich bookmakers do you select ? - we see on the table that subgroup 1.31-1.40 have a high deviation. do you think this deviation will be reduced if we add margin security (ie 5%) on opening odds ?

High deviation in some subgroups could be explained by relatively small sample size (like 70 matches, which is statistically very small). Odds are taken from football-data bet365 odds.

[ChrisB]

Re: Fair or not fair ?

High deviation in some subgroups could be explained by relatively small sample size (like 70 matches, which is statistically very small). Odds are taken from football-data bet365 odds.

i calculated all deviations by subgroups and the average deviation is 5.30. :ok how to interpret this number ? what does it mean ?

[Datapunter]

Re: Fair or not fair ? My last post ends with a question, maybe you missed that.

[v-zero]

Re: Fair or not fair ? Guys, Datapunter's point is a basic one, and one that if not true then no value bets could ever be had... Bookmaker's odds, on average, are fair. That is, if you look at 200 games with odds of 2.0 then that bet will be won 50% of the time. However, that only means that the average of the odds for all of those games was 2.0... Some may have been truly odds of 1.5, and some of 3, some of 4, some of 1.25... etc. Finding the 1.25 priced at 2 and the 4 priced at 2 is the key to making money.

[Mulkis]

Re: Fair or not fair ?

...if you look at 200 games with odds of 2.0 then that bet will be won 50% of the time. However, that only means that the average of the odds for all of those games was 2.0... Some may have been truly odds of 1.5, and some of 3, some of 4, some of 1.25... etc.

Wrong! If you take 200 games with 2.00 and they won 50% ("on average") means nothing less than all these 200 games were priced correctly! The same with tossing a coin - if you toss a coin decent number of times at the end of the day you will get 50/50 "on average". It does not mean that some of the tosses have 25% chances or 75% or anything - its always was 50/50 chance!

[v-zero]

Re: Fair or not fair ?

Wrong! If you take 200 games with 2.00 and they won 50% ("on average") means nothing less than all these 200 games were priced correctly! The same with tossing a coin - if you toss a coin decent number of times at the end of the day you will get 50/50 "on average". It does not mean that some of the tosses have 25% chances or 75% or anything - its always were 50/50 chance!

Bullshit. Look. Say half the games actually had correct odds of 4.0, and half had correct odds of 1.33, then of those bets at 2.0 the number won would still be half, but none of the bets actually had a 50% chance of winning.

[Datapunter]

Re: Fair or not fair ?

If you take 200 games with 2.00 and they won 50% ("on average") means nothing less than all these 200 games were priced correctly! The same with tossing a coin - if you toss a coin decent number of times at the end of the day you will get 50/50 "on average". It does not mean that some of the tosses have 25% chances or 75% or anything - its always was 50/50 chance!

Seriously, you could not be more wrong with this. You are comparing a situation where you don't know the actual % with a situation where you do know them. The conclusion that the prices are all correct is where the error is. Maybe they are all priced wrong. Even then you can get a correct average. Take any collection of numbers and calculate the average to be X. Does that mean each member of the group has a value of X, of course not. Take any collection of numbers and calculate the average to be X. Does that mean any sub-group will also have an average of X, of course not each sub-group will have it's own average. It really is that elementary.

[Shinigami]

Re: Fair or not fair ?

Wrong! If you take 200 games with 2.00 and they won 50% ("on average") means nothing less than all these 200 games were priced correctly! The same with tossing a coin - if you toss a coin decent number of times at the end of the day you will get 50/50 "on average". It does not mean that some of the tosses have 25% chances or 75% or anything - its always was 50/50 chance!

you may like to visit this site,which explains the danger of using averages: www.flawofaverages.com/ By end of season, the results of the top to bottom teams naturally converge together under the round-robin tournament system to reflect roughly 45% home win, 27% draw, 27% away win... on hindsight, the average odds (less bookie margin) will be seen to have been priced correctly in correlation that ratio.. However, even though you can look back and assume that the bookie is "efficient on average", you can never predict the exact WDL sequences in a particular string of future results, so it's never reliable to just depend on the bookie's opening odds as probability.. The odds merely reflect the bookie's expectation of the customers' betting action, so as to hopefully attract them to bet in the same proportion as the opening odds, & to adjust accordingly if it doesn't...