paparainbow
-
Posts
80 -
Joined
-
Last visited
Posts posted by paparainbow
-
-
They will limit you if your stakes are high and your payout high too. Are you positive that younhave a winner strategy?
why not split your bets in different bookies? There are others offering corners too.
-
These are services paid by companies. Individuals don’t bother paying them. Additionally information is in feed mode, there isn’t a ui you can log in and download live stats.
what kind of data do you meed exactly? Odds’ timeseries for instance?
-
On 11/17/2018 at 5:03 PM, laprikon said:
Variables
P(Aw) = Probability of team A winning
P(d) = Probability of draw
P(Bw) = Probability of team B winning
P(A|1) = Probability of Team A winning by exactly 1 goal = 0.191856-0.30565416*x+1.42165*x^2-1.21885522*x^3, where x=P(Aw)
P(A|2) = Probability of Team A winning by exactly 2 goals = 0.13482-0.553848*x+1.71645*x^2-1.08417508*x^3, where x=P(Aw)
TO(A|X) = True decimal odds of team A with handicap XNotes and Sources
[1] To find true odds of Team B, use the following equation: 1-TO(A|X).You can use this spreadsheet
can you re-upload the spreadsheet, please? thanks for replying
with respect to your formulas, what are those numbers? and what is the formula for TO(A|X)?
-
Do you use logistic regression or maybe random Forrest?
Also what features did u use?
-
Excellent thanks for sharing!
-
Hi, i have a question that initially I thought it was simple, but that’s not the case i guess.
Let’s say i have a model, and i can calculate the probability of all correct scores for a given match ( it really doesn’t matter the model as long as i have a probability density function f(x,y) that returns correct score x-y).
How can I calculate asian handicaps 0, 0.25, 0.5, 0.75 etc?
0.5 is easy, i take all matches that x>=y (0-0,1-0,2-0....., 1-1,2-1,3-1,..... )
similarly 1.5: x>=y+1 etc
what about the asian handicaps? Any idea?
-
Hi, i have a question that initially I thought it was simple, but that’s not the case i guess.
Let’s say i have a model, and i can calculate the probability of all correct scores for a given match ( it really doesn’t matter the model as long as i have a probability density function f(x,y) that returns correct score x-y).
How can I calculate asian handicaps 0, 0.25, 0.5, 0.75 etc?
0.5 is easy, i take all matches that x>=y (0-0,1-0,2-0....., 1-1,2-1,3-1,..... )
similarly 1.5: x>=y+1 etc
what about the asian handicaps? Any idea?
-
considering your initial input,
expected home 0.90
expected away 1.35
you have to revisit your mathematical model.
-
You do understand that the probability of picking a favourite @ 1.70 is pretty much the same right? Plus, larger column length means larger margin, hence lower profit. But it all come to this at the end of the day:
Bookmaker prices a game at 1.15 for the fav team. Do you agree with this? Is it long or short? Don't let your betting choices be driven by bookies predictions but your own.
-
On 13/2/2016, 10:42:15, Ini said:
nobody?
you should somehow convert asian handicap to another market type, where all outcomes (in fair odds) generate probability distribution (sum of prob.outcomes = 1).
In asian handicap outcomes are not mutually exclusive, hence do not form prob.distribution (like 3way winner).
-
U get value to over 1.5 because
1. you don't apply overound the same way bookies do
2. you use poisson distribution, which is know for overdispersion. Actual life variance is greater than poisson predicts, due to its limitation var= mean.
I suggest checking negative binomial to fix this, and also improve your ovr model. For example
probabilities : 20%-80%
fair odds : 1/0.20 = 5.00 - 1/0.80 = 1.25
lets suppose margin 108%
your method : 5.00/1.08, 1.25/1.08 -> 4.60 - 1.16 bookis though do not apply margin like that resulting in an odd set of lets say 4.00-1.25
they tend to trim long odds and be more generous to short ones. they have their reasons: risk management and attractive favourite odds.
cheers
-
On 18/12/2015, 7:42:55, evil_inside said:
It's from a Greek university. I can translate the most important parts if people want.
Looking at the tables and the graphs it seems that the profit is significant.
I am currently trying to reproduce the results and anyone who would like to talk about it or help is welcome.
you should provide as with the link or pdf, otherwise, whats the point.
-
Re: handball mathematical approach i guess normal distribution fit data batter, but standard deviation can be a problem. Any idea?
-
Re: Time delay correct score double system you do understand that winning home equals to all correct scores where x>y... you take into account only 3-4 out of 10+ possible correct scores.. in the long run you will loose because cs has a significantly larger overound with respect to winner 1x2. Sent from my iPhone using Tapatalk
-
Re: rugby in play algorithm for odds compiling do you think using markov chains would do it? assuming you know probabilities for try, drop goa, penalty etc. for each team
-
Re: rugby in play algorithm for odds compiling
I have noticed (or at least back in 2011 when I did some data collection on it) that there tends to be an over-reaction to tries being scored in close games' date=' the market will shorten "too far" and then bounce back a few hundreths at least all within a minute or so. Someone with a High Speed trading algo and a good model of price-over-time could probably profit off that alone.[/quote'] do you think that there is a connection in their algorithm input between habdicap/expactancy + tries? lets say 1 try for every 12 points, thats is u/o 36,6 each of too generates under/over 3,5 tries ? this is a simplified logistic though, but I cant believe that tries are a separate input in such an algorithm. Any ideas? Sent from my iPhone using Tapatalk -
Re: Ordinal Logistic Regression any luck in this?
-
Re: How to calculate first half under/over you will have to calculate the expected value of goals for that given match, lets say 2.60 and then decide a percentage of goals scores in 1st half, lets say 50%, then you get 1,30 goals for 1st half, and then you apply your model for 1,30 goals to calculate u/o 1,50. all this, supposing that you have developed an algorithm of evaluating games and producing odds. That is what bookies do, that what you should also do.
-
Re: Model of Zero-Inflated Poisson u can inflate 0-0 but you will have to do so in other low draws too (lets say 1-1, 2-2...) if you want to produce a high draw probability.
-
Re: Betting on under 6.5 goals, in play at odds of 1.01 keep in mind that bookies want those kind of bets. they dont care even if they receive 1000000 euros in 1.01 odds. also keep in mind that most of the times odds with overound that are less than 1.00, are intentionally boosted uptp 1.01 in order to attract players as your self. Last but not least, keep in mind that accordint to big number law, an outcome that has 99% of being succesfull will be 99 out of 100 times. If you place the same bet 100 times, and you actually loose one, then you are going to loose probably what you have earned previously. Imagine now that the one single loose bet happens before you actually manage to win from the other similar bets you probably have placed. Unless you consider this bet a value bet... i dont see any point in risking 1000 euros to win 10....
-
Re: Poisson. Dixon Coles approach what input are you using for your algorithm though? thats the tricky part.
-
Hi, should someone build an algorithm that calculates odds for rugby (lets say union), what would be the best way to approach it? Do you think that pythagorean expectation is a good approach. I use it, but I have a significant deviation from bookies in my odds. Is there any distribution method u think that would give a better approach. The scoring system may be the reason I default, that the minimum score is 3 and not 1, like in football,basketball, handball and other time based sports.
-
Re: double result - odds ps:thanx for your replies
-
Re: double result - odds i can calculate the h/f by using a live model. i can calculate all the possible outcomes of 1st half and then all the possible final outcoume based on the initial score of 2nd half.. nevertheless this is not a formal mathemarical aproach, more likely a simulation. results are pretty decent and consistent with the bookies outthere. i want to expand my models to other sports, like, basketball, handball or rugby, where calculating exact scores is not the basic idea. This is y i am asking you if you have any ideas
Basketball in-play modelling
in NBA & Basketball Predictions
Posted
Does anyone have the slightest idea what mathematical concept is behind odds in in-play basketball?
I mean is there a distribution or something to use, as someone could use poisson in football (with time, decaying goal expectancies over time etc)?
Even though basketball seams to follow a normal distribution, I don't know what to do with negative values (points cannot be negative).
if I model point differences (hence allow negative values) I cannot model the total points rather than the margin of victories, handicaps only...
any idea?