how to learn to analyze data?
how can you built your own predicting model in Excel etc?
is anyone who can show us how?
is anyone who can show us how?
-
-
You can look for "big data" courses @ places like udacity, coursera, udemy, khan academy et al.
Many are free and offer decent content.
Statistical analysis, big data and similar topics are part of the analytical models books used to generate odds.Comment -
thank you so muchYou can look for "big data" courses @ places like udacity, coursera, udemy, khan academy et al.
Many are free and offer decent content.
Statistical analysis, big data and similar topics are part of the analytical models books used to generate odds.Comment -
probability, econometrics, python..... maybe start with figuring out how to calculate power ratings. i may be wrong, this may not be simple.
anyway, i think power ratings are a good place to start with analytics as they are ubiqitous, easy to understand and are logically consistent with how people think simply about quant stuff i.e. rams are 7 points better than saints, but it's in NOLA so add back 4 points for saints strong HFA = 3 point rams spread.
another good place to learn alot of stuff for free is mit open course ware. thousands of free courses with lectures/course notes readily avail.Comment -
mit open courses
sounds amazingComment -
Another thing to add to your toolkit, bayesian statistics: https://www.coursera.org/learn/bayesian-statistics
Viewing statistics through the eyes of a bayesian framework can be a game changer for a lot of people. I wish I had adapted a bayesian approach in handicapping (and my real life job..) sooner.Comment -
based on my limited understanding, big data courses are for analyzing huge amounts of data, 100s of GB to TBs across a cluster of machines. you don't need to use tools like hadoop and apache spark when analyzing smaller datasets for sports analytics on a single machine.
might want to look into python, scikit-learn, pandas, numpy, scipy
tensorflow for machine learning is the hot toolComment -
Nate Silver (of fivethirtyeight.com) wrote an entire (very accessible) book on how it applies to sports betting, poker, finance, politics, weather, and even earthquakes called 'The signal and the noise'. It's a great place to start learning about statistics that is an easy read, not math intense, and entertaining in my opinion. If you prefer a more academic place to start there are a ton of cheap text books that show you the math behind the madness, such as this.Comment -
boys
give a Fuking example
a vid,a screenshot,wuteva
of how it analyzes data and gives an outcome
do not just post links and comment that they re greatComment -
I'll give you an example.
Imagine your goal was to increase the accuracy of points over/under plays. A person could write an algorithm to compile and analyze game stats over the last 5 to 10 years. Based on data, they might find saturday or sunday has a higher statistical probability of over games in comparison to mondays or tuesdays.
There might be certain days of the year when a high percentage of teams hit the over play.
The point is something like a person can't predict what type of trends or patterns might emerge until after they analyze the data. Maybe there will be no patterns or trends. But sometimes a person might find something they can use to be more profitable.
Bro. You don't know what big data is. Go derail some other thread with your jibber jabber.Comment -
There are many variables in sports, there isn't necessarily an easy way to be consistently accurate without putting in some work.
The basics would be looking at average points scored per game and avg points allowed per game as a basis for offense versus defense. Then looking at the last few games played to see if they're on a hot or cold streak. Then looking at how far into the season we are to get an indication of how motivated teams might be. Etc. Just really basic things most have likely tried and not found much success with.
There are a few really basic trends casual gamblers probably know. Like how NBA games in the playoffs have a tendency to be more tactical and slow paced, with higher pressure, resulting in lower point games than in regular season.
The idea behind big data is to take that type of observation to the next level. Its something I thought about but never really got around to doing as I've been looking for ways to make money outside of gambling.Comment -
Big data has become a marketing term, flung around by journalists without much idea what it means. Mostly when it's used they mean machine learning - although again, that's lost its meaning as a buzzword.
Something like "predictive modelling" is probably most accurate.Comment -
Comment -
Hopefully someone else will comment, that's something I know nothing about.
Machine learning has been around for decades. Its an old field that dates back to Alan Turing's day in the 1950s - 1960s. Big data is a more recent development which might include things like deep packet inspection and widespread collection of consumer and user meta data which are not necessarily related to machine learning.
AFAIK anyways.
Fantasy sports and sports gambling are banned for the same reasons state based regulation disapproves of bitcoin exchanges and escorts advertising on craiglist.
Fantasy sports, gambling, crypto and adult entertainment are all sources of revenue for small business owners in the USA which translates to economic growth and a degree of prosperity. The goal for legislators is to restrict economic growth and crackdown on independent operators in those areas.Comment -
bro, no one is going to hand you the keys to the kingdom. guys break their dikks working years to build models to find the slightest of edges. besides, we are all given this information every single fukkin day
line opens with limited information
more information becomes known
line moves
nearly 100% of knowable information is known
line closes
independent event happens
a nearly infinite number of independent events happen that prove the closing line is far more efficient than the closer
if you want to continue to reinvent the wheel, have it at. just don't come around demanding people hand you the answers. do some fukking work.Comment -
Fine, here is a solid example, step by step, of how to analyze data and pick winners.
First, assume we have Team A and Team B. Scrape the web for a lines and the past data relating to the 3 most predictive statistics or metrics (e.g. the 3 things that produce the highest variance in determining outcomes) determined by either an independent component analysis or factor analysis (I am more familiar with ICA but it is up to you). To be more specific, you need to somehow incorporate some measure of kurtosis (fancy word for the 'tailed-ness' or skew of a distribution)in which to recover the multiple source signal by finding the correct weight vectors with the use of projection pursuit (this will be important later).The kurtosis of the probability density function of a signal, for a finite sample, is computed as
where
is the sample mean of
, the extracted signals. The constant 3 ensures that Gaussian signals have zero kurtosis, Super-Gaussian signals have positive kurtosis, and Sub-Gaussian signals have negative kurtosis. The denominator is the variance of
Using kurtosis as a measure of non-normality, we can now examine how the kurtosis of a signalextracted from a set of M mixtures
varies as the weight vector
is rotated around the origin. Given our assumption that each source signal
is super-gaussian we would expect:
- the kurtosis of the extracted signal
.
- the kurtosis of the extracted signal
or
, because we know the optimal weight vector should be orthogonal to a transformed axis
For multiple source mixture signals, we can use kurtosis and Gram-Schmidt Orthogonalization (GSO) to recover the signals. Given M signal mixtures in an M-dimensional space, GSO project these data points onto an (M-1)-dimensional space by using the weight vector. We can guarantee the independence of the extracted signals with the use of GSO.
In order to find the correct valueinto a new mixture
, which has unit variance, and
. This process can be achieved by applying singular value decomposition to
Rescaling each vector
, and let
. The signal extracted by a weighted vector
. If the weight vector w has unit length, that is
, then the kurtosis can be written as:
The updating process for
whereis a small constant to guarantee that
, and set
, and repeat the updating process till it converges.
Great, now that you've extracted this information using your preferred software (personally I am a MATLAB guy, but you can do this in python as well since I know that is waaayy more popular and intuitive for most people) you only have two more things left to do before you can determine which team to place your bet on. Next to last step: find a coin and flip it. If it is heads, bet on Team A, if it is tails, bet on Team B. Last step: take some xanax and chill out. No one here, or anywhere, is going to walk you through how to simply 'analyze data'. That's like asking us to teach you how to 'do physics'. Be more specific.Last edited by peacebyinches; 07-01-19, 11:57 AM.Comment - the kurtosis of the extracted signal
-
why all this negativity
i thought this thread would be a positive one for everyone
what i want is to learn how to do this
Comment -
this is what it delivers
pretty much he applies Poisson distibution
and the fukker selling it for 50 euros
what i want to do is a way to insert all the data of a league in excel and to get the possibilities calculated by Poisson distribution and if possible Elo ratings as well,
for the coming matches
i can do it manually for a match with two teams but if i can do it for a whole league with excel that would save me soooooooo much timeComment -
Use R, import your match results, bit of format fiddling, then use the PlayerRratings package:
Gives you ELO (and various others like glicko and Stephenson) for all rows in a few seconds.Comment -
I second using R (might take a little effort to learn if you haven't used it before, but should be well worth it), and that looks like a very handy package HH. I can't believe I've never checked the R package repositories for anything similar!
Looks like there's some NBA and MLB and NFL.
holy crap... looks like I'm not getting any actual work done for a while now
Comment -
can anyone post a video of predicting possibilities for soccer matches with R?Comment -
Nivi, guys like Heehaw + Peace are doing a very good job in this thread. I can't pin-point the link, but the Poisson Dist output is a good start. Here is the basic premise:
*We need to get to the point where we project TWO NUMBERS, E (HomeGoals) and E (AwayGoals). if we assume two INDEPENDENT Poisson distibutions...then everything else falls out. Think of it like two columns:
F(0), F(1), F(2), etc. From that we get f(0), f(1), f(2) etc. Do that for both Home + Away. We can then calculate our Home Wins, Away Wins etc. This is a very good starting point.
And I echo Peace's note above. Read Nate Silver's book. He's not going to give u a license to print money...but it's a very good starting point.Where's the fuckin power box, Carol?Comment -
It's a link u posted in another thread:
Navigate those tabs on the top. Poisson is a really good basic approach for soccer game analysis.
The real question is whether your INPUT will be close to correct. The output from that link is certainly good, if not a little confusing to look at.Where's the fuckin power box, Carol?Comment -
yeah this calculator looks dopeIt's a link u posted in another thread:
Navigate those tabs on the top. Poisson is a really good basic approach for soccer game analysis.
The real question is whether your INPUT will be close to correct. The output from that link is certainly good, if not a little confusing to look at.
but i really have no idea how to use it
here is a more simple one
Comment
