Page 1
Page 1
Started By
Message

Anybody else doing ‘wisdom of the crowd’ algorithms with Twitter to make money?

Posted on 8/19/18 at 1:07 pm
Posted by CFDoc
Member since Jan 2013
2093 posts
Posted on 8/19/18 at 1:07 pm
So I’ve been reading a lot of the technical papers on the research groups using the Twitter API and sentiment algorithms to bet on sports based on wisdom of the crowd theory: LINK

So far, I have set up some python libraries that grab hashtags and return a positive, negative, or neutral score. My next task is to turn these scores into forecasting game outcomes. I’ve read several other papers that discuss this topic and the most interesting thus far use moving averages much like stock predictions to identify ‘golden’ or ‘death’ cross behavior.

The legalization of sports betting has me really thinking about trying this for a season or two just to see if it works. Anybody else trying this stuff?
Posted by castorinho
13623 posts
Member since Nov 2010
82010 posts
Posted on 8/19/18 at 1:36 pm to
quote:

wisdom of the crowd
Posted by Hogkiller10
LP
Member since Jan 2010
1529 posts
Posted on 8/19/18 at 3:02 pm to
Why dont you post some of the results you are getting so it can be tracked.
Posted by CFDoc
Member since Jan 2013
2093 posts
Posted on 8/19/18 at 4:18 pm to
I mean, I guess I could just trust George Carlin’s comedy routine to make conclusions regarding rigorous mathematical theories describing stochastic processes.
Posted by TigerstuckinMS
Member since Nov 2005
33687 posts
Posted on 8/19/18 at 5:11 pm to
quote:

I mean, I guess I could just trust George Carlin’s comedy routine to make conclusions regarding rigorous mathematical theories describing stochastic processes.

You came here to ask TigerDroppings. Watch out where you throw those stones.
Posted by castorinho
13623 posts
Member since Nov 2010
82010 posts
Posted on 8/19/18 at 5:12 pm to
Did you read the study you posted?
You're not going to be profitable doing this.
Predicting things like elections, in which the public has direct impact is one thing. Using it to predict football games this way is a different story.


Something that would be interesting would be following all players of a team and assigning a "mood" value to their social media posts preceding games and see if translates to an edge, but there's so many players that it would not be an easy task.

Something similar in tennis, especially women's tennis, would be very interesting since there's only one variable.
Posted by HailToTheChiz
Back in Auburn
Member since Aug 2010
48894 posts
Posted on 8/19/18 at 5:55 pm to
I don't understand how someone posting a hash tag on Twitter could have any effect on a football line or outcome

But Im just a nobody on Tigerdroppings
Posted by CFDoc
Member since Jan 2013
2093 posts
Posted on 8/19/18 at 7:53 pm to
quote:

Did you read


Yes. As well as several other articles from Europe where these analyses have been done for almost a decade now to predict premier league soccer outcomes. I also read a few articles from investment firms using these analyses daily to identify ‘Golden crosses’ before everyone else.

quote:

You’re not going to be profitable


Possibly not, but there’s a lot of tech papers out there that have shown consistent 8-15% returns on soccer games. It’s also shown to perform well on stock predictions LINK

Some Texas professors have gone back and looked at past NFL seasons to see how well this could work under certain circumstances and they were able to show 15% returns LINK

The interesting part of the NFL analyses is that they state it’s not in your best interest to try to predict all 256 games, but rather monitor certain sentiments of underdogs and you can find the lines that show some undervalued dogs that you can jump on before the lines move.

I know this is Tigerdroppings and all but I’ve actually seen intelligent technical and money discussions on here before. I apologize for assuming I could start another.

quote:

Something that would be interesting


I find it interesting how dismissive you were to my original post but are now engaging in the conversation exactly the way I had hoped. Regardless, I agree it would be interesting to monitor the players themselves. Golf would be another good example.

The cool part about all this is it’s free. The python libraries and twitter API costs zero. There are also tons of statistical libraries for python that are free as well.
Posted by Douglas Quaid
Mars
Member since Mar 2010
4097 posts
Posted on 8/19/18 at 9:53 pm to
I've been wanting to learn python. How should I start? Any tips to move along quickly?
Posted by CFDoc
Member since Jan 2013
2093 posts
Posted on 8/20/18 at 7:33 am to
You have any other object oriented programming experience or is this a ‘start from scratch?’

Zed Shaw Learn Python the Hard Way is a good read if you’re starting from nothing.

Other recommendations will depend on your experience.
Posted by Douglas Quaid
Mars
Member since Mar 2010
4097 posts
Posted on 8/20/18 at 8:38 am to
I've been casually doing the free code academy python training on and off for a couple of weeks but effectively starting from scratch.

No other real programming experience.
Posted by CFDoc
Member since Jan 2013
2093 posts
Posted on 8/20/18 at 9:17 am to
I like Zed Shaw better than code academy for beginners.

As you start to get the hang of the language, the Mark Lutz books are good references. There's also just the straight docs on the python website.

Also can't forget everybody's favorite, stackoverflow.com
Posted by CFDoc
Member since Jan 2013
2093 posts
Posted on 12/24/18 at 9:35 pm to
quote:

You're not going to be profitable doing this.


lol
Posted by wizziko
New Jersey Nets Fan
Member since Jan 2006
35881 posts
Posted on 12/24/18 at 9:52 pm to
Posted by ArkLaTexTiger
Houston
Member since Nov 2009
2461 posts
Posted on 12/27/18 at 8:57 am to
Be sure to include input from the refs.
Posted by Breauxsif
Member since May 2012
22290 posts
Posted on 12/27/18 at 12:59 pm to
Python is a general purpose programming language that is also used for scripting. But python can be used for everything that is done in java.

Both python and java are interpreted during runtime. The term scripting usually means writing code to automate a certain process in a specific environment (like browser).

It's dynamically typed and is advantageous for good string processing features. I utilize Python within a specific OS, not an architecture, additionally (Unix in the case of Perl, Amoeba in the case of Python).

What exactly are you wanting to do with Python once you grasp the elementary basics?

This post was edited on 12/27/18 at 1:02 pm
Posted by Meauxjeaux
98836 posts including my alters
Member since Jun 2005
39857 posts
Posted on 12/27/18 at 8:49 pm to
The crowd is generally fricking stupid.
Posted by Korkstand
Member since Nov 2003
28703 posts
Posted on 12/28/18 at 1:03 am to
quote:

Both python and java are interpreted during runtime.
I was going to do a "technically..." here, but after a little research and a lot of thought, the whole story is a lot more technical than it seems on the surface.

Java and Python programs have a lot in common as far as how they are executed. Both have an initial step where the source file is compiled to bytecode (python generates .pyc files and java .class files), then a program (the python "interpreter" or java "virtual machine") interprets the bytecode and generates machine code to finally be executed. Both Python and Java have been implemented in various ways, but the stock implementations differ in that Python's compile and interpret steps are combined into one command/program, while Java's are not. So I guess the main difference is just an implementation detail. It's the user experience.

To me, Python leans more toward "interpreted" than Java does because you can mark a Python source file as executable and run it directly as you would any other program. I don't think you can do that with a Java source file. But that's just my interpretation of the word, not really based on any technical reasoning.

The real answer, I guess, is that they are both compiled and they are both interpreted. Both languages compile source code into bytecode, and both interpret the bytecode at runtime. Further, both languages have just-in-time compilers available (Java's stock VM does it I believe, while Python has a couple available I think), which compile some or all of the bytecode down to machine code at runtime. Then you have systems like Android that compile Java to machine code ahead of time upon app install.


That's a lot of words to say basically nothing other than "there's a slight distinction but the whole situation is just fuzzy".
Posted by Breauxsif
Member since May 2012
22290 posts
Posted on 12/28/18 at 2:20 am to
I write Python scripts everyday at my job. If anything could be called a “scripting language” it would be something just listing user inputs in a batch format - i.e. something like a DOS BAT file. Even that has some extras which make it more than just a saved set of user commands - e.g. the IF and GOTO, FOREACH, etc.

In general a script is something you use to automate tasks, so to answer your question about your “interpreted” rebuttal I typically write a short piece of code intended for a few uses and then thrown away. Not something you’re going to sell, just something to make your own life a little easier. Which also means it’s likely a hack and not designed with future proofing in mind - i.e. no great influence from best practice and code quality.

Within the realm of compiling IoT data, Python is the most clear-cut way to write clear and concise code that aligns with business requirements when leveraging IoT data with my experience in cloud computing.


This post was edited on 12/28/18 at 4:29 am
first pageprev pagePage 1 of 1Next pagelast page
refresh

Back to top
logoFollow TigerDroppings for LSU Football News
Follow us on Twitter, Facebook and Instagram to get the latest updates on LSU Football and Recruiting.

FacebookTwitterInstagram