Bitcoin Forum

Hi everyone  ;D,

I wanted to release to you a project I've been working on for the past month - www.objectivebitcoin.com (http://www.objectivebitcoin.com).  Basically the goal is to keep up with what people are saying on social media and online communities, automatically classify the seniment of anything relevant, and compile those statistics into an indicator.

A negative value for the indicator means that sentiment is trending negative, while a positive value means that sentiment is trending positive.  The indicator is a z-score of the average sentment rating of the past hour, compared to the average sentiment rating of the past day or two. I did it this way because most of what I'm finding at the moment is overwhelmingly positive.

My algorithm is still fairly basic, and I plan to upgrade that as I work on it.  I'll also calculate a correlation value (with change in exchange rate) based on running the current algorithm on historic data.

I would be grateful for any and all feedback you may have!  I'm here to provide a service to the Bitcoin community.

I like the idea, any thoughts on doing it over bigger time periods?  A day or two seems pretty short.

Yes, that was a preliminary estimate of what I set it at.  I'm going to run some more correlations with price data to figure out what time period works best.  For lack of a better standard, price data is really the only measure I have to go off of to decide how valid the algorithm is...

I like this a lot.  Bookmarked.

But... while all the math, bots, spiders, & etc. operating behind the scenes appear to work very well, the page itself leaves a bit to be desired.

I'm Firefox 20.0 on Debian.  Can't see the "Price Correlation:" line at all, and I have no way to get to it.  The center box (where all the info is) won't come all the way up to size - there are sliders top and bottom - although it takes up less than 10% of my screen.

Planning an Android app?

Thanks for your feedback!

Interesting, I've tried the layout on a few browsers and they've all looked o.k. ( I'm on Firefox 19 on Ubuntu).  If you get the time, could you send me a screenshot so I know what you're dealing with?  I'm not a web designer by trade, so I'm sure there are things that need worked out.

If there's enough interest then I could probably eke out a mobile app in the future, sure.

Screenshot on the way to your gmail...

I have something similar sitting in a bottom drawer. Just Python and a numerical output, no website as that's not really my thing. I didn't find any useful (i.e. leading) correlations to make it worth developing though. I'll be interested to see if you do.

Thanks, the font was just showing up too big, and I didn't plan for that.  Oops.


I'll send you a PM in a bit  8)


I wrote a trading bot last year, and with every traditional indicator I could throw at it, its neural network couldn't figure out how to predict price changes.  I would love for this to be a forward indicator, but if not, it might combine well with other indicators.  Worst case scenario it turns out to be a mildly interesting site and I learned a lot about programming  ;D

The best strategy I found was a simple Bollinger Band - purely because the market is so sentiment-driven (a nice way of saying "bubbly") that jumping on a bandwagon as it sets off was actually worthwhile, whether buying or selling. The sad fact is though, that once so many traders (and HFT bots) jumped in it's impossible to run a regular algorithmic trading bot. When the market starts to go into freefall then the bot is trading blind too. Frustrating but there you have it.

I've also learned a lot and have a nice sentiment analysis tool sat there in case of future requirements. Plenty of plans too. I haven't got around to my planned next project of running some sort of machine learning algo on historic data to optimise parameters for the Bollinger Band strategy. I still have a feeling Twitter sentiment could figure in that too but who knows?

Are you storing the twitter feed for further analysis? Running any textual sentiment analysis or just tweet volume?

Yes and yes.  The sentiment analysis is done with a probability classifier, so items it isn't sure of get put in a separate bin so I can manually classify them later and update the classifiers with this new knowledge.

Word about Bitcoin is overwhelmingly positive (http://www.whatdoestheinternetthink.net/bitcoin (http://www.whatdoestheinternetthink.net/bitcoin)), so I have to run some statistics to see how the current mood compares to past mood.

I hadn't seen that site. Pretty much agree with the weighting though. That's from manually classifying about 1,500-2,000 tweets (so Twitter, not internet generally) as they came in which took a couple of days or so of storing them up. I bet you get that in every couple of hours now!

Yeah right now its receiving ~600 per hour, but that has jumped to several thousand when something big is going down.

Also, Twitter is definitely an (albeit skewed) representation of the internet as a whole, since the vast majority of tweets I'm seeing are linking or referring to news stories, threads in this forum, posts on Reddit, etc., from either helpful bots or users.  So I see twitter as sort of a natural starting point for this type of thing.

Now with 100% more charts!

Can you release a widget so can post on blogs.
Thanks