Hey guys, so I have a slightly more reasonable prediction than the "prepare for march 16th dip" post made a little while back.

As some of you may have seen in a post I made several days ago, I have been working on software that uses neural networks to predict bitcoin prices. The predictions can be seen at http://www.btcpredictions.com/

The short term neural network has been consistently predicting a dip from around $585 down to about $562 and then back up to around $590 starting around 4am EST, reaching its lowest price around 7am and then returning to $590 around 10am.

Obviously this program's predictions are not always right, and predictions that are further away are more difficult for it to make. However, it seems relatively confident about this dip. I believe the dip will occur, though the magnitude may be a bit off or the dip may be shifted in one direction slightly. So we'll see how well the software performs!

How can anybody say he predicted this? He just stated a truism and it didn't even come true. There was absolutely no major dip. There was a very, very minor dip two days after he said it would occur. It was about 3.5%, and that is a fact.

I mean, you can see that cosmofly is clearly trolling but I can't believe some people here actually think his "prediction" was successful...

Longer predictions are one of the things I would like to work on for expanding this project. I wrote a small blurb about it in the "Coming Soon" page. The challenge with making predictions much longer than 5 days is that they just become less accurate. If I remember correctly, I believe that the last time I tried to do a 10 day prediction it gave me something like a 7 or 8% average error, which makes the predictions not particularly useful. However I do plan on experimenting with ways of structuring and training the neural network to see if I can make some reasonably accurate longer-term predictions.

http://www.btcpredictions.com

It does a pretty good job of predicting bitcoin price trends... haven't had it running long yet but it has a pretty low margin of error.

lol okay he's definitely trolling... no possible way anyone says something like this seriously

Actually I am correct. When you talk about a dip or a spike, you don't talk about the singular very extreme prices (which is what you are referring to here). Those are meaningless unless you are staring at the price 24/7. Yes, the price went down to $590 for like maybe 5-10 minutes tops. The real meaningful price would be the average price over the course of each hour (which is going to be closer to 605). Furthermore, the price on 3/16 was not 645, it was 635 - I don't know where you are getting that information. Look at the charts: http://bitcoinity.org/markets/bitstamp/USD

Also even if what you said were true, that's an 8.5% dip, which is still very far from a "major" dip. Honestly by the time I woke up today, the price had already gone back up to over $610.

Actually on the 16th and the preceding several days it was mostly around 635 and it dipped down to about 605, which is a 3.9% fall in price. This was not the event he predicted in any way shape or form, and this prediction holds (held?) no weight. It was completely wrong lol...

The price has gone down by a minute amount and it was 2 days later than predicted. Also it already seems to be going back up a bit again. This hardly qualifies as a dip, and that definitely does not qualify as a successful prediction.

Thanks dude! Also btw it actually doesn't consider volume in making these predictions. It doesn't even look at prices either - it looks at recent percent changes in price and predicts future percent changes.

When I was first designing this, my original idea was to do like you said - input prices for a certain number of days leading up to the prediction. However, I found that this did not yield very good results. What works a lot better (and what this does) is using the percent change in price as input/output.

For the short term graph it looks at 60 hours leading up to the current hour and for the long term graph it looks at 80 hours.

The neural network is forward feed, has three layers(input, hidden, output), and uses backpropogation

I'll say it's going to stay pretty steady and then start to slowly climb back up to the mid 600s within the next several days. That's what my software is predicting anyway (www.btcpredictions.com) seems like a reasonable prediction to me

hahaha I had no idea what you guys were talking about and then I read the "Prepare for March 16th dip" thread... LOL that was ridiculous

Today is the fourth day that it has been up. The problem is that with the price being so flat right now it's difficult to really see how well it does. I mean, it has been predicting for the price to pretty much remain the same, but the real test will be to see how well it does when the price fluctuates.

Also I've been trying to test a program that simulates trading automatically based on these predictions, but the price isn't changing enough for it to make or lose any money. I included a 0.5% transaction fee so it just won't do anything if the price isn't going to change enough, which seems to be the case right now.

Ugh predictions are so boring right now... I want the price to start fluctuating so I can really test this out

I'm glad you are interested and I'd be happy to keep you guys up to date! And I'll definitely check out that book.

I use a forward feed neural network with backpropogation. There are actually two neural networks - one for short term and one for long term. They both have three layers (inputs, hidden layer, outputs) and they use the same activation function (sigmoid). Basically the only difference between them is the number of nodes in each layer, and the data I give them. The long term neural network uses 80 inputs and 20 outputs, which each correspond with changes in average price over 6 hour periods. The short term neural network uses 60 inputs and 24 outputs, which each correspond with changes in average price over 1 hour periods. Both have hidden layers with 200 nodes.

Knightcoin - That documentary looks so awesome! I am definitely going to find the full version of that and watch it!

I don't really know any other forecasting methods. I basically just used neural networks because I learned about them in a class I took a few semesters ago and thought they were really interesting. We used them to recognize hand-written digits and when I found out that they are often used for stock price prediction I decided it would be a cool idea to use them for bitcoin price prediction.

I've never heard of ARIMA but I will check that out for sure. Like I said, I don't really know any other forecasting methods but I do know that they exist. That's largely why I created this thread.

I think a lot of the major spikes or crashes in bitcoin price are probably nearly unpredictable since it obviously is very fundamentally unstable. However, I would think that it would be possible to (once a spike or crash is occurring) predict how exactly the event would play out as far as bitcoin prices go. Once a major event occurs that causes bitcoin prices to fall or spike, I feel like there must be some discernible patterns in price changes.

Or maybe not, who knows lol

Yeah, I pretty much just hold at the moment too although that's mostly because I don't have enough extra money floating around to risk buying/selling in short intervals. I think you definitely need to consider the news whether you are following some type of technical analysis or just going on your intuition. And yeah I think neural networks are very interesting (obviously)!

Bitcoin is definitely far from a normal market, which is why it's so risky to try to make money by buying/selling in short intervals. But I'm sure people do it, and I'm wondering if they are just lucky or if they have some strategies. And I'd be curious to see if there is some way to measure their performance in comparison to buying/selling based on a model like mine.

Also btw the time zone on the charts is your current time zone.

I was wondering what methods you guys have seen used to predict bitcoin prices? Right now I'm working on a software project that uses neural networks to try to predict bitcoin prices (http://www.btcpredictions.com) but I was wondering what other strategies you guys have used and how successful they have been? Do people find a lot of success just going with their gut on this?

I'm currently working on a program that will simulate trading based on the predictions of this project so we'll see how well it does. I'm testing it right now, gonna let it run for at least a few days, probably longer.

The software bases its predictions off of a giant csv file found here: http://api.bitcoincharts.com/v1/csv/

And then the software also updates the contents of the csv file with the lastest transactions every like 2 seconds or something, since those files are only updated every like 12-24 hours ish. So yes, it looks at the entire history of transactions on bitstamp. And yes, it could be used for other exchanges.

As far as predicting litecoin prices, I would need to find or create a similar csv file or data file containing exchange information of a similar format. But yeah, I don't see why it wouldn't be able to predict litecoin prices. Just might require a bit of work as far as how the program gathers data from the file. Should work for stocks too, which I'm thinking of doing at some point.

Some time in the next week ish I'm going to try to write an algorithm that will try to trade optimally based on these predictions, and it will simulate buying/selling 1 bitcoin. I'll display it's progress on the site and we can see if it is profitable