Bitcoin Forum

..just to disclose some information: me and some friends are currently working on a fully automated trading system.
we are at the moment evaluating proper indicators and setting up a environment to backtest the system.

it will probably not make use of 'complicated' indicators and stick to the common ones: MACD, SMA, EMA, Volatility

..what indicators do you trust in the most?

starfish..

Fear. ;D (Capitulation)

When I took Java, I made a program that emulated a magic 8 ball. Do you want me to PM you the code?

Here's some Ruby to do that:

class Array
  def random_element
    self[rand(self.length)]
  end
end

['Yes', 'No', 'Maybe So', 'Ask again'].random_element

its not going to work bud...these indicators are great and all until they fail, and when they fail, they will kill your account. Automated trading only works if you possess expensive and complicated algorithms which Goldman Sachs (as an example) had at one point along with having loads of money to actually manipulate the market. The trick to using all of these indicators is blending them...not taking any one by themselves. All those commercials you see on TV about these perfect trading systems that do the work for you are a load of crap. If everyone could do and afford it then everyone would be rich...think about it.

now join my charting service already  ;)

Trend following (MA crossovers) could work on a small scale.

Edit: But only until Bitcoin becomes less volatile.  Crossovers keep you on the correct side of the trend, and Bitcoin has big trends.

If you are going to go the Technical Indicator route then you should have a momentum indicator to catch when the market is trending and also have a oscillator so that you can see reversals.  Back test your trading strategy with a random trading behavior strategy.  If your strategy is not statistically better than the random trading strategy then find a better combination.  Most importantly for your trading system is to have proper money management.

Chart patterns

Your system cant do them?
Well, you could still go whith Goomboos EMAs (they seem to work out pretty well), RSI and PSAR

I personally discarded the MACD in favor of goomboos EMAs, because they tend to generate more recent indicators.

there are currently 2 prototypes in developement, one in python, one in c#.

indeed we also have own idead on indicators. basically stuff we can't trade cause we are not fast enough, asleep or emotions made us do rather irational trades instead.

waveaddict, yes you are right. fully automated trading systems are highly complicated, we are aware of this fact. but we are all also highly interested, have the base knowledge required to start such a project and are just about to launch it!

if everything fails after all at least everybody gained a lot of experience

greets

MACD is always a little late

i don't know much about chart patterns, but i recently read about them and will definetelly try to do an approach of implementing them.
but first we need to have a infrastructure with GUI and data-interface and so..

rather than implementing fixed patterns, we think of a way to do automatized future calculation on expected shape of patterns and trade with regard on this information. not clear yet, but there will be many ways of approaching a system that trades arbitrarys in expected price movement.

it will basically be trend-driven. if it catches a trend, it will stick do it until it ends and another trend is caught.

RSI is a good indicator if used right.

Please elaborate

http://bitcoincharts.com/charts/chart.png?width=900&m=mtgoxUSD&SubmitButton=Draw&r=90&i=&c=1&s=2011-10-01&e=2012-02-18&Prev=&Next=&t=S&b=&a1=&m1=21&a2=EMA&m2=10&x=0&i1=RSI&i2=MACD&i3=&i4=&v=1&cv=0&ps=0&l=1&p=0&

When the daily hourly (this is true for daily also, but the charts are in hourly intervals) RSI touches/passes through 30/70 levels, it usually means:
- the market is pretty certain on a direction of there the price will go, so we have an uptrend/downtrend
- after this first touch with 30/70 there will be a spike in opposite direction of the prevailing trend
- after this spike (if it occurs), the trend will resume and sometimes go even higher/lower than before
- and finally, when RSI exits 0-40 / 60-100 range, a reversal is likely

Study the chart above to see these correlations. Using RSI alone would be dangerous probably, but in conjunction with other indicators (like the EMA and MACD on the chart) it can be quite useful.

About current market situation:  These indicators tell me that there will be another dip down. RSI has nearly touched 30 once, but it didn't pass through. MACD is on negative and EMA is going down with current price staying below it.

Thanks!  When I profit, I'll compensate you for this.  Disclaimer:  It won't be much.

Since you are using MtGox you should take advantage of their volume data.  Use the Money Flow Index (MFI) (https://en.wikipedia.org/wiki/Money_flow_index).  It is like the RSI, but instead of using price it uses volume in the equation.

Whatever momentum oscillator you use (RSI, MFI) you should not just do the basic overbought/oversold (above 70/below 30) indicators.  You should use to to detect divergence in the price.  So if the price of bitcoins hits a new high but the RSI or MFI has a lower high than in the past then that is a bearish divergence.  Price will likely go down.  If the price of bitcoins hits a new low but the RSI or MFI has a higher lower then that is a bullish divergence and the price will likely diverge upward.

I know about MFI, but after investigating/backtracking it I found it has worse performance than RSI on Bitcoin - it generates more false signals.

I hope OP does backtesting.  I had nice backtesting results using the MFI and another indicator, but I found some that gave better results.


Reading for the OP
Technical Analysis from A to Z (http://www.amazon.com/Technical-Analysis-2nd-Steven-Achelis/dp/0071363483)

Call me old fashioned but thats far too formulaic. RSI just tells you if, relative to the period you are considering - hourly, daily etc if there have been more people buying or selling than before. This gives you an idea if a market os overbought, or over sold. It doesn't tell you the market is definitely going to reverse, because markets are irrational and can carry on going crazy in either direction.

The decision to buy or sell comes from considering a variety of factors. To me RSI just defines buying or selling opportunities (buy when its oversold, sell when its overbought) the decision as to whether to take a given opportunity must still be made. If you do trade, you then need to have the balls (and int the case of bitcoinica, the leverage) to see the trade through.

The tricky part comes when RSI moves but price doesn't. You then face the difficult decision of whether to cut your losses, because if you bought at RSI <30 and you are subsequently sat on a loss at RSI >70 (and the longer the period, the stronger the potential reversion to the mean) you are probably going to want to take that loss, for fear of incurring a bigger one. Thats hard, because nobody likes crystallising losses.

backtesting a.. let me call it a 'shitload' of indicator-combinations is our main goal besides automated trading without worries while we sleep ;)

and another thing backtesting shmacktesting. the market is rarely if ever identical. forward test, see how many scenarios you go broke under to figure out risk, then compare this to your potential rewards. Thats all you can ever do.

Assume you are going to lose, and make sure that loss doesn't wipe you out. If you win, rinse and repeat.

That is partly what backtesting is for.  To calculate the drawdowns in the past and how that impacts your money.  If you don't think history repeats itself then you probably don't want to be using technical analysis.

"gut feeling" - the best indicator.  ;D  But seriously, the technical analysis + graphic shapes + sending large player + count words on this forum (such as "down" 15 times "up" - 2 times, "dump" - four times, "Rally" - 0 times, "rocket" 1 times  = TREND MOOD DOWN) is already in use I think  ;)

This sounds like a really bad idea...

Edit: I'm not at all opposed to automated trading on principle, even in bitcoin. But this particular instance sounds like it will end in tears

LOL

we also thought about doing a mood indicator also, but we thought about a more empirical way of doing it ;)

like: our bot discloses a selection of very strong signals to the public (if the bot works, i guess it will work even better if many ppl read the signals) which only burden in order to optain the signal is to tell us how they feel. you tell us your emotion on the market, we tell you when our bot decides to switch trends.

Good luck gewure. I have written an automated trading system and it works pretty well. The benefits of using an automated trading system is that it removes the largest bias that will affect anybody's trading methodology... human emotion.

If you design a system that is in effect correct 51% of the time and you code something up to trade on it consistently; let it run and let the easy money flow into your accounts without you having to move your fingertips.

Those who typically design automated trading systems will aim to be correct >51% of the time which isn't too difficult. There are hundreds of different methods and tactics that can be used to gain small margins here and there a few of which have already been talked about in this thread. Add some code to a handful of them and you will have a very effective trading system which will have the capability of pulling small profits upon each trade.