So far I'm very happy with my java code.

TradesLikeAPotato: I am still looking for a collab, if you want to discuss your ideas ...

IMHO - very bad idea. Bitcoincharts is well known "stable" service. This days I have a satanic problems as my bots must load historical data from it, but have 404,502 and more shit.

Maybe you build a really better service, but I don't need this.

I (as a user) need a standalone bot, which:
 
can sit on some VDS and keep all it need in local environment
have trading algorithm which can be modified or completely rewritten without head ash
have interface to major exchanges API (bitstamp,btce,btcchina)
have modular structure, which allow it to be real open source, but not the one more programmer's toy (old,without eye and paw on github's attic)  

Something like QtBitcoinTrader or MT4,  but concentrated on easy implementation of different trade strategies on different exchanges. Based on node.js.




  

I am back guys.

I fixed the email & underscore errors: pull the latest version and `npm install` to get it working.

(Thanks avert for the email fix suggestion: implemented it the same way. For underscore: they removed a function I was relying on so I am now shifting towards lodash)

I'll check the code fixes posted asap and merge it in. Remember guys: Pull requests are always welcome!

C:\Users\Mark\Desktop\gekko-master>npm install
npm http GET https://registry.npmjs.org/moment
npm http GET https://registry.npmjs.org/bitcoincharts
npm http GET https://registry.npmjs.org/btc-e
npm http GET https://registry.npmjs.org/mtgox-apiv2
npm http GET https://registry.npmjs.org/prompt-lite
npm http GET https://registry.npmjs.org/lodash
npm http GET https://registry.npmjs.org/line-reader
npm http GET https://registry.npmjs.org/async
npm http GET https://registry.npmjs.org/bitstamp/0.0.4
npm ERR! not found: git
npm ERR!
npm ERR! Failed using git.
npm ERR! This is most likely not a problem with npm itself.
npm ERR! Please check if you have git installed and in your PATH.

npm ERR! System Windows_NT 6.2.9200
npm ERR! command "C:\\Program Files\\nodejs\\\\node.exe" "C:\\Program Files\\nod
ejs\\node_modules\\npm\\bin\\npm-cli.js" "install"
npm ERR! cwd C:\Users\Mark\Desktop\gekko-master
npm ERR! node -v v0.10.18
npm ERR! npm -v 1.3.8
npm ERR! code ENOGIT
npm http 304 https://registry.npmjs.org/mtgox-apiv2
npm http 304 https://registry.npmjs.org/bitcoincharts
npm http 304 https://registry.npmjs.org/prompt-lite
npm http 304 https://registry.npmjs.org/lodash
npm http 304 https://registry.npmjs.org/line-reader
npm http 304 https://registry.npmjs.org/btc-e
npm http 304 https://registry.npmjs.org/async
npm http 304 https://registry.npmjs.org/moment
npm http 304 https://registry.npmjs.org/bitstamp/0.0.4
npm ERR!
npm ERR! Additional logging details can be found in:
npm ERR!     C:\Users\Mark\Desktop\gekko-master\npm-debug.log
npm ERR! not ok code 0

I never tested anything on BTCe, but for mtgox:

I'm assuming 1% fee. Without treshold, poorly (minus 3%, on my May to August data).  With -0.2/0.2 treshold, pretty good, +30%. But lots of trade signals (25). Too much for my taste, I guess.

Extremely useful, thanks! This is what finally got me started using gekko myself, doing some proper backtesting of EMA methods.

My search starts from the assumption that (a) more recent history is more relevant for present day prediction than older history, (b) overfitting of the method must be avoided, so results need to be compared across different time frames, (c)
trading is slightly more expensive than assumed by default in gekko, because of signal delay & slippage

I'll post my results when I'm done with the testing. I have to do it partly manually because I want to control the treshold parameter, so I'm not searching through all EMA combinations, but only those that previous posters have mentioned as promising. Maybe someone else can fill me in in case I forgot a promising combination.


A quick question for Goomboo

Looks like we're turning this thread into a discussion about backtesting EMA strategies, finding the optimal parameters, etc. If this is starting to become too tangential to your original idea of this thread, i.e. you think we're hijacking your thread, let us know.

Hmm that's very strange, I did heard some similar things before though.

Unfortunately I am not able to reproduce it, could you please send me the candle file you used + the EMA settings? That way I can debug the behaviour more easy without staring in the dark.

Thanks. Ok I believe I see the reason now - bitcoincharts reuses the closing price of the last datum for the opening price of the next, whereas I only included trades within the range t and t+i (where t is the unix timestamp at the start of the candle and i is the candle interval). Both ways make sense - bitcoincharts gives the theoretical price for a trade if you were to do it at t, whereas whydifficult's and mine give the price for the first trade - however our way conforms to the official use of the term 'opening price' http://www.investopedia.com/terms/o/openingprice.asp. Except for the opening price, most of my data agrees with that of bitcoincharts', except for where the opening price is the same as the high, low or closing, in which case my values differ in about 5% of cases.

Make of this what you will. I don't think it makes too much of a difference anyway - backtesting is an approximation, and during real trading your actual trade prices are likely to be different from those given by the candles anyway. All of these datasets seem to be accurate, albeit inconsistent with each other, but these inconsistencies seem easily explained.