Chrome Browser extension: MtGox trading bot

[Lucko]

Do you fill like sharing? I would relay love to have some simulation tool so I can setup bot more easily...

[sukiho]

the market cannot be predicted, all you can do is follow

How about making some educated guesses?

I think the only educated guess you can make is that the current trend will continue until it doesnt, which is what this trading method is based on

[rikigst]

Do you fill like sharing? I would relay love to have some simulation tool so I can setup bot more easily...

https://dl.dropboxusercontent.com/u/13037778/share.7z
Download, edit pippomacd.java (uncomment one of the 2 commented sections to get a step-by-step simulation OR find the best threshold values)
eventually change parameters in pluto A = new pluto(price,time,12,26,9,60); (last one is timeframe)
 complile

Code:

javac pippomacd.java

run

Code:

java pippomacd

[rikigst]

Since macd version of the bot sold too early this week, is it possible to have buy and sell thresholds proportional to the macd value? E.g. sell threshold should decrease (increase in absolute value) when macd is positive, and viceversa. Same way, buy threshold should allow easier buys when macd is positive and should not allow to buy w/ macd is negative

[Lucko]

Do you fill like sharing? I would relay love to have some simulation tool so I can setup bot more easily...

https://dl.dropboxusercontent.com/u/13037778/share.7z
Download, edit pippomacd.java (uncomment one of the 2 commented sections to get a step-by-step simulation OR find the best threshold values)
eventually change parameters in pluto A = new pluto(price,time,12,26,9,60); (last one is timeframe)
 complile

Code:

javac pippomacd.java

run

Code:

java pippomacd

I was away so this is the first time I see this... Thanks for that.

[sukiho]

Was working fine but when I changed from 15 mins to 30 and then back to 15 its calculating the moving averages wrongly. I removed and reloaded but still doing it.

edit: just now I changed to 30 and back to 15 and its now working properly, altho I had done that several times before with no effect. I also tried an older version of the bot and it was showing the same wrong moving averages and still is, so its all a mystery to me. maybe it was using some of the 30 min data to calculate the 15 moving averages?

2nd edit: restarted my computer and it was doing it again, changed to different sample intervals and it eventually fixed itself. something to watch for I guess, always turn off trading if you are making changes and check that it is calculating the moving averages correctly before starting trading again

3rd edit: problem back again and wont go. Im using long moving averages, like over 80, maybe the data collected doesnt go back far enough to make a long moving average accurate?

[EmBison]

Still would love to see a setting to not trade until Gox's 0.6% commission has been cleared. As Goomboo's said, without volatility, day-trading profits on Gox are quickly lost cross-over trade commissions.

[Gimmoroy]

Is there any particular reason to limit sample fetch with 200 ?

[whydifficult]

Is there any particular reason to limit sample fetch with 200 ?

The only reason for the sample fetch is to create initial EMAs, the older the sample the (exponentially) smaller the impact on the first EMA becomes. So if you increase the sample size you are just hammering Mt. Gox for a lot of trade data that impacts your EMAs only by a tiny amount.

This image shows the importance of older data versus newer data in EMA, the specific numbers depend on the weight factor (ie 10 and 21).

[iTrader]

neat idea building a bot as a browser add-on! will have to test it out!

[avtogen]

I want to use that bot, but little bit confused - my transaction fee on Mt.Gox is standard 0,6%. What threshold should I use for not losing money? Default - 0,25?

[whydifficult]

the 0.25 threshold is for adjusting how flat a trend can be. This strategy doesn't take fees into account because it only tries to predict trends and it doesn't know how far a trend will go (and thus how much profit you'll get).

[Gimmoroy]

I want to use that bot, but little bit confused - my transaction fee on Mt.Gox is standard 0,6%. What threshold should I use for not losing money? Default - 0,25?

This have nothing with Mt.Gox fee. Buy threshold and Sell threshold shows not a percentage of price, but only a relative strength of trend. Method, implemented in this bot, suppose that one can choose a trend value to make a decision to sell/buy and get the profit. It can work only when you don't pay any fee at all.  

Also suppose this: price make a big rise. You see a huge trend. You pay all your money (as this bot do). Next day price jump upper and stabilize at high value. Congratulation: you loose all your money !  I call it "Zhopa".

Definitely, to avoid this shit, bot must have two more options:

1. Thresholds in percentage of price between consequent trade operations
2. Limit of money for single trade operation in percentage of amount

 

  
    

[rikigst]

Just an advice for those who are operating the bot: if you want to reduce the time sample, while having the same (slow) ema/macd, just multiply all ema parameters by the factor you've divided the time sample with.
E.g.: macd 1h 12/26/9 is the same as macd 5m 144/312/108, same as macd 1m 720/1560/540
This should guarantee the same behaviour, plus you don't have to trade only @ multiples of the (old) timeframe. You'll have to mess around with options.html and MaxSamplesToKeep before, though.
Plus, i've updated the simulation thing, now its output looks better:

Code:

Output of: pluto A = new pluto(price,time,144,312,108,5); A.profit(0.56,-0.56,true)
2013-05-03T18:00:00.000+02:00 BOUGHT @ 92.59999 BTC =10.734342411916026 0.5654848609507588
2013-05-05T08:10:00.000+02:00 SOLD @ 112.91154  USD =1204.7589458210525 -0.573699710766984
2013-05-07T22:05:00.000+02:00 BOUGHT @ 110.03   BTC =10.883671654513552 0.5648415180770835
2013-05-14T22:25:00.000+02:00 SOLD @ 113.7      USD =1230.0486263154817 -0.5613634113015287
2013-05-15T15:15:00.000+02:00 BOUGHT @ 114.5    BTC =10.678326066005143 0.56118574111735
2013-05-27T20:20:00.000+02:00 SOLD @ 125.6      USD =1333.1505673669044 -0.5822878506110936
2013-06-09T19:25:00.000+02:00 BOUGHT @ 100.3537 BTC =13.20480727485369  0.5821903208364937
2013-06-13T23:05:00.000+02:00 SOLD @ 102.33522  USD =1343.2089563845743 -0.5688030244850157
2013-06-15T12:45:00.000+02:00 BOUGHT @ 103.0    BTC =12.962618472293853 0.5688525282485604
2013-06-21T13:55:00.000+02:00 SOLD @ 108.311    USD =1395.5702043365038 -0.5621901633269558

Download: https://dl.dropboxusercontent.com/u/13037778/pippo.zip
Download, edit pippomacd.java (uncomment one of the 2 commented sections to get a step-by-step simulation OR find the best threshold values)
eventually change parameters in pluto A = new pluto(price,time,12,26,9,60); (last one is timeframe)
 complile

Code:

javac pippomacd.java

run

Code:

java pippomacd

[ewibit]

Code:

.
.
.
warning: ./org/joda/time/ReadableInstant.class: major version 51 is newer than 50, the highest major version supported by this compiler.
It is recommended that the compiler be upgraded.
8 warnings

is this a problem?
TIA

[rikigst]

warning: ./org/joda/time/ReadableInstant.class: major version 51 is newer than 50, the highest major version supported by this compiler.
It is recommended that the compiler be upgraded.


is this a problem?
TIA

try upgrading your jdk to the latest version

[sukiho]

Just an advice for those who are operating the bot: if you want to reduce the time sample, while having the same (slow) ema/macd, just multiply all ema parameters by the factor you've divided the time sample with.
E.g.: macd 1h 12/26/9 is the same as macd 5m 144/312/108, same as macd 1m 720/1560/540
This should guarantee the same behaviour, plus you don't have to trade only @ multiples of the (old) timeframe. You'll have to mess around with options.html and MaxSamplesToKeep before, though.

this bot seems to have some problems when you go above about 90, so I would be careful using those 5 or 1 minute time samples

[rikigst]

this bot seems to have some problems when you go above about 90, so I would be careful using those 5 or 1 minute time samples

Problems? I didn't have any, what do you mean?

[sukiho]

this bot seems to have some problems when you go above about 90, so I would be careful using those 5 or 1 minute time samples

Problems? I didn't have any, what do you mean?

it seems to calculate the moving average slightly wrongly and erratically, assuming that bitcoincharts and others are correct, I posted about it a few pages back

[Voodah]

I want to use that bot, but little bit confused - my transaction fee on Mt.Gox is standard 0,6%. What threshold should I use for not losing money? Default - 0,25?

This have nothing with Mt.Gox fee. Buy threshold and Sell threshold shows not a percentage of price, but only a relative strength of trend. Method, implemented in this bot, suppose that one can choose a trend value to make a decision to sell/buy and get the profit. It can work only when you don't pay any fee at all.  

Also suppose this: price make a big rise. You see a huge trend. You pay all your money (as this bot do). Next day price jump upper and stabilize at high value. Congratulation: you loose all your money !  I call it "Zhopa".

Definitely, to avoid this shit, bot must have two more options:

1. Thresholds in percentage of price between consequent trade operations
2. Limit of money for single trade operation in percentage of amount
    

How accurate is what this guy is saying?