Quote from: cpusher on March 29, 2014, 09:39:48 AM

0992 is a big improvement for me, much less memory usage so i can do a lot of backtesting/tuning thanks for this!

just updated to 0993, and 1 configured bot is missing now (2014-29-03 10:24:57.398 |ERROR| There is an error in XML document (79, 38).)

from last 12 hours i have the following issues:

on CEX.io i started with only btc available in my wallet with only HTS trading:
I had 4 bots running (all the available btc pairs)with a static amount. for nmc i now have 3x trade NMC amount in wallet, for ghs i have 2x trade amount in wallet. and had to cancel 5x buy order ( 5x trade amount) and 2x sell order (2x trade amount) that went out of control. (I did not used the new timeout insurance) Something is seriously wrong here Huh

(already experience same behaviour on cryptsy a while ago with HTS)

on cryptsy i had fee overcome active with a small loss (this is fixed with 0993?)

deactivating bots until 0994...

I am not really sure what has happened to your trade-bot. We did make some changes to the save system and its behind automated tests to. I did not expect to see anymore trade-bots that gets lost. (please note, inside your TradeBots -folder there is almost always a backup of your trade-bot in case you lost it. You only have to rename it back to .XML to restore the trade-bot.

At the current moment we are not (yet) planning version 0.9.9.4. The current version is looking to be working smoothly for now (except Hpotsirc below here, but that will be solved). So we already started to continue development on the new exchanges and features over here.

For Cex.IO i have 2 good recommendations. First of all try to trade with low amounts like 1 or 2 coins per bot. Secondly i can recommend to use the CancelOrder and set it to 3 minutes (or 4minutes). By default we are at 10minutes and this is to long for Cex.IO at the moment.

With these recommendations is it still not fully sure if the orders get trough, but it gets you really close. I have the same issue here and this is the solution as far as i have seen.

Version 0.9.9.3 has a solid feeovercome insurance. You can even test it yourself, it just a matter of setting up a trade-bot with the BlindSignal indicator and the insurance that you want to test. Then run a backtest and see what happens. (its for 99.9% sure that you see earnings/profits with this test setup, but still beware of the BlindSignal indicator, it does not give momentum. It only gives a blind signal to trade, so you are trading on the fee alone with this setup.)

Quote from: Hpotsirc on March 29, 2014, 02:36:26 PM

With the new version i still have problems with bitstamp. The bot executes the orders and eventually tells me they were filled and changes the coin position but my wallet amount doesn't change at all. When i log in to Bitstamp i see that no order was received at all. At least the message that my trade amount exceeds my wallet amount is gone.

I have also got some problems with the backtesting function. It seems to give me completely random results. I let the backtest run, clear my statistics, switch back the coin position to its original state, change no settings at all, let the backtest run again and get completely different results. The results range from nice profits to huge losses so i really can't trust any of these results at the moment.

Another problem is with the auto setup of indicators: For some indicators this works, for example MACD. I add the last x prices and press auto setup, the graph and the tuning of the indicator changes. For others, e.g. Aroon the graph does change once i press auto setup but the indicator settings won't so i can't really comprehend what the auto setup actually did.

Can you please contact me per PM or support ticket about bitstamp? I would like to see why this does not work or whats happening at your computer.

About the backtest... there are 2 reasons why you might see different results. First of all the start position of the coins (you saw this one already) and secondly the last record buy and sell prices. However, if you press Clear, Backtest and Clear again then the next backtest should show exactly the same results. This output must the equal.

About the auto tuning... this technology works very well, but its not said this technology is prefect. What the auto-tune does is trying to make the indicator active so they become responsive within 24-hours. Once there is a profitable buy and sell signal the tuning itself stops. However, sometimes this is just not possible. Maybe the market does not move enough or the settings of the indicators would become invalid. I that situation the auto-tune fails.

Just to tell... the auto-tune as it is right now works well, but we are still only at 33% of what is possible. We still have a few tricks to make this tuning even better that it is right now.

Quote from: krach on March 30, 2014, 08:23:42 AM

Quote

in 0.9.9.3 I see big cpu usage in comparison with the previous version

maybe try to change the buffer size in the server settings.

It is possible to see a little higher CPU usage if you compare version 0.9.9.2 and 0.9.9.3. If just the fee that has been read/verified each time. But this difference is very, very small. I did not even register it in the tests. (i need to say i have not updated my Atom here yet)

If you have performance problems then i recommend to:
- Reduce the memory usage (see server settings)
- Reduce the cache buffer size (see server settings)
- Disable the exchanges that you are not using (see server settings)

But i do not expect this would be needed at all, unless your memory is very limited or unless you want to reduce internet bandwidth.