Re: Haasonline Simple Trade Bot For BTCe and Bitstamp[Main topic]
These last days I've been spotting a room for improvement when it comes to MACD. I don't know if this is very financial as I'm not really deep into stocks, but I believe it would improve overall behaviour and that it makes sense.
Take for example this MACD right out of my bot:
http://i.imgur.com/HGLqMEq.jpg?1
I have highlighted what I would call 'Phantom signals'. That is, places where there is either a sell/buy signal that only lasts 1 cycle, and that eventually cause the bot to sell/buy when in reality it's not certain where the market is heading. What I would like to suggest is to have a configurable amount of 'consecutive' signals that you want to require before the bot takes the final sell/buy decision. I can't guarantee this results in more profit, but I have the feeling it might have. And if it's optional, it would just be an extra thing we could try, and the default would remain like it is. I believe the bot has done some poor decisions at times because of this, you only need to look at the last one I have highlighted, that's 5 diferent buy/sell decisions so it's clear that with that data we can't really make a buy/sell decision.
Also, did you read my previous suggestion?
Take for example this MACD right out of my bot:
http://i.imgur.com/HGLqMEq.jpg?1
I have highlighted what I would call 'Phantom signals'. That is, places where there is either a sell/buy signal that only lasts 1 cycle, and that eventually cause the bot to sell/buy when in reality it's not certain where the market is heading. What I would like to suggest is to have a configurable amount of 'consecutive' signals that you want to require before the bot takes the final sell/buy decision. I can't guarantee this results in more profit, but I have the feeling it might have. And if it's optional, it would just be an extra thing we could try, and the default would remain like it is. I believe the bot has done some poor decisions at times because of this, you only need to look at the last one I have highlighted, that's 5 diferent buy/sell decisions so it's clear that with that data we can't really make a buy/sell decision.
Also, did you read my previous suggestion?
I would like to suggest Stephan to turn that option (ignore raise/drop safety) into 2, separate ones, one for Buy and one for Sell. That would allow to ignore only one of the 2 safeties, in my case the one I want to ignore is the buy safety, and then the bot would always try to make profit. If the market reversed strongly, the stop-loss function could work in conjunction and prevent losses.
This would prevent having to find 'the perfect' settings with ignored safeties like right now, because honestly, I am quite not succeeding, and I tried a fair amount of combinations
This would prevent having to find 'the perfect' settings with ignored safeties like right now, because honestly, I am quite not succeeding, and I tried a fair amount of combinations