Stat Help - anything that works?!
#1
Hey guys,

I know we are all in the same boat of trying to game a unpredictable market. Im not looking for a one size fits all promised return solution.

With that said I have been running backtesting on every strat I can find and NOT A SINGLE ONE have shown ANY positive returns.


Can anyone help point me in a direction? What am I doing wrong here?
  Reply
#2
Have you changed the default values? You just need to keep testing with different values, different coins and in different market conditions.

There is no one size fits all strategy, changing a single value even by a single number can vastly change the returns.

There a few depositories for strategies and brute force testing around.
  Reply
#3
Every back test strat i run produces negative results. I am convinced I am doing something wrong. Its like if I could just invert the buy/sells the back test would actually show positive results.

So confused. can someone help?
  Reply
#4
If you want to invert the buy/sells then go into the strategy and edit it?
You don't need to be an expert coder to pick up the basics, try the BBRSI strategy in the repository, you should be able to produce a positive result with that in various market conditions.
Start at around a month's worth of data (try and get a mix of bull and bear trends) and start backtesting the candles starting at 1 min and working your way up to an hour in 2 or 3 minute increments.
  Reply
#5
(05-22-2018, 03:37 PM)Ballsy Wrote: If you want to invert the buy/sells then go into the strategy and edit it?
You don't need to be an expert coder to pick up the basics, try the BBRSI strategy in the repository, you should be able to produce a positive result with that in various market conditions.
Start at around a month's worth of data (try and get a mix of bull and bear trends) and start backtesting the candles starting at 1 min and working your way up to an hour in 2 or 3 minute increments.

Sometimes I feel like its just me and you here ol' buddy! Smile

Can anyone show me a stock strat- modified or not that will back test with a positive result?

ballsy - not sure how to add new strats from the repo yet - just getting start here Sad
  Reply
#6
A member on here has made a great folder of strategies.

https://github.com/xFFFFF

Go to the link above, then goto the strategies folder, in there you'll find a whole heap of strategies, in each folder you should find a *.js file which goes in the strategies in gekko and a *.toml file that goes in the config/strategies folder.
Now when you goto backtest you'll find the strategy in your list and adjustable parameters (from the *.toml) will be in the window to the side.

It's just a case of try different strategies to suit, try the BBRSI and the bull/bear add strategies. You should definitely get some sort of profit from them but it depends on the parameters you choose, this can take a while to backtest, backtest and backtest some more. Nobody will want to do your backtesting for you, or provide you with a finally tuned winning strategy.

On a side note you need to remember that backtests will only provide examples they don't predict future trends so even though it gave out 100000000% profit for the past week doesn't mean you'll get it the following week.
  Reply
#7
(05-22-2018, 07:43 PM)Ballsy Wrote: A member on here has made a great folder of strategies.

https://github.com/xFFFFF

Go to the link above, then goto the strategies folder, in there you'll find a whole heap of strategies, in each folder you should find a *.js file which goes in the strategies in gekko and a *.toml file that goes in the config/strategies folder.
Now when you goto backtest you'll find the strategy in your list and adjustable parameters (from the *.toml) will be in the window to the side.

It's just a case of try different strategies to suit, try the BBRSI and the bull/bear add strategies. You should definitely get some sort of profit from them but it depends on the parameters you choose, this can take a while to backtest, backtest and backtest some more. Nobody will want to do your backtesting for you, or provide you with a finally tuned winning strategy.

On a side note you need to remember that backtests will only provide examples they don't predict future trends so even though it gave out 100000000% profit for the past week doesn't mean you'll get it the following week.

Yeah im not expecting or looking for anyone to share their 10000% profit strat.

I got the repo of strats installed. a bunch of them do not work. they error out looks for one file or another that doesnt exist.

I spend 3 hours last night running LTC back tests and not a single strat made any profit at all no matter what data set i used or which strat I manipulated. 

I am just looking for someone to share a back test. any backtest that actually proves a profit possible. I am losing hope at this point.
  Reply
#8
@pingywon,

I had that same feeling when I first played around with Gekko late last year after figuring out how to install it. All the included strats (w/ exception of PPO, I haven't tested that too thoroughly) perform poorly. Your objective here is to figure out how to get BBRSI or RSI Bull Bear working on your computer instead of trying to backtest the included strats to make them work.

Once you get them working, you should be able to get backtest results that are profitable.
If it isn't crypto, it isn't worth mining, it isn't worth speculating.
https://www.youtube.com/c/crypto49er
  Reply
#9
(05-23-2018, 06:49 PM)crypto49er Wrote: @pingywon,

I had that same feeling when I first played around with Gekko late last year after figuring out how to install it. All the included strats (w/ exception of PPO, I haven't tested that too thoroughly) perform poorly. Your objective here is to figure out how to get BBRSI or RSI Bull Bear working on your computer instead of trying to backtest the included strats to make them work.

Once you get them working, you should be able to get backtest results that are profitable.

Thank you for the advice. I am not complaining or look for anyone to GIVE me their strat but I have literally spent hours doing backtesting (with limited knowledge admittedly) and I have to see ANYTHING that provides a positive result. 

I haven'teven seen anyone (all over the net, not just here) boast about their strat actually providing them with profits. Backtests, sure but that could be just dumb luck in manipulating the backtest till it provides something.

I am beginning to think that Gekko is a great framework and idea - but not actually making anyone any $, and that is not fault of gekkos - the more I look around I dont think there are any bots (even paid ones) that a really performing great.
  Reply
#10
Well, at least you didn't spend a ton of money on Gekko like you could have with those paid bots.

I actually had RSI Bull Bear ADX make profitable trades for me. But I was trading $200 so I wasn't going to brag about $10 gains. I since changed the parameters because I was using the UI and it doesn't save the candle size (that affects performance a lot) info and it has been making losing trades since. I think the original settings were 8 minute candles but that made it impossible to check if the bot is performing properly with a free TradingView account.

I switched over to paper trading now via command line, on 5 minute candles, still saw it made some bad trades. It seems like the bot was trading in bull trend when it was bear but I can troubleshoot it now because it is on 5 minute candles. I literally turned on debugging for the strat as I type this. Hopefully I will get more info on the issue.

Ultimately, I personally feel like I learned a lot with the Gekko Trading Bot. I always seen it as a tool to trade for me when I'm not at my computer. It also doesn't have emotions and doesn't trade earlier (I do that too often) than I'm supposed to. But making it work brings up a whole new set of challenge that I'm only beginning to tackle.
If it isn't crypto, it isn't worth mining, it isn't worth speculating.
https://www.youtube.com/c/crypto49er
  Reply


Forum Jump:


Users browsing this thread: