[SHARE] Simple RSI BULL/BEAR strategy

[embezon]

Open the PDO file to see if it is a valid indicator file.

You can also delete it unless you need it for another strat.

Thanks, solved.
var SMA = require("../SMA");
changed to
var SMA = require("./SMA");


Greetings!

[mvangoor]

I've been playing around with genetic algorithms on the adx strategy using gekkoga... the analysis is still running so these figures are just a work in progress, but thought I'd share my results so far.

The risks with GA is always that it tunes very specifically to the data that you have... I still need to test these settings on data in different time frames - they're just sooo slooow to download! If someone has the data to run this on similar timeframes as tommies stunning results in his first post I'd be really curious to see the results.

EUR:ETH
GA Epochs: 342 (~5min per epoch)
Exchange: GDAX
Data: 23/5/17 - 16/2/18 (9 months)
Market Gain: 392%
Simulated Profit: 983%

Candle Length: 17 Minutes

Parameters:

Code:

# SMA Trends
SMA_long = 500
SMA_short = 42

# BULL
BULL_RSI = 12
BULL_RSI_high = 82
BULL_RSI_low = 61

# BEAR
BEAR_RSI = 11
BEAR_RSI_high = 42
BEAR_RSI_low = 26

# ADX
ADX = 2
ADX_high = 62
ADX_low = 50

Image: https://imgur.com/zsBusoU



I've had some other promising results but they've been tuned over smaller timeframes so need more work.... I want to see what's possible on smaller, more volatile altcoins - BCD and XLM might be next on the list, but I haven't researched too much so I'd be happy to hear suggestions.

Mega thanks to AskMike and Tommie for all their work!

hello

can you share your gekkoga config ? have been trying but after a few rounds it crashes. Also no trades are being executed by gekko when running in gekkoga

Hi Gryphon. (and to all testing...... and testing) 

Just a simple confirmation.
Although your settings appeared to be a bit odd, I backtested several times over different timespan and various assets.
This one is over 1 month on BTC/BNB at Binance.
This is about the average result. Some tests ran slightly below whilst others fairly above.
I'm gllad I tried your settings. The results are way above my expectation. 

[crypto49er]

Let's all be honest here. Backtesting in Gekko is like being a kid and going to a candy store and having anything you want. The caveat is you own the store. Every trade in backtesting executes exactly at close price 100% of the time. That never happens in real life. That also means the insane amount of gains that RSIBBADX can make in backtest also will never happen in real life. It's time we get some more realistic backtesting in Gekko.

So I made some modifications to papertrader. You guys can see the changes I made to try to make it a lot more realistic. The gist is, your trades won't get execute at close price 100% of the time, and it's not just some random number generator that determines if your trades get executed (but adding that would make things more realistic than what we have currently).

Watch here: https://youtu.be/DaUyFLEsMxs

[smiley]

@jack will you post your updated papertrader.js to your github, as retyping off your youtube is a bit cubersome.
Thanks for your hard work on this.

[crypto49er]

@jack will you post your updated papertrader.js to your github, as retyping off your youtube is a bit cubersome.
Thanks for your hard work on this.

You're welcome. Here you go.

https://github.com/crypto49er/moddedgekk...rTrader.js

[martyc]

Let's all be honest here. Backtesting in Gekko is like being a kid and going to a candy store and having anything you want. The caveat is you own the store. Every trade in backtesting executes exactly at close price 100% of the time. That never happens in real life. That also means the insane amount of gains that RSIBBADX can make in backtest also will never happen in real life. It's time we get some more realistic backtesting in Gekko.

So I made some modifications to papertrader. You guys can see the changes I made to try to make it a lot more realistic. The gist is, your trades won't get execute at close price 100% of the time, and it's not just some random number generator that determines if your trades get executed (but adding that would make things more realistic than what we have currently).

Watch here: https://youtu.be/DaUyFLEsMxs

Nice work crypto49er,
Video and code make perfect sense. Did you think about or try to to automate the process of determining the average volume?
Something like this http://www.sqlitetutorial.net/sqlite-nodejs/query/

Cheers,
Martin

[tommiehansen]

Let's all be honest here. Backtesting in Gekko is like being a kid and going to a candy store and having anything you want. The caveat is you own the store. Every trade in backtesting executes exactly at close price 100% of the time. That never happens in real life. That also means the insane amount of gains that RSIBBADX can make in backtest also will never happen in real life. It's time we get some more realistic backtesting in Gekko.

So I made some modifications to papertrader. You guys can see the changes I made to try to make it a lot more realistic. The gist is, your trades won't get execute at close price 100% of the time, and it's not just some random number generator that determines if your trades get executed (but adding that would make things more realistic than what we have currently).

Watch here: https://youtu.be/DaUyFLEsMxs

This is of course true.
Good work with the mods.

[TheBoef]

@jack will you post your updated papertrader.js to your github, as retyping off your youtube is a bit cubersome.
Thanks for your hard work on this.

You're welcome. Here you go.

https://github.com/crypto49er/moddedgekk...rTrader.js

Im running a backtest for 4 days now with this paperTrader. Last six months USDT / BTC pair and still no config found that has more profit then the market :o

[tommiehansen]

@jack will you post your updated papertrader.js to your github, as retyping off your youtube is a bit cubersome.
Thanks for your hard work on this.

You're welcome. Here you go.

https://github.com/crypto49er/moddedgekk...rTrader.js

Im running a backtest for 4 days now with this paperTrader. Last six months USDT / BTC pair and still no config found that has more profit then the market :o

4 days?

Well... that's your problem.

[PGTART]

Hi Tommie Hansen
Letting you know that I used your code and extended it and shared it publicly.
There was allread the extension _PINGPONG and I added _PEEKTREND
So it got a long name now ) the added idea is to have added a stoploss mechanism that follows the market
( if market goes up so does the stoplevel, market falls stop level stays and will eventually be triggered).

https://github.com/PGTBoos/GekkoStrategies
a bit more about it here : https://forum.gekko.wizb.it/thread-57996.html