[SHARE] Simple RSI BULL/BEAR strategy

[tommiehansen]

Just out of curiosity -- how many of you, or are any of you, running this strategy and finding it profitable, with live trading?

I'm running it since a few weeks now on ETH Poloniex, it's giving mixed results, but generally has made a few % profit.

Do you mean overall after a few weeks its profitable by a few %, or it's had a few profitable trades, but is not making a profit overall?

I hope it's the first, and that "mixed results" means there are a few (or even quite a few) loss making trades. That's to be expected!

Yeah after a few weeks I had a few % profit, but it makes a single "mistake" and that's all gone then. I'm having trouble figuring out why it doesn't take better advantage of sharp rises and falls, for example this short period:



Trying to see why it fails to sell when the fairly steep dip happens around 6 AM..

Just out of curiosity -- how many of you, or are any of you, running this strategy and finding it profitable, with live trading?

I'm running it since a few weeks now on ETH Poloniex, it's giving mixed results, but generally has made a few % profit.

Do you mean overall after a few weeks its profitable by a few %, or it's had a few profitable trades, but is not making a profit overall?

I hope it's the first, and that "mixed results" means there are a few (or even quite a few) loss making trades. That's to be expected!

Mixed results are always to be expected as you say. If one backtests long periods it's easy to see that there sometimes do occur a couple of bad trades in a row etc.
No risk, no reward.

Agree! And I'm fine with that, but I find that testing with a long period skews the results of the short term last period, market is behaving very differently the last month vs a year ago vs last December. I've seen results with 1000's of %'s of profit but applying those strategies to for example just this year gives you a negative result!

You're talking about the strategy as if there is a fixed setting that "just works"?
This is not the case, quite frankly -- this isn't even remotely close to the case.

It will always fail to do what's the absolute greatest thing to do, every single time.

It's not about timing a specific point in time perfectly, it's about doing more good then bad in the long run.

-

I've done basically 10's of thousands of backtest. Just for the XRP/USD pair i've done over 15 000 backtests (via my backtesting tool GAB) for the period 2016-01-02 > 2018-02-18 at 5 min candles

From all these runs and for all runs for all other coins the summary is basically this (5 min candles):

1. Large coins are worse e.g. BTC/USD and ETH/USD (but ETH often fares much better)
2. A shorter long SMA is usually better but often also has a higher risk. A general rule of thumb here is a value of around 300.
3. Short SMA of 10-30 usually works better.
4. RSI length of 15+ usually works better
5. ADX length of 10+ usually works better
6. 5 min candles simply yields the best results (but i haven't done 10 000+ runs for all time series)
7. etc..

Here is the average for the most profitable runs out of 15 034 runs XRP/USD @ Poloniex between 2016-01-02 > 2018-02-18 where it tops out at 12 trillion % and the lowest is at 1 trillion %. Note that this is by using 5 min candles:

Code:

[SMA]
long = 245
short = 16

[BULL]
rsi = 22
high = 76
low = 48
mod_high = 7
mod_low = -8

[BEAR]
rsi = 21
high = 53
low = 35
mod_high = 17
mod_low = -8

[ADX]
adx = 19
high = 75
low = 23

Do also note that this is simply the average settings for all those 100 top performing runs.
One wouldn't want to directly use these values but rather round them to more sane values e.g. replace RSI 22 with 20 or 25 etc.
Basically use it to get an idea on what works better and don't.

These are the averages for the top 50 sorted by win% where it tops out at 8 trillion % and where the lowest of these 50 is 1 million %, also at 5 min candles.
The average win% (number of winning trades vs number of losing) has an average of 86% for this toplist:

Code:

[SMA]
long = 220
short = 21

[BULL]
rsi = 29
high = 76
low = 31
mod_high = 7
mod_low = -10

[BEAR]
rsi = 27
high = 55
low = 24
mod_high = 21
mod_low = -7

[ADX]
adx = 17
high = 75
low = 26