[SHARE] Simple RSI BULL/BEAR strategy

[crypto49er]

Answers.

Q1: Yes
Q2: Yes. For sticky orders, which is the only type that is available right now. https://gekko.wizb.it/docs/gekko-broker/...order.html
Q3. Yes. Some one needs to code a market order, so that it either executes or gets cancelled if it isn't filled.

I wish I have the coding skills to do it.

[jasonbtc]

Has anyone actually tested this strategy on bear market ? I am experimenting a bit but I am not getting good results yet.

[crypto49er]

I have. I ran it for about a month in July and August. It is too frustrating a strat because the win rate is low. It usually does something like this.

1st roundtrip: -2% loss
2nd roundtrip: -3% loss
3rd roundtrip: -1% loss
4th roundtrip: 8% profit
5th roundtrip: 2% loss
...

I still believe it has potential, but you literally can't look at the results as there are way more losses than wins. It also needs the following adjustments:
-Buy only when RSI is going up, not just hitting a set number (often it buys when RSI is going down)
-Buy only when a candle hits a certain volume (I think someone else suggested that earlier)

I'm working on other stuff now, but I will eventually add those to the strat and see how it works. They are not that hard to implement, anyone who modded a strat previously should be able to do it.

[mark.sch]

Wow, here is a really active discussion about the RSI BULL BEAR strategy. I am taking a wider look into indicators and strategy concepts these days. Since this strat has such a popularity, are there any performance figures using it on ETH (vs. EURO or USDT) from the beginning of this year until let's say 2018-09-07 ?
To get a reference when things are really optimized using this strategy. E.g. if it gives a ROI of +80% then another strategy approach needs to beat this. Of course this does not mean it will perform well on another coin also.

@crypto49er (small) losses in july / august also disappointed me, so I added the Triple EMA (TEMA) indicator yesterday to get a smooth trend line and to measure trend strength using the current and previous candle TEMA (taking the TEMA of two weeks). This improved the existing strategy a lot since I have a chart segmentation between strong bull, strong bear and sideway trend. Long term MACD or RSI did not perform nice for me for this approach. Right now I am trading on strong bull only - this stopped the frustrating july/august trades. Will now look for a suitable strategy (probably using another candle size) to handle the sideway trends and pick gains during this time - I consider to use Bollinger Bands for this.

Also worth to look at:
https://medium.com/@deandree/can-you-pro...bc97a5f960

[martyc]

I have. I ran it for about a month in July and August. It is too frustrating a strat because the win rate is low. It usually does something like this.

1st roundtrip: -2% loss
2nd roundtrip: -3% loss
3rd roundtrip: -1% loss
4th roundtrip: 8% profit
5th roundtrip: 2% loss
...

I still believe it has potential, but you literally can't look at the results as there are way more losses than wins. It also needs the following adjustments:
-Buy only when RSI is going up, not just hitting a set number (often it buys when RSI is going down)
-Buy only when a candle hits a certain volume (I think someone else suggested that earlier)

I'm working on other stuff now, but I will eventually add those to the strat and see how it works. They are not that hard to implement, anyone who modded a strat previously should be able to do it.

I'm doing a bit better than that but I have tweaked the code a bit.

1. Added a couple of lines that check for msSlow/maFast and if they are too close then I ignore any signals. (Theory is that there is a high liklihood of getting whipsawed back out again)
2. Calculate the MA of volume and if the volume is < the MA of the volume I also ignore signals. (Again theory is that I want a 'real' change in sentiment to be what drives my signal)

Cheers

[crypto49er]

@Mark.SCH

I think there are some backtest data for ETH/USD trade pairs around page 4 or 5 of this thread.

I like your TEMA idea. I also agree that in this bear market, we can only trade on strong bull bounce (and then get out ASAP as soon as it starts falling). I basically created a scalping strategy that do just that. Still working out the kinks, but at least it doesn't lose more than 0.5% per trade (not a hard stop loss, it is based on the indicator) when it does lose.

@martyc,

Nice! I think you should have it send a message to the console when it ignores a signal so you can then check after a few candles if ignoring the signal really saved your bacon. I'm sure it did but it's just a positive reinforcement on your modifications.

[mark.sch]

@Mark.SCH

I think there are some backtest data for ETH/USD trade pairs around page 4 or 5 of this thread.

I like your TEMA idea. I also agree that in this bear market, we can only trade on strong bull bounce (and then get out ASAP as soon as it starts falling). I basically created a scalping strategy that do just that. Still working out the kinks, but at least it doesn't lose more than 0.5% per trade (not a hard stop loss, it is based on the indicator) when it does lose.

Thanks @crypto49er, I found some backtest results at your mentioned pages, unfortunately they are from 2017. @askmike mentioned a strategy contest soon (for Gekko Plus), this will be very exciting to see, to get an idea of strategy potentials.

A further gain beside TEMA was the implementation of inverse fisher transformation (CCI) indicator for Gekko, however, it does not show the same potential as it does on Tradingview. There are several things left I am going to implement/test...but some figures on ETH/EUR right now (from 2018-01-01 till now)...at least losses are pretty reduced

2018-09-10 21:08:58 (INFO): (ROUNDTRIP) REPORT:
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) entry date (UTC)   exit date (UTC)   exposed duration P&L profit
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-01-02 02:00 2018-01-11 00:00 9 days           474.19 47.49
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-01-12 19:00 2018-01-14 13:00 2 days           78.91 5.37
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-01-28 05:00 2018-01-29 20:00 2 days           47.53 3.07
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-02-09 18:00 2018-02-10 16:00 a day           -62.67 -3.94
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-02-14 12:00 2018-02-18 07:00 4 days           95.48 6.25
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-04-08 05:00 2018-04-09 21:00 2 days           1.50 0.09
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-04-10 01:00 2018-04-10 08:00 7 hours         -11.82 -0.73
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-04-10 14:00 2018-04-16 12:00 6 days           453.74 28.27
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-04-18 21:00 2018-04-25 09:00 7 days           482.27 23.46
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-04-26 23:00 2018-04-30 21:00 4 days           53.33 2.10
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-05-03 03:00 2018-05-06 09:00 3 days           253.54 9.81
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-06-03 05:00 2018-06-04 10:00 a day           55.11 1.95
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-06-19 11:00 2018-06-21 16:00 2 days           -1.04 -0.04
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-07-02 14:00 2018-07-03 23:00 a day           -68.95 -2.40
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-07-05 05:00 2018-07-05 15:00 10 hours         -43.77 -1.56
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-07-08 00:00 2018-07-09 13:00 2 days           -41.90 -1.52
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-07-16 10:00 2018-07-18 20:00 2 days           104.29 3.85
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-08-28 14:00 2018-08-30 02:00 2 days           -24.12 -0.86
2018-09-10 21:08:58 (INFO): (ROUNDTRIP) 2018-09-01 15:00 2018-09-03 05:00 2 days           -57.21 -2.06
2018-09-10 21:08:58 (INFO):
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) start time: 2018-01-01 00:00:00
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) end time: 2018-09-06 00:01:00
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) timespan: 8 months
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) exposure: 0.20480453403748308
2018-09-10 21:08:58 (INFO):
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) start price: 639 EUR
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) end price: 195.73 EUR
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) Market: -69.36932707%
2018-09-10 21:08:58 (INFO):
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) amount of trades: 39
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) original balance: 1000.00000000 EUR
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) current balance: 2725.90770746 EUR
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) profit: 1725.90770746 EUR (172.59077075%)
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) simulated yearly profit: 2541.8269382434 EUR (254.18269382433996%)
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) sharpe ratio: 19.024849238650347
2018-09-10 21:08:58 (INFO): (PROFIT REPORT) expected downside: -2.284959663620867

[jasonbtc]

Added insane backtest results for ETH-USDT for RSI_BULL_BEAR_ADX,

Market: 17950.26650%
Simulated profit:  4449031.09720% (yes, that's over 4.4 bloody million percent)
screen: https://i.imgur.com/L2iPfHV.png

Do note that this is just with default settings as well, no optimizations has been done in order to improve the result.
This extreme performance is so insane though that one has to ask about the accuracy of the test; it's just too good to seem plausible.

Nice.. but the settings ?? Are you sure they are the first ones you posted?
Can you just copy paste them again here?
Thanks

First one for the ADX-mod, yes.

---

The ADX-mod has been updated with a TOML-file and some fixes (e.g. removed this.settings in scripts since no longer needed and some cleanup).

---

Edit: Added an even more insane (implausible?) backtest for ETH-USDT:
Market: +17 950%
Simulated Profit: +83 573 824% (yes, over 83 million percent)
Screen: https://i.imgur.com/bB33jX4.jpg

It is interesting that these settings do not produce good results anymore. I backtested USDT-ETH with these settings for 1,5 years backwards from now (Sept2018) and the results were poor.

[crypto49er]

@JasonBTC,

You're right. I didn't want to believe at first but it is pretty obvious. Running at every 5 minutes is a perfect way to drain any profits as the strat rides it to the bottom. The strat also has a flaw where it attempts to buy during the bull trend after a peak. In a bull market, that will probably work great as its more a respite for the bulls before they push to new highs. But during bear market, it's like throwing good money after bad by buying as the market turns bearish and then it ends up holding until the it hit bear high, at which point, you are already down 5 - 10%. The end result is it performs worse than buy and hold.


  Screen Shot 2018-10-02 at 9.10.49 PM.png (Size: 53.94 KB / Downloads: 169)

There are ways to fix this as others have mentioned. But without applying any fix, you can probably increase the candle size to 1 hour or higher and run new backtests to find the correct parameters.

[jasonbtc]

Do you use any automated tool for backtesting or just manual attempts?
I use an automated backtest tool , but then again to fully explore all possible combinations of parameter settings requires a lot of computing power and a lot of waiting time.
Any hint on how to manage this situation ?