Flawless victory strategy (Patreon)

Introduction

Hi there, and welcome to another blog repost where I will test and analyse a trading strategy. This time I will test a strategy that I found on the internet called the “Flawless victory” strategy.

Strategy information

This strategy was originally published by Bunghole on Tradingview and the Author of the original Pinescript strategy is Robert Roman.

Now the strategy is also described in the Tradingview information section and goes as follows:

The buy signal is triggered when close price is less than lower Bollinger Band at Std Dev 1, and the RSI is greater than a certain value (now clue at the moment what this value is though).

The sell signal is triggered when close price is greater than upper Bollinger Band at Std Dev 1, and the RSI is greater than a certain value (again, no indication of this value given here).

The author also states that this strategy is interesting because the parameters were optimized by machine learning and finally gave the best Sharpe Ratio when backtesting. He left his computer on for about 28 hours to fully backtest 5000 EPOCHS and get the results. Now I wonder how he found these optimized parameters and burned his CPU to its core…

Anyway…

The actual code for backtesting on the Freqtrade trading bot was actually developed by Sean Nowotny and I will use his code to see how this, spot trading only, strategy will perform. The code Sean made was also actually the reason I found this strategy by the way.

I also did a quick glimpse to his code and he programmed this strategy to work on the 15 minute timeframe. Like the Tradingview setup. He also has set his stoploss to infinity and did not provide an ROI point here.

So, please know that all kudos go to these people I mentioned just now and if you think it’s all worth it, then give them a compliment or a thumbs up or something else that shows your appreciation to their work.

So let’s get started and do some backtests to see if I can confirm on my personal setup that this strategy.

Oh, and before I forget. Some additional information here. The strategy was created to only trade on the Bitcoin USDT pair, but as you might already know, I have 50 pairs configured. So keep that in mind when watching the backtest results later.

Finally, I also had to change the original code of the author a little bit to make it work on my machine since the trading bot has made some slight alterations in the way the strategy file should look to make it work on the current version.

Initial backtest results

initial backtest results

So, after my initial backtests I got the following results back. As you can see the 15 minute, 30 minute and 1 day timeframe are the only timeframes that got some positive results. All with around 60 to 65 percent winrate.

On the other hand the drawdowns are dramatic on all timeframes and the ratio’s are also not that great at all. In my case only 44 percent of the coins was susceptible to this strategy. But you’ll have to remember that this algo was primarily build for bitcoin trading and no other pairs.

I should then expect to see that Bitcoin/USDT would show up as one of the main winning pairs.

But unfourtunately here it ended way down in the bottom losing pairs instead. Not the greatest loss, but also not a winner according to the strategy.

The best 10 performing pairs are shown here. All altcoins.

The profit plot is noteworthy here. This strategy managed to start its drawdown on almost the peak of the last bull market. And you can see that from the moment the drawdown starts, the underwater plots get larger and larger. There was a point that this strategy got some real quick and nice gains up to 15k. But all those gains are largely given back to the market. The pairs traded are also quite largely divided here between pairs that have losses and gains.

What I find the most remarkable though its is degree of parralellism. From the moment the bot starts, all 10 slots for trades are used and there is almost no point in time where this lowers. This could be an indication of overtrading and a weak spot with this strategy.

But maybe I am a little bit too pessimistic here and should give the strategy another optimization again. Although it is stated by the original author that is already is optimized, It is worth a try too see if the optimization should be actualized to match the latest conditions.

So, let’s cranck up the PC and let those cores burn once more…

Optimizing the strategy one more time

After 500 epochs and letting my PC burn through it’s cores around 29 hours I had another set of optimized parameters too try and see how the strategy would perform with these settings. Mind you that we are now entering overfitting territory so these results have a high chance not happening in the future at any time. So be warned about this. However it should be interesting to see where the limits of this strategy are and if bitcoin trading will be better with these adjusted settings.

Endresults

So I backtested one more time and got the following results back from my backtesting engine. Again, be aware that these results have a high degree of overfitting and have not guarantee of future performances.

Optimized backtest results

However, optimizing the buy and sell parameters did manage to make this strategy perform better than the original, and already optimized, settings. Sharpe has been increased as with the winrate and the amount of profits. But still the Drawdown is very very high and the applicability to multiple pairs also got worse.

Trading Bitcoin still does not perform well with these settings. And I wonder how the author got his original results back then.

The most noticable thing here is the profit plot again of this optimized strategy.

I have never seen this kind of chart before. If I take a look at the peak of the profits, there is almost 40k of profits but in the end it is all given back to the market again and you would end up at around 5000 hypothetical dollars.

The way the pairs react to this strategy is also something I have never seen before. At some point all the pairs are done trading to a specific level and gains or losses are kept there and make a horizontal line.

But according to the parralellism chart there is still some heavy trading and I’m not sure which pairs are traded there because there seem to be no gains or losses to all of them.

A very strange behaviour and certainly in need for more investigation if you still like the strategy itself.

Strategy league

Nowadays a strategy really has to perform well to end up at least in the top half of the strategy league. And with all those really good strategies tested in the past it is no wonder that with these scores the Flawless victory algo will not and up very high in the league.

As far as I can say with the current results, the strategy has a possibility to make you money. But it comes at high risks and I’m not sure if I want to take such risks with my money.

End of the strategy league

I can imagine why, at that time the author of that strategy was very optimistic. At some point the original settings had over 15 k of profit. But from that moment on the strategies edge was gone and it only gave the gains back to the market.

I think that I can conclude here that I found a very clear example of a trading strategy that had an edge once but lots it and the gave almost all gains back to the market again. Easy come, easy go as they say…

Files and other backtest output

Nonetheless, If you like this strategy and want to use this on your own system, then please pay the original author’s Github page a visit or see the Tradingview pine script for more information.

All the output files, including the strategy are available in this earlier post: https://www.patreon.com/posts/84663753

Thank you for reading this post.

And I’ll see you in the next one…

Goodbye!