Search
  • Alphachain Capital

ETH Trending Alpha ST 2.0

New performance optimsations to the ETH Trending Alpha ST model on Token Setss.


Firstly a big thank you to all of my set buyers to date. Since it’s inception on February 25th 2020, just three weeks ago, the Eth Trending Alpha ST set has raised $200k in capital and become the best performing social trader set in ETH terms since inception, up 125% at time of writing, while also becoming the current highest capitalised social trader set on the platform. And what’s more, it hasn’t even got started.

As mentioned previously, our team are continuinously researching optimisations to our strategies in order to provide you with the best possible strategy at any given time, while maintaining the core model functionality for which the set was originally intended.

I am pleased to provide you with some important performance enhancements to the ST model which have gone live today. These performance enhancements have made the model much more adaptable to varying market cycles, enabling it to assess the current market structure (trending or sideways) and adapt its trading parameters accordingly. Before we go into the detail, let’s look at how it has affected the back test:

ETH Trending Alpha ST 1.0 (before optimisations):



ETH Trending Alpha ST 2.0 (after optimisations):



So what changed?


  • Net profit increased to 25,159% from 10,095% previously (2.5x increase)

  • Strike rate (percent profitable trades) increased to 73.33% from 67.39% previously

  • Profit factor increased to 8.5 from 5.4 previously

  • Drawdown increased to 18.1% from 10.4% previously

  • Smaller number of trades with longer holding periods

We observe notable improvements across the board with the exception of the max drawdown, which increased from 10.4% to 18.1%. From a risk-adjusted returns perspective, supported by the increased profit factor, the back test shows overall performance has increased significantly. Net profit had increased 2.5x while drawdown had increased only 1.8x, while profit factor increased almost 60%. A higher strike rate also shows that more of the models trades end up profitable than before.


PLEASE NOTE THAT THESE ARE BACK TEST RESULTS. ACTUAL PERFORMANCE MAY BE CONSIDERABLY DIFFERENT TO WHAT THE BACK TEST SUGGESTS. ONLY RISK WHAT YOU CAN AFFORD TO LOSE.


Breakdown by year:

More important than the back test results themselves, the adaptability which the model now shows should help it perform much better over the longer term in varying market cycles — this is a difficult thing to achieve!


So what problem were we trying to solve?


As mentioned in previous posts the problem with short term trending models is that they can often get whipsawed in and out of trades during sideways markets causing losses. It’s impossible to remove these completely, and while the profits of the larger trends generally more than covers the losses of the whipsaws, we wanted to find a way we could limit these losses during these sideways moving markets while maximising the profits from the larger trends.


We have two goals which are at odds with each other:


1. Limit losses from sideways markets


2. Maximise profits from large trends


One way to limit the losses from sidways markets is to have tighter trade exit rules. This means losing trades will close out earlier reducing the losses from those trades. This sound great, however, having trades that close out earlier will also close trades earlier on large up-trends when there might be a small correction before the trend then continues. This correction would stop out the trade early and then wait until a new signal appears once the trend has continued, resulting in a new entry price higher than the recent exit price. This results in lost profit we could have made by just holding the position through the whole move, but helps us reduce losses from sideways action.


To show more clearly what I mean see the chart below from Jan 2020 showing how the previous ST model worked:


While this was a very profitable period for the model, we can see that it entered three different trades during this up-trend. The model misses out on the profit inbetween the previous trade closing and the next trade opening. In large trends, this can result in quite a large amount of profit that the model misses.


Here’s what that same period looks like with the new optimisations in ETH Trending Alpha 2.0:


In the new model, it makes just one single trade during that entire period. In terms of PnL, the single trade is 2x more profitable than the three trades combined from the old model. This can make a huge difference to PnL, as we have seen in the back test results.


Conclusion


By making the ETH Trending Alpha 2.0 model automatically adaptable to different market structures, we have enabled it to maximise the profitability from the larger trends while also being able to close trades early during sideways market action, limiting the potential losses from false trend signals. When the model generates a buy signal while detecting that the market is not currently in a long term up-trend, it will use more conservative risk management rules (closes trades earlier). When it detects a longer-term trend emerging it will adapt its trade exit criteria to give it more room to exploit the profitability of the larger trends.


We believe we have gone some way to achieve both goals and vastly improved the overall performance of the original ST model.


The performance optimsations will be applied to the current ETH Trending Alpha 2.0 as of today and there is no action required from any of the set holders.


Thanks again to all the current set followers and I look forward to sharing the continued growth of the sets with you all.


For any questions you can find me on twitter Adam Haeems or https://t.me/alphachaindefi

48 views

Subscribe to get the latest DeFi fund performance updates and reports

Join the official Alphachain DeFi fund Telegram Group

FAVPNG_telegram-logo_CnYrMEdY.png

© Alphachain Capital Ltd