Many traders focus all their efforts on finding the perfect algorithm that will perform well under all conditions. We believe no such thing exists. Our approach is to generate different horses for different courses: to validate an array of strategies and tactics in a rigorous, procedural way, and read the market conditions to decide which one to deploy at which moment.
Firstly, we track factors and indications which suggest the likely state of the market over a given period. The list of indications is constantly refined to improve validation. Determining the market state will serve as the first layer to the overall strategy utilisation.
A trending market = Imbalance or price disagreement
A ranging market = Balance or price agreement
Once we have determined the market state, we look at our ever-evolving catalogue of strategies and pick the strategy best fitted to the market state. Some strategies work better in ranging markets, and some in trending markets, so fluidity and adaptability are the keys to successful trading. We modify our approach in relation to the ever-changing conditions of the market, to perform well — and deliver alpha — under a greater range of conditions.
This approach is different to the traditional method of back-testing a set of variables and then unleashing it in an ever-changing fluid system. Risk utilisation is also altered under this framework unlike a static approach.
Does this mean we do not back-test? Of course not. We have a team of analysts and automated systems back-testing a multitude of different variables, tactics, and historical theme patterns to gain insight into what works and what doesn’t work. Where we are different is in our adaptability in identifying changes in the current market and changing our approach accordingly.
23rd May — 8th July 2021
Balanced State Strategy 1: 31.9%
Bitcoin Performance: -18%
Above shows how one range-bound long/short BTC strategy we are developing would have performed against BTC over a certain period. (Note, these results do not take into consideration DEX mechanics, and only show results through the initial testing phases on centralised exchange data. Given the complexities of shorting on a DEX, we are likely to offer long-only strategies until we have nailed the long/short strategy: a core focus of our research efforts now.)
Before we include a tactic or strategy in our sets, we begin with back-testing variables and patterns. We then tweak them until we find the optimal combination parameters. Then and more importantly, we live-test them in real market conditions to see how they really perform. This live-testing phase takes longer than the back-testing phase, as we gather real trading data and indications to refine the parameters. Once we are satisfied with the refinement process during the live trading results, we present it as a working, alpha-producing strategy, and deploy it onto the DEX, exposing it to the unique dynamics and mechanics of a DEX. During the live testing round in the DEX environment, we get a feel for the liquidity, speed, and spread issues which we can’t capture in the backtests. Once the strategy has passed this round, then it is finally presented as a finished product.
Our approach takes longer, but is more enduring and hopefully more rewarding. We could back-test a set of variables over a dataset and ‘show’ that it produced results in the past, and then send it into the present hoping for the same results. However, we chose a more situationally-aware technique, carefully choosing the right ship for current weather conditions, to steer you to the true alpha where we can find it. The metrics we use are sophisticated, involving indications from on-chain data as well as CEX data, Quant methods as well as human intuition, and we spend time looking for patterns in history as well as identifying patterns and changing patterns in the current state.
Below is another backtest result in ETH. This one is a bit different in character to the one we showed above. The one below is what we would call a ‘tactic pattern’, and should be transferable across different coins with some tweaking of the core parameters, whereas the first one above is exclusive to BTC.
Here we show what the net profit/loss would have looked like over a sample set of 30 trades (1st chart), and what the individual trade profit/loss breakdown was (2nd chart).
We did not overlay the ETH performance over the same time period, as this 30-trade sample set was produced in a two-week trading interval. (This is a relatively constant trading tactic). The MREN tactic is long/short; however, we are making tweaks for it to be long only initially here as well. We will take the core variables of this tactic and tweak them for each coin we use it on. Again, the data above does not take into account DEX variables so the final data will look a bit different to these initial results. We have more of these tactics in the pipeline and are excited to develop them further and share them with you in future blogs.
While you are here, please go and subscribe to ourYouTube channel, we need 1000 subs in order to livestream and are getting close. Less than 50 to go and then we will do something fun!
Stay up to date with all the latest SingularityDAO news, join our Telegram Community and Announcements Channel, follow us on social media,Twitter, Facebook, YouTube, Instagram, Medium and Subscribe to our newsletter.
