Above – Dataswarm returns vs S&P over 24 months
DataSwarm has used its signal to trade in the US markets for 2 years, from December 2019 to November 2021.
The results (see above) were that we managed a 63% return over the period, at a low relative volatility/risk rate – Sharpe ratio of 2.7 vs the S&P 1.4, and a Beta of 0.44 holding the S&P at 1
(The NASDAQ Beta is c 0.9 over the period).
There have been 4 distinct phases in the system trial:
1. “Covid Crash”
The trial started in December 2019 and the system was slowly gaining on the indices. (Note – we set the system up to trade medium – large market cap stocks on the NYSE/NASDAQ so we used the S&P 500 as the general market benchmark).
The Covid crash saw the Zeitgeist signal turn strongly negative – fast – so we sold our investments quickly and early, as shown on the graph – avoiding the worst of the drop.
2. “Continuous Improvement” in CovidWorld
In and between Covid lockdowns over 2020, a lot of industries traded relatively flat after the crash. Some specific sectors, mainly dealing with digital transformation and healthcare/pharmaceuticals – rose very strongly. The system picked up on these quickly and the system gained value. Operationally we were continually learning about how the signal best works in the market, as well as building out new capabilities to support our market analysis and trading activity. In this period we developed the core of DataSwarm system’s ability to integrate trading the markets using Zeitgeist combined with market financial data.
3. “Interesting Times”
From January – c July 2o21 we entered what we called “interesting times”. There were two distinct phases:
January – February: a sharp market rise and then rapid crash. This was the period of extreme “meme-stock” manipulation, the system went up with them, and came down again. We were happy with it going up, less happy that it didn’t exit fast (more on this below)
March – July: Flatlining – the system was struggling to find a consistent Zeitgeist signal, and the result was a flatlining performance over the period. This required us to do a deep analysis of what was going on, what we found were a number of areas for further system improvement, mainly:
- Market sentiment was very confused in the period, the system needed improvement to deal wth that
- We also found we needed to flex the system’s limit functions to cope with shifting overall market situations, also the various prediction factors have shifting impacts in different conditions.
- Managing risk – if you want big gains you need to accept more volatility but it bites in a downwave as the system does not exit as fast. (The Memestocks also showed “interesting” characteristics that we needed to correct for).
At any rate, by end July we had put in place a number of changes and the system started to find a useful Zeitgesit signal again and started to respond positively to the market again.
4. Develop and Test New Systems
In addition to the analysis work, over the first 6 months of 2021 we had carried on building the Dataswarm software system and had started work on a highly automated backtesting and prediction capability. Over July we started to introduce the initial automatic predictive algorithms to trade in the market directly. In addition we started to automate the actual trading process on a different trading platform to the one we were using, so we could run an end to end automated system.
By the end of November we felt we had proven the enfd to end systems, and it was time to cut over to the new platform. Also, we wanted to do quite a bit of root and branch rebuilding of what was now a 2 1/2 year old system. We had learned a huge amount about the market and the technology and wanted to make major adjustments to the core system. But this would be hard to do while also trading (and our resources are not infinite). So we decided to end the test at 24 months of trrding and port the system to the new platform and fully test it first. We will write about that when we start our 2nd live test.
(Update 14 March 2022 – we have started the 2nd experiment, some notes are over here)