KenTyde: operational and technological savvy

Back in 2007, the largest CTA in the Netherlands, Transtrend (a full subsidiary of Robeco Group at that time), hired Elmi to set up KenTyde, a joint venture between Transtrend and Robeco Alternative Investments (RAI), which focused on applying trend models to niche and frontier markets. 

KenTyde Diversified Fund was a multiple award-winning CTA. During the generally challenging post-GFC period, it delivered a Sharpe of over 0.4 and a correlation to the SG CTA of below 0.60 between October 2009 and September 2015, using fully automated systems. Unfortunately, the parent companies (which also underwent some corporate activity) did not develop a distribution strategy to raise third party capital on top of the USD 20 million of seed capital from Robeco, and the program was closed in October 2015. (The Hedge Fund Journal’s subsequent interviews with Transtrend over the years suggest that they decided to trade the niche, alternative and frontier markets inside their flagship program rather than in a separate fund).

Setting up SysCat

At KenTyde the trio ran an efficient and automated operation. Waaijers built out KenTyde’s technology, Elmi developed its research and trading models, and Ringeling worked in both areas. The trio decided to launch their own firm in 2016, in part because of the operational and investment performance, but mostly because of strong team chemistry. The firm is based in Noordwijk in the Netherlands. “Working together in the office helps with brainstorming,” says Elmi.

There is a silo between research and operations and both areas follow thorough Dutch style processes of data gathering, research, testing and quantification. Ideas can be generated from the performance of current systems, event studies and external research papers. The testing process leads to a steady and gradual implementation of ideas with careful impact monitoring before they are fully integrated, and some fail the tests. 

From medium-term to short-term trading

SysCat initially found that the natural way to raise assets was to continue running medium-term trend following strategies in niche markets for some strategic clients including one in Switzerland. A few years later they set up a fund in November 2019 that combined medium-term commodity relative value as an alpha engine, with long volatility as a convex engine, to balance out the potential negative skew of the medium-term strategy. 

By 2020, SysCat determined that their competitive edge was not in medium-term trading (either of trends or relative value), where there are hundreds of CTAs, but rather in a differentiated short-term trading strategy that is designed to provide more predictable and reliable “crisis alpha” than a traditional trend following CTA. If KenTyde had a 0.4 correlation to CTA indices, SysCat Convex Alpha has averaged rolling three-month correlations of just plus 0.1 to SG Short Term Traders, and minus 0.1 to SG Trend Following and Alternative Risk Premia indices. “We have a much lower correlation and that is consistent with the objectives. We are not following medium-term trends and, if required, change our positions several times a day. Our short-term trading system is more proactive and reacts to direction much faster than large term trend followers,” says Elmi. In theory SysCat should be more correlated to short-term traders, but they display wide dispersion anyway with lower correlations inside the space. A negative correlation to the alternative risk premia index is no surprise: “We avoid including any trading concept with a negative skew profile into the strategy, like shorting volatility or trading liquidity premia,” says Elmi.

Whereas CTAs do average near zero correlation to conventional asset classes over very long periods, it fluctuates in a wide range from year to year. In contrast, “SysCat’s aim is to create a steady state low correlation to provide more consistent diversification,” says Elmi.

SysCat steadily increased the SysCat Alpha strategy allocation to the short-term trading, and in February 2021 they decided to completely retire the medium-term strategy and focus solely on short-term trading. SysCat’s offering has had three names in various places over the years: Niche Quantitative Strategies, SysCat Convex Alpha Fund, and the SysCat Alpha strategy, but there is currently only one BVI fund, administered by Apex.

Hyper selective high-speed trading

SysCat’s investment universe includes 7,000 markets, made up of 6,000 US single stock and ETFs and 1,000 futures, including single stock futures on non-US equities. Yet the strategy is hyper selective, only trading about 0.5% to 1% of the markets at any time.

Between 30 and 70 markets are selected, based on activity filters that are updated every minute or more often. SysCat are looking for an explosion in trading activity before selecting a market to trade and will exit when activity settles back down. 

Some markets, such as Tesla, are regular fixtures, while others, such as European grain markets, got picked up during 2022 as the Russia/Ukraine war led to more trading activity. Recently in 2023 bond markets, including corporate bond ETFs, flagged up on the screen.

There are no minimum exposures to any asset classes or markets and the selection is completely opportunistic. “We are blind to the names and labels. We have developed a common language to compare all markets, with data normalized to their own histories to define higher activity levels,” says Elmi.

The optimal trading frequency varies between markets. “The average holding period is 4 hours, though this can vary in a wide range between markets. Five minutes for the Nasdaq is slower than 4 hours for Paris rapeseed. Markets with wider spreads such as natural gas are traded more slowly,” explains Elmi. (Clearly however SysCat are not doing high frequency trading in milliseconds or microseconds.)

SysCat dynamically adjusts the latency of signals and their trading frequencies. The accuracy with which they predict optimal trading frequencies for markets is a key determinant of profitability.

SysCat do not find that typical systematic strategy labels, such as trend, countertrend or pattern recognition are very useful for their short-term models. Elmi explains, “We seek a local autocorrelation and increased trading activity, which might sound like short-term momentum, but we can also easily trade against medium-term trends”. 

(Incidentally, SysCat are not arbitraging mis-pricings of ETFs as some Dutch market makers do and prefer it when ETFs are accurately priced.)

Positive skew

Building a positively skewed strategy with any positive expected return at all (let alone a high one) is a considerable challenge. Many tail risk hedging strategies, and long volatility approaches, have a very low or even negative expected return.

SysCat is distinguished by seeking, and delivering, a positively skewed return profile, which has been much more positively skewed than other CTA indices, or conventional asset classes. Between November 2019 and April 2023, SysCat has shown a positive skew of 0.8 against minus 0.2 for the SG Short Term Trader index, minus 1.3 for SG ARP and minus 1.4 for the SG Trend index. This has been attained with a near zero correlation to equities, bonds or commodities. 

Risk controls and long volatility strategy

“Diversification is an excellent return stabilizer but it’s not a free lunch,” says Elmi, not least because patterns can break down and correlations can spike up during crises. “Correlation forecasting may work under normal conditions, but during extreme events, such as February 2018, uncorrelated markets can be hit simultaneously,” he points out. 

SysCat believe that a long volatility profile is the best defence against tail risk, and it avoids making potentially unstable assumptions about correlation, Value at Risk, or the ability to execute trailing stops. The long volatility strategy also harks back to Ringeling’s time making markets at Saen Options, where he worked with Govert Heijboer, one of the founders of the True Partner volatility arbitrage fund.

SysCat relies on a long volatility strategy for tail risk management, which is designed for market shocks that have not been seen in the past couple of years. “They could include extreme events like the Japan earthquake and tsunami of 2011 which led to grain markets being shut for several days,” says Elmi.

SysCat also pays careful attention to stress testing downside correlation between engines, and performance under extreme events or sudden regime shifts. Liquidity stress tests include open interest, volume, price ranges and bid/ask spreads. Exposure can be reduced if the systems forecast a spike in volatility based on historical statistical parallels. Volatility scaling considers not only standalone volatility but also spread volatility and downside volatility. 

The allocation to long biased volatility arbitrage was larger when SysCat ran medium-term alpha but has been downsized because the focus on short-term trading already has a naturally more convex and positively skewed return profile. The actual exposure to long volatility dynamically fluctuates with exposures in the alpha engine.

The long volatility strategy identifies underpriced out of the money options and delta-hedges them to gamma trade and potentially extract profits from realized volatility exceeding implied volatility. The objective is for realized volatility profits to offset the costs of the options. 

Challenging periods

SysCat expects to deliver predictable and reliable crisis alpha in most crises, but might fail to do so if their models misjudge the optimal trading frequencies, and if they over-forecast intraday activity. “There also needs to be some detectable pattern in trading activity. The strategy will suffer if any observable trace of autocorrelation or trading activity disappears,” says Elmi.

Performance drawdowns feed lessons back into the evolving research process, though this involves discretion rather than machine learning (which does not enter any part of the process). “We look at strategies like the engines of a car. We want to know our engines better. Like any other engine, they need to be modified, updated and fine-tuned,” says Elmi. 

Execution and systems

Automated systems clean and verify data, construct portfolios within constraints and limits, execute trades and monitor trade reconciliation. SysCat uses a mix of vendor and proprietary order management systems and execution management systems. Interactive Brokers are the brokers.

SysCat acts as liquidity providers, leaving limit orders. The limits are determined by short-term liquidity estimates and can be modified during the order execution process. “This means that trading slippage is well under control. But for us, the slippage is more than just the trading slippage. We incorporate the latency of our trading signals into the execution process. The system slippage, defined as system prices versus achieved prices, is carefully monitored,” says Elmi. Incidentally, and perhaps surprisingly, the system slippage was much higher for the medium-term relative value strategy than for the short-term strategy. 

Strategy development 

The investment universe will expand as assets and service provider relationships grow. “We are actively working on creating a reliable market access to commodity futures in China, South Africa, and Brazil, to name a few examples,” says Elmi. It is not currently practical to trade OTC markets, and some larger contract sizes. “European electricity or coal contracts are too large for the combination of our risk profile and the current AuM to include,” says Elmi.

SysCat could be open to acceleration capital deals but would not want to offer heavily discounted fees for what is a scarce, capacity constrained and distinctive alpha strategy.