Analytics has recently been in the news from a retail and institutional trader perspective. Tradefeedr launching an FX algo forecasting suite enables retail clients to access accurate and independent data to inform their algo execution strategies better, while institutions are investing heavily in technology to improve their trading process and data analysis.
No central source
So, how can retail FX traders best use data to give themselves a trading edge in a highly fragmented market with hundreds of platforms, venues, and liquidity providers generating market data in real-time, 24 hours a day?
Unlike listed markets, there is no consolidated tape or central source of data, and the availability of executable prices differs across market participants. While fundamental and sentiment analysis are essential tools for traders to optimize decision-making, using technical analysis based on statistical time series analytics is still the standard in retail FX trading.
“Time series data which focuses on mining historical and real-time data to analyze trends in search of the repetition of well-known chart patterns and other technical factors are the lifeblood of technical analysis,” explains Rich Kiel, the Global Head of FX solutions at data analytics specialist KX.
The Entrance of AI
As we have previously reported, ChatGPT (an artificial intelligence chatbot developed by OpenAI) has recently garnered positive coverage. AI technologies such as machine learning have already had a considerable impact in trading data analysis observes Will Carter, the Head of Trading and Analytics at trading solutions developer MahiMarkets.
Will Carter, Head of Trading and Analytics at MahiMarkets
“Technical analysis has been around for a long time to assist traders in identifying patterns, but machine learning has been the most significant innovation in data analysis in recent times,” he says.
“Applying machine learning to data has now become plausible for the sophisticated segment of the retail community.”
Given that the direction of travel in retail is high-frequency trading and that machine learning consumes vast amounts of data, traders first need to ensure they have access to data at high frequencies.
Affordable access to data is a key consideration for retail traders. Kiel observes that not only is data readily available, but it also comes in all flavors with everything from APIs delivering ultra-low latency real-time information to varied degrees of delayed market data as well as the availability of historical market replay streams and data downloads.
“Emerging technologies such as cloud computing facilitate the storage of huge data sets at lower cost, making this data available to a wider range of market participants,” he says. “Additionally, platform operators and technology providers are in an arms race to provide the broadest set of capabilities to remain competitive. Access to market data systematically through retail brokers and platform providers has now become standard.”
Data Hungry
Public sites such as Yahoo Finance offer decimated data buckets/bars in many assets going back a long way. However, according to Carter, machine learning is very data-hungry, and data at a 100-millisecond granularity level or more is crucial for a research environment.
“Public data resources are not currently good enough – traders need to capture, cleanse and store the data themselves,” he adds.
While quantitative trading based on time series data has long been the domain of FX trade decision-making, many institutional investors have also employed decision trees, including fundamental and even sentimental analysis, to form a more holistic trading strategy now often referred to as quanta mental trading.
Fundamental data, such as interest rates and commodities prices, is well suited to quantitative prediction of FX movements and is combined with other market data, such as primary market indices and currency rates, to learn and create a forecast for FX, explains Yaron Golgher, the CEO & Co-Founder of I Know First, a developer of AI-based algorithmic forecasting solutions.
Yaron Golgher, CEO & Co-Founder of I Know First
“The AI algorithm generates the forecast signal value,” he adds. “At each time horizon, we measure the price deviation from what the system considered fair – that is the signal. A positive signal is up, negative is a down signal.”
More than 20-time forecast points are used to map the trajectory of the forecast price. These are compressed by averaging into six-time horizons – three days, seven days, 14 days, one month, three months, and one year.
For each point, the system generates predictability, reflecting (inversely) the level of unpredictable noise. The higher the predictability, the higher the confidence in the forecast.
“Each forecast point is a weighted average of tens or even hundreds of independent predictors and each predictor module is comprised of several inputs,” says Golgher. “Thus, every module provides an independent forecast because it is based on a different set of market data.”
To undertake quanta mental analysis effectively requires both computing power and sophistication using traditional big data analysis combined with emerging capabilities such as machine learning.
“This isn’t in the domain of most retail traders, but as brokers and investment managers continue to expand the availability of trading algorithms to their clients, systematic execution based on quanta mental principals will continue to become more prevalent and accessible going forward,” says Kiel.
Carter agrees that the barrier to entry for traders creating a quanta mental research environment is lower than ever before and that implementation comes down to a number of factors including breadth of research and basic technical skills as well as data access.
“It also raises questions about how brokers manage the new alpha-seeking community of retail traders using these sophisticated technologies,” he says, adding that brokers can no longer operate under the assumption that the retail trader will permanently lose and continue to take the opposite of the trade (known as ‘B-Book execution’).
“Assuming you have a broker facing a successful quanta mental trader who is consistently extracting alpha, if the broker always B Books that flow the alpha comes from the B book,” concludes Carter. “Instead, it needs to go through a more actively managed portfolio so that the alpha comes from the market.”
Analytics has recently been in the news from a retail and institutional trader perspective. Tradefeedr launching an FX algo forecasting suite enables retail clients to access accurate and independent data to inform their algo execution strategies better, while institutions are investing heavily in technology to improve their trading process and data analysis.
No central source
So, how can retail FX traders best use data to give themselves a trading edge in a highly fragmented market with hundreds of platforms, venues, and liquidity providers generating market data in real-time, 24 hours a day?
Unlike listed markets, there is no consolidated tape or central source of data, and the availability of executable prices differs across market participants. While fundamental and sentiment analysis are essential tools for traders to optimize decision-making, using technical analysis based on statistical time series analytics is still the standard in retail FX trading.
“Time series data which focuses on mining historical and real-time data to analyze trends in search of the repetition of well-known chart patterns and other technical factors are the lifeblood of technical analysis,” explains Rich Kiel, the Global Head of FX solutions at data analytics specialist KX.
The Entrance of AI
As we have previously reported, ChatGPT (an artificial intelligence chatbot developed by OpenAI) has recently garnered positive coverage. AI technologies such as machine learning have already had a considerable impact in trading data analysis observes Will Carter, the Head of Trading and Analytics at trading solutions developer MahiMarkets.
Will Carter, Head of Trading and Analytics at MahiMarkets
“Technical analysis has been around for a long time to assist traders in identifying patterns, but machine learning has been the most significant innovation in data analysis in recent times,” he says.
“Applying machine learning to data has now become plausible for the sophisticated segment of the retail community.”
Given that the direction of travel in retail is high-frequency trading and that machine learning consumes vast amounts of data, traders first need to ensure they have access to data at high frequencies.
Affordable access to data is a key consideration for retail traders. Kiel observes that not only is data readily available, but it also comes in all flavors with everything from APIs delivering ultra-low latency real-time information to varied degrees of delayed market data as well as the availability of historical market replay streams and data downloads.
“Emerging technologies such as cloud computing facilitate the storage of huge data sets at lower cost, making this data available to a wider range of market participants,” he says. “Additionally, platform operators and technology providers are in an arms race to provide the broadest set of capabilities to remain competitive. Access to market data systematically through retail brokers and platform providers has now become standard.”
Data Hungry
Public sites such as Yahoo Finance offer decimated data buckets/bars in many assets going back a long way. However, according to Carter, machine learning is very data-hungry, and data at a 100-millisecond granularity level or more is crucial for a research environment.
“Public data resources are not currently good enough – traders need to capture, cleanse and store the data themselves,” he adds.
While quantitative trading based on time series data has long been the domain of FX trade decision-making, many institutional investors have also employed decision trees, including fundamental and even sentimental analysis, to form a more holistic trading strategy now often referred to as quanta mental trading.
Fundamental data, such as interest rates and commodities prices, is well suited to quantitative prediction of FX movements and is combined with other market data, such as primary market indices and currency rates, to learn and create a forecast for FX, explains Yaron Golgher, the CEO & Co-Founder of I Know First, a developer of AI-based algorithmic forecasting solutions.
Yaron Golgher, CEO & Co-Founder of I Know First
“The AI algorithm generates the forecast signal value,” he adds. “At each time horizon, we measure the price deviation from what the system considered fair – that is the signal. A positive signal is up, negative is a down signal.”
More than 20-time forecast points are used to map the trajectory of the forecast price. These are compressed by averaging into six-time horizons – three days, seven days, 14 days, one month, three months, and one year.
For each point, the system generates predictability, reflecting (inversely) the level of unpredictable noise. The higher the predictability, the higher the confidence in the forecast.
“Each forecast point is a weighted average of tens or even hundreds of independent predictors and each predictor module is comprised of several inputs,” says Golgher. “Thus, every module provides an independent forecast because it is based on a different set of market data.”
To undertake quanta mental analysis effectively requires both computing power and sophistication using traditional big data analysis combined with emerging capabilities such as machine learning.
“This isn’t in the domain of most retail traders, but as brokers and investment managers continue to expand the availability of trading algorithms to their clients, systematic execution based on quanta mental principals will continue to become more prevalent and accessible going forward,” says Kiel.
Carter agrees that the barrier to entry for traders creating a quanta mental research environment is lower than ever before and that implementation comes down to a number of factors including breadth of research and basic technical skills as well as data access.
“It also raises questions about how brokers manage the new alpha-seeking community of retail traders using these sophisticated technologies,” he says, adding that brokers can no longer operate under the assumption that the retail trader will permanently lose and continue to take the opposite of the trade (known as ‘B-Book execution’).
“Assuming you have a broker facing a successful quanta mental trader who is consistently extracting alpha, if the broker always B Books that flow the alpha comes from the B book,” concludes Carter. “Instead, it needs to go through a more actively managed portfolio so that the alpha comes from the market.”
Read More
Paul Golden