Unlike other actively managed funds where investment decisions involve human judgment, this fund has machine learning algorithms as its core drivers for investment strategies and portfolio construction
Here is the advent of Artificial Intelligence (AI) and Machine Learning (ML) in fund management. Tata Mutual Fund has launched ‘Tata Quant Fund’, an AI and ML powered fund. The fund employs a proprietary quant framework that combines multiple rule engines and predictive models to create investment portfolios. The new fund offering period closes on January 17.
The fund, a pure quant-based MF scheme, is the result of Tata Asset Management’s decision last financial year to put together a team specialising in data science that ventured into quant models. The same technology that suggests songs you love listening to in your music streaming app and estimates the time needed for your cab trip will essentially pick stocks and build a portfolio. RupeeIQ takes a detailed look.
When a machine reads and understands a large amount of data to act on the same or make decisions, it is called AI or artificial intelligence.
Machine learning systems are ones in which machines independently analyse historical correlations and hidden patterns in available data and can recommend/infer the outcome or impact of future situations of a combination of factors.
The new fund offering has an active multi-factor investment model with embedded AI modules. The AI modules dynamically change factor strategies once every month based on prevailing market conditions.
Unlike other actively managed funds, where investment decisions involve human judgment, Tata Quant Fund has machine learning algorithms as its core drivers for investment strategies. The fund would invest in a portfolio of stocks selected from the BSE 200 & equity derivative list. The fund would use proven factor strategies like value, quality, momentum, and market-cap for rule-based stock selection and portfolio allocation. The fund portfolio would be a 30-50 stock basket.
The machine learning modules that select factor strategy deployment for the forthcoming month will self-learn and adjust to changing market dynamics. This self-learning and realigning of the model will happen every six months (i.e. every six months incremental data will be fed to the system). Typically, AI and ML systems over time become less efficient unless new data is fed and the system is refreshed. So, Tata Quant Fund addresses that problem.
The machine would take market direction calls as well. This would help in protecting drawdowns, thus minimising loss of investment capital and helping in more consistent capital accretion, Tata MF says. All this for what? The Tata Quant Fund strategy is aimed at outperforming market returns over the medium to long terms while trying to minimise the probability of loss in a range of investment horizons.
There are two parts of how a portfolio will be made.
First, predicting the optimal strategy for the forthcoming month and portfolio creation. Each month during the past 20+ years, each stock of BSE 200 is scored on a select set of Factor Models – Value, Quality, Momentum, Value + Quality, Quality + Momentum, and Market Cap. The scores consider 1-year data of the underlying variables for individual stocks. A portfolio comprising of top-scoring stocks is created for each factor model for each historical month. The machine is provided 20+ years of macroeconomic and key market-related data along with data on monthly returns from the above set of portfolios. The machine learning algorithms analyse hidden correlations and patterns in historic data for identifying portfolio that give the highest returns. Every month the investment strategy framework with embedded ML engines take in the latest available macroeconomic and key market-related data as also (historic momentum values of each Factor Model portfolios. Basis the learnings of the machine, it decides on the factor strategy that will perform best in the forthcoming month.
Once the optimal factor strategy has been identified and a portfolio of top-scoring stocks created, the process then moves on to predict the direction of return during the next 30 days. This is the market direction prediction. The second algorithm predicts the direction (positive/negative) of portfolio returns. This predictive engine independently learns and predicts the next 30 days directions for portfolio returns independently based on similar variables as the previous model. When predictions turn bearish, the model recommends taking a hedged position to reduce net long equity position to as close to zero as possible. This is done by using derivatives to hedge the currently held long portfolio. A stock level hedging strategy is employed. When predictions turn bullish the portfolio created basis optimal factor strategy chosen by the first model is used to rebalance or create a cash long position in identified stocks.
Tata Quant Fund employs a systematic machine-driven investment strategy that minimises human biases in the investment decision-making process. Human expertise is largely limited to design, validation, and finalisation of the framework used in the investment process.
Portfolio creation and regular rebalancing are driven by machines with embedded machine learning algorithms.
Of course, human expertise is involved in the execution of recommendations generated by the framework (i.e. actual buying and selling).
The fund manager i.e. Sailesh Jain, however, reserves the right to intervene in extreme situations that may severely impact fund performance and are driven by events very different from ones that occurred during the model training period.
The expected benefits of removing human intervention from the investment process are to eliminate various behavioural biases like confirmation bias, loss aversion, recency bias, etc.
The fund model has a high average monthly turnover ratio. Tata MF says the back-tested performance results for Tata Quant Fund show that costs relating to portfolio churn do not significantly impact the alpha generation ability of the strategy.
The success of the model is based on a systematic investment approach and therefore it may not be able to leverage short term opportunities available in the market from time to time.
There is no guarantee that the quant model will generate higher returns as compared to the benchmark.
If the machine has not learned something in the past or has not witnessed an event in the past, it will not be able to account for the same in its prediction.
The fund benchmark will be S&P BSE 200 TRI. The fund has an exit load of 1% of the applicable NAV if redeemed/switched out on or before the expiry of 365 days from the date of allotment. The minimum application amount for this fund is Rs 5,000/- and in multiples of Re.1/- thereafter and additional investment of Rs 1,000/- and in multiples of Re 1/- thereafter. Tata Quant Fund is available with Tata MF’s digital channel partners Paytm Money & Groww App.
Prathit Bhobe, MD & CEO, Tata Asset Management said, “Machines have massive computational power needed to process very large data sets, spot patterns and correlations, make decisions faster, objectively and without human biases. In the current world, computers are powerful enough to solve problems, a lot of data is available and we strive to use this data in combination with algorithms to its best”.
Actively managed funds benefit from human intelligence that learn, comprehend, and respond to different market challenges in complex manners. Passively managed funds, on the other hand, are rule-based and they avoid pitfalls of biases that accompany human judgment. Their strength lies in increased objectivity and elimination of human errors. These traditional styles have contrasting pros and cons, according to Utpal Sarma, Head – Business Analytics, Tata Asset Management. AI and ML add a third dimension to fund management. They enable machines to mimic human judgment to a certain extent while retaining the benefits of disciplined rule-based investing.
In one of the existing quant funds in the market, rules are fixed and the machine does not learn. Also, the algorithm to take portfolio or cash call is absent and thus the model does not hedge. In another quant fund, the stocks are first eliminated from the universe and factor strategy is picked by human intervention. Again, the machine does not learn and the model does not hedge. Tata Quant Fund does all that.
Tata Quant Fund is classified into the Thematic category as defined by SEBI. Quant funds should be treated as another Investment style or strategy. Only investors, who understand quant fund strategy, can invest a part of their portfolio in quant funds to diversify their risk and their portfolio. The accuracy of Tata Quant Fund’s AI & ML model holds the key.
Disclaimer: Views expressed here in this article are for general information and reading purposes only. They do not constitute any guidelines or recommendations on any course of action to be followed by the reader. The views are not meant to serve as a professional guide/investment advice / intended to be an offer or solicitation for the purchase or sale of any financial instrument like quant fund mentioned in this article.
Subscribe & keep learning!