I started 4 months back with 0.1% of my portfolio. Now, I have scaled it up to 12%. The plan is to increase it to 30% over the next 6-12 months.
I reinvest the entire money on the reset date. I am not taking out anything from the portfolio.
I am using the 3 strategies that I have built now. The first one is the one I have mentioned here which is a quarterly strategy invested monthly. The other two are monthly strategies invested monthly. All 5 monthly transactions are with equal absolute sums of money. So, in percentage terms, the quarterly strategy takes 60% of the capital and the other 2 take 20% each. All the strategies are on Nifty500 stocks. I am also building a yearly strategy which is purely fundamental based, but it is in the absolute beginning stages.
Hi Abhishek
2013-2017 was good for small caps
And post 2017 is good for good mid - large caps
I think in next 1-2 years this quality bubble will take a hit and will not give much returns or may turn negative (as most companies are trading 2x more PE than growth and Peter Lynch considers such situation deadly ) .When everyone has same opinion (super high PE to quality )then I think most likely it will turn wrong sometime in future .
If this happens ,will it affect your quantum returns .
Just curious as 30% is large part of pf and it may turn risky
Itâs just my view and I am sure you must have put some strategies to protect /hedge
May be risk is less as itâs month on month basis and huge swings in one month are less possible
Have u thought about this and whatâs your risk strategy
Great to see that you are increasing allocation slowly. Probably you are tweaking your algorithms as well. Do you exclude any on the companies from the nifty 500 universe using some criteria like known red flags/fraud suspects or sectors ?
@maheshkumar Of course I have thought about it. The whole idea of having a quant strategy is to exactly keep myself away from such forecasting without any concrete data points. I have back tested the systems over the last 12 years including periods of 2008 crash. The max drawdown I have seen is around 50% which is okay because everything else falls around the same. These strategies tend to outperform significantly during good times and perform okayish or poorly during bad times.
Bottom line is it has equity risk similar to any other discretionary portfolio of hand picked stocks.
@james_kerala No. This is non-discretionary. I do not tinker with the output at all. The idea is to incorporate checks and balances in the strategies so that potentially problematic stocks do not make it to the list.
Does you system processes fundamental data every quarter and then from the short list choose every month based on the price action? Any free and reliable source of historical fundamental data youâve found or you source it from a vendor?
It will combine multiple rule engines and predictive models to create investment portfolios that are aimed at maximising returns during up-trend markets and minimising losses during down trend.
Broad and persistent factors of stock returns are used for building rule engines for portfolio creation.
Each rule engine uses scoring to create concentrated portfolios of 30 stocks.
I do not know anything about this fund, but I am very very wary of anyone who markets AI & ML. Whatever I have studied till now, and with help from a close friend who is a ML professor in a leading US university, my takeaway is that ML is not very useful in stock picking at this point in time.
What most of the âquantâ funds are doing is essentially correlation and regression analysis, Most of the time not even that. Rest is all marketing BS
Another interesting point I found missing in their presentation is their backtest results. If they are not giving it for a quant strategy, I will not even look at investing in it. One of the biggest advantages of a pure quant strategy is that it can be backtested and as an investor I can see and verify how it performed under different historical circumstances.
I have been learning the ABCD of ML also, but right now donât think pattern fitting in data will help generate better returns. There is very little evidence in financial research which says that stock price performance can be predicted beyond a very short duration and even then is prone to many errors.
Nevertheless I am happy that someone is thinking differently and adopting to newer technologies.
A simple strategy with 20 -40 % appreciation in price in last 6 months and price correction of 10%. We can start with this and backtest the results. Backtesting of various strategies is the key to success.
One of biggest strenghts of ML is that it can work on large datasets with stable relationships. for example ML is used for image recognition, medical imaging where once you train the algo, new data fits into it. with financials data, the relationships change with investors acting on the data and change in constitutents. Running an ML algo on data from the 90s is worse than useless.
i agree with you âŚfor all this talk around AI/ML is mostly marketing BS outside of some very narrow applications
There are very interesting ways in which AI and ML is being leveraged to aid fundamental research and decision making at large Investment firms. While a supervised ML model to determine an investment worthiness of a stock at a current point would be mostly useless, if these additional signals generated are used in conjunction with all other available information it significantly improves the decision making and outcome.
There are AI startups working on reducing cognitive biasâ in investment, harnessing the power of alternate data using real time web scraping some of the outcomes are really mind boggling. Having worked and advised a large asset manager in Europe I can definitely say AI and ML already impacting institutional investing in a big and bold way.
I am sure it is. Just not sure what is underneath the hood of AI and ML. I have also spoken to a few folks from such institutions who claim to use AI and ML and found what they are doing at times is simple correlation analysis or sentiment analysis.
Long before the HFTs were popularized, suspect a few people minted money. I had heard of one particular trader called Paul Rotter. It was rumored that he gamed order flows. My guess is that he was of the earliest trader who used algorithms to predict the order flow. Below is a popular interview of his:
Thanks. I am personally not very interested in high frequency trading or very short duration trading. I am more keen on slightly longer time frame - monthly, quarterly, half-yearly strategies. HFT is a difficult game and I, as an individual, cannot compete with the large institutions who can afford co-location and very expensive processors & computers. With an elongated time frame, I am not at any disadvantage with respect to anyone else.
I use a large number of indicators, both fundamental and technical, that I use for testing across strategies. I have also made a few indicators for myself, which is what I am using to get the alpha.
Both fundamental and technical data are collated from multiple sources including local data providers who supply data for chart packages like MetaStock or Amibroker. The biggest challenge in India is to get clean data. A lot of programming and effort is there to get the data and clean it.
No system is too basic. Also, the concepts are useful when thinking about new strategies or whether a strategy is likely to work or not. Sometimes, if a strategy is not working, it helps in understanding why it is not working.
Fortunately, like in any other pursuit, there is a large number of books, articles, blog posts available on the net. I will also post any good articles, books etc I find on this thread.
A simple way for those (like me) who are not too tech minded nor have enough resources is to look at stocks making and sustaining 3 month/6 month/12 month/all time highs (preferably 52 week highs or all time highs) and make a list and try to study this list to find out companies where we feel valuations are reasonable inspite of run up and there are more triggers lined up for the stock price to go up.
Essentially you put the technical hat first and then from the ideas it generates, put on fundamental hat and zero in on the best of them.
Will urge those interested to read The Next Apple or William O neillâs How to make money in stocks. These are very relevant books for the current market scenario.
You are absolutely right. I have actually tested a strategy for this And it gives quite good CAGR results. I took those stocks which were making both 3-month and 6-month highs and reset them every quarter. Picked the top 10 based on my personal filtering indicator. It gives a 35% CAGR returns over a 10-year time frame ( taxes, slippage, brokerage etc not considered).
Also, looking at different quant or technical parameters to identifying stocks for further fundamental research is a very powerful concept and one which I have learnt from you and now taking baby-steps in.
Sounds good. But what if this 3 month/6 month/12 month/all time high is the top? .
Another approach that comes to my mind is buy a decent company at 52 week low. ( Of course the assumption is there are no corporate governance issues, has earnings visibility) .I havenât done any back testing. My observation is that this can be very rewarding. But very difficult to implement though. One doesnt want to buy something not moving.
And it gives quite good CAGR results. I took those stocks which were making both 3-month and 6-month highs and reset them every quarter. Picked the top 10 based on my personal filtering indicator. It gives a 35% CAGR returns over a 10-year time frame ( taxes, slippage, brokerage etc not considered).