Systematic Equity Investment Portfolio Performance Tracking

[ncube]

March crash in portfolio has been recovered completely. Well done brother. Best Wishes.

Thanks, The DD for this strategy was recovered in April itself as I started reducing the equity part of the allocation when the down trend was confirmed and later increased it when the base started forming.

I believe its more important to get the asset allocation correct and understand the returns profile of various strategies that we run and design the portfolio accordingly rather than focusing on increasing the returns from any one strategy. Objective is to increase the overall returns of the portfolio for a given unit of risk by reducing the returns co-relations between constituting strategies as much as possible.

I am posting here as I wanted to track online the performance snapshots of strategies that I run at regular interval and think its easier to track it as a thread in a forum which can be accessed from anywhere. So currently just exploring this option over hosting and managing a website myself.

 

[trendtrade]

Thanks, The DD for this strategy was recovered in April itself as I started reducing the equity part of the allocation when the down trend was confirmed and later increased it when the base started forming.

March time was really stressful for Equity Portfolio Holders. At that time there was just total gloom and doom kind of scenario in the global markets and no one actually had any idea regarding how this Corona Issue will impact Global Economy in the coming days. Most guys were staring at erosion of 30-50% in their portfolio values and that was very nervous time indeed!

I believe its more important to get the asset allocation correct and understand the returns profile of various strategies that we run and design the portfolio accordingly rather than focusing on increasing the returns from any one strategy. Objective is to increase the overall returns of the portfolio for a given unit of risk by reducing the returns co-relations between constituting strategies as much as possible.

This is really insightful ncube. But how does someone actually calculate the risk/return profiles of their Swing Trading Strategies which are somewhat Discretionary in nature? If a strategy is automated, then running the back test is easy. But how would one actually back test a strategy which evolves some elements of discretion based judgements? Is it possible to just input all the required data columns in a database and then step forward one day at a time and note down all the buy/sell trades that one would have taken according to his discretion based system. Which stocks he would have bought and sold and in which quantities, and how would he have changed the overall allocations in his portfolio.

Can someone please guide in this regards, how to do such kind of manual back testing. What are best practices for such exercise, so that we can get clear picture about the performance of our subject trading metohd ?

I am posting here as I wanted to track online the performance snapshots of strategies that I run at regular interval and think its easier to track it as a thread in a forum which can be accessed from anywhere. So currently just exploring this option over hosting and managing a website myself.

This is really good idea. You would have a clear picture about the performance of your portfolio at any given time. It will document all the ups and down nicely. Plus, other friends are able to learn from it and share their views as well.

Wish you the very best.

Regards

 

[ncube]

March time was really stressful for Equity Portfolio Holders. At that time there was just total gloom and doom kind of scenario in the global markets and no one actually had any idea regarding how this Corona Issue will impact Global Economy in the coming days. Most guys were staring at erosion of 30-50% in their portfolio values and that was very nervous time indeed!



This is really insightful ncube. But how does someone actually calculate the risk/return profiles of their Swing Trading Strategies which are somewhat Discretionary in nature? If a strategy is automated, then running the back test is easy. But how would one actually back test a strategy which evolves some elements of discretion based judgements? Is it possible to just input all the required data columns in a database and then step forward one day at a time and note down all the buy/sell trades that one would have taken according to his discretion based system. Which stocks he would have bought and sold and in which quantities, and how would he have changed the overall allocations in his portfolio.

Can someone please guide in this regards, how to do such kind of manual back testing. What are best practices for such exercise, so that we can get clear picture about the performance of our subject trading metohd ?


This is really good idea. You would have a clear picture about the performance of your portfolio at any given time. It will document all the ups and down nicely. Plus, other friends are able to learn from it and share their views as well.

Wish you the very best.

Regards

1. Gloom, doom, euphoria etc are natural human behavior and is applicable in every activity that we take part in. Just that in the financial markets the impact will be bit extreme due to the heightened greed and fear. As one spends more time in the system, he will gain the required experience to read the situation and act accordingly. Hence its important to focus on the long haul and be in the markets. If one is confident about his strategies, trading/investing process and believe in the long term growth of India he can easily see opportunities in every situation.

2. When I say returns profile of the strategy its not some calculated fixed risk/return value that we measure for the strategy as the risk/return for a strategy keeps changing . It is about understanding the strategy behavior itself by answering questions like, why we think our strategy works, what is the edge, under what market conditions it performs well and when it does not...etc. Once we know this its not difficult to link it with some measurable parameters or reference it with some index and change allocations/stop trading the strategy etc based on our historical understanding. This is where our journals or trade records and its performance will help.

3. I had tried to show the process in action in this thread, to give an example, the strategy I mentioned here is long only growth/momentum based strategy, so based on its historical performance I know how it will perform under various conditions. In this case I know that if the index is below X day ema, this strategy will perform poorly. Now how to use this knowledge is when we identify this trigger we reduce the allocation to this strategy and next step is to know what is the reason for this current condition, it was quite obvious that its due to covid-19. Once we know this the next step is to identify the sectors which will benefit from this and slowly start increasing the allocations to it.That is why if you checked my earlier posts I had higher allocation for Pharma sector and the allocations kept changing as the situation evolved.

4. About posting my journal here, its just temporary as I am exploring some permanent solution and the information I will be posting there are for my self reference and may not help others. A good friend in this forum today recommended wordpress and it looks interesting and think it will meet my needs, Hence will plan to move there in coming days.

 

[ncube]

Date: 06-08-2020: Mean-Reversion Long only Strategy Status:
1. Sharing the return curve of how a mean reversion long only strategy looks like.
2. This strategy was performing average over most of 2019 and was also impacted by the 2020 March crash, but overall it has doubled the capital in about 1.5yrs.
3. We can never know when the strategy will start performing, however having confidence and will to sit through the drawdown duratiion and systematically executing the strategy will definitely yield results over time.
4. As many doubt, wanted to show that running multiple un-correlated strategies help normalize the return curve and over time generate exponential returns.

For those who find it difficult to believe attached below are the performance charts for this strategy that I run in one of my trading accounts. Also it is possible to generate good returns by our-self and not a prerogative of only popular PMS & fund managers. The argument that PMS/Fund managers manage huge AUM does not matter when someone is trading at higher timeframes as there is enough liquidity absorb these high volumes and can be manage with execution strategies.

Hence focus on finding few good uncorrelated concepts which is proved in the past and will continue in future and build a systematic strategies around it with good money management and asset allocation rules to see the difference.

Performance Metrics:

 

[ncube]

10-08-2020: Pharma on Steroids: Highest single day profit in portfolio ever:

 

[ncube]

11-08-2020: Booked profit at open, Always take some profit off the table when there is euphoria, if required one can add back again later.

Booked some more profit from GMM before close of session: (Check before selling sometimes BSE prices are higher than NSE)

A very satisfying day, Generated enough cash for new opportunities and remaining GMMPFAUDLR & IRCTC ensured the PF does not go into red.

Looks promising...Sentiment model in combination of profit booking ratio model seems to be working fine...need to test further.

https://zrd.sh/p/RAzQzf1Y

 

[ncube]

Date: 24-08-2020: Holy Grail System?

1. Had been testing a new systematic long only strategy for last 1 month. The results are really impressive with 17 continuous wins and no losses so far.
2. Performance details:
Duration : 1 Month
Total trades: 28
Trades closed:17
Trades open: 11
Avg Win %: 6.5%
Avg Loss %: NA
1 Month Returns: 12.08%
3. Trade details:

5. Strategy uses mean-reversion factors trained with AI deep-learning algorithms. The model is self learning based on recent events and though I have developed it over time it has become a black box to me too, Hence I will not be able to explain how its working. But in the last 1 month it has given an impressive 12+% return on capital and its continuously learning.

6. As its my birthday today, I want to try an experiment and share the trade details here in real time even before I take the trades. Since it is EOD based trades I will be posting the trade details at end of day for the next day.

7. Disclaimer: Through I am posting my trades here, its just experimental and for educational purposes only to show how a systematic approach works as even I dont know how the model will perform in future, if anyone plan to follow it, its at their own risks, I will not be responsible for any of your profits or losses.

Trade details for 25-08-2020 : Will buy LT at open tomorrow

 

[Mavala]

Many many happy returns of the day ncube ! May god bless you.

 

[ncube]

I think it is not fair on my part just to post the trades without giving some idea about the strategy framework. I usually do not prefer strategies which are semi blackbox, this is the first time I have developed such a strategy and want to see how it performs. Anyway it has given good results in 10+yrs backtest and last 1 month walk-forward test.

Strategy Details:
1. The stock universe is all NSE stocks with some filters for penny stocks and low liquidity.
2. On this stock universe I run a monte carlo simulation with 1 year look-back to generate 1 month forward forecast probability distribution.
3. I then pipeline the forecast to a tensorflow framework which uses Bayesian optimization with Reinforcement Learning algorithm to identify stocks which has higher probability to trace the forecasted return curve. (This is where the model becomes black box and I do not have control how the AI algorithm learns and selects stocks)

I think it is slightly technical, in simple language I am just taking 1 yr returns for the stocks and run multiple montecarlo simulation to forecast how the stocks may perform in future. I then feed this forecast to an AI algorithm to learn the patterns and select stocks which is most likely to go up in near future. The output is a ranking list for the stocks and I select the top 10 from the list.

Portfolio construction:
1. There will be about 10 stocks in the portfolio, 90% capital equally distributed, 10% cash as buffer and profits reinvested.
2. If a stocks is sold it will be replaced with a new one if available.
3. Current open positions in PF:

 

[ncube]

Date: 25-08-2020
1. Bought LT @999