Teach a Man to Fish – "Rise of Machines" version

[marimuthu13]

Thanks "Marimuthu" for sharing your thoughts. What you are describing is a mean reversion system and looks to be built on fundamental principles. It is also good to read that you have established non-subjective rules. The idea of this thread was to demonstrate 1 of many ways to create a systematic trading system. I wanted to highlight that challenges to making money from market is not the system - but ourselves. You are proving to other readers that how system trading could be profitable if we have the right mindset.

I am glad that filter thought was useful for you. I agree that the system highlighted here requires automation. But perhaps a lot of my research focuses on automated trading and thus I tend to research ideas that can be easily automated. Having said that, I think your idea is also easily automotable and testable. Have you ever run a backtest on it or do you trade it live?

You are very right about it was our MIND , not OUR system that makes money.. Discipline to follow system during larger drawdown is also.key.

I have backtested it from 2018 to 2019 , started trading it with real money..

Problem here is trying to add filter or modify for better results which results in below average performance. ( call it as trader errors)
..

 

[ag_fx]

You are very right about it was our MIND , not OUR system that makes money.. Discipline to follow system during larger drawdown is also.key.

I have backtested it from 2018 to 2019 , started trading it with real money..

Problem here is trying to add filter or modify for better results which results in below average performance. ( call it as trader errors)
..

After some time you have to stop trying to improve it. Because there is little benefit in trying to capture every small variation of the system. I use a filter to avoid false breakouts as in a trend following system those happen a lot of times.

In a mean reversion system, I would say filter might not be of great use. After all you are trying to short in a rising market and going long in a falling market. Even if you choose to place limit trades instead of market trades, it would help.

If you were satisfied with your 2018, 2019 results on a Sharpe Ratio or a Calmar Ratio perspective, then I'd suggest stop trying to improve. Just follow it with small capital deployed for 3 months at least. Drawdowns are a part of system trading and we have to digest them. My positional trading system (not disclosed here) was short on Bank Nifty on 04 Jun. Markets on 5 -6 June kept going up, but my system did not revert to Long. It was tough two days. Then today markets opened gap down and I was in a good amount of MTM. Through the day I saw my +ve MTM of more than 1300 points evaporate. There was an itch to override the system and close the positions. I did not. But this discipline comes from practice and from backtesting your system well.

Btw - What sort of exit conditions do you use for your mean reversion?

 

[marimuthu13]

After some time you have to stop trying to improve it. Because there is little benefit in trying to capture every small variation of the system. I use a filter to avoid false breakouts as in a trend following system those happen a lot of times.

In a mean reversion system, I would say filter might not be of great use. After all you are trying to short in a rising market and going long in a falling market. Even if you choose to place limit trades instead of market trades, it would help.

If you were satisfied with your 2018, 2019 results on a Sharpe Ratio or a Calmar Ratio perspective, then I'd suggest stop trying to improve. Just follow it with small capital deployed for 3 months at least. Drawdowns are a part of system trading and we have to digest them. My positional trading system (not disclosed here) was short on Bank Nifty on 04 Jun. Markets on 5 -6 June kept going up, but my system did not revert to Long. It was tough two days. Then today markets opened gap down and I was in a good amount of MTM. Through the day I saw my +ve MTM of more than 1300 points evaporate. There was an itch to override the system and close the positions. I did not. But this discipline comes from practice and from backtesting your system well.

Btw - What sort of exit conditions do you use for your mean reversion?

Yes ..filtering may not work in mean reverse system as you mentioned..have to backtest it for a period

For me, good learning from you is not to try and fine tune the system to the core..keeping this mind very strongly...

For entry , its limit order

SL i.keep at 50% of the range , tgt is 2R, thats full range of first 15 minute candle....have not tested for partial profit, letting winners to run... This is also key takeaway from your thread

 

[ag_fx]

Yes ..filtering may not work in mean reverse system as you mentioned..have to backtest it for a period

For me, good learning from you is not to try and fine tune the system to the core..keeping this mind very strongly...

For entry , its limit order

SL i.keep at 50% of the range , tgt is 2R, thats full range of first 15 minute candle....have not tested for partial profit, letting winners to run... This is also key takeaway from your thread

Since you are dealing with gaps, you should have tgts and SL based on the gap and not just ORB. 0.5 Range as SL and 2xRange as TP gives you a 4:1 Reward:Risk ratio. My guess, without backtesting this system is that it might be too tight at times. But again this is just a guess and not robustly tested.

How do you identify stocks that have gapped more than 2% on a day? You have a screener ?

 

[marimuthu13]

Since you are dealing with gaps, you should have tgts and SL based on the gap and not just ORB. 0.5 Range as SL and 2xRange as TP gives you a 4:1 Reward:Risk ratio. My guess, without backtesting this system is that it might be too tight at times. But again this is just a guess and not robustly tested.

How do you identify stocks that have gapped more than 2% on a day? You have a screener ?

No scanner for this stock selection..

https://www.nseindia.com/market-data/pre-open-market-cm-and-emerge-market

Just using NSE site, as its downloadble in excel , very easy to sort it out..by 9.10 am will be able to ready with 3 stocks for the day..

By default nifty 50 will come, i will select security in F&O in dropdown list...

 

[ag_fx]

No scanner for this stock selection..

https://www.nseindia.com/market-data/pre-open-market-cm-and-emerge-market

Just using NSE site, as its downloadble in excel , very easy to sort it out..by 9.10 am will be able to ready with 3 stocks for the day..

By default nifty 50 will come, i will select security in F&O in dropdown list...

Oh so you are only using pre-market open. If a stock is up/down by 2% in pre-market? I thought you decide on the basis of Open price.

 

[marimuthu13]

Oh so you are only using pre-market open. If a stock is up/down by 2% in pre-market? I thought you decide on the basis of Open price.

Pre -market opening closing price @ 9.07 am is open price @ 9.15 am ....

 

[trader.trends]

My thoughts on Filters:

Filter calculation cannot be arbitrary. You should look at all the failed trades and check their MFE (Max Favourable Excursion). That may give you an idea of the optimum filter that can be used.
Using filters is a double-edged sword. Let me give an example.
Say you have a 1000/- stock and you use a filter of Rs.3/-. You will be using this filter for both entry and exit, I assume. You will buy if it touches 1003 if 1000 is the high of the setup candle. If the low of the setup candle is, say, 990 then you will exit if it touched 987. By adding the filter, you have increased the risk to 16 points. If your intention is to book at say 2%. You will exit at 1023. If your win-loss ratio is 40-60, then you make 40*20=800 points on the winning trades. On the losing trades, you give up, 60*16=960. You end up with a net loss of 160 points. If your win-loss ratio is 50-50, then wins are 50*20=1000 points and losses are 50*16=800.
Recalculate the same without a filter. Win-loss ratio of 40-60 - Wins 40*20=800, losses 60*10=600. Win-loss ratio 50-50 - Wins 50*20=1000, losses 50*10=500. In both cases you will see that you are better off without filters.
In most good systems, losing trades are more than the winning trades but the winning margin being higher makes up for the losing trades. When you have a greater number of losing trades having a filter increases the loses in the losing trades and decreases the wins in the winning trades.
So while the idea of the filter is appealing when we see false breakouts, it may not be so appealing on a large number of trades in the long run.

 

[trader.trends]

-deleted

 

[marimuthu13]

My thoughts on Filters:

Filter calculation cannot be arbitrary. You should look at all the failed trades and check their MFE (Max Favourable Excursion). That may give you an idea of the optimum filter that can be used.
Using filters is a double-edged sword. Let me give an example.
Say you have a 1000/- stock and you use a filter of Rs.3/-. You will be using this filter for both entry and exit, I assume. You will buy if it touches 1003 if 1000 is the high of the setup candle. If the low of the setup candle is, say, 990 then you will exit if it touched 987. By adding the filter, you have increased the risk to 16 points. If your intention is to book at say 2%. You will exit at 1023. If your win-loss ratio is 40-60, then you make 40*20=800 points on the winning trades. On the losing trades, you give up, 60*16=960. You end up with a net loss of 160 points. If your win-loss ratio is 50-50, then wins are 50*20=1000 points and losses are 50*16=800.
Recalculate the same without a filter. Win-loss ratio of 40-60 - Wins 40*20=800, losses 60*10=600. Win-loss ratio 50-50 - Wins 50*20=1000, losses 50*10=500. In both cases you will see that you are better off without filters.
In most good systems, losing trades are more than the winning trades but the winning margin being higher makes up for the losing trades. When you have a greater number of losing trades having a filter increases the loses in the losing trades and decreases the wins in the winning trades.
So while the idea of the filter is appealing when we see false breakouts, it may not be so appealing on a large number of trades in the long run.

Very valid and logical points..

But What if, filter itself playing key role in increasing win-loss ratio?

In this particular system, which is ORB in first 15 minutes, many chances are there where price just move beyond by few rs from ORB due to early morning momentum , then come back to the range , hit SL....

anyway thread owner will have his saying as he has done backtesting...