THE INVESTMENT BLUEPRINT Podcast by RAIN Technologies Episode 4 Lakshmeesh Rao

Show video

Hi, welcome to the Investment Blueprint. I'm your  host, Raghu Kumar, from RAIN Technologies. The   Investment Blueprint is a weekly podcast where you  can find me interviewing real world quantitative   traders and investors, hedge fund managers, and  industry experts in the field of quantitative,   data driven trading and investing. My goal is  simple: I want to help you make more consistent,   profitable trading decisions in a data driven and  meaningful manner. Today's topic is on: how to do   algorithmic trading in India, with a focus on the  different types of algorithmic trading strategies.  

We often tend to generalize trading strategies:  if it's not an investing strategy, then it's a   trading strategy! How many ways can I slice  and dice the universe of trading strategies?   Well it turns out that you can slice and dice them  quite a bit! Today, i talk to Lakshmeesh Rao Chief   Investment Officer at RAIN. Lakshmeesh joined  us a few months ago, along with Bharath Rao, to   take over the CIO duties at RAIN. Lakshmeesh used  to run a company called Alpha Matters, where they   built and licensed proprietary trading strategies  for India's capital markets, the US markets, and   all sorts of asset classes in between including  cryptocurrencies and forex. Lakshmeesh is very   diligent and methodical with his backtests,  and it's been really fun to kind of get to see   how he thinks about solving real world problems  in a daily manner. Today, we talk about different  

types of trading strategies, how they differ,  and as a quant, what you should know about them.   So, Laksheemsh, welcome to the podcast. Thanks a  lot, Raghu, it's a pleasure to be here. Awesome,   so you know, I thought we could start  today's conversation by talking about   your background with algorithmic trading. How  did you get started with this whole field? Well,   I finished my engineering in India in the field  of electronics and telecommunication. After that,  

I had a brief stint with Infosys for a year and a  half, and then I went on to do my Masters from the   Delft University of Technology in the Netherlands,  where I specialized in Control Systems.   So Control Systems was a multi-disciplinary  field wherein you would take any system and   abstract it into a mathematical model and then  we used to design algorithms to control them.   So when I was graduating, I also  got into this project called IGEM   which was mostly to do with biology,  where we try to do some funny stuff with   cells. But so there, I had a teammate, so he  was into machine learning and he used to tell me   a few things about the markets and all that. So  once I graduated, I had to pick a field wherein   I would implement all these learnings,  whatever I had done in my Master's,   so I thought like, okay the market seemed  interesting to me at that point of time   and I just wanted to see like,  okay where I can go with it,   so then I approached a bunch of people along with  Bharat. So Bharat was one of the people to respond   to me. And yeah today, I got started with Bharat.  So that began my journey of algorithmic trading,  

from there. Got it, so, generally  your experience in the Netherlands,   I think you've told me once about this but I  totally forgot , just generally how was your   experience in the Netherlands and how is it  different from your upbringing in India. Well,   it it was very different back there, I mean there  you have to be independent, all on your own,   here at least teachers everyone will follow  up with you. But there's no such thing, right,   and there it should be purely driven by you, what  you want to achieve, your goals, your dreams,   right. Your ambitions. Right, yep, got it. And  it was mostly a project based education in a way,   so you got it great. I think that helped me a  lot. That's really cool, yeah, and one thing I've  

noticed is you have very peculiar sleep habits,  you know, everybody at RAIN generally works   relatively normal hours, and you kind of  come out of left field and you stay up until   I think like four in the morning, and then you  start at like nine in the morning or something,   right, so how did you kind of get into that habit  and is there any reason for that? Well , I think   designing algorithms is more a creative process  in a way, right, where you would need a lot of   peace time. I mean, alone time, where you need  to think a bit and try to come up with ideas,   I think that entire process happens to me good  in the night time, where everyone is sleeping,   and there's no one to disturb me. That's  amazing, I can totally attest to that, you know   when I used to build algorithms you know, on  a daily basis, I used to also have very odd   sleeping hours, I would work at night, like  completely in the night and then I'd go to bed at   like three or four in the morning. Yes, and then,  yeah totally I totally get that. So you know,   today's topic is on the different types of trading  strategies, right, so obviously, you know, we have   the general bifurcation of investing and trading,  but within trading strategies, I feel like there's   oftentimes a generalization that happens, and  all trading strategies kind of get clubbed   together. So I thought we could kind of dive into  that today, so you know, from your perspective,   generally speaking, what are the different types  of trading strategies that a quant should know   about? Well, according to me, there are primarily  two sorts of trading algorithms which you can   build. One, which is based off of momentum,  which means, let's say something goes up,  

right, you tend to buy it higher and you tend to  sell it even higher, or you trying to short it,   and then try to buy it back at a lower price.  Right. So that's more to do with momentum, there   you're saying like, okay, if something runs up you  will have enough momentum going forward to capture   that extra alpha there and still try to make  something out of it. Right. So that's one,   and then there is value trading, right, or  mostly mean reversion strategies. We are saying,   once something goes down, it's most likely that  it's going to come back to its mean. Similarly,   if something goes up, again, you're expecting  that that will also come back to the mean. So   broadly, if I try to define it, or  from at least from my experience,   these are the main two categories of  algorithms which you can come up with.  

Got it. But there are also some other ways  of looking at this entire process, like,   what we do in the US, for example right ,so there  we do, we trade volatility based instruments,   there we try to identify some more fundamental  factors and then combine it with machine learning.   That doesn't fall into these two classes, but  there we try to identify somehow the underlying   phenomena due to which you can produce returns, so  that's another class of algorithms all together.   Got it, got it. So basically, the second  class that you described is basically like a  

mean reversion kind of. Yes, that's a purely  mean reversion sort of a strategy. Got it,   interesting ,okay so for example  like an arbitrage strategy   that would fall I guess into the second  classification, right. Again, two ways of   doing it. One you can either look at the pure  time series itself where you're saying, okay,   a time series itself is going to be in revert.  For example, SPY in the US. So there, we had a  

mean reverting strategy on the SPY itself,  so there we did not compare it with any other   instrument or anything, so there we worked off of  purely the time series of SPY itself. So there,   you say that, OK, if SPY falls x percentage,  or depending on some quantitative factors,   then we say, OK, it's fallen enough sufficiently  down, so now you can buy and expect it to go back   up. And similarly, when it goes up sufficiently,  then we say okay, it's got run up, the run up   is done, now it's time to short it. Got it, so  that is one based off of the time series itself,   and then you can also look at it in terms of the  cross section. Okay. You can compare SPY with   let's say the NASDQ or something like that, but a  good example in India would be HDFC and HDFCBANK,   for example, like a pairs trade up. So there,  you can say like, OK HDFCBANK today has gone up  

x compared to let's say HDFC, and then you say,  OK, I'm going to short HDFCBANK and I'm going to   go long on HDFC, expecting both of them to come  back to the relative mean. Right. So that is   another type of meatier version, so there  you're not looking at the time series of   one individual instrument, but you're looking  at the time series of multiple instruments.   Correct. Now, the same thing can further be  extended into, let's say, a sector level, also,   so that's where comes your stat-arb strategies,  right, there you're saying, OK, instead of just   comparing two stocks, now I'm going to look  at the a bunch of stocks in a sector itself,   and I'm going to go long on a bunch of sector  on a bunch of stocks in that particular sector   relative to their value with respect to the  other stocks in the same sector. For example,   we might take a look at banking. So in banking, we  might say, OK, like let's say, SBI HDFCBANK, all   of them are slightly undervalued today, whereas  your ICICIBANK or AXISBANK are overvalued,   so I'm going to go short on a bunch of stocks there, and I'm going to go along on a bunch of   stocks in the same sector. As an extension to that,  then we can also look at the entire universe of  

the stocks itself, right, so there we don't need  to limit ourselves to a sector, so then we can go on   to, say, like we can do something like an index, we  can compare it with respect to the index. An index   can be something which you come up with on  your own, or it can be the index like a NIFTY   or something, and then you're going to say like,  okay, Nifty has done this, some of the stocks have   gone, have outperformed compared to NIFTY. Some of  the stocks have underperformed compared to NIFTY.  Got it, got it. So you know, for a newcomer to kind  of start building strategies of their own, right,    what types of trading strategies, or what category  of strategies, would you recommend for a newcomer   to kind of get started with, and why? OK, to get  started with, I would say, even before he gets into   actually designing a strategy, right, it becomes key  for him to understand what he's trying to trade.   For, example is he trying to trade NIFTY, so then, he  has to understand what are the characteristics of   that instrument. From at least, from my general  experience, NIFTY and BANKNIFTY are mostly  

trending stocks, so he could start experimenting  with something, like a simple moving average,   and then go on to more complicated ways of  extracting trends. Right. For example, then, if I go   to SPY, if someone wants to trade SPY, so then again,  he has to understand how SPY behaves, I mean, you   cannot go about implementing trend sort of systems  on SPY because you know that it will not work out.   Got it. All this is related to the time frames, but I'm  mostly speaking about a daily time frame here.   So then depending upon the particular  characteristic of the instrument, for   example if it's a simple trend it's a simple  momentum based strategy that he wants to try out   they can start off experimenting with a simple  moving average and see how it behaves and he can   also use any of the publicly available indicators,  from my experience most of them don't work, yeah,   okay but yeah given his understanding and if he  knows when to exactly trade a certain indicator I   think he can get it to work, but I would suggest  just start off with a simple moving average,   and see where it takes you from there. Got it, which  is probably one of the most common, you know   indicators that people use right. Yes, so in  terms of like just kind of challenges that they   would face right in terms of building a momentum  strategy you know versus a mean reverting strategy   what are some common things you've noticed with uh  with either of them any sort of biases or anything ? Yes, okay that is one thing which as quants,  we always need to take care of. The way we back test it.

The data that we backtest it on.  For example, I mean, it's better if he just splits his data set into two parts where he is going to say, OK, I'm going to test only on a particular   data set, and then I'm going to just leave out  the other data set as a test bit. Right,   where I'm not going to, so, let's say I'm going to  optimize a bunch of parameters on my training set,   I'm not going to go back and then re-optimize  those parameters to again fit the test set.   Right, yeah, because that is just for your testing,  to see like, OK, whatever theory you have is valid   or not. So once that gets invalidated, there's  no point right for you to trying to optimize it   over the entire time frame and then say like, OK,  something is working, let me start running it with   capital and all that, because more often than not,  that will lead to problems. Right. The other problem   is, you adding too many rules into one system. It's  better if you can keep the system as simple as  

possible, but come up with multiple systems. Yeah.  So that way, you're trying to hedge out your risk,   rather than trying to come up with one super  optimized system, saying, OK, I'll add a bunch   of rules, and then I'm going to be working just  for the sake of it working, and then you know that   most often than not, it's going to fail in  the forward test, or in your live environment.   Yep yep yep, that makes sense. Yeah,  so one thing I'll say about it,  

yeah, another thing is the data quality which  you're dealing with. Yes, that also makes a lot   of difference. For example, one thing we've noticed  is, are the open values right? I mean, I can run a   simple trend following strategy and say like, OK, I'm going to take the opening values of any day   with assuming the day level data is available to  me free of this one, I mean, it's available freely   in most websites, and you can go and download it  from YAHOO, but then you know that those numbers   are not realizable in your real life, right, that's  and again another problem which people have to   understand that, data matters. The quality of data  matters, the quality of data which you're going  

to backtest on is very important. Makes sense, so  you know, let's kind of pivot a little bit into   something a slightly different topic, but you  know still related to the topic of today,   so you know one of the kind of introductions  you and Bharath made to RAIN is this idea of   running multiple strategies in tandem, right. Yes.  Can you kind of elaborate on this whole concept ,  you know, why do you think it works, or why does  it work, I should say, and how does one kind   of think about that, you know? Yeah, and so, this  goes back to my point of, like, not trying to   over optimize something. There are only a certain  ways you can extract alpha from an instrument   yeah and any other way you try it will give  you something or something there about itself   but then you're trying you're better off adding  different types of strategy into your portfolio   than trying to say okay i'm just going to run  only a trend following on the same instrument   but i'm just going to change my moving  average settings right i'm just going to   make it from 100 to 50 or 25 yeah and this way  i'm going to come up with 10 different systems   but then if you can add a mean reversion strategy  to just one single trend following strategy   i mean the amount of diversification benefits that  you get will be humongous yeah just to give you an   example from the crypto markets yeah so crypto  markets we know mostly are trending markets   you know that okay once uh these are mostly  called reflexive markets when you know that okay   when something shoots up the trend is going to  continue for some time yeah instead of just adding   one strategy for example one of our strategies  had like a drawdown of close to like 15 percent   annualized to produce a return of 25 annualized  unlevered and then what i do i come up with   the same trading strategy but on let's say on a  different instrument for example let's say we take   bitcoin and ethereum as an example we run the same  strategy on bitcoin the same strategy on ethereum   and then we try to diversify across different time  frames right you can run strategies at a d level   data or at 1 hour 15 minutes these sort of things  and then you try to combine different philosophies   of reversion also in this from what we've  observed at a very lower time frame the   crypto markets are mean reverting yeah so  what happens yes and then you can also add   some intraday strategies yeah yeah on the same  instrument as well right so what this does is   eventually we were able to bring  down the drawdown from let's say 15   to now let's say it's close to five six percent  annualized but to produce the same sort of returns   yeah which would be like uh let's assume on a  conservative estimate like somewhere close to   50 to 70 percent unlevered right but then you see  the beauty of it right you're able to produce to   seventy percent with the five percent uh draw down  right as compared to running a single model which   would have suffered fifteen percent to produce  similar returns yeah so so technically if you   had a higher threshold for your drawdown you  could even double down on leverage if there was   leverage available right exactly yeah and then  yes so then that that comes to my next point   so what is the advantage of doing this yeah so  then you can start levering your instruments up   so let's say you were willing to take the  same 15 risk but now for the same 15 risk   when you know that your radon is 5 you can hit  a leverage of 3x and that automatically boosts   your returns up to like what 200 210 percent  that's amazing yeah yeah so i think that's   where the diversity helps one is you bring down  the drawdowns yeah which brings down your risk   and it can also help you take leverages uh  knowing okay what you're dealing with yeah for   example with the model which has like let's say  15 drawdown there's no way you're going to hit a   2x or even a 3x leverage because at one point  of time i mean the numbers might say anything   but when you're running something live you know  that you'll start panicking and you'll do that   right yeah it has to be realistic too exactly  yeah yes yeah it has to be yes the numbers are   one and then you also have to think about your  life trading experience exactly yeah i mean will   you still have confidence in the strategy after  something you've lost 15 percent of your capital   right it's something you'll have to think about  so that's why what we're trying to do here at   rain right now is to have uh to come up with for  each portfolio come up with multiple strategies   so that we are able to cut down on our risk one  thing which we've learned is returns will come   if you're able to manage risks properly right but  once you blow out capital there's no way you can   get back into the game exactly yeah that's the  number one rule right to not lose money is money   it's not to make money it's to not lose money  right yes returns will come yeah if your models   are good enough and given the market conditions  are right the returns will surely come right but   you need to have the holding power which is what  the diversity of strategies will give you right   all right so you know obviously you and have have  made strategies for other markets as well not just   india's capital markets you know cryptocurrencies  forex and also the u.s markets right  

so you know when you guys make strategies for the  u.s market specifically right u.s versus india   because obviously different asset classes have  different you know things going on for them but   specifically for the us and india how are the  strategies different you know when you kind of   uh you know your general approach towards the the  environments and how are they similar to yeah yeah   yes yeah so we don't try to differentiate between  whether it's a what sort of a market it is okay we   first try to identify okay what is the underlying  behavior of any instrument and then we go about   look for example nifty the index in india you know  that a trend following strategies will work there   the same thing will not work in the us over a  long term there we know that okay mean reversion   strategies at least for us has worked very well  yeah if you're able to handle the risks that   come with it the then the other aspect in u.s is  that's a more deeper market compared to let's say   india or your crypto currencies so that's again  a challenge which you'll have to deal with yeah   you'll have to start looking at probably lower  time frames and those sort of things to deal   with these issues for example a simple strategy  which works on even a daily time frame in india   would most likely not work out in us okay just  with the depth of the market and the liquidity   that's there in the market so there what we do  we try to go go at lower time frames and see like   okay what's the kind of alpha that's available  there okay that's interesting so you're saying   something which works for the daily market on  ohlc data for india would probably not work in   the us you're saying it'll not for example  one simple strategy i can tell you is the   fair trading strategy i mean that works very  beautifully on the day level data in india   but the same thing will not work in the us i mean  if you just got it at least from our experience   it hasn't yeah yeah so there then we try to  move to lower time frames and see like okay   if it's working there or not yeah got it got it  and what about like different asset classes so   you and bharath have built strategies for crypto  as well you know yes we're looking to launch soon   at rain as well but uh it is generally you know um  kind of you know from a high level perspective how   do cryptos kind of differ from stocks uh cryptos i  think are highly volatile compared to stocks right   right but then that's where you can also produce  your returns for example one simple model which   you are running in india a simple time strategy  will produce was producing close to again like   let's say for a drawdown of annualized max dot on  like 15 volatility okay it used to give us close   to 20 to 25 annualized returns the same strategy  for the exact same risks in the crypto markets   is giving us close to 80 to 90 percent annualized  wow so that's the major difference there   right but due to lack of competition or is it  because it's a new asset class or both yeah   it's a new asset class plus it's not as deep  again right so the moves tend to be large   so that's again where you need to handle  your risks i mean the there also again i mean   you're maxed out on can say anything but there's  nothing stopping bitcoin from like going down 30   in a day just like that right right right right so  i think we have to have a different conversation   on just crypto you know i'm realizing this on  the spot so we'll definitely do that and you know   so kind of looking at you know the behavior um  of of of a typical quad right especially you   know someone was kind of starting out building  their own strategies for the first time you know   what are some common mistakes you know quants make  when they're back testing and building strategies   okay some of the common mistake i would  say are biases which you introduce   it's some sort of look ahead bias right i mean you  know that okay you shouldn't be taking the closed   data of that day for example but you'll still  make a mistake of saying that okay just the data   points that you take you'll have to be careful  that i just do whatever i'm doing and data which   i know that okay it's not in the future yeah right  so that that is one point which you need to avoid   and then again you'll have also have to take care  of the way you implement your stop losses and all   that so that you know that okay most of the times  your stop losses are not going to get realized   right just because of the gaps and all that right  so that you'll have to be very careful of right   and also you'll need to think about okay  just the inherent uh biases which exist and   let's say let's say i'm trying to do a a startup  strategy on the universe of stocks in nifty today   but before today also there were other stocks  also in the nifty which have gotten taken out   for various other reasons right survivorship bias  survivorship bias exactly yeah so that is again   what you need to consider like okay something  has worked but i also need to know that okay   what hasn't worked right they also need to go  and analyze like okay why it hasn't worked and   tomorrow if such a situation occurs how am  i going to handle it right especially when   the bias you know is is implicit right so for  example um if you're talking about a strategy   where you're building you know a portfolio of of  stocks which are performing well right exactly uh   and then you have survivorship bias with an index  right that can make a huge difference because   yes but those two things are correlated right but  maybe in a stat or situation maybe not so much but   still i would say it's something you should look  out for right yeah yes yeah yeah you always have   to yes yeah that becomes even more relevant  in terms of pair trading and all those things   right and you also need to know more importantly  like when your strategies will not work   right right more than when your strategies  are working you also need to understand   okay when can the strategy fail for example a a  momentum or a trend following sort of a strategy   even if there are no trends in the markets  the market can go flat for quite a long time   a classic example would be 2017 in india  right or 2018 in the crypto markets when   let's say the bitcoin didn't move at all yeah  so there you're going to get whipped a lot yes   so then you'll need to consider those  things and say like okay these are the   situations wherein this strategy will fail  right and that is where again i would say   like so instead of running one thing you tend to  diversify yourself at a strategy level at a time   frame level as well because i mean there might  not be momentum let's say at the one day level   but there might be momentum at a low at some  lower time frames so if you can create that   as well along with your existing that'll also  help you cut your losses during the bad times   right right right yeah and basically covering  all the different market regimes right i think   yes yeah so so yeah i would say as a beginner you  would need to consider all these things so more   importantly as to why a strategy is working yeah  if a strategy is working you also need to make an   analysis as to like why it's working also right  i mean it just can't be just purely numbers right   right yeah that understanding has to be there  as to why something would work why something   would not work it makes sense yeah so that they're  conditioned yeah it is a never ending process yeah   that's a beautiful thing right because the markets  are evolving right see if anything you think is   true today you know might not hold true so these  concepts probably matter more than you know   any sort of uh concrete you know strategies or  anything right um so um yeah cool actually so last   question for today you know you know i mean if you  had to give some advice to any quants um listening   to this who are building strategies for the  first time because we do have many of those guys   listening to this uh what you know like what would  be your advice apart from the things that i've   said yeah one thing i would like to say is if it's  your money that you're trading you're still okay   yeah because it's your personal money but at least  at an organization level we know that people's   monies are involved i mean they're hard earned  money so we cannot be reckless with our strategies   right i mean we cannot choose strategies which  you know are going to undergo huge volatility   right and you cannot take unnecessary risks i  mean you can't take high leverages yeah right   so you need to be careful in your risk management  when you're dealing with at least an organization   right and not dealing with public's money  basically right so you need to understand   that okay i'm dealing with someone else's money  so let me ensure that i'm as safe as possible   with others money yeah and i designed my  strategies accordingly makes sense all right   man well thank you so much that was a really  engaging uh conversation and uh i'm sure we'll   have many many more well i mean we have ours all  the time but uh and i'm sure we'll have another   one of the podcast as well yeah thanks a lot yeah  it was a pleasure speaking to you yeah cheers bye you

2021-04-29

Show video