Interview with Marsten Parker
Alright, everyone. Welcome to another episode of Line Your Own Pockets. We have very special guest today with, Marston Parker who is the creator of the RealTest software and a market wizard. That's a big deal to be into. I love those book series.
Michael:So it's a huge, huge honor. 2 episodes ago or yeah. 2 episodes ago, we talked about all the different backtesting platforms that we used. I mentioned that I kind of exclusively right now use RealTest, which is, which is Marson's product. So I think this is gonna be a great episode to kinda get you started and and learn about how someone who builds these programs thinks and how someone who is a systematic trader and, again, a market wizard thinks.
Michael:So high level introduction. But for anyone who's just hearing about you for the first time, just doesn't hasn't come across the name before, Marcy, can you tell us just a little bit about yourself?
Marsten:Sure. I'll keep it brief. But, yeah, I originally was pursuing a career as a classical violinist, but I was also a self taught programmer since high school. Once I figured out I wasn't going to be the next Jascha Heifetz, I moved back to the Boston area where I grew up and managed to get a job as a coder, worked for a couple of different startups. Then by the mid nineties, one that I was an early employer of went public, Segway Software.
Marsten:And that's one that Jim Simons was the chairman of. And that left me with the problem of having capital that I had to figure out how to invest and an interest in the stock market, which were both first time events for me For some reason that I'm still trying to understand, about a year later, end of 97, I decided to actually leave that company and focus full time on trading. And, hence the chapter of the title of my chapter in the book. And as it happened, I did well. The results were good, but I was supporting my family with that, so I didn't compound any wealth.
Marsten:But but I produced really good stats that were good enough for Jack, so I ended up in the book. And, yeah, but then about 15 years later, I had a coincidence of a of a really serious drawdown coinciding with some very unusual expenses due to family health and house maintenance issues and sort of a perfect storm. So that left me with trying to figure out what to do. And I ended up it wasn't like near bankruptcy or anything, but it wasn't gonna work to to just trade for a living. So, after trying a few ideas, I said, well, I have this software that I've been using myself and had a few people using it for for decades.
Marsten:And why don't I overhaul it and and see if there's a market for it and and, you know, give it and that that was actually, like, 2017. And then, coincidentally, I that the somebody, Mark Ritchie, the second contacted me because he was trying to help find people for the book, and someone had given him my name. And Jack talked to me and decided to include me. And so that helped with the publicity for for Realtest, gave me the confidence to really put it out there. And, so been marketing it since the start of 2021, and and I'm really enjoying it.
Marsten:You know, I was a software developer before, and, like any developer, I I I'm motivated by, you know, being able to build things that people like to use and, and that are that are useful for other people. And I I'm just enjoying the the support aspect as well as it feels like a collaborative effort where where there's a really nice community. We're up to about 800 users that are, that are that all contribute to the to the product, you know, in terms of ideas and priorities. I just happen to do all the coding and documentation and support. But so so so, yeah, here I am, and I I actually have not been trading since June.
Marsten:The 4 strategies I had been running that did spectacularly well in 2020 and 2021, but that isn't saying anything because just about anything did so. Kinda ran out of gas in 2022, and they just went flat. I think that the edge just went out. I didn't have a drawdown. It just went flat.
Marsten:So, you know, at some point, I thought, well, I'm just not gonna keep putting putting risk on. So, and, yeah. So that's where I'm at that with that. At some point, I expect to get back into it and come up with some new strategies, but I feel like so much has changed in the market. It's hard to even know what what period is relevant for backtesting and so on.
Michael:Yeah.
Marsten:It's also hard to find find time to do the strategy research when I'm a one person full time software business.
Dave:So, Marson, that's a great overview of, your career. I wanna go back to actually the very beginning when you started trading. And, you know, a lot of people that come to me, they have been successful in another area of their career, like, maybe completely different from trading. So they come to trading, and they have a sense that they're gonna be really good at it or it's not gonna take a lot of time. Then all of a sudden they see this that it's harder than they think.
Dave:Right? Did you have a moment like that being a, you know, high level musician at the time coming to this field that, did you have confidence? Did you were you able to get over a hump? Was there any part of that process that you remember?
Marsten:Well, the you know, at the beginning, I just kind of I opened up my 1st brokerage account and, got my restricted shares, you know, unrestricted and transferred over there. And and so then I had had some capital, and I started, you know, watching some stocks and trying things almost at random and started to read trading books and so on. I didn't have a goal of becoming a trader at all, so I didn't have any preconception that I would do well at it or that I would fail at It it just was kind of a very present centered, you know, oh, I seem to be interested in this and and sort of dabbled in it. By the end of 1997, when I left my job, I had really mostly just done ad hoc discretionary trades, and I had lost about $50. And on that basis, I thought, oh, I might as well.
Marsten:But I had kind of turned the corner of saying, I need to make this systematic. That's the only thing that's going to work for me. And I had met a guy named Gary B. Smith online who was writing about a systematic approach to trading that was working for him. And I was corresponding with him a lot and decided to just adopt that approach.
Marsten:And we basically became trading partners. So I had a kind of a plan. So I guess I had confidence at that point based on the fact that I had met someone who was currently successful and was willing to to work with me.
Michael:So Well, I so I think that brings up a good point with with Gary that networking is something that, you know, me and Dave talked about before. And and and it's I think it's super important because everyone kinda looks at the market from a different lens and a different, different spectrum. Do you do any of that now with you know, you said you weren't trading recently, but I, for 1, I'm I don't post much on the real test forms, but I'm on there all the time. I get the email, I think, once a day that says, here's everything you missed, and I usually go and and read through that. Do you find that helpful at all that you have now?
Michael:You said 8 800 users, but you've got dozens of people in there talking about, you know, different strategies you're gonna try, and they're gonna be all different experience types. There's gonna be some who are gonna be really advanced when it comes to this and some that are are brand new. Do you ever get any idea or inspiration from just kind of reading your own forms at this point?
Marsten:Oh, a lot. I mean, if anything, it's too much information because, I mean, of course, I don't know what most people's strategies are. Real test is a piece of desktop software where your files are on your own computer. So I don't have access to user scripts, obviously. But a number of users share things with me or send me things or whatever.
Marsten:And I've listened to a whole number, a bunch of trading podcasts, and I'm exposed to the fire hose of information that we all are. So there's definitely no lack of ideas. It's sort of like what to focus on. And part of it with me is my own psychology has always been I think I'm somewhere between the 2 of you, just as I am on this screen, where I'm not a day trader, but I've never liked to hold a position for more than a week. So I've been just happiest in that sort of 1 to 5 day time frame.
Marsten:And it's probably because that's what Gary Smith's strategy used. And so I that's how I traded for my first few years of success. So I it probably just got embedded in me as this is what I need to do. So, like, that's the main way I tend to limit myself is that, oh, I have to find strategies that work on this time frame. I've just never been able to get my mind around, you know, like classic trend following or any anything with 0.5 MAR ratio, for example.
Marsten:I just don't see how people tolerate that.
Dave:So that's kind of where I'm at with that. Yeah. So, Marston, you mentioned that there's a you're gonna get a fire hose of ideas. I think a lot of traders, they find something that works, and then they kinda get stuck with the one strategy.
Marsten:And That was really
Dave:I've always said that the only thing harder than coming up with one strategy is coming up with the second strategy.
Marsten:So Right.
Dave:And it's it's a little ironic because, you know, like you said, there are a lot of ideas you can work on. It seems like we're inundated with ideas. So I like what you said about, you know, restricting it to a certain, subset of ideas that will work for you. Do you have any other advice for figuring out what strategies to work on given you got a whole bunch of ideas that might work, but they're each you're gonna have, you know, require a lot of your time to dig deep and and figure out what to do? Do you have any ideas for, prioritizing ideas and, you know, figuring out what to work on next?
Marsten:Not really. Because, see, my my personal approach has has always been kind of ad hoc. And and that's part of why Realtest is the way it is. It it emphasizes a very rapid feedback loop where it's you know, I work hard to make backtests run as fast as possible in it. And from the beginning, I've always, I should mention, focused on I'm trying to back test and train a model where every day you scan the entire stock market for some signal trading signal or whatever trading rules.
Marsten:So there's a fairly large amount of data involved. So so basically just sitting in front of a tool like that and iteratively trying stuff. You know, sometimes what I try would come from an idea, but but for, you know, mostly, like you you said earlier about people get stuck on their on one approach, and that that was certainly me. You know, I had this long short breakout strategy that had mostly worked for more than 10 years. And at at it, I was aware of other like, I was kind of vaguely aware there was this thing called mean reversion, but and I played with it a little bit.
Marsten:I just didn't like the feeling of it, you know, the catching a falling knife thing. So but I finally when the other systems clearly had lost their edge, I dug deeper into meaner version and convinced myself it would work and initially had really good results with it. Too good, because then I sized it up too much, and that was what led to the big drawdown. So basically, I've only ever used to well, no, there was a third idea, which was recent IPOs that came a little later. So basically I only ever used 3 ideas in all my trading short term breakouts and breakdowns, short term meaner version and trading recent IPOs at their all time highs.
Michael:And and while we're talking about multiple systems, right, that that was the thing that gravitated me to real test. I actually heard about it from, the chartist. I I think, you know, Nick Ratch. He he mentioned because I've been following him for a while, and I really liked his concept of these are a bunch of different, strategies that I I combine, and and that's where my edge is. Now, you know, in the book, well, that I just finished again listening to had, you you mentioned that your goal would be if you were if you were, you know, back doing this and doing it full time is essentially 4 strategies.
Michael:Right? To trend following maybe to mean reversion, something like that. And it's something that I think I focus on a lot and I think is really, really helped my trading is this multiple strategies approach. But in your testing and in your support and everything with the software, have do you have any idea of how many strategies you think is optimal? Have you done any looking at that?
Michael:Like, do you know clients that are running 20 strategies versus 5 versus like, is there any is there ever a point where it's kind of too many, assuming you have the buying power to to trade them all?
Marsten:I've seen the whole range of, the whole spectrum. On one end, you have one strategy on one symbol. And although I don't see that too much in real test because if you're just trading one symbol, you don't really need what it can do. But I think maybe it was more to test the software, but I've seen someone get over a 100. It gets pretty slow at that point.
Marsten:I bet. But, yeah, I've seen people who who like to have kind of 20 or 30. I mean, it's typically not there there aren't 20 or 30 different trading ideas. It's more like variations of the same strategy so that you make them more robust with respect to parameters. But some
Michael:So you kinda mean, like so someone, if they had a trend following strategy, they may look at a 10 day breakout and a 20 day breakout and a 30 and just test those as different strategies just to see what would happen there. Yeah.
Marsten:Right. Include them all as different strategies. Yeah. I have an example like that that does trend following on the S and P 100 constituents with 3 different look back lengths. You can clearly see how just that amount of diversification and also adding the constraint that they can all trade the same symbol really helps because we all know sometimes market cycles are longer and sometimes they're shorter.
Marsten:So that way you're kind of covering all cases. And I know that's quite common among trend following triggers. But I like with the 4 that you mentioned, I like the idea. I think of it as a sort of a 2 by 2 grid where you have long and short on one axis and meaner version and breakout on the other. And that kind of really covers all the bases, at least except that, as I mentioned, I limit myself to the same kind of time frame for all of them, at least for now.
Marsten:I have experimented with adding oh, like, you know, monthly NASDAQ 100 rotation or that S and P, trend following example that I talked about or other things like that, which can yeah. But then you get into the question of of, you know, capital usage. Like, with with short term shorter term things, you can you can combine 4 short term strategies that are each using 100% of your capital, especially if all of them are rarely fully invested. They're all the type of strategy that is opportunistic. Yes.
Marsten:It takes an unusual market. Basically, it takes a crash, for example, for a long meaner version strategy to get all of its entries at once. Yeah. Given that, I mean, with meaner version, it's kind of like the more extreme pullback you're looking for, the better the expectancy, but the fewer the entries. So you're always kind of looking for the knee of the curve of quality versus quantity on the trade count.
Michael:That becomes hard. Yeah.
Marsten:But when you combine it with other things, that becomes an advantage if it doesn't trade so frequently, because then you can use that capacity for other strategies. But when you start putting in longer term buy and hold ish things, you know, they just they just are consuming capital. Now the the the one probable or possible next direction I wanna go in is is is adding futures trend following to shorter term stock trading because that would be a really efficient use of of of capital.
Michael:Yep. That's a very Jerry Parker esque with the the future trend following.
Marsten:Well, he does stock trend following in futures. He's, you know, t trend following plus nothing. Right?
Michael:Yeah. That's his whole motto. Yep. So
Dave:one of my favorite quotes is, if you torture your data hard enough, it'll confess to anything. So
Michael:Yeah.
Dave:How do you how do you, ride that line between knowing whether you've tortured your data too much, you tortured your strategy too much versus optimizing. How do you think about that, and, you know, how do you how have you addressed that over the years?
Marsten:Well, in the in the early years, including some of my most successful years, I really had no clue about over optimization. I literally was quite guilty of using too many parameters and over optimizing them. I mean, it was all within the same strategy concept of short term breakouts. But there's a target percent and a stop percent and a time stop. And how much does it have to move today to be considered a breakout?
Marsten:And how many times over average does volume have to be so on? It's easy to come up with 10 parameters, and I would optimize all of them. And I would tend to leave them alone until I got into a fairly good sized drawdown, like maybe 10%. And then I would, and I had a rule to stop trading if I hit 20%. And then I would re optimize.
Marsten:Okay. How could I have avoided that drawdown? And then I would start again. And at any given time, like, from 1998 through 2010 or so, I was running strategies where the back test said something like 100% return and a 10% max drawdown. And and I was actually getting getting more like 20% return and a 20% max drawdown or 25.
Marsten:So during those years, I actually averaged, like, 27 a year. But, anyway, you know, and I I just kind of thought that's how it worked. I was fine with that. I I I just learned through experience that your actual results are not gonna be anything like your back test results. And I'm like, okay.
Marsten:Fine. But I still thought, I was aware of that problem, but I still thought, I mean, would would I rather be running a strategy with better back test results or one with worse back test results? It made more sense to me to run the one with better results, so that's what I did. You know, I would do a fair amount of not too much, but sometimes, you know, the thing that I know you've talked about, Dave, of looking at individual trades as a way to try to refine your rules. How could you have filtered out not one bad trade?
Marsten:I mean, that's ludicrous, but how could you have improved your stats?
Michael:Well, and I think that comes to
Marsten:but let me just finish that. In more recent years, though, I've really moved away from that. At some point, I just became too self conscious that that approach didn't really make sense. And I started first of all, I limited my parameter granularity. So I would just say, gave myself a bias of, like, whole number of parameters.
Marsten:Or like, if I was going to optimize, I could only pick 3 or 5 values or something. And so that's kind of how I went. I mean, because you have to do some amount of because part of what you're doing is discovering the nature of the market. And part of what you're doing is curve fitting to bad data, to pass data.
Michael:Well, and that becomes a lot of the one place, even for someone who's fully systematic like yourself and and fully systematic, like Dave, where I'm not quite there. I'm still semi discretionary. But the, a lot of that discretion comes into what do you curve fit and when. Right? You know, that becomes something that,
Marsten:you
Michael:know, I think that the human still has a very important job on because you're right. If you take especially in real tests, you can just run an optimizer and says, hey, you should be using a, you know, a 34.5 moving average. And then it takes the human to look at it and say, well, why? Right? Why does, you know, not the 20 or the 50 or, you know, why do these whole numbers don't work when this one random value does?
Michael:So I think that's just an important something to mention to the audience. Right? Is is, yes, optimization can be useful when you're doing it from kind of an exploratory nature of, for example, you know, do I want to trade more volatile stocks or less files or does this work better on more expensive stocks or cheaper stocks? But when you find that there's just one piece of data that just pops out and says, oh, yeah, this this random assortment of indicators ends up working where, you know, if I move the indicators kind of 1 tick down or 1 tick up on on a numerical scale, then all of a sudden everything breaks. Right?
Michael:So that's a, I think, a great level to make sure that that, that you're doing that with some sort of discretion even if you're not gonna add any discretion to your trading. But that was one question I wanted to ask because you went from kind of fully discretionary to just right off the deep end and said I'm gonna be fully kind of systematic forever. Do you ever attempt are you ever tempted to kind of come back a little bit and use systems to help find the entries and and using that back test, but using just a little bit of technical analysis or anything like that to for trade decision?
Marsten:Not recently. Yeah. I mean, my yeah. But when I was just trading completely ad hoc discretionary, I proved to myself fairly quickly that that was stupid. And then I met Gary, and our approach was kind of semi.
Marsten:It was mostly systematic, I would say. We had one discretionary step of reviewing the charts for our entries, but then the exits were completely systematic. And the entries, there was a scan that produced fairly few candidates. So that part was also semi systematic. But I would never try to write a fitness function that produces a score and say whatever parameters have the best score, I automatically use those.
Marsten:So I've never been kind of a pure quant in that sense. I'm 100% mechanical trader and a pretty much discretionary researcher. So so I'm a big believer. Like, even if you run an optimization, the, you know, the next step is to try real test shows all the tests in a nice window where you can double click on a road to see see the equity graph and then just hit the down arrow to go through them. And, and you get a feel for that beyond the stats by looking at different graphs of the tests.
Marsten:And on that basis, you kind of choose just the parameters that feel the most reasonable and sensible to you.
Michael:Yeah. And I I I know Dave's probably itching for a question, but I did wanna hit on that. I love the I love that quote of I'm a researcher. I think so many traders try to look at themselves as I'm someone who I'm going to hit a button and hopefully make money. And then they don't look at themselves as that idea of, you know, when I'm when I'm back testing systems, when I'm optimizing systems, that that that's the kind of hat or mind frame that they should get into is I'm I'm researching the market.
Michael:Like you mentioned, I'm studying the market. I'm trying to figure out how the market moves in response to x, y, z stimuli, the same as, you know, anyone would do from sending any sort of scientific endeavor. So I I did just wanna hit on that because I thought that was a great way to kinda reframe yourself, and I think that would help a lot of people if they just go in and say, hey. I'm just I'm researching. If x happens, does y happen more often than not?
Marsten:Totally. I've I've actually never really identified as a trader. I, yeah, I I I I design trading strategies because that's how I wanna approach in investing is with short term trading strategies. So you do you when you evaluate ideas
Dave:or when you had evaluated ideas, is there some sort of minimum threshold that you looked at before you thought, okay. Yeah. This one is worth looking at further and doing more research on it. Do you ever remember any sort of, rules of thumb that you had for yourself or whether an an idea was, you know, worth looking at further?
Marsten:Yeah. I would say I mean, first of all, it has to be pretty simple. I'm I'm I'm pretty skeptical of of, you know, really complicated indicators and stuff. The and in terms of stats, you know, typically, I I'm generally just kind of looking at the the rate of return and the draw downs and and the expectancies. And, you know, I have a there's a kind of a ballpark area that I that I'm looking for or have been typically, which although it's getting harder to find, which is, your rate of return over 20%, max drawdown under 20%.
Marsten:So MAR ratio of greater than 1. And and expectancy high enough so that, you know, you don't have you don't have to worry about execution too much. So so, like, I don't know, half a percent expectancy or higher or something like that on a per trade basis. There's one rabbit hole that many people go down with systematic researches is the illusion of the very high trade count strategy, very short term, very high trade count without factoring in slippage or whether all your limit orders are going to get hit. And you can see some ridiculous looking stats and equity curves.
Marsten:And when you try it in real life, it definitely doesn't work.
Dave:Yeah. That's, you know, a lot of traders will do back testing, come upon something that looks great, and then they feel like they're done. Right? But that really, they don't realize the work starts once you start making trades and, like, figuring out how it compares to the back test. Now I was struck with your I was struck with your, you know, your back test strategy show a 100% gains, and then you were expecting 20% eventually once you had some experience in dealing with it.
Dave:Do you feel like that's a good expected return when comparing the backtest to real life. And I know that's something that a lot of traders will sort of all of a sudden, they blame the back test. They're like, man, this back tester sucks. Right? Because it doesn't reflect reality.
Dave:Surely, you get some of that, feedback. And how do you, how do you approach people through that process of figuring out reality versus theory?
Marsten:Well, my usual rule of thumb 100% was kind of extreme. And that was partly because that was like back in the late 90s, early 2000s, where it was very easy to find a strategy with stats like that. Now it's pretty hard to even hit, like, 30%. But my old rule of thumb was just expect half the return and twice the drawdown. If you know you can only tolerate a 20% drawdown in live trading, then probably don't trade a system that thinks it will be more than 10%.
Marsten:But it depends on the kind of strategy. There's an important distinction there to look at, which is out of sample results versus in sample results. In other words, what your what your live trading going forward is compared to what the back test was before you started. That's the thing where I where I said, you know, twice twice the drawdown and half the return. But then there's also for the period you've been live trading, you've got to run the back test, and and that should be pretty close.
Marsten:I mean, if if that's not close, then there's a flaw in your back testing software or a flaw in your executions or or you didn't account for enough slippage or whatever. So those are 2 separate metrics that have to be considered.
Michael:Now, you know, when you mentioned a lot in the in the chapter in the book to that and we talked a little bit about here that sometimes you had strategies, they worked fantastically, and then all of a sudden the edge disappeared and, you know, you you were left kinda looking for that that next thing. This is something that I'm always trying to figure out an answer for, and I'm I'm sure there's not a perfectly correct one. So we'll we'll have to just pontificate a bit. But what are your suggestions to someone who is a systematic trader, who is kind of going through this on trying to identify what is kind of a normal drawdown for a system versus maybe an extended drawdown for the system versus this just isn't working anymore, and I've I've gotta I've gotta go do something else.
Marsten:Yeah. That is the big question. Name conference that I recently, spoke at. We had a panel on that, and it and nobody had a really clear answer. But, you know, I mean, it's it's sort of a judgment call.
Marsten:I mean, some of the things I would say are are you have to have some concept of what the source of your edge is, like, why your strategy worked. And so the first thing to check is might that have changed. And back in 2005, when that was the most dramatic example from the book, where the end of day entries, my edge just seemed to fail. And my usual approach of reoptimizing and finding something that would have worked better. I couldn't find any parameters that would have worked.
Marsten:So I kind of ended up being going on for a while like that. So I kind of concluded that that didn't work anymore. And then if there's an external thing you can incorporate like, I was talking to some people who knew somebody who knew, like, an institutional trader who worked for a mutual fund and was telling him about these new algorithmic execution platforms that they had so they no longer had to do big block trades. They could cut them up into little pieces and be much smarter about that. Okay, that was a big part of my edge.
Marsten:All edges are like being slightly earlier than somebody else who's trading larger than you. That's the only way to make money in the market. I used to like to say the only edge is being early. So if you can identify that you're no longer early, that's a very good reason to stop a strategy. I mean or another like, now my, IPO strategy stopped working for obvious reasons.
Marsten:That's an easy one. There aren't any IPOs, and and they're all bad. Yeah. I mean right. They're getting faded more.
Marsten:I mean, the fact that the fact that IPOs dry up in a bear market was a feature of the strategy, not a bug because it was kind of like a built in regime filter. Mhmm. But but I don't know what's going on now. I mean, may maybe it'll come back at some point. You know, I I I check them periodically to see how they're how they're doing.
Marsten:And the and then the weirdest thing is that that strategy that failed in 2,005, which there's kind of a version of it in in the real test examples folder called MHP classic. And, it suddenly started working in 2022 when all my others failed. So
Michael:Well, yeah, that that was actually gonna lead that leads me perfectly to my follow-up question of if a strategy dies or lose its edge, and I I I probably already answered that already, you don't just delete it and move on and never check it again. Like, is there do you have some sort of process to say, hey. You know, maybe once every few months, I should dig up the old strategies that I killed and just see maybe they're coming back to life in in some way. And have you ever kind of brought any back from the dead like that I that idea that's kinda come back? If that if that continues to perform for, you know, a number of months or years or something, do you ever consider dipping your toe back in it?
Marsten:Yeah. I do. Yeah. I'm actually running now in a paper account, a version of that old long short strategy with smarter position sizing rules than I used to use and in combination with a long short meaner version. So that gives me my 2 by 2 grid that I like.
Marsten:And it's, the expectancy is a little low for my taste, but it's, it looks promising. We'll see see how it goes. I should just mention that I don't think running I think the main use of an IB paper account is to make sure your automation works because it does not give you a reliable sense of execution at all. If anything, if you get good stats in an IV paper account with short term trading, you'll probably do better in life because their execution modeling is very poor. You basically have to wait for actual trades in the market with sufficient volume at your price before they say your field.
Marsten:So it's not like you're participating in the market. But anyway, that's Probably
Michael:it's probably better that way than the other way around, I would guess. Yeah.
Dave:Right. Yeah. That's true. I've I've experienced that same thing.
Marsten:Yeah.
Dave:So one thing I'm struck by just this whole conversation is, you know, here we are talking about automated trading. We're 2 developers. You know, we're used to writing code if this then that. And yet a lot of what we're talking about here is discretionary decisions in the research process or that, you know, whether to turn strategies off, whether to turn them back on. I think people will be surprised to see just how much discretion there is with fully automated trading.
Michael:You mean I just don't go to the beach and and, and then take naps all day and make $1,000,000? That's not the way it works?
Dave:Right. Believe it or not, that's that's not how it works.
Marsten:So, you know, as
Dave:a developer yourself, did you did you set out and you said you said you went to automation pretty quickly. Did you think that, hey. I'm gonna have this completely automated where I don't do there's no hardly discretion at all, or did you have and did you you know, tell me more about how your thought process there over the years.
Marsten:Well, the first thing I automated a little was kind of collecting stats, because from the beginning, it's probably part of why I succeeded is I had an instinct to always be very accurate with recording my stats and review them and try to learn from them. And then when I was doing manual order entry, I realized I would make errors a little too often. And so that's the most logical next step with automation is to just do the orders. But I never had a goal of some people who are attracted to algorithmic trading have a goal of they're going to build a perfect system and turn it on and then kind of come back in a year and be rich. That to me has never made any sense.
Marsten:And I know that even in, like, big quant hedge funds, there are staff who are sitting there watching the thing and supervising it and intervening sometimes. Not often, but sometimes there's always a level of discretion in this. So each person has to find the right balance. I mean, I'm aware of some real test users much more automated. They're, like, content to have something running, and they won't even check it, like, all week, but it'll email them if there's an issue, you know, that sort of thing.
Marsten:I've always liked the manual process of, like, clicking a few buttons every day. Like, I like to manually tell Norge to update my data and then manually tell Realtest to generate my orders and then take a quick glance at them and then manually press the start button on the automation app. And Yeah. Watch it watch it a little to make sure it's doing the right thing. And and, you know, but there have certainly been times when I've been unable to be at my desk at that time or whatever, and then I might set it up to run automatically, but I prefer to kind of be involved a little bit.
Marsten:But Yeah. There is if it takes only 10 minutes a day, that's fine. That's good.
Dave:Yeah. There is I I agree. There's value in like, there's some parts of my process I could automate that could be complete automated, like like writing to my journal and keeping keeping exact stats of, each trade could automatically be done. But I I do value doing that after each day, entering the trades, you know, not quite in an automated fashion, but it's something I have to do. And I have to see it and watch those stats coming in and and assign the strategy to the trade.
Dave:So, yeah, there's a lot of value in that. I I I totally agree. So here's a question that, you know, you know, chat g g GPT is on the scene. It's been around for seems like it's been a long time, but it's only been a few months now, actually.
Marsten:Yeah.
Dave:Have you worked that into your development process or any of your trading process at all yet? And how do you think about the chat gpt and just LLMs in general?
Marsten:Yeah. I use chat gpt and also Claude fairly frequently. Not for, like, you know, very big things. I mean, I just I use it to try to save me time. Like, I tend to use it instead of Google a lot nowadays.
Marsten:Or, I mean, I've used it for some development tasks that aren't on real test itself, but, you know, like write Python code that talks to these various APIs, like, that that Fast Spring uses for my sales and Lime LM uses for my license management and blah, blah, blah. Integrate all these. It's a huge time saver as opposed to having to look up and learn all these different rest APIs and write the code for them. Yeah. Especially if you want to write some Python.
Marsten:And I'm still not super fluent in Python. And I had a batch script example in real test of how to automate something in PowerShell. And I just told chat GPT to translate it to Python, and that worked well. So stuff like that. Just looking things up, getting information.
Marsten:And also, like, recently I've been working on making sure my support for trading futures where the currency is different from your base currency is done and accounted for correctly, which is a little complicated. So I'll ask it to work through the accounting of a trade like that or even to just look up what's the tick size and point value of Nikkei Futures in yen, stuff like that, rather than having to try to find the exchange website or whatever. That's the level at which I'm using it. I'm not using it for, like, you know, find me give me a new trading idea or things like that.
Michael:No. And I, and I think I've already put this in the show notes to one episode, but I'll put it in the show notes to this one as well because I've created my own real test GPT where I fed it the, the user guide, download the user guide in PDF, and I upload it there. And as someone who's not a coder and and not a developer myself, I found it hugely beneficial because, as, someone who's played with a lot of the back testers, I I settled on Realtest for a couple of reasons. But one of it, it it was simpler for, again, someone like me who's not as as savvy as you guys, but I find it does an amazing job for us, Lehman, getting us unstuck. I don't know how many time I've just taken my real test code and just paste it in there and go, what did I do wrong?
Michael:There was one time where the results just looked too good, and I couldn't figure out what I did wrong. And I was like, there's no way this thing's gonna make, what it actually is. And I just paste it in. I said, you know, what's going on here? And I had some look ahead bias in in the code.
Michael:So it was basically looking at the next, and it was able to identify that yet the it's you're looking ahead at the next bar. So, you know, in the real world, you'll never get this. And I I just thought that was that was a huge moment for me where it's like, yeah, as opposed to someone who's, you know, a layman and may maybe in coding and in trading and would see the results that I got, which, again, is, like, 250% a year with, like, a 5% max drawdown. So, someone who's really new could say, oh, well, I just hacked the matrix. I figured it out, and so risking the real money.
Michael:With me, at least, I knew there was something wrong with it, but I just couldn't figure out what. But yet, Chesh Ept was able to take it and say, yeah. You're looking ahead at the next bar in this line right here. And then I changed it and all of a sudden came back to reality. I was really excited for a second.
Marsten:Yeah. That explains why you hardly ever ask questions on the forum. But, yeah, I'm surprised about the look at I mean, it's very hard to look ahead in real test. The only way is to deliberately use a negative bar offset or to, like, calculate a bar offset that ends up being a negative number.
Michael:And that's what I was I was trying to use some sort of anchored view app strategy. Yeah. Yeah. And I was trying to find a low point, I think, in the chart and anchor it to that. And it was the way I did it is it offset it so that it it knew the day before.
Michael:But, yeah, I was just very, very impressed that it it so, again, if you're a Realtest user, I'll put in in the show notes here the the link to that. Feel free to use it and ask questions like that if you ever
Marsten:I'll play with that. Yeah. That's I I've had a few users, you know, approach me about that or show me things that but I I haven't at some point, I should really put it on my website and make, you know, make it part of what's available. Probably save me a lot of a lot of support time.
Michael:Yeah. I'll I'll send it over to you for sure. It's been it's been good. And and so far, it only has the next exercise to try to get the support form, which, again, I'm a huge fan of. Me and Dave talked about this.
Michael:Our backtesting video isn't out yet, but, we talked about how I tried Amni broker and the forms there were, hostile to to say at least. And then I came over to Realtest and I again, by reading the forms, everyone's super nice and everyone's super helpful. So I would like to, at some point, feed that GPT all of the support forms, and I bet you it gets even it gets even smarter after that.
Marsten:Oh, yeah. I can probably get the discourse software to to dump it out as a big file or something. It'd probably be easier than trying to do a screen scrape of all the posts.
Michael:That'd be great. Great. See? Look. We're we're doing a podcast.
Michael:We're chatting, and we're we're building stuff at the same time. It's fantastic.
Marsten:I am so I'm going on the forum as chat MHP.
Dave:So one more question for me. I was talking to I I was given a talk at the local university here yesterday to some students, and they were asking about trading. And they said, you know, what are the what's the hardest thing about trading? And I think the hardest thing is having to reinvent yourself basically every so often, like, completely throw out what you've done in the past and reinvent what you're doing. And it seems like looking back at your career, you've had to do that and done that successfully a few times.
Dave:Do you have any advice for people that are going through that or are getting ready to go through that and they don't realize it or, you know, adapting over time?
Marsten:That is a really good question. I mean, I'm kind of at that point now where I I feel like I need to reinvent how I trade, you know, going forward. And, you know, I've you you were talking earlier, about going back or maybe Michael was looking at your former strategies. And I typically, when I when I hit a rough patch, that's what I've started to do. You know, I I have, actually, in the older old one point o version of RealTest, there was a a window called the system window that had a whole list of strategies, and you could kind of just go down and then run them one after the other.
Marsten:And, so I just kind of do that to see has anything been working well recently? And just going through, like, if you run 20 or 30 old versions of a strategy, even if they're kind of the same concept with different parameters, And then just notice how many have had good results and how many haven't. That gives you a good feeling for whether the concept overall is broken touching back into that topic. But it's it's kind of I mean, it's the same. It it's I mean, personally, my answer is is always always comes back to iterative experimentation.
Marsten:I mean, that's kind of how I approach everything. Maybe it's because of trying to be a violinist. You just practice every day, and you keep trying stuff. And you have to get into a feedback loop where you say, okay, this wasn't good for this reason. Let me try this instead.
Marsten:So it would be the same thing with research. You know, I just kind of it's kind of a combination of persistent repetition and and keeping your mind open. So so you introduce some ideas from left field while you're doing that. But I don't know how to turn it into, like, a checklist or a process or a set of guidelines. You just kind of have to show up and do it.
Marsten:And you kind of know it when you see it. At some point, you find something that feels like it might work. In the end, it's about getting to a point where you're confident enough to put your money into something that it's more likely to work than not.
Dave:Yeah. I love your idea about going back and looking at the strategy that was working at one point. It's always good to go back and review charts, I think. You know, look at ones where you made the most money, look at ones where you lost the most money. Doing a deep dive and comparing fills and really doing hard research into something that has worked.
Dave:There's always there's always good that comes out of it. There's always something you learn from it. And, yeah, I think that's a great that's great advice.
Marsten:Yeah. You just have to always be aware of, sample size in that exercise because if you get too fixated on 2 or 3 trades, generally, it's random what happened with those trades. But I think it's useful to kind of quickly flip through a lot of trades also. Or if you run a whole series of tests like an optimization, quickly flip through a lot of equity graphs and then come back, cycle back around and go a little slower and zoom into something, focus on something. But, yeah, certainly studying the trades of a system on charts that show the entry and exit points and maybe plotting a few indicators, if you want, on those charts and thinking about that can be helpful.
Marsten:Most of the time, we don't really know why. You can never know why any individual trade worked. That's kind of the paradox of it. Each individual trade is just part of the sample. It's really about understanding the sample.
Marsten:But but somehow studying the trades is an important input to that regardless.
Michael:Well, yeah, it's like you you can't predict the next coin flip, but if you do enough of them, then you get a you know, you'll you know what's gonna what's gonna happen. But that doesn't give you an edge Alright. On knowing what's gonna happen.
Marsten:My my thumb just didn't have quite the right feeling.
Michael:Well, listen. This has been absolutely fantastic. I I had a lot of fun. I I hope you'll you'll come back. But before we we leave people is where where can everyone reach out to you?
Michael:Where where can they find your find your work, social medias, you know, whatever you wanna point people to?
Marsten:My website is mhptrading.com. That actually has links to everything else. So that's that's the best place to to learn about Realtest and and see all links to wherever else I have a present. This podcast will appear on there once it's released along with all the others I've been on. So yeah.
Marsten:I'm on x formerly known as Twitter as
Michael:Also always call it Twitter. Yeah.
Marsten:Mars 10p, m a r s 10p. Because I came from Mars when I was 10. And those are the main main places. I I don't ever tweet very often, but it seems like other people like to tweet about real tests so I can be on there so I can like them.
Michael:Awesome. Well, again, thank you for coming by. I appreciate it. I know the the users appreciate hearing from, another systematic trader and a market wizard. So, definitely, the more systematic traders we can get on, the better because I I think a lot of people think it's just me and Dave sometime because everyone else out there is is just talking about their fundamental thesis and what happened in the news and and all of this.
Michael:So it's great to show the people that, there are some other nerds out there like us that just look at it kind of as a giant research experience.
Marsten:For sure. Yeah. Yeah. I really you guys are doing a great job on this, podcast series, and I hope you keep at it. And I'm really, really glad to participate.
Michael:Oh, I appreciate it. We have a lot of fun. We'll we'll be we'll be around for a long time, I'm sure.
Marsten:Yeah.
Dave:Thanks for the comments there, Marcin. That's great. Great to have you on and, a great conversation. Thank you.
Michael:Thank you. Talk soon.
Marsten:Bye.
Creators and Guests
