213: John Roberson – An Obsession with Edge, Powered by Automation

Chat with Traders is brought to you by Trade the Pool. Did you know that every decade the market reinvents itself? Online brokers opened the doors, mobile apps made trading seamless, and commission free trading erased barriers. Now a new era has begun. Meet, trade the pool, limited risk trading. And now you also have unlimited time to reach the profit target. From now on, your trading risk is capped, and your trading opportunities are limitless.

Trade the pool funds home-based stock traders with up to$200,000 in buying power. That means you can trade larger positions and scale your strategies without risking your own savings. It's time to trade with more capital, making it truly worth your time and effort. Ready to trade the pool? Click the link in the description and join the stock trading revolution today. Ladies and gents, this is Chat With Traders episode 213.

Today I speak with John Robertson. And if you ever speak with John yourself, well, maybe he will tell you he's simply a hick from Texas. However, I will tell you he's a well-experienced equities trader who's been able to harness the power of technology and leverage the skills of his team to maximize his trading operation. John started out in two thousand seven and has been involved in various groups since. Today though, he is with Kirschner Trading Group in Austin, Texas.

Although John originated as a discretionary click trader, over the years he began to automate more and more. To the point that he now runs a portfolio of eighty plus automated models that do everything from news and earnings to mean reversion and momentum to arbitrage and microstructure. In our chat, John tells of his early days, which are particularly interesting because it was one tiny edge he stumbled across almost a year in that enabled him to start making money consistently.

and it afforded him more time to discover other trades or other edges. But the main part of our chat is centered around automated trading and lessons from his years of practice. Make sure you stay tuned for the part where John tells a frightening pull the plug story of when a stop loss order malfunctioned while trading Chevron stock.

I have no idea exactly how much this cost him, but within seven seconds his account had been drained to just$6,000. I'll post a visualization of the trades from Nanax in the show notes. It's wild. Uh the talk of automated trading leads into machine learning and John's use of such techniques. Please note, at one point I asked John what resources he found helpful for learning and understanding machine learning.

He read several books on the subject but couldn't think of the titles while we were talking. He's since sent me the links to all of these and they can be found in the show notes at ChatwithTraders.com slash two one three To take us out, John talks on obsessing over your edge and why he adheres to the concept of thinking really, really small. And that's it, my friends. Here is John Robertson for episode two hundred and thirteen.

That's a very simple question to get things rolling. How did you get into trading? Completely by accident. Just absolutely completely by accident. And uh in college I studied philosophy and I wanted to be a philosophy professor. So I was I was gonna do my PhD. I was super excited. I had the guy, uh inexplicably the guy that I wanted to to study under said yes and and that was great. But I got married.

And so I got married and you know, that's that's great. It didn't change my plans, but I said, Well look. Um uh somebody with a philosophy PhD is never gonna make any money and we have a small amount of debt, so why don't we work for one year? We'll work for one year, we'll pay off our debt, and then we'll be uh and then we'll we'll continue in school. So we said, okay, that's great. Uh my wife is a she was a school teacher.

Um, and so that was great. That was more than enough to pay our bills. And then the goal was I would find some job um and hopefully make enough money to pay off all the debt. Well, so as it turns out, nobody wants to hire a philosophy major. So I couldn't get a job anywhere. I mean, I even did I did a bunch of stuff in the philosophy of math and somehow I couldn't even convince some random private school to to hire me teaching a little bit of math. Um and so there was this

So then there was this trading shop that one of my roommates in college was trading at and he said, Hey, come over here. I said, Okay, well there's nothing else that I can do. So I'll go over there. And that was how I got started trading just completely by accident. Now this trading shop he was at, that was a prop firm, I take it.

Yeah, so there was well the the details of it are a little weird and complicated because the regulatory treatment was changing at the time, but it was regul it was retail and then prop. Um so I was I was sort of Yeah, I I I was retail for the first few months and then I became prop after a few months. There was all the regs were changing in the market and that kind of stuff.

So from the sounds of things, it was your your goal in this year, which you were taking off from school to earn some money to pay off your debts. How did you go during that first year? Did you earn enough money to pay off your debt? Well, that's complicated. Yeah. So sort of. Um so so I started in July of two thousand seven. And uh and from July two thousand seven through March two thousand eight, um I tried everything, man. I mean, I just I talked to everyone.

who would talk to me and and I I listened to them. I had this I had this sense that I wasn't gonna steal anyone's trade. So so a couple of people told me some strategies that actually would have worked. And I mean, looking back on it, I was being stupid. They were they were they were inviting me to to try out some of their kind of lower tier strategies helped me get my start. But I I was too stubborn and so I I just

I just did anything for about nine months. Um and then in March, by by the end of March. I started with five thousand bucks and I was down to eight hundred dollars actually. So I was down from five thousand to eight hundred dollars and I mean I was I I I was pretty stressed. It was like this this year's gonna be a failure. Um I won't have made any money, I won't have anything. And so I went back home. Um and uh and it was a long weekend. I think it was Easter weekend maybe.

And so so it was a long weekend and I had these really, really, really extensive journals. Um these journals that I had kept. uh for that whatever it is eight months. And I looked through the journals and I tried to figure out, look, out of all these different things that I've tried, did anything actually work? And I was guided uh by this this idea I'd read in the in the Market Wizards book, uh the first one I think. He says, um I think he's Jim Rogers.

He says, My trading style is basically I stand in the middle of a room and I look around till I see a pile of money sitting in a corner. Then I walk over and I get the pile of money and that's it. And so I said, look, somewhere in this journal, there has to be a little pile of money for me to walk over, pick up, and come back. And so sure enough, I I did find something. There was just this one tiny trade that I found, but it worked incredibly consistently. Um I I I can tell you

more about it um in a bit, but I I don't want to distract from the narrative. And so that I went from trading all these different things and trying all these different things to I just had one strategy that I could do most days. basically one time. I basically had enough capital um to be able to kind of go all in once on a pretty small arbitrage. And so sure enough, I mean somehow as luck would have it,

Uh that very next month in April. So I had 800 bucks in my account and in April I made eleven hundred bucks. And that month I only had one down day as I recall and I lost six dollars. So it's like, wow, you know, maybe this, maybe this market wizard guy really knows what he's talking about. Um, I should try it again. So sure enough, then in May the next month, made a thousand bucks again. All right. So I figured, well, might as well double my size.

So I got more aggressive. I doubled my size. And then in June, I made two thousand dollars. Then four, then eight, then sixteen, then thirty-two, sixty-four, one twenty-eight. Technically after twelve months, I think I had just broken even. But within a few months after that, I think I had made like three hundred and fifty grand, something like that.

by the by the end of two thousand eight. So I'd gone from, you know, eight hundred bucks to three hundred and fifty grand. I'd I I wish that I had a strategy that was that good again so that I could make um Those kind of returns again. If I do, Aaron, I'll I'll share it with you. And so you you and I can both trade it. I'd appreciate that. You're welcome.

Well I guess uh it's gonna be no surprise what my next question is. You know, can you talk about what that strategy was? Yeah, sure. So so it it doesn't work now. And in fact, it didn't work after just a couple of months. It's a little bit complicated. Uh so I'll try to hit the high notes. So in two thousand seven, two thousand eight, there were some meaningful structural changes taking place in the market.

Um, I couldn't remotely have articulated it to you like this at the time. I I I I just sort of knew what worked and I found I find that that's kind of how my trades always work. I just sort of find what works. And then I can articulate five or ten years later why it worked, but it I just saw that it worked. And so what happened? There was Reg N MS that was passed in 2007. And so you have the hybrid market, this kind of stuff. And so it went from like before that.

Uh let's say you were just trading like some of the bigger stocks, like the big ETFs. Um there weren't very many ETFs at that time. But so maybe you trade, you know, you you trade the big ones. And at that time, as I recall, they all traded on nicey. Okay, which is weird. Um, none of them trade on Nicy anymore. But so I mean they were just a handful. And so they were on Nicy. And so what would happen was in you could just you could just put these bids out.

like meaningfully in the money. Um so, you know, it's trading FXI is trading at two hundred bucks a share. Uh this is before all the splits and stuff. So FXI is trading at two hundred bucks a share. um the bids like you know 200 by 250 cents and then you got your bids starting 50 cents outside of that, you know, and you put them on Nicy, and then they would get swept.

occasionally some guy comes in seeking liquidity and he just hits market on Nicy. And so you're you know you get the free money that fifteen, thirty, eighty cents in the money and then you immediately uh get out on ECNs. Okay. So you used to be able to do that. And then they put in Reg and MS and that kind of disappeared relatively quickly.

But then what they did was they go from that to uh they started converting everything in the middle of the friggin' um financial crisis, they started converting all the ETF. Which was really small at the time. And they started adding a bunch of ETFs. So that there are all these new ETFs to trade. Um, there's a lot of volatility, a lot of action. Everybody is way too busy doing so much other trading, um, trying to make sure that, you know

Lehman Brothers is the end of it and not whatever your favorite investment bank is. Everybody's so busy doing that that they hadn't really figured out the technical rules of ARCA yet. And so on ARCA, they had all these they all they had all these special rules for the closing auction, um the way that they did the closing auction. And so what I what I figured out, so is it was it was also really weird to get the feeds for it. Um so the the feeds would be like

Uh as I recall, I I couldn't I couldn't get it on any trading software. So I had to go to the actual ARCA website and hit refresh like a million times. I had to hit refresh on the ARCA website just over and over and over and over and over um in order to see any of the data. And so in the final minute at that time, the ARCA closing auction Would just it would just say like FXI trading at 200 bucks and it's gonna close at 203.

Uh and sure enough, it would do it. Um exactly as it said that it was gonna do. And I think it was because there weren't any other ARBs in at the time, right? Um and so so I I was doing that. And then by the end of the nine months my margins had started dropping a lot on that trade. And then a year after that first April I mean the the trade was it it was it was morphed at best, but the alpha had decayed just massively.

So what was that like for you? Because this is like the one trade which made you profitable. You know, where did that leave you now? What was the next step from there? And yeah, good good good question. You obviously think like a traitor. Um so then so then two thousand nine two thousand nine was a little weird for me. Uh it worked out Great though. I'm I'm grateful that I did it and all that. So I so the first several months of two thousand nine went very well.

Um, I knew so little about the markets and that kind of thing. Gosh, I just I I look back and I think about. in March of two thousand nine, why didn't we just buy all that there was of Las Vegas Sands and like Invesco and Pimco, the the closed in funds that were on the market and some of the bank preferred So I didn't do that, but I I did, I'd figured out a couple of new trades. And uh so I so I was up to like four or five total trades that were.

uh at that point, but I was still pretty reliant on the on the first trade. And so I think I was up like 150 or 200 grand a few months in to the to the year. And then uh There was this there was this kind of event in early in early June. It wasn't horrible, but it was the first time that I put in a really big all-in bet. And it was this it was this momentum trade that was based on some stuff that was happening with the banks. Um, and they were just the banks were going wild.

And I being an idiot. bought only one bank. I only bought JP Morgan. Just as I was about to exit and it was gonna have my biggest day ever, I think. Uh it's gonna be my first six six figure day. And uh and I was I was really excited. I was just having so much fun. Um and then JP Morgan somehow leaked its stress test. Uh so JP Morgan leaks its stress tests. I think I went from like up one twenty or one forty to like down twenty or something like that. So it was it was it was

A huge loss for me of kind of confidence and and a lot of other things. But it was like John, you gotta you gotta work on really building out your strategy portfolio. Um so I started doing that for the back half of that year. And so for the back half of that year, uh I made some money actual trading, but what I really focused on was I had found a couple of trades that would work, but it was like You could do forty or fifty different stocks at a time in these trades.

And so I figured out that some people that I knew were really, really fast on the keys and they were great with decision making and stuff like that. And I I wasn't that good at that. But I knew that computers were a lot faster than I was. So I got this guy. I don't even remember how I found him, but he was a software developer, and he knew just enough about how trading worked. uh that he could that he could write me some stuff. And so I wrote my first automated stuff that year.

Um so it was it was basically these sort of sophisticated stop loss orders. Uh there were there were this there were these kind of very specific um set of things that all had to kind of align for me to want in a to be in a position. And then the moment that they backed out, boom, I just wanted to get out immediately, right? Um and if I did that right, I was basically risking eight or twelve cents to make like

40 to 80 cents. So it worked out really, really well. But if I missed it, like if if I missed my outs, sometimes it could just turn against me so fast. And so so I got this guy to write some stuff for me. And predictably, he wrote exactly what I told him to do because he was good at his job and I was not good at my job. And so what I told him to do lost me a whole bunch of money.

Because there were all these little exception cases and things like that that I hadn't thought of. I mean it was I I had two of my biggest losers ever, uh, right after he put that Uh right right after we rolled out the first one. It was because I kept trusting it was gonna work the way that I wanted it to in my mind, but it it it didn't. Um and so I went so for a few months, for for a few months I'm sort of working on that, I'm fixing that out. And then after a few months, it does work.

Right. So now I've got it to where and so the the year after that I made I had a new best year ever. And then the year after that, I had a new best year ever best year ever. And the year after that, I had a new best year ever. And so at every, at every point, I'm kind of adding new strategies, new ideas. By the end of it, I had gone from these automated exits.

to then i added automated entries and then i said well let's just put them all together you know let's make this uh let's make fully automated stuff And so that's kind of that that's kind of the next few years.

Okay. Yeah, this is really interesting. I mean, this is one of the things which uh got me excited to speak with you is because, you know, your origins you started out as what you could call a discretionary click trader. Um, but you've sort of evolved into very much an automated trader, um, still trading uh intraday, et cetera. This trade which you found where there was kind of forty to fifty stocks you could be trading at uh various times throughout the day.

What was that trade? Is this the trade I'm just trying to put two and two together here'cause you sent me through a a few things um prior. Um, is this the trade which uh you would see kind of price bouncing from large orders in the depth? Yeah, yeah. So there was so one of the bigger parts of it was that uh at that time and they they don't do it nearly like they did. I mean it's not even it it's nothing similar. So at that time, really often in the order book.

There would be at whole numbers. So at, you know, 150 bucks or 25 bucks or whatever. There would be these just massive, massive orders for no reason. Um, I don't know, it was some mutual fund was taking profit or I I have no idea. But so you'd see randomly 400,000 shares there. And so you could try to get in front of it. And then if it works, or if if it doesn't work, then you you kind of get out.

um as that as that order is is uh getting eaten up. Right. So at at that time they would go slowly enough. There weren't all the HFTs and things like now. Um so you could actually do it by hand uh if you were good enough. Now those those big orders get they they get played on by HFTs a whole bunch now. And there there are some interesting trades you can do with the order book now. I mean obviously HFTs make a lot of money on order books. Yeah, so you would

You would find there were kind of some some specific elements to a setup that would make one really, really good. And then you could use these giant offers or sometimes bids as your backstop. Okay. That makes sense. And so the automated exits which you built into this strategy, so you would get in uh manually yourself. And then you had these automated exits to with the sophisticated stop loss orders to to get you out if um it didn't work.

Yeah, so I so I would get in every morning super early and I would I would scroll through literally every stock in the market that traded any meaningful amount of volume and I'm just looking for I'm looking for my setup in you know, 3,000 stocks every morning. It took forever, man. And then in whichever ones that I found any potential setups, I would just put a bunch of orders out and then get my automated stop set up and then hope and pray that they work that.

You know, as someone who started out as a hand trader or click trader, whatever you want to call it, you know, was it always the goal for you to um to automate your trading? Or was it just sort of something which uh came about as a almost of a need, uh almost as like a necessity.

Yeah, it it it just emerged out of evolve or die. You know, and I I wasn't My judgment wasn't as good as other guys. My uh my speed on the keys was slow. My uh all of that kind of stuff. But I I understood uh the power of technology and I I was able to do that at least.

Okay. So what would you say was the most difficult part in the beginning once you did start to automate some things? So sounds like the first step towards automation for you was automating your exits, which was mostly your your stop loss exits. uh in that scenario you gave just before. Um, you know, once you kind of moved on from that and started to do uh build this out a bit further and automate more things What proved to be rather difficult? I mean, you already sort of gave that example where

Um, you'd given a bunch of instructions to the developer and you'd done exactly what you asked for, but there were things which you hadn't thought of. I mean, was there anything like that which, you know, went on to to cost you money? Yeah, so very very frequently stuff like that would happen. There was one. Oh gosh. Yeah, I guess it's.

It it's really related. I was the the silly way of putting it is that I I am a conspiracy theory on zero hedge. What was it? It was September twenty thirteen. In September 2013, so I had this. This was actually a generic stop loss. This wasn't even um this wasn't even a particularly interesting stop loss. And so this stop loss. Got triggered. Um, I was in Chevron. I don't know. I wasn't even in that much. I was like, Mm, I think I was in like twenty five thousand shares or something.

So I'm in like 25,000 shares of Chevron in uh in September 2013, and this this stop loss gets hit. And so I had always been obsessed with risk. That's that's probably the one thing that I'm really good at in trading is I I I can take a loser, like I've got a real strong chin and uh and I'm really good about

being obsessed with like what's your edge, what's your risk reward? Um, what what are your potentials to have a a big tail loser? Um, how do you mitigate it? That kind of thing. And so I talked to the folks who were in the software. about what the risk limits were. And I'd set my risk limits really, really low. I mean it was like

I wanna say it was single digits percent of my account or something like that. And it was like whenever I lost that much, it was just supposed to completely shut off and I wasn't supposed to be able to trade. Um, unless I sort of called them up and said, let me trade. So um my stop loss gets hit and the stop loss Started acting wrong. I started acting wrong for sort of some complicated reasons, but the but the basic but the basic net effect of it was that.

uh I think I was short and so what it would do was it would cover the short okay but then it would mark it by the same amount again So I'm short. 25,000 shares and now not only did I cover that 25,000, but now I'm locked. Then it decided, oh, well, I'm in a new position, so I got to get out of that. So it sells the 25,000 shares and then short sells them again. So now I'm short 25,000 shares. So it does this recursively as fast as a computer could do it for 7.3 seconds.

During which time it lost a pretty significant amount of money. Um, I mean, it basically zeroed out my account. I think I had six grand left in my account or something like that. And it was I'm very grateful the the guy sitting next to me, I'm like, I'm like looking at my screen and my screen is just doing weird things and the guy next to me is just like, he's like, pull the plug. Uh and so we uh so I I managed I I managed to stop it only because my friend sitting next to me is just has the

what's the word, the the composure uh to tell me pull the plug, you know? And so so I did that. So that was that that was a big loss. I'd had a bunch of those. Um and the and the basic lesson that I figured out was if you're a trader, what you really want is you want to control every single thing that can possibly affect your PL negative. Right. So like I don't mind if

somebody else, if let's say I have some interface, right? Like I have a backtesting interface. If somebody else writes all of the code to print the chart on the backtesting interface. then I'm he can write that and I'll just barely test it and whatever, because that doesn't actually directly affect my PL.

Um, but if it is a risk limit that should have been triggered, right, the moment that I hit whatever percent of my uh of my account, the moment I had lost that, it should have stopped, but it did not. It didn't, it didn't work right, right? So I figured out all of that has to be a hundred percent under my control. I can't trust anyone else. To write that.

And so I think I think that became the most important part. And so from there I started figuring out how can I write everything for myself? How can I write my own system? the whole way down, uh anything I can't write for myself, like, you know, you can't create your own data feeds. Um you have to buy them from someone. Anything like that, how can I put in all these tests, all these redundancies to make sure that they're not messed up?

to make sure that if there ever is an error, we're, you know, we might lose some money, but we're not going to lose our account and we're certainly not going to lose the firm. Okay, so just for some clarification here, this stop loss which went rogue and was buying and selling just as fast as it could. What was that stop loss? That wasn't something that you had you had built that was a like an an order through your trading system or your your platform or

Yeah, yeah. So part of it I had had built for me and then part of it was just sort of built into the platform and I was just kind of stuck with that, right? And so the thing that I had built had mistakes in it. Uh, however, I thought you know, foolishly. I I thought I thought that when when we finally hit that mistake I have so many other things in place um that it'll be okay. And so it so it had a mistake and then there were mistakes there were mistakes at like nineteen different levels.

Okay. And uh and they they they all kind of It was sort of like I think I've counted it out and yeah, it it was in the teens, the number of things that all had to go wrong simultaneously. Um if any one of them had not failed, then uh Then at some point the the risk would have been stopped and I would have just lost, you know, 20 grand or 70 grand or whatever. Okay. So you just had to ultimately pull the plug and that stopped it at least.

Yeah. Yeah. Yeah. The guy sitting next to me, he's he's great. I love that guy. Yeah, well thankfully he was there. Otherwise who knows what might have happened. Have you ever watched a stock explode and thought, if only I had the capital, or sat on the sidelines because your account balance felt too small to matter? Good news.

With Trade the Pool's limited risk platform, you don't need millions or even thousands to start trading the US stock market. Bypass the PDT and tap into over twelve thousand US listed equities. From penny stocks to big caps, ETFs, even the newest IPOs, and short anything you like, with zero locate or hard to borrow fees.

Start your evaluation, get funded with up to$200,000 in buying power so you can go big without risking your own savings. And now you can also have unlimited time to reach the profit target. It's a game changer. Not ready to trade yet? Trade the pool offers a free demo and educational resources. Practice on live data, master the platform, and build confidence risk-free before you even pay a cent. Click the link in the show notes to start trading with Trade the Pools Capital.

I mean, where do we go from here? I I guess one of the things I'm really curious about is well maybe this might be a good question to ask you, is is what was the first strategy which you built which was fully automated. So from entries to exits, you know, all the components of the strategy. Yeah, what was that strategy which you first built out, which was a fully automated strategy? And and also how long did that take? Yeah, so that's a so that's that is a

It's a good question, however, it's not one I can answer. And the reason is, I mean, the first dozen or so that I made were all failures. Um, so I thought that they were gonna work and then they did that they were just sort of all failures and I don't have So I mean I was running actually dozens simultaneously by the time I ever actually had any that ran reliably.

Um so I couldn't I couldn't kind of go back chronologically and pinpoint which one sort of started working first. It was definitely the the curve definitely worked such that I was making more money with this kind of gray box trading um where I would either enter or exit using a script uh for a very very very long time.

And I was taking all of those profits and using those to build the automated system because I was just so confident that it could, that it could do so much if it ever started working. So this might be a better question to ask you is with the experience you have now, you've been doing this for many years, uh the the automate automated side of things, why do you think some of those earlier strategies turned out to be failures?

Okay, so trading or any kind of strategic behavior really is it's a it's a complex system, right? And you have to get and all of the things multiply together. So it's like You know, instead of it being if you have ten different parts that all have to work together, it's not like two plus two plus two plus two plus two. It's like two times two times two times two times two. And so if If any of the elements was doing badly

uh then it it it could cripple the whole thing. Um, or if any of the elements I hadn't built out. So what I really had to do was I had to just kind of fail. at every different spot that I could, fix that, keep shoring it up, keeping it keep improving it, and then finally sort of all the pieces came together um and worked together. I still can't. I still can't think of what would have been

kind of the first one that that really worked. But I know so I so I I know that at some point I I started I started rolling out a lot. And the thing that I really needed And this is something I'm really bad at. What I needed was I needed to build a toolkit. Okay. Um I I I was I was fortunate. I kind of stumbled into somebody else's uh sort of primitive toolkit. And now at this point we built out, I think it's a pretty cool toolkit.

But so I needed I needed a suite of more sophisticated backtesting stuff. So with with the back testing stuff, I could do a bunch of back tests. Um I could look at him and say, oh, what actually what are actually the good signals for a winner? What are actually these weird tail risk cases that my code does not get me out of? what are those kind of things, right? So the back testing system I could do that. I need a good uh back testing visualization. The other thing that I use a lot now is uh

Well and and then I needed sort of scoring systems. So I I needed to figure out, oh, okay, it turns out that most trades don't just rely on uh is this one value high or not? it relies on, well, do all of these things sort of come together at at at the right time? Um and then if so, How do you get out when you have a loser? How do you get out when you have a winner? How do you add in those sorts of things? So I've added in I've got a bunch of back testing stuff and that is

so helpful. The the big next step I took, and this was really, really helpful. This this was just huge, was machine learning.

So all of the stuff that I had written, that that was the problem of it, was that I had these really complicated, uh contingency plans in my mind that was like, well, you need this and this and this together. And then if this ever happens, here's how you deal with that. If this happens, here's how you deal with that. And so if I would have actually written down all of those different pieces.

I mean it would have taken me more than a year of my own time just to get one out, just to get the first out. And so so I I couldn't really do that. I needed uh of a programming language that would be simpler for me to be able to write in.

Um and I needed I needed to be able to can to get to really rich decision making uh without having to write it all out kind of in really specific detail. And so what we did, I've got um we had we've we've built up this toolkit over the years, but uh we put in a bunch of machine learning and the way that it'll work is Uh

I tell it all the things that I know will be significant for the trade. Right? It's like this and this and this. They'll be significant. And then the way that they interact with each other, that's going to be significant. And then all of this other stuff. And then we run giant backtests, we get giant data sets. And then we put those data sets through a bunch of different machine learning algorithms. Um and they have all this cool stuff where they do these. We we hired um

We hired some folks from the the local university. They were PhD candidates. Um PhD candidates. will work for much cheaper um than than PhDs, professors, that kind of thing. And they also love the fact that I have these data sets that no one else has access to, right? I have real financial data sets. Um, and that's very, very hard for them to get access to. So they got access to my financial data sets and they got some money. Um, and then they built out all this cool math stuff. It's like

I I I would know I would know the trader's perspective of here's what I want the machine learning to do. Here's how we can direct it. Here are ways that it's probably gonna go wrong. Um it's gonna be l looking for sort of straight up normal distributions here. And the distributions are not gonna be anything like normal. They're gonna be wacky financial distributions.

And so they built they built all this cool stuff. So it'll it'll go, it'll examine the data, then it'll figure out how predictive is each kind of different element of this data. It will build us a bunch of scores. That scoring system will work. But a lot of different algorithms. And the different algorithms will also have uh different kind of settings and hyperparameters inside of them. So then we'll have we'll have a bunch of scores.

Uh and then it it builds out this thing where you functionally get out-of-sample results, uh, which I think is super important. I've seen way too many people with these back tests that look perfect, you know, oh my sharp ratio is ten or whatever. But it it's just not gonna work because it's it's not truly out of sample. And so we do all that and then we say, oh, okay, well

what actually makes for a good score in this. And then we do the good scores and uh and then we plug that into the uh into the strategy logic. And then it's gonna use that to make its decisions when to get in, when to get out, when to add. That kind of thing.

I don't know if any any of that made sense. It was kind of complicated. No, it does make sense. And um this machine learning part i is quite interesting also. So I have naturally a few questions around that. Um this might not be the right terminology. I'm kinda blanking on what exactly you'd call it, but Are there um... Which machine learning algos have been or are you most using? So there's like uh, you know, r uh random forest is one. Um I can't think of any others off the top of my head, but

Yeah. Yeah, so so we've looked through a whole bunch and we've tried a whole bunch. And what I've figured out Is that based on my actual stuff? Okay, so this is not uh other people probably have they have datas and stuff that works differently. Like I know that so I know that Google can use a neural net to do image recognition. Okay. I can't use a neural net. I think that my data sets are just not nearly big enough.

But the things the the things that have worked for me are number one, the absolute most basic, simple thing that there is, which is called ordinary least squares, OLS. It is just, it is Excalibur. I had a um, and then the other is uh versions of modified XG boot. So XGBoost is at as far as I'm aware, it's it's the most hailed uh machine learning algo. And then OLS is kind of It's it it's amazing. And anybody anybody who's in school will think, well, an undergrad at least.

Um they're like, oh, OLS is boring, it's okay boomer kind of thing. Um but then you actually use the thing, and it's just wow. And and only if you use it right. Okay. It I I I couldn't get it to do anything. Okay. I couldn't get it to do anything at all that was any good. And then um

These guys from the university came over and they had all this really cool math stuff to do. A small amount of that was my ideas. Most of it was like, these are the best practices in the industry. This is the way you should do it. Um, and then I said, Oh, we should modify it in this tiny way. And then a couple of them, they came they came up with themselves. So they were just they just did such a good job. And uh and yeah, so once once you actually use it right.

And I have yet to find somebody who can do the other stuff uh that really, really, really beats it. Um so I use I use some clustering stuff. Uh so k means is the best known. And what that means is more or less. So imagine you said that there are mean reversion traders and momentum traders. If you have if you have one group of mean reversion traders and one rook group of momentum traders,

A lot of their behavior, you should just split them out. And you should say, oh, okay, well, whatever a mean reversion trader does, it's probably like They make money on most of their trades and then occasionally they take big losses, for instance. Momentum traders, right, they they tend to be the other way. Um they tend to have some giant winners and then a lot of losers.

And so if your if your algorithm were just trying to look at all traders together, it would say, oh, okay, well there's not really a pattern as to whether you have big losers, big winners, frequent winners, frequent losers, whatever it is.

But then the moment you know, well, it's really in two clusters, it's momentum and mean reversion. The moment that you know that and you split them up, you say, oh, okay, people in this mean reversion cluster, their stuff follows this way, and the the momentum folks follow this. So the the K-means has been effective, OLS is the best, and then XGBoost has been really great for us because it gets

We we'll use it for kind of a second opinion kind of thing. Um so the OLS may say like, oh yeah, this thing is It's a 72 out of a hundred. So, you know, it's it's pretty good, but it's not great. Or it's a ninety ninety-nine out of a hundred. It's awesome, you know. Um and but then occasionally uh we train several XG boosts in in different ways. But so occasionally one of them, one of the XG boosts will come in and say, yeah, it is really awesome.

But if you look at this kind of thing, it's also a lot riskier than you think. There's like a 35% chance that you're gonna suffer a gigantic loss. And so yeah, absolutely, take the thing, but size your position smaller than you would normally think to, because you don't wanna, you know, you you don't wanna take this gigantic loser that's just gonna mess up your psychology. Once you take these strategies which have been optimized with the use of machine learning.

When they are actually trading, do you completely understand why they are making the decisions they are making? So no, I I definitely do not completely understand. However, to me it's very important that I have a pretty good grasp of it. So I'll I'll go into the algorithms. Or the algorithm the algorithms that really get to make a lot of decisions. There there are some that can say no, but they can't say yes.

If you know what I mean. Like there are some that are just purely risk mitigation and they're so complicated. I don't understand them in the end. Um, but with those. that doesn't affect that can't create tail losers. So when when I say like control your PNL totally, um, it it's okay with me if it's like I lost some opportunity.

Because I didn't I didn't fully understand that. But all the ones that can just say yes and that's the end of the story. I've looked through every feature, I've looked through the through the distributions of everything, um, I've looked through the history of it, all of that. And I've I I couldn't I couldn't run the calculation myself, but I'm absolutely confident that yeah, this this makes sense.

Okay. So when you started out trading way back in two thousand seven, I think you started to get into the automated side of it. around about two thousand eleven, you know, and now you're you've implemented uh machine learning techniques and all that sort of thing. What's been some of the the most helpful resources to kind of get uh familiarize yourself with machine learning?'Cause obviously it's a pretty steep learning curve uh to begin with, I'd imagine.

Like how did you learn about it? How did you get familiar with it and comfortable to start putting it into use with your own trade and So a friend of mine is gosh, he's really smart. So he he recommended a couple of little books to me. And well, they were they were little to him. He he was so experienced and so good at everything. uh that I that I looked at those and I was like, wow, this thing is eleven hundred pages and it expects that you know.

all kinds of industry jargon and this kind of stuff. And so instead I went to the Amazon recommendations based on this book. Okay. And so I went to those and I looked through those. And then I went sort of three deep on it. So we had the book that he recommended to me, and then there was. Uh, the book that Amazon recommended based on his recommendation. And then another recommendation based on that recommendation. And that one was some kind of like machine learning for dummies. Okay.

Uh and I read that thing and I read it in like 20 minutes. It was like a hundred something pages, and I just blazed through it so fast. Um, I just I I loved it. It was it was like I I read it and I was like, yes, this finally makes sense. And so I I read that one. Um and then there was another one that was called uh that I moved to after that. It was called Data mining.

I don't know, top top ten algorithms for data mining, I think it was called. Um, it was a little it was a little hardback book and that was they didn't mean data mining in the C F A way. It's uh Uh that was a a term that they gave to some versions of machine learning. I don't know if they still do, but um or at least some stuff that's related to machine learning. I don't know all the details.

But so so that book it it had ten algorithms in it. And that includes all of the algorithms that I use now existed in some form in that book. Um, I haven't been able to use most of the algorithms in that book, but all of the ones that I use were in that book in some form. Um and so I read that. And based on the first book, I had enough of the vocabulary that I was able to read this book and say, oh, okay, yeah, we could, we could kind of figure this out.

So I figured those out. And then absolutely the the most important thing was just hiring people who actually know their stuff. Right. So I I would I I would hire who whoever it was. I mean, I think at this point I've got 17 different people who are doing stuff with us right now. And it's just when whenever you find somebody who truly loves something, and like I like machine learning, I don't love it.

But you find somebody who loves machine learning or algorithms or some specific type of trade, it doesn't matter what. I mean a a conversation with them is just exhilarating. It's it's just

fantastic. Um I'm I'm not nearly the interviewer you are, Erin, but um I I can I I love to hear anyone talk about anything that they love. And so for me, that's by far the number one uh source is I had I had to develop a little bit of vocabulary and then I started talking to people who actually know what they're talking about. Right.

I know some people are gonna be listening to this and kinda curious as to what some of those books you read were. So I mean that was my motivation for asking, but maybe uh We'll we'll link up after this and um you know if you can think of what any of those titles were, then we'll just stick'em in the show notes.

If we look at your current situation today, as I understand it, you are running a portfolio of around about eighty automated models. So your day-to-day is kind of trading these models, overseeing managing these models. I guess one of the things I I sort of want to get to is Like what are these types of strategies? How do you categorize them? Yeah. Not exactly the parameters, but How would you categorize these strategies? Because You've come from

a click trading background, right? So I I I feel like you probably think about strategies quite differently to how someone who might be uh just a pure quant who might be using Uh or someone who might, you know, if you think about like a a systematic trend follower, like their strategies are very kind of I I feel like the strategies you use on an intraday level uh are quite different and

It d is that making sense? Like I'm I'm just interested in sort of how you categorize the strategies which you run today. Yeah, yeah, yeah. So so I run several I think of them as families of strategies. Um, so I I run several different families of strategies uh and they're pretty much all gonna be based on some anomaly um that I or someone else has found. Uh, and then we're we're we're sort of working on trading it together. Right. So you'll get

Uh let's see what what would all the families be? There are some that are mean reversion. Okay, and so those are gonna be I have no idea why or how they work, if I'm being honest with you. Um, I mean I I like I like like I understand. I understand the concept of why they work. What I don't understand is why are they not already crowded out by like big market making firms and HFTs and stuff. Like I don't

I I don't get that, but somehow somehow they do. So so to me, they're relatively straightforward. It's like the things that you would expect in a in various little mean reversion strategies. Can I just rephrase the question a little bit? Have you automated things that like good click traders would go after?

You know, like like the things that you would trade by hand, are they the things which you've automated? Or are these more things which I don't know if this has just confused the matter or not, but Oh, yeah. Like there are guys, that's just a very simple example, who will short uh stocks which go parabolic. And you could categorize those as mean reversion types of strategies. But, you know, have you ridden algos to short parabolics, for example? Or are your strategies a lot more niche than that?

No, yes, yes, absolutely. Yeah, I've I've written algos to short parabolics. Okay. Or to buy parabolics. You know, both ways. Yeah, so I mean I've got mean reversion families. I've got momentum families. I've got stuff that trades based on earnings or conference calls. Right. I've got stuff that's it's like you know that news is gonna come out in something.

And you have some expectation of what's going to happen when that news comes out. And then the moment the news comes out, it actually does start to go up, you know, and you knew it was going to go up or down. And so it just buys. It's read the news. It saw one of the keywords that it was supposed to. And so then, you know, it buys on this breakout of whatever sort that it was.

There are what else? I mean there there's there's a bunch of weird anomalies. Like there are things that are like at this time of year uh for whatever reason investment bank all put out a bunch of ratings. And so weird stuff happens because of ratings. Or at this time of year, you know, all the big the CFOs decide that they're gonna put money into

looking into acquisition targets. And so people start to get really excited about potential acquisition targets. Um and there starts to be some momentum in them. And so we're going to trade on that anomaly or whatever it is like that.

Just before we move off the subject of automation and we'll we'll wrap this up shortly, I just there's just a couple of other things I'd be curious to ask you about. You know, for someone who's currently uh a click trader, uh discretionary trader. And they're doing well, they're making money, they're they've reached a point of consistent profitability and they see that they they're attracted to the idea of implementing some sort of automation.

How would you encourage them to take the first steps towards that? Like where would be a good starting point? Okay, so I think really what you want to do is you want to find someone who has a good platform that's already built out for you. Um, if I'm being just giving you the advice that I think is the best. It may not be the easiest.

'cause it it may be expensive to get access to somebody's stuff. But I mean I I can tell you absolutely from personal experience, if you can if if I had had a full suite of tools twelve years ago, uh when I first started building stuff, I would have been really blowing and going within eighteen months max. And

I mean, my my PNL would have gone parabolic. It it would have just been amazing. And so instead, um I built up all these different things kind of slowly over time. Um and they Uh as a result I had these kind of parabolas that were punctuated by giant crashes, right? As as I would make some some giant mistake. Uh and so so at this point we have we have kind of the full the full suite.

Uh and so that so that feels a lot better. Uh but I I would strongly recommend if you can uh to find somebody who's who's got it fully built together. And so so you know, may maybe there's some third party out there. I I don't know what it is. A second option is to partner with someone who knows what they're doing. I mean I do this is probably not that helpful to most of y'all, but the uh like every every summer I bring in a bunch of interns.

And so I I bring in a bunch of interns and the idea is y'all get exposure to this world, y'all get to see what I do, um, what we do, how how we do it, how we make money, uh, and then maybe a few of you will really click here. And then whoever does I mean we're we're gonna get it's like

All of this stuff is already built up, right? So so one of you who's already a really successful hand trader may just be saying, Yeah, John, an internship isn't what I want. And that's That's great, but maybe you can find someone else who's already built the stuff up who could similarly partner with you, right?

It may not be as an intern, it may be as a true partner. But yeah, I'd I I would strongly recommend it it's such a complex system. There are so many things that multiply rather than add uh that you should really you should really try to get something that has the full ecosystem where it's a third party uh or it's a partner. Okay.

Just a couple quotes of yours which I'd like to pick up on, uh and this is just from the the few things you sent me prior. Uh One line which I picked out, which I thought was really interesting, you might have said a little earlier too. I should obsess over what my edge is. So I guess the question I'd like to bounce off that is

How should a trader think about and how can a trader truly maximize and optimize their edge to the extreme? Cause I think this is something which you have done really well. That's a good question, man. And I I wish I had even I I wish I had much better answers than I do. Um I I can tell you that that really helped. Uh

turned my trading career around, I I remember I I kept hearing this voice in the back of my head that there was this guy that there were these two guys who were talking. Um and one of the guys was like, oh yeah, I'm putting I'm putting on this trade, I'm gonna do this. And so this other guy is like 哇 And the other guy tries to explain it to him and he says, Well, why would you do that? What's your edge? He said, What's your edge? And so I'm always, I'm always defining my edge.

Whenever, wherever, however I can. And then the idea is you just want to make the absolute best that you can off of that. Um so the way that I would say that your edge evolves. Okay. So first of all, like I said, I've everything that I do, it's just, I noticed that it worked. And I couldn't necessarily articulate the whys of why or how it worked. Um but I definitely understood this happens, this distorts the market in the following way. And so you can either

kind of get in before it has distorted the market, you can kind of predict that it's about to distort the market. Or you can get in on the back end once it's kind of blown up the market. Um, and you can you can get in there to pick up the pieces, right? So I see this pattern that'll happen and I'm gonna follow it. Uh I mean to to give an obvious example and here's okay, I'm actually gonna give a strategy to y'all. This is just the easiest. I've done this like four different times.

Whenever Bitcoin finally makes a move. Then Ethereum makes a move after it. Then kind of the third tier coins make a move, and then the altcoins make a move after it. So if you look in the past, whatever it is, three months. Um I was sitting in Bitcoin. Bitcoin makes it its move. It goes from 10 to, I think it went up to 30 something. And then it was very easy. You could just go buy Ethereum.

So you can buy Ethereum, I think it was at like 300 bucks at the time. And then you buy your ether at 300 bucks, and then the thing goes up to 1500, 2000, that kind of thing, right? Well, sure enough, after ether makes its four, five, six X move. Then all of the sort of next tier coins start doing this crazy stuff. Um, I mean, the most obvious one being Dogecoin, but I think that had its own catalyst.

I mean the point of telling you this thing about the crypto, like that's just obviously this is not investment advice. This is just a pattern that I have noticed. And if you go look in the past, you can see that the pattern has definitely happened and you could have definitely made money on it, right?

And so so I don't have to fully understand that. I can understand it in kind of increasing depth, but I can keep making money on it. Okay, so so with the edge, here's here's the thing. Um it was a thing Th this this is a thing that that I thought of. I I think that's a lot. Um A lot of people say, well, I couldn't be a trader, I couldn't be an entrepreneur, I couldn't be whatever, because that's just I I don't think that big.

Um it it's just it's too big, that's too much, that's too whatever. Um I I I think that it's the exact opposite. I think that actually the reason that very few people make it as a trader or as an entrepreneur is whatever. is that they don't think small enough. So in this case you have to be willing to say My edge is this tiny stupid thing. In my case, it was this one tiny little um anomaly that I picked up and barely fully understood and only lasted a few months.

13 years ago and I had this one little edge and it worked. Well, and so what did I do? Instead of saying, well, I have to be, I have to be better than that. I can't just I I have to be better than the market. I have to be better than stocks. I have to be whatever. No. I just took this one tiny edge that I had and I worked from there. And so what you want to do with your edge is you keep you just keep growing it 1% every day if you can.

Um, I know so many people that it's like, oh, how'd how'd you do last year? I made 70 grand. Well, what if you just put something automatic in there that increased your size by one percent every day? Okay, and then they do that. How did next year go, wow, I made 300 grand. How did next year go? Wow, I made a million bucks. You know. Um and it it can happen So fast.

And so with with your edge, I mean I I think the thing is you're constantly as long as you're constantly trading it, you're seeing all these different elements and cases and specifics, and that helps you. start to sort of triangulate, what exactly is it that's really making this thing tick? Is it like this trade goes in cycles. You know, I mean like the the the the hot stuff. The hot stuff lately has been these meme stonks and the spacks, right? So those have been those have been just so hot.

And those things have been completely cyclical, right? So the the meme stonks it's like it's like they go in they go in these little waves where they're sort of only work on the momentum side for a little while, and then they kind of turn for a little while, and then they only work on the mean reserve reversion side for a while. So, you know, may maybe what you figure out is, yeah, I I need to figure out what part of the cycle I'm at.

Um, so now you've expanded your edge. You had one edge, which was maybe you knew the momentum side of it. But then now you figure out, oh, okay, I need to know where I am in the cycle, and then I need to be aggressive or conservative based on that. So then that's this kind of next layer that you've added to your edge. And so if you can add those little layers, one percent a day.

you know, twenty percent a month, whatever it is, um, and you keep adding them, they compound very quickly. And so your edge goes from being this this tiny thing that everyone would make fun of because you just found this. this tiny thing that barely even matters in the market to all of a sudden you can make a significant amount of money on it.

I don't know if that's helpful. Is that helpful? I think it is helpful. I mean, that's exactly how you became profitable in the first place. And I mean, who knows where you'd be today if you hadn't have kind of discovered something like that.

'Cause it sounds like you were almost you know, in those that first year you were almost at a point where you'd run out of money and were gonna go look at other options, but you'd found this little edge which you were able to continually um, you know, build upon and increase your size and enable you to really grow your account and to, you know, start building out into other edges and maximizing those.

Gosh. Yeah, I mean when when you put it that way I'm just I'm I'm so incredibly grateful for everything that kind of come together at different points. Grateful for all the people that I met. I'm I'm ashamed that I made as little of it as I as I could have. But I'm I'm just I'm so excited, man. It's like I wake up in the morning, like I'll wake up at three o'clock in the morning. Uh just hyped. I'm just

Trading trading stocks is the best, most wonderful thing. I can't believe I'm so privileged that I get to do it. I I just I love it, Aaron. I feel ya. I feel ya. Yeah. Some d some days you've got to be um yeah, you feel very fortunate to be able to do what you do. So On that note, John, let's uh let's call this a wrap. Uh if someone is interested in kind of connecting with you somehow or following you online, I know you don't have a great online presence, but I think you are on Twitter.

I don't know, maybe LinkedIn. Do you want to share anything uh like that? Yeah, sure. So I mean, technically I have a Twitter, but it's it's mostly for myself. Um I I think the main place I think I'm on LinkedIn uh somewhere. They sent me enough. emails that was like your friend wants you to to join, but I I I have a I have an account on Quora uh that I used to post on a decent bit. Yeah, you you could reach out to me there. You could find me on LinkedIn um and send me a message.

So I'll put links to all those three in the show notes and uh those show notes will be at chatwithraders dot com slash two one three, episode two hundred and thirteen. Uh John Uh, needless to say, really appreciate you coming on the podcast and sharing what you have. Uh, it's been really nice to speak with you. Yeah, Aaron, thank you so much. Thanks for what you do. And man, go out in the markets and crush them. All right, we'll talk soon. Thank you.

You've reached the end of this episode of Chat with Traders, but rest assured there are more. We'd love it if you'd leave our-