Amibroker: The Adventure Continues

Topics include:

• Batteries Data Not Included - the best way to get data into Amibroker
• The code that adds custom columns in Amibroker is resource-intensive
• Quirks with databases in Amibroker
• Your computer hardware matters with Amibroker (get more RAM)

Dave's hardware specs for his backtesting computer.

---

Before you go!

The next time you find yourself in a drawdown, I want you to stop what you're doing and head on over to betterbacktesting.com to sign up for my free backtesting email course.

I hope to see you there!

Creators and Guests

Host

Dave Mabe

Host

Michael Nauss

What is Line Your Own Pockets?

A weekly show for systematic traders who want to make more money from their trading strategies.

Michael: 00:00

Hello, everyone. Welcome back to another episode of Line Your Own Pockets going down the road of Amity Broker with me. I wonder we should do, like, a gray hair count the the more the more I go down Amibroker. But this wasn't I don't think we were planning on making this a full on series. But like we talk about, we always, you know, read the comments and everything, and it was like overwhelmingly, I'm going through the same thing and yeah, can you talk more about this?

Michael: 00:28

And a lot of that was going on. So you guys want something and we'll deliver it. So yeah, we're gonna talk a little bit more about how things are going today.

Dave: 00:39

Yeah. There was a lot of feedback on my newsletter from people that were like, oh, this is perfect timing. I love this. Like, keep going. Keep doing this.

Dave: 00:49

This is really great. So, yeah, let's keep going. What what happened this week, Michael, with your Amber broker

Michael: 00:57

Oh, almost journey. Launch launched a couple computers out Windows and and things like that because what I was getting was so switch from Polygon to IQ Feed. So then it was the whole I think that was the most discouraging bit, which is why I really encourage people that are doing this to just go with IQ Feed just right off the bat, because it was like, you know, I got a little bit ahead in a race, and then I had to go back to the starting line, and then I had to start it all over again. And that was, you know, really, really discouraging. I've gotten some data in there and I need to figure out more, and that's something that we'll chat a little bit about.

Michael: 01:44

But I was able to get, you know, the plugins downloaded and set up, And after we recorded the podcast last time, I sat down with Dave for a second on getting the Symbolist in, and then I was able to backfill about a year's worth of data with a huge struggle, which I'll talk about how I was able to fix that in in after.

Dave: 02:06

Yeah. So with Polygon, I think the way they have it set up the way they make it so they make the entire day available of minibars for all symbols, and that's one file.

Michael: 02:22

Mhmm.

Dave: 02:22

I I and the way it's set up to do the import, it's it takes a long time to import into Ambroker that way because it has to switch between every like, the little database for every symbol to import the bars each that way. So I think you could make it a lot more efficient by reordering the data in the file and making one big file so that it goes one symbol at a time. That's a that I'm kinda curious if that would work and make it significantly faster. I think it would, but I'm I'm curious now. I'm kinda curious to try it, but kind of a rabbit hole that I wanna avoid.

Dave: 03:08

Maybe somebody has explored that and can give me some feedback on that.

Michael: 03:13

Well, and and I'm kicking myself because it's what I think Polygon or Massive or whatever they're trying to, it was like $30 a month and IQ feeds like $100 a month. So, you know, the amount of time that I wasted, it was years of paying this different in fees. So it's one of those things, and I do it all the time, where I'll buy, there'll be two things online, I'll buy the slightly cheaper one, then I'll break it, and then just and then have to buy another one. So it's just another one of those lessons of just get the best stuff off the bat. And, you know, obviously, a $100, if it makes or break, it's what you're trying to do with trading, then then you're just not there yet.

Michael: 03:56

Right? You need to, you know, save more money or, or start with a free product or, you know, do there's a lot of things you can do. So just huge kicking myself in the butt for that. Yeah. Part of the problem, there was like all these little snags in the way, but like we talked about, is that IQ feed doesn't just there's not a just fill everything.

Michael: 04:17

There's a multi step process. First, you have to get a list of things and say, these are the symbols I need data for, and then it goes out and grabs it. So that was a huge bit, but you've got some great stuff on your website that helps with that and and all of that.

Dave: 04:32

And then So I think the one important thing I wanna point out here is there's different lists of symbols for US equities. You might think, Oh, let me just go get the list. There's one list. Well, there's not. There's all sorts of different lists.

Dave: 04:51

There's all sorts of symbology around different symbols, which ones to include. Do you include OTCs? Do you include warrants? Do you include ETFs? Do you include ADRs?

Dave: 05:08

The best way to get the list is to use the one that your data feed provides. And each data feed will provide a list. I mean, they're coming up with the Symbolist of the feed that they're sending you. So they're gonna have a list. Very conveniently, DTN, which is the company that makes IQ Feed, provides a list that's updated daily.

Dave: 05:34

And you might think, oh, okay, well, this is a small text file. Let me just download it. It's like several 100 gigabytes large because it's got every symbol, every option. So it's actually a massive file and, well, maybe I shouldn't use the word massive since polygons changed their name to that, but a very large file. But I've got a script that will download that file, parse it in exactly the right way that's very convenient to import into in a very convenient way.

Dave: 06:11

So there's a script on my site for doing that. And you're really touching on something that seems like it's a downside of Amibroker, but it's really an upside because it's very flexible. You can import any data you want. Since they're not providing the data, then it's very flexible. You can provide you can do anything you want with the program.

Dave: 06:41

So there's a lot of flexibility there, but there's a little bit of a DIY element that you're not used to.

Michael: 06:47

Well, and maybe so maybe I shouldn't blame Amity broker for this. Maybe it's not real tests that I'm spoiled by. Maybe it's Norgate. Right? Because what Norgate does a really good job of is they parse all of the data for, you know, survivorship bias and all of that, and it's only equities and they strip out the options.

Michael: 07:05

And yeah, so maybe it's not the difference between Amibroker and RealTest this time, it's just that the Norgate plugin is so, so easy because I don't need to grab, I can create my own symbol lists if I want to. But by default, it just pulls in absolutely everything and goes, here you go. But that was a big thing. As soon as I got that set up, now things make a little bit more sense. There was some funny things where, you know, the all symbols, something that Dave showed me, because there's an option to apply and download all symbols.

Michael: 07:41

It doesn't mean all symbols, it means whatever you have selected. So like little things like this that there's just no way unless you read everything, and as soon as you figure it out, it's like, okay, it makes perfect sense. It's just everything that I've selected. But as soon as I got that done, the problem was I would do this big import, and I'd get that everything would look good, and then I would run a backtest on a very simple product or very simple thing, and at the end of every single backtest, the program would just crash. It would just go.

Michael: 08:08

And, you know, I took the error codes, and I put them into ChatGPT, and I went back and forth, and I put it in a whole bunch of places. And, you know, it was just wrong. And it was just it was talking about all kinds of nonsense that that didn't actually end up happening. Figured it out. So for the listeners there, when you set up a database, there's an option for you have to program when the start time and end time of the day is for you.

Michael: 08:33

If you have that wrong, and you're doing something, I think this was just purely coincidental that we were talking about the first thing I'm testing is an opening range break. The fact that I live an hour off center to the market. Right? So the market opens at 10:30 and closes at five. That was just destroying everything.

Michael: 08:50

So I think what was happening is the end of day close, it was just breaking somehow, and then the whole program would just crash. And as soon as I went and deleted everything and rebuilt the database with that time change, things started to work for now. First, for the one year database. We'll talk about the five year database soon. That's still giving me problems.

Dave: 09:13

Yeah. So, you know, one of the many things I like about Amabroker is the fact that you can add custom columns into your backtest. But when you do that, and you run a backtest, you get this progress bar that's really accurate. It's a multi threaded backtest. So you can see the number of threads for the number of cores you have on your machine.

Dave: 09:37

It's pretty accurate for how long it's gonna take to finish, but it's gonna finish the first phase of the backtest, which is like the main part of the backtest. There's a phase at the end where it's adding the custom columns. And during that part, it's going to look and feel like the program is completely locked up. It looks like it's not responding. And even you said to me, Dang, something's wrong with this.

Dave: 10:05

It's not working. That last phase is single threaded. It's not multi threaded. So I've actually talked to Tomas about this exact situation. He's going to try to put something in there that gives the user a little bit more feedback about what's going on.

Dave: 10:23

But if you have a lot of trades in there, it's going to take a long time And it's just going to feel like the program is not doing anything, but it is actually churning and doing this. And the funny thing about this is there were earlier versions of Ameren Broker where this phase took 20 times as long as it does now, if you can imagine that. So there were times where I would start a backtest on Friday after the close, let it run, it would get to the 100%, which would be fairly quickly. And in that last phase where it's adding the custom columns, it would take like all weekend. So literally over twenty four hours to go through that.

Dave: 11:07

It would eventually add them, but it would just take a while. And you have to be patient and understand what's going on. Otherwise, your first reaction is like, wow, this thing's a piece of crap.

Michael: 11:20

Well, and you know, a lot of this will just be growing pains, right? As soon as I have a process down of right, making sure that everything is good and testing it against a small list and and all of that to get started. That's that'll be fine. But yeah, this wasn't this wasn't that. This wasn't taking a while.

Michael: 11:38

This was full on, like, error screen. The only button that's highlighted is exit program, like the whole whole thing crunched. Now, it could have been, and this is what I haven't had a chance to try the five year database again, because like you mentioned, that takes a long time, so I have to wait till some other work is done, which actually leads me to a question I'm gonna ask you a little bit later. But the other thing is I had Norgate installed on because I I use Norgate for real test, and I think somehow it snuck its way into Amibroker. So I wonder if and I was able to go through.

Michael: 12:15

I found I asked Chatchi Dee, it said, you know, take these DLL files and just rename them so that it you know, the plug in, it like hides the plug in, which Dave Lasswell, that's probably the wrong thing. Yeah. But anyway, I went I went through and I just I just said you can't see Norgate anymore. And that may have been part of it as well. But, yeah, getting the data into the system, that was the the the hardest part because it took like literally it's been a week.

Michael: 12:43

So we record these once a week, and it's been a week, and I've just been trying to, and I've only got a one year dataset, which is okay for for testing whether or not the strategy is working correctly and everything. But I want to get that, you know, five years plus for optimization. And, know, I don't want to optimize off a small dataset, I want to optimize off a larger one. But yeah, just, you know, if you're getting into this, just understand that the, you know, the data is going to be for me anyway, that was the biggest hurdle. I hope I figured it out now, but, yeah, that was that was a huge, huge hurdle to get through.

Dave: 13:18

Yeah. Well, one thing that it doesn't register with some people is creating a database in Amber Broker is super easy and you should do it. Like, you should not hesitate to create a new one. That's no big deal. Like, I think a lot of people think, okay, I've created my database.

Dave: 13:37

I gotta make this one work. Like it's

Michael: 13:39

I've I've deleted. Well, just so you guys so I deleted I kept thinking it was the database that was the problem. So I kept going in and just deleting the whole database and, you know, deleting like a 100 gigs of data, and then going back in and doing it again and again. So I created tons and tons of database. I actually went through and I removed Amni Broker entirely because I had the same install because I tried this thing like probably a year and a half, two years ago, and just, you know, LLMs weren't good enough, didn't have the time, didn't have the energy.

Michael: 14:13

So I was worried I had like some legacy stuff that I had messed up. So I uninstalled the whole thing. I deleted everything that was associated with it and the program files and and everything is, like, completely clean, like it never existed on my computer. And I reinstalled it and set up a new database, and that still only took, like, an hour or two. Right?

Michael: 14:32

So it's not Yeah. You know, the the import might take a while, but the actual work to create a database, it's like three buttons and you're good to go.

Dave: 14:39

Yeah. So what I think you should do when you you know, you're testing with a one year database and now you wanna test with a a five year database. Mhmm. What one thing to point out is you can't convert a database to have more bars in it once you've created it once. Yeah.

Dave: 15:00

So but that's fine. So you so what I would do here is create a new database. You can always switch back and forth between databases. Go ahead and do your testing. It's probably a good idea.

Dave: 15:12

Test on your one year database, make sure things are working there. Create a five year database, run the exact same code. In fact, you don't even have to, like you can have all your analysis windows up and just change the database, point to your new one, run the exact same code and the exact same analysis window that you did with your one year tested on the five year. So there's a lot of flexibility here. And like I said, I could tell you're getting your bearings.

Dave: 15:44

It's very interesting to me because I went through this fifteen years ago and I've forgotten all the sorts of things that I learned. It's helpful for me to see what you're going through. When I get these Slack messages from you to say, Hey, what's going on here? This thing's broken. All that's very helpful for me to see.

Dave: 16:07

A little bit of a stroll down memory lane about, Oh yeah, I forgot about that. I spent a weekend trying to figure out what was going on there. It's fun to see you going through the same thing.

Michael: 16:18

Well, and I've actually I've actually feel like I am, like, 1% better at understanding the code behind it. Because so far, everything I've done is I've taken your demo file. You know, I've been plugging in a chat GBT. I've been having it it do things. But, you know, certain things like finding out how the parameter section works, that's been huge.

Michael: 16:37

So I go through it and change certain things there. It's not I can see that there is actually a lot of similarities between there and real tests, it's just not set up, I think as cleanly, I think as part of it. So when you get over that hump, think you'll be somewhat in the same area, whereas with RealTest, there's a very clear, like this is your import section, this is your data section, this is your parameters, like it, and everything is separated and you can't kind of combine it where in Amibroker you can just you can do whatever you want. So by taking those guidelines off you, you end up just doing the same thing anyway. It's just I'm creating a note says this is the parameter section and putting everything under there, where you could just throw it anywhere you want it in the code.

Michael: 17:21

So, you know, you kind of, you're taking the guardrails off yourself, but at the same time, you can just put them back on. So that's what I'm doing is I'm just kind of forcing myself to put them. So I'll probably end up with a sample script that will have just the notes of this goes here, this goes here, this goes here, that goes here. And then the other thing that was a bit of a problem is the whole adding of the columns, because I created my own thing. I didn't realize that there had to be notes in certain sections in order for the columns to go in.

Michael: 17:51

So that was a bit of a battle, but I I got over that as well. So, you know, it's it's been a lot. Like, this isn't something that I think you go, oh, I'm gonna learn Amnibroker this weekend. Right? I'm dedicating like probably a couple hours a day.

Michael: 18:07

A lot of that is just waiting for back tests to finish and all of that. But I'm slugging through, but I'm using it as like a I don't really want to look at anything else until I've done this because it'd be like reading multiple books where you'd forget kind of where you were in book one when you're reading book two. Like, I just really want to focus on this. So I think you should almost write off a month of just, okay, let me get so I can build very basic scripts in that period of time and then pretty much nothing else. And soon as I can do that and I could kind of sit down from a blank page and build an opening range breakout kind of strategy, then I won't feel confident that I have any idea what's going on.

Dave: 18:50

Yeah. Yeah. Well, you'll get there. And one thing to note, as you're talking, reminded me of a point I wanted to make. So you might think that Amber Broker looks old.

Dave: 19:04

It's this downloadable program. It looks sort of like it was created in the late '90s. You might think that's a bad thing, but it's actually a good thing. You don't want your software that you're relying on for a serious I mean, this is serious stuff we're doing. We're making a lot of money trading the stock market.

Dave: 19:26

It's an important part of our lives, of our well-being. So what you don't want is software that's jumping on the latest fad, the latest user interface style and to reorient the program completely. You have to go through a similar process to learn what they've changed. Amber Booker is not like that. This is a program that's been around for thirty years and the only changes to it, I can tell very quickly that the person that's making changes are adding things that traders need.

Dave: 20:06

And you can tell that he is a trader just by watching the stuff he adds in a very systematic and intentional way. He's not taking requests from the latest faddish requests that come from traders that think they know what they want. He's really thinking carefully about what should be in this program. And I think it shows, it's been around for thirty years and people are still using it and using it to make a lot of money. So I think it's been a success, but you may have to look under the covers a little bit to really Either understand

Michael: 20:47

an LLM that has caught up and is really good with it, or better error coding or both for mine if you're talking to it. But this is and again, this is one of those things that I I think about that eventually there will be an LLM that will understand this well enough that it will be able to to push. When we had Marcin Parker on talk about how I built my own, right? I have a real test GPT, and you guys can reach out to me if you want access to it, because I've been told not to just put it on the store. But it can, you know, very quickly, last night there was a scan that I wanted to run, and I can just type it in there, it can spit it out, and I can change some stuff.

Michael: 21:29

So, you know, you always look at it and you say these AIs are as dumb as they'll ever be at this point. They're only gonna get better. At some point that will help, but I I want to I want to be able to still do a lot of this myself so I can kind of fact check it and go from there. Yeah. And one of the things that's helped me out huge is just looking at other people's codes, like just Googling, I want to build this strategy and seeing someone else who had a blog post and they shared their code with it there.

Michael: 21:55

And then just, you know, side by side, this is what I did, this is what he did, and try to figure that out. And really, it's been one of those, you know, I loved math in school, and it's been one of those as well, where you're trying to figure out like a formula or an equation or something, and you smash your head against a wall until you're incredibly frustrated, and then you get that breakthrough and you go, okay, now I understand. So it's, you know, just every time kind of pushed me along, but I did have a question just about the more of the internals of kind of how this works and just for my own my own system. So part of the problem that I have is that, like, right now I can't go and build a database, a new database in the Amibroker for myself because I got all this stuff I gotta do for StatsEdge trading tonight with RealTest. Right?

Michael: 22:40

So I I can't have the system just running. It's not it's not doing any trading. I'm still in backtest world with it. But the way I have things set up now, just so everyone knows, have my MacBook, and this is a like a $7,000 crazy, crazy MacBook, super, super fast. And that's what I do the podcast on, and record my videos, and it does all the day to day work.

Michael: 23:02

And then I have a PC that I remote into, and that does the crunching thing. The reason I got that PC instead of some people saying, well, just run Parallels or something like that on your MacBook, a virtual machine, is because the amount of RAM that RealTest needs is huge. Right? So I I need like 32 gigs of dedicated RAM just to to build it in RAM. But my chats with you showed that Amibroker isn't that RAM hungry.

Michael: 23:31

It's more, I think, processor and memory hungry.

Dave: 23:35

Well, you will need even more RAM than you're using with RealTest. It's been many, many years since I've only had 32 gigabytes of RAM on a machine that I run Amibroker on. I've got 96 now, and it does handle AMA broker pretty well. But there are so in the last phase of the backtest we were just talking about where it adds the custom columns, that's where it's gonna use an extraordinary amount of RAM. And and if you look at your memory as that's happening, it's gonna be bouncing up and down dramatically like no other program I've ever I've ever seen.

Dave: 24:20

So that's the part that needs all the memory. And I I believe that some of the crashes you've experienced are probably because of a limited amount of RAM and running a back test that has a whole bunch of trades in it. That's the part that's gonna create it's it it just has to go through. And if you can look at the sample code and see what's going on, it's it's keeping track of the, the columns and then actually adding them to the the backtest in the last phase so you can see how that uses a lot of RAM. The more trades you add, the more RAM it's gonna use, and it'll use every bit of the of what you got, certainly if you're only using 32 gigabytes.

Dave: 25:10

You're muted.

Michael: 25:14

Sorry. I had to yell dogs during the middle of that. But the what I was because what I was gonna propose is, like, splitting them. If I I was under the assumption that I wouldn't need an insane amount of RAM for Amibrokers, so I'd have, like, my real test machine and then use the the parallels for the Amibrokers so I could kind of separate those. But looks like looks like no.

Michael: 25:34

But Yeah. So you're gonna you're gonna make me buy a go buy a gaming computer next.

Dave: 25:39

Well, you know, it's funny you mentioned that because I I think a lot about the trading computers I get for this. And we've done, I think, at least a couple episodes on

Michael: 25:51

Well, we went through your I remember your your your painful journey.

Dave: 25:56

Well, there's a so in in that, I I I made a post to the newsletter and linked to exactly the configuration that I just bought for my trading computer and for the computer that runs Aaron Broker. And it was a very popular post. I got more feedback on that than I had multiple traders contacting me saying, Hey, yeah, bought that exact configuration. One guy in Chicago said, Yeah, I called the store. You bought this from Raleigh.

Dave: 26:28

They won't ship it to me. Is there someplace else we can get this? So I'll put a link into the show notes link to exactly the configuration I have that I'm happy with. But yeah, I think it's important to have a multi thread CPU and a lot of memory to use Ambroker well.

Michael: 26:51

Which we are as we're talking, we're in, like, the big memory crunch where I

Dave: 26:57

I'm

Michael: 26:57

sure you've seen the chart of RAM prices just through the and then I live in Canada, and we, you know, me and you are supposed to be mad at each other, I guess, for some reason. So getting computer parts up here is a little bit harder, but that could be part of it as well, because that was actually going to be what I talked about, it's probably actually going to be the difference because I now have this one year database running pretty smooth. I'm confident enough now that if I run a backtest and there's just no result to the end or something just goes kaput, that is the code and the script inside the backtest and not any other part. And that's where I need to be, right? I need to be to the point where I'm like, okay, this is, you know, everything hardware wise and database wise is correct.

Michael: 27:41

If it's not doing what I want it to do, it's like my fault in programming, right? But I was still having problems with the five year. So I'd run the five year, the last time I ran the five year thing, it got kind of hung up, but I was able to hit stop and then wait half an hour and I could see the data was there. But every time I ran a backtest on it when it had columns, it would it would just cut out. So it just seems like I'm I'm probably hitting a hardware limitation there.

Michael: 28:11

So it might just be, I guess, incrementally increasing Like you said, it's very easy to make database, so instead of going from one year to five year, maybe I take the same script and then I go to two years, and then I take the same script in three years and the same and just try to find where my upward limit is and just live within that until I can get a gaming PC shipped to my

Dave: 28:33

Yeah. So so with my old trading computer, I would set up different databases for four year chunks. So for example, like, have data going back to the year 2000, one minute bar data across US equities.

Michael: 28:49

Which is yeah. No. I'm not trying to get that far. I'm trying to go to thought I was saying the year 2000, but no. I'm trying to go to 2020.

Michael: 28:56

But yeah. So you're you're twenty years further back than that.

Dave: 29:00

Now I don't need to waste resources and do a backtest across all that data for every backtest I do. Right? That would be So what I did was I would set up databases from, you know, 2,000 to 2005, 2005 to 2010. And that way, when I come across a strategy that I think will benefit from having more history, I can go run it just on that database. And it's not doing the entire database, which would require more resources.

Dave: 29:34

It's just looking at a specific time period. That worked pretty well. And I liked it. I can segment it off and have my main backtesting database just be the time period that I know I'm gonna be backtesting on for most strategies. But then I do have the option of going back in the event that I need to and to use a larger back testing window.

Michael: 30:01

And kind of like that too, because, you know, I know you're not the largest fan of like a walk forward testing or in sample out of sample stuff, but you're kind of doing that by default anyway, I think with doing that, which is one of the things I like about it is the idea is that I want to get to a point where the strategy, at least the base strategy, is somewhat profitable on its own on the one year database without optimization and things like that. As kind of a, okay, now this is worth your time to go back and to, you know, completely and go from there. So I do like that where it's like, okay, yeah, it's kind of earned its right to now run it against a larger database. But I guess, yeah, I need to go back and I need to figure out, A, can I just upgrade the RAM on that little computer I got? Or, do I have to go and buy a whole new beast?

Michael: 30:58

And, you know, like, I think it's good that we have this dichotomy because I'm a Mac guy, not because it's simple. I go and I buy the thing and I plug the thing in and the thing works, right? So it ends up just being a different way of looking at it in a different beast entirely to go, okay, well now I need to, you know, figure out, I need to get a computer that can do this thing and do all that. But this is important, think, the audience, because you got to think about these things, right? If you're sitting there on like your ThinkPad laptop with eight gigs of RAM that your work gave you, you're not going to be running giant, you know, giant back tests and things like this.

Michael: 31:38

You're gonna have to go through and go, okay, well, I need to upgrade, I need to find a dedicated thing, which again, you should, right? If you're, again, if we're trying to make a bunch of money on this, it makes sense that we should go and do this.

Dave: 31:53

Yeah. I always come back to being comforted by the fact that this stuff is so hard. I mean, most people aren't going to go through the steps that you're going through for a variety of reasons. And it also becomes really clear why it's so hard and why so few backtesting softwares support backtesting across the entire market. Because as you're saying, it's a very technically difficult and resource intensive problem to solve.

Dave: 32:26

It doesn't seem like it should be, but when you do the math, look at the day that you're having to crunch through, it's just very resource intensive. So it really makes the good software providers and the the backtesting software that's worth using really stands out because you can tell that somebody didn't make this just in their garage. This is somebody who's actually trading and understands what's involved. Go back and look at the episode we did on back testing software. We talk a lot about this, but it really becomes very clear who knows what they're doing and who doesn't.

Michael: 33:16

Should like this. And I don't know if I made this analogy with just you, Dave, or if I did it on the on the podcast, so I apologize if I do it again. But like, I like game. Me and both my wife, we play video games, right? We enjoy.

Michael: 33:30

And for me, my favorite part of it is learning the new mechanics, right? So it's like, okay, you're playing this game and it wants you to do this, and then this is how, and you're kind of, I don't like ones that really hold your hand through it, but you're like learning as you go and you're kind of figuring it out. And I would say, as incredibly annoyed as I've been at this process the whole time, I think you kind of have to enjoy this part of the game because wherever you are in your trading journey, there's a wall there, right? And you know, the way to make more money is to move past that wall. And the thing that I like about this as opposed to, you know, discretionary trading is discretionary trading is like, okay, I just need to be more of like a Zen monk, or I need to be more of, you know, battle my emotions, and oh, I did that stupid thing.

Michael: 34:21

And, you know, it's it's great for them, but it's not a tangible, this is my problem, and then this is my solution to that problem. Right? Where for me it was, okay, I gotta get this goddamn data in this goddamn system. And that that is that is a tangible problem with a tangible solution. And when that's done, it's like, okay, have get more data in this system.

Michael: 34:40

And I know that every step of this will lead to eventually I'll have, you know, a system that's a pretty basic opening range brake system, and it will make a non zero amount of money over the long run. Right? It's not gonna make money every day, but it will. So knowing that that is there and knowing that, you know, all of these steps to get there, it kind of does feel like that where it's like I'm, you know, I'm leveling up on the way to the process, and eventually I'll have a giant loud gaming PC that the wife will hate, and it will I'll be able to run this data, and then okay. And then the question is, do you get to that point and then you're dissatisfied and you dust your hands?

Michael: 35:15

But like we talk about it's like, no. It's like, now that I've got that, now that's on rails. The amount of work I have to do for this opening range brake system is near nothing. I'll have an SOP. It will probably take a minute, couple minutes a day to get everything set up and running, but then you just don't go to the beach.

Michael: 35:32

Right? You you keep you keep pushing, which is why I love the more I got into systematic trading, the more I loved it is because there's always a wall in the way, and you're always trying to figure out as soon as you get that, there's never a step back, which is what I think, you know, for people who are listening who are discretionary traders, it's where a lot of them fall into problems where they are, you know, they'll end up doing some good things, and then they'll, like, have a bad day or or do something they shouldn't or do something stupid. And there's no forward progress there. You're stuck inside this cycle because it's your brain you're battling as opposed to an external beast that you're you're you're conquering. So that's the big thing is I feel that with this, like, I have I have a monster that I have to kill, and it's you know, I know there's another one behind it, but at least it's right there, and I can see it, and I know when I'm done.

Michael: 36:22

And I know, you know, that it's it's gonna be time to do something else.

Dave: 36:26

Yeah. And unlike a game, this isn't, you're not doing this for fun. You can see, okay, if I solve this, if I can get through this, I know there's strategies I can use that will make money. I mean, that's the whole point. So, I got a question from a reader on my newsletter the other day asking about that they wanted to learn Python and they were asking about the best way to learn it.

Dave: 36:55

And they said, okay, here's this online tutorial site that you pay for, like this course. Do you think that's worth using? And my answer was no. Like, you're not ever going to be as motivated to learn something unless actually makes your life better by learning it. Coming up with arbitrary examples like some online courses do, you're not going to stay motivated to do that.

Dave: 37:26

It's way more motivating to have something, okay, here's a trading problem I have. I have this idea I want to test. What's the best way to do that? Starting from that standpoint rather than, hey, I wanna learn something just for the sake of learning it. You're gonna learn it way faster by solving a problem.

Dave: 37:49

All of a sudden when you solve it, your life gets better in some way.

Michael: 37:53

The and I can totally second that as someone who's gone through that, I've downloaded all like the little apps and like the little like learning games and the right? You know, they give you stars and blah blah, but you're like, okay, that's whatever. That's fine. But no, a 100% agree that, you know, by you need to know that by the end of it, this is this is the goal. And and for me, I I just looked the the way I started this mission for myself is like, I want a fully automated, like opening range brake system.

Michael: 38:23

And I gave myself a time limit, which I know is arbitrary and whatever, but it's just to push myself to do it, saying I want to do this, and I want to run it against this prop account that I want to do that in this period of time. So it just makes it to the point where every time I'm like, you know, screw this thing, which has happened a bunch of times where I'm like, know, hey, I've got my swing trading, it's going well, I had an amazing year last year. But by setting that kind of achievable goal and just pushing yourself to it, you'll get you'll get way better at whatever it is you won't go through and and, you know, you won't push yourself harder than if you just, okay, this is the mission. And then again, look at it like every block, just look at the thing that's in front of you right now. Like, I'm not thinking a million years down the road, it was get the data in.

Michael: 39:15

Okay. Now back run a simple backtest right now, add columns that simple back, right? And just, you know, and that I think helps with it as well, where I would set myself, I want to do this one thing by the end of the day. And sometimes, you know, I'm being a good dad and taking time off, putting the kids to bed, and then going and sitting until midnight to try to figure it out. And then sometimes I figure that finish that task at noon, and I get to do whatever else, but just put like small I think that's part of the problem too is if you sit and just looked at the end goal and say, how do I get there?

Michael: 39:46

You've just screwed yourself. Right? What is what is the steps that I have to do to get there?

Dave: 39:51

Yeah. It's a really good skill to have to be able to break a problem, you know, large problem down into solvable steps. And the better you are at doing that yourself, the better you can harness an LLM to do that for you. To separate the problem, it's just two different steps, solve certain parts of it. It's definitely a skill you'll get better at it over time, but it's really valuable to be able to do that.

Dave: 40:23

I mean, that's literally how you solve problems. You can't tell an LLM to create a profitable strategy for me. Like, there's so many steps in doing that. Like, the problem doesn't even make sense. Mhmm.

Michael: 40:38

Well, and to go into LLMs, again, it's not great with Amibroker, but I have found, wife probably thinks I'm crazy, but I found like the voice mode on them have been great, where I'm just sitting here and I'm just talking to the thing, like it's, you know, an assistant kind of like helping me out. And it's been great just to say, okay, what about this? And I did that a lot with your demo code. Because I have some experience with, you know, when I saw that ref thing, you'll know what I'm talking about in trade ideas that I was doing with. And it sometimes works the same and sometimes work different.

Michael: 41:15

You know, I kept typing with it and it'd give me these big long explanations, but I found the voice mode really good just to say why and just have it kind of rant at you. So yeah, just understand that you have these assistants, and if they are small problems that you can get through, they're really, really good. But the the big problems, again, especially with Ameren Broker, it's just awful Yeah. For now. For now.

Dave: 41:41

Yeah. Well, LLMs have never made a trade or never had to live off money they've traded, so that's that's a very important difference.

Michael: 41:50

Yeah. So I am kind of embarrassing to say, but I am still in data mode. So I guess I you know, it's one of those things I'll continue to plug away. I think the kind of takeaways I have from here is instead of going from one year to five years, it's probably I'm hitting a ram limit. And I'll go from one year to two years and I'll build that database and run the Now that I have a stable working opening range break short strategy that is different from yours, that I took yours and I kind of built myself.

Michael: 42:20

Mhmm. And it looks okay. I haven't done any optimization, so now it's just expanding out either a year or maybe six months at a time, more and more databases until I get to the point where I'm starting to hit that limit and getting those errors, which will probably take the whole weekend, but that's okay. And once I get there, I'm like, okay, this is the current physical limit to the computer that I have, And let's kind of go from there. Because even if I could get two, three years for a one minute candle system for data, that's that's good enough, think.

Dave: 42:54

Yeah, you can a can lot of confidence in in that, for sure.

Michael: 42:59

Yep. So then that might be in it's, I don't know, it'd been ages ago that we did this podcast where we talked about buying something with your trading profits. So maybe that will be the goal is that the first bit of money I make from this particular system, I'll allocate to some monster, again, gaming PC with, what, RGB lights, and it'll be loud as hell, and the wife will hate it. So I'll get I'll have to, like, put it in in my little server room or something instead. Yeah.

Dave: 43:28

I love it. Yep.

Michael: 43:30

But yeah. So that's that's the update. And, you know, again, as long as you guys are are interested, we'll continue to make more updates. I don't know if we'll be able to do it like every podcast like we've been doing, but I think maybe a monthly check-in or something on on Miami broker process will be cool. And, you know, ask your questions.

Michael: 43:49

So again, we did this additional episode because of the YouTube comments and the emails and everything that we've gotten. So, you know, continue to ask away and what you want to see and what you want to look at, because it makes our job super easy when we're just like, hey, you got this cool email, let's talk about it. And we can just turn on the thing and away you go.

Dave: 44:11

Yeah. The feedback's been great and, yeah, continue to do that. I've got a whole bunch of questions from readers that I've got queued up and I'm trying to work my way through. So it's always good to get questions like that because it it just makes us think about new things to talk about and new ways to help people.

Michael: 44:31

Well, and like you mentioned, right, me going through this process now is like shaking things loose in your brain from when you used to, when you did things before, right? It's like understanding that, you know, it's you, when you get good at something, you kind of take for granted some of the details that just become like second nature to you, right? They're just like breathing, and you don't understand that, hey, oh yeah, someone could get stuck there, but I've been doing it for twenty years, so I just kind of know that that error means this, and I need to get past that point, and I need to move on. And yeah, so when you guys bring these questions to us, you've seen it in some podcasts, they turn into like pretty interesting debates, even though sometimes the initial reaction seems obvious, that you can go back, oh, no, this actually now that I think about it, and you start to sometimes expand your own mind to answering other people's questions.

Dave: 45:27

Yeah. Love it. Alright. Well, I'm I'm glad we got to catch up on this. And, yeah, I'm looking forward to the next next phase of your Amber Broker adventure.

Michael: 45:37

Yeah. Hopefully, a little bit past importing data the next time if we catch up next month. But as always, I'm Michael Noss.

Dave: 45:44

And I'm Dave May. Talk to you next week on line your own pockets.