QuantConnect Open Algorithmic Trading Meet-Up #5

Lexie: okay um we'll go ahead and get started um hi everybody my name is Lexie and i'm with the marketing and PR department at QuantConnect thank you to everyone for joining us tonight welcome it's your first time and welcome back if you're returning this is our fifth monthly meetup um we hold them on the last thursday of the month at alternating times to accommodate our friends friends around the globe the schedule for tonight will begin with one of our engineers giving a quick tutorial on generating data for back testing strategies on your local machine we hope you enjoy the presentation and if it sparks any research ideas or trading hypotheses we hope you join us on quantconnect.com and with that colton i'll let you go ahead and get started Colton: hi guys uh my name is colin sellers i'm an associate software engineer here at quantconnect and today i'll be giving a presentation on our a tool of ours called the random data generator so this tool in particular was created as a way to give lean users an option to test their algorithms without needing to purchase or download real financial data and to reflect market or realistic market data we use a brownian motion implementation to generate the data so with that let's go ahead and get started oh sorry lexie do you mind turning on screen sharing for me thank you it Lexie: should be good to go now Colton: all right do you guys see my virtual machine Lexie: yep we can see it Colton: so i actually wrote down an instructional to kind of help you guys along if you want to follow along no pressure though otherwise mostly for uh followability later if other people want to try this on their own i'll be going kind of fast but all the steps are included so i will put that in the zoom if i can well it looks like they cut me off on my instructions and you know what this chatbox might not be best for this i'll still include it all there but i'll also give these instructions to lexie so she can include them in the video recording sorry about that one more step so i broke it up into four sections just to make it uh as simple and as easy for people to give it a try on their own uh the first section is all about getting yourself set up with the lean in visual studios i've already completed all these steps for setting it up particularly because it takes some time to download and get the repo in so i'm going to skip the first section on lean open-source setup but the gist of it is get lean from git you can either use our Github and download the zip directly install git on your machine and use git clone instead and then use vs to open the project in this tutorial i'll be using vs 2019 and the only two things you'll need to do it are vs 2019 and the .Net 4.6.2 developer pack so i'm just going to go ahead and open lean and show you in vs here so in the lean repository the solution file is the quantconnect.lane.sln if you open that

with vs it'll open up everything for you nice and easy so the very first thing we have to do just before we can use this tool or even look at the toolboxes we'll have to build lean because of as of right now the repository is clean it's just the source files so if you open up lean go ahead and build your solution Member: colton yeah may i ask a question Colton: of course Member: first thank you for your time and putting this together so all of the just visual studio 2019 is this the installation is there a virtual machine that we can get with everything on it or some kind of docker image Colton: we actually don't have that for vs we do have a docker image that you can use in conjunction with your local machine but there isn't a dedicated vm for development that is something that we should look at though that is a good idea and that's actually what i'm using right now as you can see um Member: right yeah so i was just curious if this image was available instead of having to install that stuff Colton: you know that's a good point if i can package this up let me let me check in with uh jared and see if there's somewhere that i can put it out there but that is a good idea i'll take that into consideration Member: okay thank you and this this is the free version of vs 2019 right or do you have to get a license for it Colton: uh this is vs 2019 community the free version Member: all right thanks Colton: no problem all right it looks like we successfully built the lean project now the first step we're going to have to do is to use the toolbox to generate actually first i'm going to pull up the documentation on it just so you guys can see where that is and i'll also include a link so as our documentation states you have to go find the executable for a quantconnect toolbox and use the parameter to define the application we're going through is the random data generator so this page is perfect and straightforward so first i'm going to open up a command prompt and get us to where that exe lives i like powershell people use bash command line it all works the same so i know that our toolbox lives in the toolbox directory and because i just built in debug form i'm going to go to the bindibug file and look at my binaries and compiled files now this is filled with a ton of files uh all the dlls all the dependencies it needs everything but for our case we're just looking for the toolbox exe so if you're in bash you can use grep but i'm going to use get child item with powershell and i'm going to look for the file so we see it was created at 6:06. just a way of confirming to myself to see that it's there i didn't want to run it and have some errors get thrown because you can't find it but now that i know it's there i'm going to write a command to generate some data so first call the exe itself and then we're going to feed the params the parameter to use the random generator is the app param type that out then some of the parameters that are optional as we saw here include the resolution the from date the to date the market that it's included in the density of the data that it creates if you want to include coarse data and then the amount of symbols you'd like to generate a lot of these have default params so you won't have to worry about defining them all and for the sake of this tutorial i'm just going to give a really simple one and it is we're going to go resolution first i'm going to set it to daily so we can get some daily data a start date which let's choose 2018 january 1st and an end date 2019 january 1st so we're going to generate a whole year's worth of daily data market usa to make it straightforward and then the amount of symbols that i want to generate which is a required param let's just do three and enter so this goes through a couple different processes none of which really concern you as much but to give some more insight into what's happening it might be helpful to know first it will go and create the data for these tickers that it randomly generated once that data has been created including random events like ticker changes uh dividends splits all the above all the stuff that you'd expect to see with a real uh ticker or a real security then it takes all the information that we created in the daily data and it creates coarse fundamental data to reflect the fundamentals of the securities that we just created so not only are we creating trade bar data for every day we're also creating events like realistic events and also the fundamentals that we can analyze for universal selection now that we have created this data actually this is a good time to take a stop and ask for any questions before i move forward Member: so colton i'm curious about the because i'm brand new to quantconnect i think i've gotten through 25% of the boot camp so you have to forgive me my newbie questions Colton: no worries Member: so Colton: every question is a good one Member: why what the the need for this is because i don't want to use the back testing in quantconnect Colton: i would say for your purposes if you are comfortable in the web ide because you don't have access to the rest of the engine the quantconnect data is real and good you can stick with that but for the sake of being able to step out and work on the entire engine like an open-source contributor or maybe you just want to be able to work in your own ide environment the random data generator is just a way to save some money and not have to purchase from our data providers through our website Member: oh okay so you're saying that if i'm all web-based in your ide then the cost is it doesn't cost me Colton: nope yeah it's all uh all our back testing data is free and available through the web ide Member: okay so now um if i want to do local development then i need this Colton: yeah it so our repo comes with a set of data that we're allowed to distribute on our repo and it's specifically for testing well mostly for testing um you could use daily data like we include with spy like all our spy data is included in the repo if you're if you download lean you'll have access to all that uh for free for daily but more finite like minute and second and tick data it's all very limited uh in the open source distribution so in order to have enough or large enough data set or a versatile enough data set like options and all that you'll want to use this tool to be able to back test your algorithms locally instead of through the web ide Member: okay and that just speeds up the process is basically what you're saying Colton: yeah i think it's just a cool one you don't have to you don't have to purchase any extra data and you can still have a realistic trading environment when i say realistic i want to make be clear that loosely realistic because as we all know the markets are strange but the idea oh go ahead Member: so loosely realistic meaning that this data that's produced here isn't just like random gibberish Colton: exactly okay so it follows a a brownian model which is actually it applies to a lot of things in nature but it's like a uh you know let me pull up the information i had on it because i was looking at this myself just earlier to make sure that we could just talk about a little bit Member: yeah because i mean when i hear that i i just think you're gonna take like maybe some real daily data and a couple points there and you're just going to do some kind of regression that gives reasonable data between those points and that's like this is all based off of real data it's just that you regress the rest or what is tell me more about what brownian is doing Colton: so it's not based on real data but rather that through the brownian model we can reflect what we see in actual markets and if you were to look into this uh the brownian motion model you'd see that it actually is very common for analyzing market behavior it's random in a sense where you won't be able to predict what's next but you will know that it's within a range it's not going to be at one moment your ticker is at $90 and the next minutes at $1 right you'd expect to be a smooth regression between the two right and uh like a stochastics okay Colton: thank you colton and please let me know if i'm asking too many questions i'm just such a i'm a curious fella little gregarious but i'm very interested Colton: no this is great i i appreciate questions it makes it easy for me to slow down a little bit that's why i asked so thank you um and Lexie: you have one more question um in the chat colton Colton: oh it looks like we do have a question it says can you generate random data on the qc web ide i believe that answer to that is no if i were to double-check it might be possible in the research environment but i'm gonna i'm gonna default on derek's answer who is a quant developer here at quantconnect as well and uh he said that it is not available uh in the web id only through the toolbox thank you Derek Member: so so colton if if i wanted to um actually just do local development and i needed to backtest i would use the lean model i and i could pick whatever language i want at this point i mean like between python or c# it looks like c# makes sense for this visual studio one i haven't written python and v and visual studio in 2019 of use of my other ide but can you tell me the comparison between the web ide and this visual studio ide Colton: you know the biggest difference is just that with the visual studio ide you have the entire engine laid out in front of you all the files everything that's going on whereas with the web ide you're able to debug your algorithm directly but you won't be able to look at anything else and for most people that's plenty plenty enough unless you're trying to do open source contributing or make changes to the lean engine for your own repo then it would be a little bit nicer to well i guess you'll have to have it right you'll have to be able to see the changes but you can do python in vs it's not as friendly to me i use vs code as a personal for for using the python algorithms but you can do everything at the algorithm layer that you do here as this is the same as the web ide the only difference is that if you want to go any deeper than the algorithm you will have to use the actual distrib or the actual repo locally Member: okay and if i'm a test-driven developer is there a testing framework i can use in both worlds the web and the ide or the web and local dev or is it only in local dev i can use a testing framework Colton: it would be in the local dev that you'd have to use a testing framework although i would suggest for those of for those of you that are looking to oh it's like testing or like research uh the research environment in the web ide through a jupyter notebook is actually probably the best tool for algorithm design specifically now if you're looking for testing specific parts of lean then yes you'll have to do it locally with art we use a end unit for our testing framework Member: but a jupyter notebook does it have its own testing framework Colton: it doesn't not not at all in the same way of testing as you would think traditionally uh i only suggest that as like a research alternative right or not alternative research a research environment Member: see colton i've been doing dev for a long time but i just haven't become familiar with all these tools Colton: there's so many huh there's so many Member: they really are and i'm and you know when you're a newbie you really would like to have a guide and i'm looking for somebody who'd willing to give me an hour to just to help me get past that first 20 percent right and um do i just reach out to this group and say who would give me some time or Colton: yeah definitely i don't know if you're on our slack as well uh i'd say the slack is a great place to reach out to people and get some advice on where to start and where to go and i think too like with with this specific instance i've i've given only one other presentation before at these meetups is that i'm trying to get that information out hopefully as uh seamlessly as possible and have a recording that people can follow that take the steps through so yeah just reach out asking people i'm i'm on the slack as well uh when i'm working i'm i'm around i don't mind giving pointers or uh answering some questions especially i like to see people uh learning learning lean and i myself have a lot to learn with lean even now as i get further into this into this job Lexie: for sure and we also have tutoring if you go onto our community forum page you can see the tutors and you can book times with them and they can go through all of those questions with you maybe for the sake of time um colton you can maybe continue your presentation um nick maybe you would have some questions at the end maybe colton will answer them in his presentation um is that okay with you guys Member: yeah i'm sorry lexie i'll Lexie: no it's totally fine totally fine we appreciate the learning experience for sure Colton: all right well with that i will uh move on to the next phase so that that first part i believe if you guys saw my instructional that was everything for data generation now what we generated is important and i'm going to get a little rundown of what was created and how it can be used with lean or what it means to lean is a better way to put it so i just realized that i closed my powershell which showed me the tickers that i created which is a little sad but i do know that we can create oh yep so we see this at 6:11 these were all created so that's how i'll know what my tickers are i should have kept the window open but that's on me so the first thing that uh of the four important data files that were created when we ran that random data generator is the data the trade bar data itself so we'll go to to daily and then we can look at the sets of data we created for this ticker foom we can see that it's first emittance of data was at 2018 february 15th and the format of these csvs are actually defined in our readme so if you're ever looking at these wrong readme if you ever looking at these and you're curious what they are uh they're right here so with daily we have our date times our open high low close and volume and for the open high low and close these are in deci-cents so if you want to get the dollars divided by 10,000 and it'll be in dollar format now another piece of important data that it creates is our factor files and this is lean's way of dealing with uh splits and dividends and we're going to stick with foom for this just to make it simple but with whom we see that on 2018 january 2nd we have a price factor of 0.75 now what this tells me is that on the second we had a dividend paid at this price factor now the actual math behind that was handled by lean you don't have to worry about it but it is one minus this price factor multiplied by the current value and that'll give you the dividend that they paid for for that date next we have the map files so map files are interesting and tricky because the name of them changes from time to time if the ticker itself changed names then the map file would have a new name and also have the previous names in it so we actually know that fume was c f v before uh 2018 uh february 14th so that was just another random generated event was that the ticker changed or the the security sticker changed its listing and last is our fundamental coarse data this is the information i was talking about in the last section that is important for universal selection via fundamentals uh there's ways and you'll see this in the tutorials you can filter based on fundamentals there's coarse and find fundamentals and the coarse uh the coarse includes very high level technical indicators are not technical indicators but technical values that you can filter by in lean again all of it's handled for you but i just kind of wanted to give a dig down so you guys could see what actually is happening after you create these now i know that fume was first trading february 14th under that ticker so i could pick any date in here that is after that and i should likely see it so let's just do the 21st here it is so this is the interesting thing about the ticker changing after it was generated the ticker is fume but because when it was originally listed it was cfv the security id which is this string will be the original ticker cfe and with that that is the all the data that was generated and kind of how it plays into lean so let's go on to using the data the last section and before i switch does anybody have any questions Question: Are you guys thinking about making it possible to choose from different distributions as a param? different distributions as a param uh sorry denny said are you guys thinking about making it possible to choose from different distributions as a param i'm not 100 sure what you mean by that do you mind clarifying Question: Distribution as part of the generation process of the random prices that's all right oh thank you oh i see you're saying we don't have other implementations of our random value generator but yeah there is there is information on the random data generator that talks about the interface that we created it's called the i random value generator interface and if you want to extend that class and create your own model for generating this data you can and it's really good to look at the one that we currently use just as some information but yeah we don't want to get too deep into the code there but definitely take a look at it Question: Thanks Colton: no problem all right so i chose uh the basic template algorithm because it's just a really good place to start to make it uh nice and easy to set up you could create your own for this but you definitely don't have to now the ticker we were looking at was foom so i am going to modify this algorithm a little bit to make sure it has the data i'm looking for uh we did daily so make sure that your equity uh your ad equity call includes the proper resolution and then we'll pick a date that we know we have it i know that we have it after february 14th so let's just start there and because we generated all the way to the end of the year it should be good to go to the end of the year but we probably won't even have to get that far something you want to check before running algorithm is to check your config i actually came in here earlier and made this change myself but by default it is basic template framework algorithm uh change it to basic template algorithm to point lean to the right class to run if you look at this there's actually options to change to python all sorts of different environments but just for backtesting all you need to do is change the algorithm type name to the file that you're using the class name you're using under the c# algorithm project so it looks like we have everything set and squared away i put a breakpoint here in on data so we could see that first piece of data enter uh enter the algorithm and with that i'm going to hit start and i see we have a question from lee as well i will address that as soon as i finish this section so it looks like we got our piece of data that we were expecting which is great now the object containing the data is complicated there's a lot of parts to it you know we're moving futures change options chains all that but because in my algorithm i only subscribed to fume daily data all we expect to see in these bars is a fume bar and there it is as we can see this bar was created the day of 2013 to 20 or uh 2-13 to 2-14 so february 13th to the 14th it has all the data we expect then just like all the tutorials show you if you set set your holdings you can define that you want to have 100% of your portfolio to this symbol um i picked this specific algorithm just to make it easy to add the data and look at it but definitely give it a try look at something a little more complex there's plenty of options here under the c# project to try it with lots of ones that even i don't know how they all work but give it a try so this concludes a very basic use of the data we created so that's all i have for you today but i'm going to take some questions please take a look at the resources i provided and give it a try so lee said if we're going to install lean locally with linux and a docker is lean with linux faster or slower than lean with windows in terms of performance although linux does use mono for lean in my testing and my experience running it in the docker versus running it in windows and in the docker it also uses mono the performance is pretty much the same negligibly different uh i'm i know that we're moving to .net standard and in the future here and i've actually heard that so it might bring some uh performance improvements so i guess we'll see how that reflects but the nice thing is linux will also have net core and uh won't have to depend on mono anymore no problem uh anybody else Member: are are you saying colton that this build of lean uses .net core Colton: currently we are .net 4.6.2 but we are in the process of migrating towards uh .net standards so we can use .net core in the linux environment uh use whatever .net

framework uh is applicable to all environments oh so right now right now we are on 4.6.2 still Member: so you're really just i mean natively you're with windows unless you use mono Colton: yes we do have our docker file that has mono built into it and as i said i i don't see any downside i'm pretty sure i ran some benchmarks when i first built my new pc for uh running it in the docker image through mono and running it just directly through vs and the results were the same Member: thank you Colton: no problem thanks nick all right i think that's it Lexie Lexie: yeah thank you so much colton for presenting um does anybody have any lost questions for colton nope i'm okay well i don't i think if nobody has any more questions um that's where we can wrap it up tonight thank you colton for presenting with us today um and all of you for joining us as a reminder we hold these meetups on the last thursday of the month we hope to see you again at the next month to develop and deploy your strategies please visit quantconnect.com and to stay up to date on new events be sure to follow us on twitter @quantconnect okay everybody have a great night and have an even better weekend all right Member: hey colton yeah um i don't know i don't want to take people's time but i have some more questions if i could ask you Colton: yeah i'm i'm totally down for that i'm here Member: um so the first question is who do you know that's making money with this is it is it like like i you know i'm kind of asking myself some questions about you know do i want to do this or not and invest my time in it i mean i i enjoy computers i enjoy being able to do algorithms and backtesting it all looks neat but the end of the day i'd like to have an automated process that'll make me something Colton: right Member: so talk to me about your anecdotal experience with that Colton: yeah well what you're getting to is a is the hard point of algorithmic trading is that there isn't a guaranteed path to making money right it's Member: right right Colton: it's about yeah it's i'm no math major and in that i'm aware that my strategies can't be as well they can be with the right help aren't as technical as those that have that background and from what i know is that the people that have like funds and uh have the help that they need to create those strategies is that that seems to be where it takes off and when people actually start to make money either a fund comes in and builds their own strategy and sells it as like an etf we have some people that do that or you get a talented quant that comes in and builds strategies and starts leasing them to other people as an alpha stream and that's another avenue for quantconnect i think that's actually a pretty big one especially just considering that individuals can come onto the platform learn how to develop these strategies and then sell them to other people is uh probably i would say is the fastest way to make money right unless you have money already um but yeah it's it's difficult i know as as an engineer i'm good with writing c# code i'm i well i'd like to think i'm good with writing c# code i'm good with understanding the process of building the architecture of lean but when it comes to writing strategies i am just as green as most people are it's it's a different beast i would say than just regular computing Member: okay well i would like to hear if like the first thing i'd like to hear is is there a reasonable hope of um well i guess i'm it i know it's a hard question answer and i'm trying to put i guess as i think about my first question the one i guess is does the um simplest algorithms you know i'm trying to look at occam's razor that says the simplest thing i mean i have seen algorithms that like nobody will understand except for the writer right and i am just asking if you have seen either simple simpler wins because what i just heard you say is that highly complex wins Colton: yeah i guess i i don't want to say that it is limited to only complex algorithms because i have seen simpler strategies perform well and i think what you're getting at with occam's razor is is true yet it's so difficult to try to measure the effectiveness of one strategy because at one point in time it may be amazing and then another it's not you know the the most basic algorithm we have in here it's hard to call it an algorithm is just but i hold spy you know and that works during certain sectors of time but one of the ones that i would point you to as just a good one to look at that i like as a as a straightforward not super complex yet strong is the price earnings anomaly strategy let me see if i can find the link to that real quick here what it uses is the fundamental data on price earnings for stocks it filters all of all all tickers that it has access to by their fundamental price earnings and then those that have a really what is that it's a really low price to earnings right would be undervalued in theory according to what what most people would think right right and it can select those stocks and purchase them and i actually ran that through a backtest once on our platform and it was it performed pretty well but in the last year you know things have changed like our market is so volatile it's i noticed that when it got to that time zone wasn't so great it's it's very interesting but that is a a prime example of one that's it's not that crazy yet it performs uh let me see if i can find that price or anything Colton: sure i'd really like the link and i'm just enjoying this so uh if it's okay i'll just keep asking my questions but i don't i just tell me if you have to stop so Colton: you're good um Member: so like the is there means is is there a place that people post open source algorithms that are currently making a certain percentage i know that on the forums there will be a lot of people who will share their back test results and talk about their strategies and work together in terms of releasing strategies that perform really well i would think that most people might hide them but i have seen i mean not too long ago after a quantopian shut down we had some people talking about an in and out strategy that a lot of people in quantopian were using and that's on our forum straight up available open source right there built for quantconnect and if you look at those strategies it's uh they're good they're good strategies and it's like okay so there are people out there that are putting it out there i'd say your best bet is to look at the forums uh talk to some people and get some advice because a lot of people out there know a lot more than i do about these strategies are about how to strategize and how to be an effective quant and i just linked that price earnings anomaly as well and it looks like derek posted our entire strategy library as well Derek: yeah nick you might want to take a look at that that's just a large collection of strategies built on our platform that uh they display how they've performed in the past and then you can take it and you can change it however you want and you can even live trade them Colton: some written by derek himself actually Derek: yeah exactly Member: derek i mean i don't know how to ask this but like you know have you seen people make money with this that that i mean i mean just like anecdotally Derek: i mean when you when you look at the wikipedia page for automated trading it says that 75% of the trades placed in the market in 2014 were placed by automated computers so i don't know i just think if there's that many players in the game that's just kind of your evidence that somebody's making money somewhere um but yeah as colton was saying it's it's a hard game to it's hard to get these strategies to work because um there's a trade-off between overfitting and having just a very simple strategy that works that's just something that you guys already touched on Colton: right i think i think the overfitting part is huge if you if you are looking at one range of time and you are trying to force this beautiful return you are probably going to overfit you're not going to be versatile enough to fit in in any environment and that's why with our alpha streams we rank them by their sharpe ratio because we can have it's the best metric that we can put on it to say this is a good bet you know this this is a good idea to put your money into versus those that don't have good sharp ratios Member: okay so i know nothing about like i remember you had mentioned uh the quantopion that went out of business and so are there competitors to quantconnect i know nothing about any of your competitors Colton: oh yes it's it's actually pretty big i think for the longest time quants in quantitative trading was pretty much kept to banks and banks alone you know i mean they they ran the scene and they knew that what they were doing was effective and that if they kept it in-house that it would be beneficial but at this point quantconnect with quantopian falling off is the lead in the algorithm trading space there are other people building engines and trying to build like tools like pieces of tools but in terms of comprehensive like having access to all these different things even though lean can be complicated it that's because it has all the tools available that you'll need i truly believe that it is the leader in algorithmic training Member: okay and why is that Colton: pretty much time time sank into developing engine and then the open source community that we have and the people that we have working with us as well as just the versatility of the dual i mean just now i think in the last month one of our engineers put in future options and i don't think anybody else has that being able to expand and go to more brokerage more brokerages and more securities is pretty much the way that we are encompassing everything that we can we can be that trading platform for all types of uh securities Member: yeah i mean and nick um this is eugene lights but listen to the conversation Member:colton can you hear him Colton: kind of cut off there for a second are you there Member: great um actually it's going into direction with our ability to do um from crypto to options to um you know the regular equity asset class i don't think i i've looked and i couldn't find anything that's the reason why i was on quantopian i looked everywhere but quantconnect is the only place you can find all that and i hope that they're going to expand some of their brokerages um that's still i guess the thing that they're talking about but expand their list of brokers would be awesome too Colton: for sure that's definitely a priority uh to figure out which ones are the best to go forward you know the hard part for us to work with these brokerages is that we have to work through their interface right whatever api they give us is our limitation and if there's issues uh building our interface around it even though we built it to be as modular as possible working through them convincing them that they want to partner with us and all that is is the uh is the hard part you know Member: yeah i was disappointed with the alpaca on falling out because they're building um a brokerage that's basically for algorithmic trading right for you know so they have the tools but they're they're utilizing they're in silicon valley so their mindset is more the silicon valley big tech i wanted to take your data and sell it and take it all Colton: yeah yeah i'm actually really disappointed too because i have an alpaca brokerage account now i'm unable to live trade uh through that that interface and yeah i think we're uh looking for alternatives to kind of fill that void especially with alpaca being the zero commission brokerage that they are but yeah that's part of the challenge Member: that's this is great i enjoy the conversation because it's one of the most valuable parts so tell me about the people that work there at quantconnect and why really when it comes down to many companies are the the people that make the difference so what is it about quantconnect that's going to make it keep going and uh be successful i mean you you just told me because the time they put in but i'm more i'm interested with the people Colton: yeah yeah that's uh it's interesting because we are small you know there's not a lot of us but what we do have is that every single individual seems to have their specialty as you might call it and me being an associate software engineer i'm on the lower rung i actually uh and pretty new i started last year after graduating from uw uh in computer engineering and just the amount of skill that comes from the people i work with is insane actually and i think about why like how did they get started at quantconnect i thought about why i got started with quantconnect and a lot of them it's funny when i look through the forums i see posts from them when there were users so i think that's what makes us a little bit different was that because we went for the open source approach and having people in a community like on the forums it created the environment where people who were interested in it also could get a job and that's what happened with me i uh i was looking for work and during the covid crisis i was really concerned that it wasn't going to be possible and one day i just happened to be on the quantconnect lean github because i was using the platform in school or i was i was trying to use it not for school but i liked the idea of algorithmic trading and i saw the listing and i applied and it's like wow you know the fact that i could go from being a user to a worker to back it and just the amount of knowledge that has come from working with them is huge and i kind of see that through most of our team honestly Member: cool that's um that's encouraging and so the main places for interaction learning you've got your tutorials i've got the person i can sign up with you've got your slack channel which i i've got to get through the rest of my training to get access to it Colton: oh yeah and at least 50% i think i'm 50% through but um and then i can but those are and you've got your your channel on youtube is there any other places besides youtube and slack and obviously the community that you recommend Colton: definitely the community if you from what it sounds like you're looking for the the forum the uh discussion boards is the way i know that they've put a lot of work into the discussion format boards too lately to try to get people to uh communicate effectively like we added uh the reddit type stickers that you can tag onto things just we're trying to create an engaging environment for people to work together because that is that is the core of our our business really right so if there's anywhere i'd recommend it's the uh quantconnect community board on the discussions uh i guess i'll just put a link here too yeah i mean i i'm uh i'm there i've got i'm about halfway through the boot camp and there's and it's been really good so i uh appreciate all that and so um i i see some of the pieces and now i just have i have two somewhat unrelated questions but i really would like your opinion on so the first is um what is the deal with what happened with um is it robinhood and all this trading of game is there any algorithmic part that's like i uh like what was that uh Colton: you know i don't have an answer to what that was but i think i know what you're getting at and uh Derek: the reddit users Colton: the reddit users wall street bets right they they short squeezed the heck out of this because it was shorted so much by institutional investors and i don't know i've seen i've seen plenty of memes already i've seen plenty of jokes being made but if you're asking what kind of strategy that i would think would work with that i mean i might not have the best idea but in my head the first thing that pops to mind is volatility strategies that is an extremely volatile uh event i mean jumping 300% in a single day is insane and i don't know what the strategy would be built around that but that would be the thing that would catch my algorithm's eye if that if that makes sense that's what i would filter by some large spike in volatility very abnormal behavior and then trying to capitalize on that is up to the the strategizer Member: okay and so across these algorithms i can sell options i can buy options i can i could do all the things i could in a commodities trading i i'm it's just all automated right Clton: yep yep i mean you do have to have you know the the groundwork of deciding that behavior but yes it can be it can be all automated and if you build something that is universal enough to be able to be used for any stock that just fits some sort of parameters that you set some sort of restrictions then yeah Member: okay i can't imagine doing this without doing test driven that's just Colton: yeah i mean that's what that's what the back testing is all about i mean if you build an algorithm and you queue up a back test and you run it you can see what's going on you can see what you're missing where you're not hitting if you add the debugger in the web ide you can really like take a step and say like okay this is where my algorithm is making the wrong decision and now i need to refine that behavior and make it better Colton: all right last question here is unrelated but are you familiar with warren buffett's bet with um uh but uh it was it was it was with the hedge fund manager he basically bet a million dollars that vanguard 500 would beat the hedge fund you familiar with that Member: i am not familiar with that that's kind of funny Member: so let me post this npr show i was just curious if anybody who had heard of he basically said okay 10 years i bet you i can make more money in just vanguard 500 than you can with all your your uh your algorithms and you can listen to it but um it i was just curious if you'd heard of it and if because it kind of like flies at why do this and just use a a very simple um low low fund mutual i mean low uh low-cost mutual fund yeah so i was just curious if you'd heard of it not that this is the end all of things i just thought it was applicable Colton: yeah i mean it's something to look into that that's the thing is about in the world of the markets there's always i mean the gamestop thing is crazy everybody's still reeling over that but like tomorrow will be something different i mean it even elon musk tweeting today about uh what did he tweet about cd projekt red sent the shares up 13 it's like little things like that i don't think there's a way to perfectly capture all the stuff you need to to turn it into an algorithm but the idea is to start searching and keep an eye out for those for those things Member: okay um it's ten o'clock and i know your your meeting's over but colton you've been very kind to take my questions and derek thank you for jumping in there and other people i just want to know if i really how much of my time that i should invest in this Colton: you got this i'd say it's worth it i'd say put the time in it's it's gonna be work but if you have the motivation to go that far with it then it will work for you and that's that's the goal i mean that's the goal everybody's looking for right is that you can do something that you can leave passively for you and adjust from time to time with the right work but uh not have to just day trade directly i mean nobody really likes sitting in front of a screen like that Member: right all right well um thank you for your very valuable time today and for everybody who stayed on i appreciate you listening Lexie: yeah thank you for your questions nick yeah you know i mean this is possibly something that would really uh um help uh you know many people i mean heck yeah if you could have a you know an automated money machine but all right thanks so much for your valuable time i'll talk with you later Lexie: yeah thank you guys have a great weekend everybody

2021-02-03 19:46

Other news