If you enjoy our teaching approach, we're considering developing a course dedicated purely to Machine Learning and AI-based trading strategies. Edit: Thanks for the support. We have released our ML/AI course. You can find it on our courses web page. Make use of the early bird offer.
Yes please, I and couple of trading buddies are interested. We can venture out if you can give us a good foundation and /or foothold. This approach is quite difficult to get into, may not be for you, is for us.
I would be 100% interested if the course was based on Futures (NQ, ES, GC, CL, etc) DAY TRADING along with using Prop Firms when automating (as well as using Brokers like AMP or other). Heck, I'd buy it this very moment if it was available since it doesn't exit! (I did sign up via the link in your description box BTW). There are already some courses out there - but they're for Forex, Options or Stocks. NONE for Futures only!!! This would be unique. Once ready, you could promote it at places like TopStep or other for a whole audience that would be foaming at the mouth to have a course like this for Futures Day Trading (as opposed to just downloading daily historical from Yahoo - which is a joke). If you could teach in your new course how to get the historical data from the broker (programmatically), for example in different TimeFrames to run through ML and determine which has the highest success rate - this would also be unique. Everybody and their mother is using Yahoo as an example of analyzing data. Ridiculous. I do recommend creating for beginners to Pro. Just saying "Here we use Regression" isn't helpful to those who don't even know what that is ;) One last recommendation - be available for student questions. Most teachers think they're so amazing that what they said was perfectly clear (to themselves, of course) LOL. But most teachers don't remember what it was like when first learning and skip CRUCIAL pieces of information. Which all this is fine as long as the teacher is available to answer every question students have via email, Discord, etc. But NOT rely on other students answering peer students' questions. I never understood why teachers are OK with the blind leading the blind. HAHA Hoping you put this course together with these suggestions! I have my CC already in my hand LOL
It depends, if the price is great, definitely can promise atleast 10 people including me from my side. Something like this is really important, but the price point of each course already available is very high for India like countries.
Not banks, but brokers used to do that all the time. They still may be doing so. Basically they took opposite trades of their clients. 90% new traders lost 90% money within 90 days. Even AI can't be that precise. It was literally free money.
@@mitesh8utubeThe fact that 90% of traders lose their money doesn't mean you can win by putting the opposite trade because, like in the casino, there are commissions. 90% of casino gamblers also lose their money, but this doesn't mean that you can go to the casino yourself and beat the house by placing your bets on the opposite color than other gamblers
@@mitesh8utube that's right, what also seems to be true is that reverse trading isn't a good strategy. A hedgefund did some tests, theres a video up: "unveiling the truth: why reverse trading fails" channel name ukspreadbetting
A few questions 1. When is the ml ai course going to be released? 2. Like what's the main difference between the quant strategies and the ml ai models 3. So like which is the better one or which one is the ongoing hype 4. Which one is more effective as per you?
Thanks for the videos! I have math degree, and I have some experience in coding but it was years ago (C++). I trade now manually (as you mentioned using patterns etc). Now, I want to break into this field of automated trading based on machine learning or whatever technique. I just need some training in modern tools available now that you just described (programs that pull data together, code your idea, testing, link with my broker, etc). Could you create a simple course which explains how I can go from trading idea to its automation in a simple way just using my PC ?
My background was also in C++ back in 2007 period. Python will be a piece of cake for you . So i doubt it will be difficult for you to learn. The execution was the hard part in past few years. But now with quanctonnect and its integration with python and many famous brokers it has become easier. Because of your math background i really think you could pull off ml/ai trading much easier than many
Thanks for the video! AI in trading is, at to some degree already present for some years. Look at programs like NeuroShell and others who already did strategy optimization using genetic algorithms and backpropagation neural networks. However with singularity we are facing now, allows more powerful tools. I dont think manual traders will be obsolete, for instance I love to manually trade, to be in control and feel the flow with the market but, as trading is a bit like an art, it cen become quite indiviual and singular. I believe the main challenge for machines now is avoid those colossal drawdowns (20%, etc..) and survive long term in this hyper complex / fractalized maket.
Thanks for watching mate. I agree to your point. i personally enjoy doing manual execution as well. Especially in my value investing based trades. From a a pure personal joy it is extremely satisfying and thrilling to do manual trades but there in comes other issues like fear and FOMO and overconfidence etc.I think eventually as people learn more and it becomes more easier to implement these ml algo models easily for retail traders more will transition. 5 years ago i couldn't imagine doing all these ml models with so little code and at this price. As time goes by it will get cheaper and more convenient to implement algotrading with ml
Very inspiring video! Though I suspect that stellar results on the test set mostly come from the fact that you are exploiting the data leak. quantiles that you use to open positions are calculated using the future data (of the entire test set). In real trading, you don't know upfront what will the quantiles of the future predictions spanning years ahead be. Also, it's a good practice to turn any duplicated code into a procedure to reduce possible copy-paste errors and easу maintainability. Good candidate for that here is the featurization code. Also, best practice for combining preprocessing with models is to use pipelines. Plus, PCA needs prior data normalization, therefore, optimal way would be make_pipeline(StandardScaler(),PCA(),your_model). Another necessary step is comparing your model with a dummy model (DummyRegressor in this case) and checking whether your model is having a real gain over random guessing.
Thanks for watching. The testing data and training data are completely separate. Quantile calculations are also separate. I can confirm that there has been no data leaks. You can go through the code and run it yourself to confirm the same.
there is lots of alpha and ideas that can be generated but It is not practical for retail traders. like for instance the data can be expensive, level 2 data is like million of data in just 1 day. as in the order book the bids and offers keep on changing, then there is market orders . for all these data you will have storage issues. Now if you want train all this data for a ml model then youll need a lot of computational power. Now to obtain all this data alone will be significantly expensive let alone working with computational power and storage. Not to mention the work involved in just cleaning the data Some people have come with simple non ml models like see whether the bids is higher or offers is higher currently or buy market orders is higher rectly or even footprint charts and work form there discretional trading style. eg jigsaw trading. But from a ml/ai side its not that easy due to reasons above. I'm pretty sure many of the quant funds who are in high frequency trading/market making are doing it but they have a lot of financial power to buy all this systems and quants not to mention the tech infrastructure. For retail traders, i doubt there will be an easy solution to apply level 2 to ml and ai at least any time soon. Maybe in the next couple of years, if we do get any of these to be cheaper then yes
I have an SMC profitable strategy but I wanna learn how I can do it with AI trading. I hope that in your course you explain how to do with AI an SMC strategy where you have different parameters such as market structure, zone selection, entry models etc
Most algorithmic strategies apply to quantitative methods to extract data and form models such as: mean reversion, statistical arbitrage, trend following, momentum, unexpected events or economic data. SMC concepts are difficult to quantify due to the subjective nature therefore. “CISD” is simpler than say a market structure break because you’re essentially quantifying a candlestick pattern (e.g. low price of the day below previous day and price closes above - simple example).
@@sliealwhufcyes I know that the hard part is the subjectivity. I am a SMC trader myself, and for me I only like trading with price action not using indicators. Maybe with AI or reinforcement learning there is a way of teaching AI to use some subjectivity
Einstein was a physicist not a coder 🤣. Python is the easiest programming language out there. Anybody who puts in an effort can easily learn it. Yes and that too in just 1 week. This is not C++. And you dont have to oearn everything like web development or flask . All you need is th emsot important things neccesarry to get you started in trading and finance. However for people who dont put in the effort, I dont know,
That's the plan! So far we are done with strategy codes in pca, regression, decision tree and neural networks for the course. Will be adding more before the final release.
@@quantprogram Is there a way of adding a machine/deep learning into an existing algo strategy to improve it? Or doesnt work that way? I have heard that using neuronal networks and that stuff you could teach the AI to learn discretion and learn from mistakes. So I was wondering if that possbility really exists... Great stuff!
It does work that way. Ai/ml can be used to improve an existing strategy. Yes neural networks works are used to take in information for multiple days and find patterns and ember them. Especially ltcm. Machine learning or ai is basically learning from mistakes hence the word training.
If we were actually able to build something like this, could we not package it and sell it as a product? Or is this only for personal use? I love this though, I am avid about AI and ML and this looks like a fascinating project. I worry about my math and coding skills, but intend on using any and all resources available to pull this off. You are a great teacher. Thanks for making it understandable
You can. But every trader will want their own input variables. Here we used rsi. I might prefer using volume or obv or add or even another index or interest rates . I might also want training data to be different. Instead of pca some other statistical model could be used. There are many combination so then the package will be different
The course was officially released 3 years ago with 5 strategies. Now there is 13 which has been released during regular intervals including this one which was just released this week. Note that the course is not about the strategies. The goal of the course is to teach you skills to become a full time quantitative and algorithmic trader. The strategies serve as a foundation for you to create your own strategies which are even better than those. Please contact our email that can be found in the about page for further details
Could you create a video, on how to learn about algorithmic trading from your youtube channel/ other resources, whichever you feel is great. That actually can help us a lot , as beginners Mostly face this as the biggest issue. Also i feel that can increase the traffic.
It’s already there in our channel as showed in the video. It’s a 3 hour full tutorial on algorithmic trading in python created for beginners and another 1.30hr tutorial on quantconnect which focuses on execution.
It’s not a summary. It teaches the viewers the necessary coding skills in python and quantconnect. This should equip you to get the foundation to learn further. Algorithmic/quantitative trading field is ever evolving. There are new updates everyday. Now with ml/ai it’s changing everything. Summarizing won’t be enough
Two questions. 1. I know ML ,AI should i go for prometheus or wait for your ML,AI course 2. If i chose prometheus. All stuff is included in that or i have to do other two as well ? Know about all the things you told in the video and was on the way towards learning that but i think choosing your course will help me save time towards learning that as you’ve already gone through that stuff.
Answers below. 1. It’s up to you. If you purely want something that focuses on ml or ai based trading content. Then it’s better off to wait for the ai course. But if you want something that’s purely quantitative trading ie using normal quant models using optimization, forward test, Monte Carlo etc ie excluding ml/ai then Prometheus should suffice. This strategy codes is also included in prometheus. 2. If you choose Prometheus which I personally don’t think you should as you’re expecting ml or ai content, you will get a discount when the new course get released. The Prometheus wont contain the codes or contents of the new ml ai course when and if it gets released. Just one more thing as stated in the video. It is lot of work creating content especially ml content as we will have to break down the math into simple explainable terms just as we did in the video. So the time to create that course might take a while. There is no specific schedule decided.
@@quantprogram True that good things take time, but I request you to do it. anyway, thanks for your honest reply. I'll sign up for your Prometheus course until you work on it not because of the strategy or money but because of the knowledge I'll get, for a bigger all-round picture. Hope it meets the expectation 😊