Predicting Price using Lorentzian Classification with Justin Dehorty - Trading Bot Sessions (EP 010)
HTML-код
- Опубликовано: 10 июн 2024
- This is episode 10 of Trading Bot Sessions where we talk with traders who have built automated trading strategies and trade everything from stocks and options to futures and crypto. In this session, I speak with Justin Dehorty, a software engineer with a background in computational chemistry and biophysics.
Most recently, Justin wrote an open source machine learning indicator that was chosen as an editor's pick on TradingView.
The indicator seamlessly merges key concepts from Lorentzian space, a component of general relativity, with a nearest neighbors algorithm to predict future prices in markets.
We cover Justin's work in Pine Script, as well as the application of machine learning in technical analysis, how kernel regressions are an overlooked and powerful alternative to moving averages, and the limitations and future of Pine Script.
JUSTIN'S LINKS:
Patreon: www.patreon.com/user/about?u=...
Discord: / discord
Twitter: / justindehorty
LinkedIn: / justin-dehorty
TradingView: www.tradingview.com/u/jdehorty/
THE LORENTZIAN CLASSIFICATION INDICATOR:
• Lorentzian Classificat...
www.tradingview.com/script/Wh...
KERNEL LIBRARY:
www.tradingview.com/script/e0...
www.tradingview.com/script/AW...
_________
✅ Automate Your Strategy and Indicator Signals from TradingView and TrendSpider and try TradersPost free for 30 days.
traderspost.io/pricing
_________
💬 Join our Discord Channel for updates on automated trading on TradingView and TrendSpider. Or find us on social.
traderspost.io/community
00:00:54 Intro
00:01:25 Justin's Background in Biophysics and Computational Chemistry
00:05:08 A Brief Intro to Lorentzian Distance
00:07:59 Connecting with His Mom - A CPA and Investor
00:09:31 Using Machine Learning and LLMs for Other Investment Ideas
00:14:09 Intro to the Machine Learning Lorentzian Classification Indicator
00:24:37 Interpreting the Win Rate and Other Metrics
00:25:46 Using the Backtest Adapter for More Comprehensive Testing
00:26:35 Experimenting with Other Indicators
00:30:18 Releasing the Next Version v1.5 and an Open Source Philosophy
00:30:56 How Machine Learning can Impact Technical Analysis
00:32:25 Using Kernel Regression over Moving Averages
00:38:31 Limitations of Pine Script and Exploiting the Built-in For Loop
00:41:40 The Future of Pine Script
00:45:35 Testing the Lorentzian Classification Strategy
00:48:49 Using Lorentzian Classification as a Trend-Following Strategy
00:55:09 What's Next for Justin
00:56:52 Getting in Touch with Justin
DISCLAIMER:
Trading in the financial markets involves a risk of loss. Content and strategies produced by TradersPost are for informational or educational purposes only and do not constitute trading or investment recommendations or advice.
It's beyond sad that a channel like yours has 1k subs and all those garbage 100x guaranteed trading channels get millions. Subscribed, keep up the good work lad.
Seriously. Videos titled "THIS AI TRADING INDICATOR PROVIDES 100% ACCURATE SIGNALS" get 100k views, and this dude breaking down how that indicator actually works has under 1000. It's sad and I'm glad I found this channel after watching one of those videos.
What a great time hearing Justin! Thanks TradersPost
Our pleasure!
The indicator and TradersPost has given me true passive income. Thank you so much for the both of you. You have changed my life.
appreciate you sharing your experience. do you mind sharing the timeframe and setting you found useful?
Thank you! You are welcome!
Amazing stuff!! Beyond my mind but so greatly appreciated. AWESOME young man. What a future!! WOW
Great interview. Thanks for sharing.
Glad you enjoyed it!
Okay, ...Justin in 10X smarter than me
Me too 😂
1000X more than me
He’s a smart one!
Amazing content. Wish you more subscribers for the authentic work. Justin you are just brilliant
Thank you!
Awesome interview! Following both of you.
Awesome! Thank you!
Great indicator and great interview about it. Thanks :)
Glad you enjoyed it!
Tremendously amazing💖
Thanks!
2 months ago, Justin mentioned that thought process around the 12 minute mark, which occurred for me as well. Although the difference is, playing with it, and not. Hopefully you both find some interesting stuff. I'll have to keep an eye out for the research that's been done.
woahh transcribing the news and correlating it to a potential trade sounds super promising.
Right?
wow great stuff
Glad you liked it!
Apologies if this is a dumb question, but is it possible to use the signals on the LDC in conjunction with webhooks for traderspost? Great interview, Dehorty's work here is amazing and greatly appreciated!
Not a dumb question at all. Using Justin's Backtest Adapter (www.tradingview.com/script/Pu38F2pB-Backtest-Adapter/), you can add alerts to the strategy and connect them to TradersPost to make it fully automated.
@@TradersPost thanks!
Yes
Which are the best time frames for this indicator
Does the Lorentzian indicator repaint ???
Hi Guys. Great video I truly enjoyed it. I would greatly appreciate any ideas on an AI screener I can use in junction with the program for day trading stocks. Any recommendations?
Have you tried to build something?
Is there a way the kernel can be used as a scanner..I have been experimenting with Justin's Lorentzian classification indicator for quite sometime and its really wonderful. Just wanted a way to identify potential trades by using it as a scanner.
Unfortunately, screeners on TradingView only work with the built-in indicators they provide, so Justin's indicator is not available for screeners. You would need to set up the indicator alerts on multiple symbols that are already pre-determined or pre-screened.
I overlaid this on 20 ema and it moves very identically!!
Hello, Is there a pine script back tester strategy already written for this?? Where can I get a copy so that I can backtest the settings???
Have a look at Justin's backtest adapter: www.tradingview.com/script/Pu38F2pB-Backtest-Adapter/
Tried it for a while - the indicator is just not that accurate
Have you tried to code your own pinescript strategies yet?
Is it repaint?
Pinescript uses token based system and compiles into Java on the backend.
Interesting! I wouldn't have guessed Java.
@@TradersPost It's certainly not the fastest way or most scalable but currently it works. Pine will need to really step up their hardware game if they seriously want more technical ML based systems to be adapted in their platform though. Because even after certain optimizations you will start hitting the brick wall of loop computation limits or memory exhaustion. Although memory part mostly solved with dumping everything to libraries but computation still requires serialized tokenized system.
Oh btw I have coded more than maybe 200 private or public strategies with very complex logic along with some statistical approaches and after a while like Justin said here it's extremely hard to go forward with complexity even if you optimize the shit out of your code because you don't have access to their compiler or hardware. Because you are using their data you are locked to their closed backend which is understandable. Pine and TV is very convenient for moderately complex stuff for now but you can't go haywire on full DL or ML for the moment. We will see what will happen in the future though.
I was not expecting Java either.
I wouldn't have guessed Java either, rather maybe LLVM. Then again, Java does make sense. Java bytecode is quite flexible, and JVMs are very very optimized execution engines. And you basically get safe execution for free.
Interesting. However not sure I got what to do with it
easy just use it to find the best place to yolo your life savings into 0DTE OTM call options
Start experimenting with pinescript strategies and backtest different ideas!
@@TradersPost that is obvious. However it is to abstract for me personally to do something with it. Anyway any new idea is welcomed. Maybe other can find inspiration with it
34:50 🔫🤫