Support Board
Date/Time: Mon, 11 May 2026 19:24:40 +0000
Post From: Using AI in conjunction with Sierra Chart.
| [2026-05-09 18:18:20] |
| User61168 - Posts: 479 |
|
Frankly, in these last few months, I have become a big fan of AI. Sharing my journey so others (non technical users) can benefit. I switched from using just ChatGPT as a chit chat tool to Codex and Claude Code (still a bit concerned with using desktop or cowork). Historically I have being with Sierra since Oct 2018 when I moved to futures. With over 1000+ algos created and backtested across 300 Sim accounts and across all Symbols, it was impossible to keep track of performant algos with over 256K files generated in my SC folder. I was shocked to see that many trade logs lol With this background, I used codex and with 30 mins (with a little bit of back and forth) it scanned all the trade logs (I made sure to exclude my live account trade logs), generated top 20 performing trade stats based on my success criteria, drew equity curves and additional metrics that statistics tab does not support etc all into a single excel file. This was phase 1. Impressed with its capability, moved to phase2 Phase2: I moved to ASCIL. Within a few days, I could create dozens of advanced strategies mostly the things I could not achieve with simple alert trading systems. To achieve efficiency, I tagged codex with Claude to “peer review” generated code. I was fascinated with the issues each of the two were able to find and fix on their own before I could test myself. Long story short, my experience compared to early last year improved 100 fold. Last year, I could not even compile the damn code and now it knows my version of Sierra and generates error free code (still takes few iterations but gets it right). Bottom line: this has been a blessing for strategy development. Phase3: OHLC optimization. Wanted to see if it can scan ohlc data and generate some positive expectancy strategies on its own. It did! I used 2026 data, ran on 2025 and 2024 with random start/end dates and most were positive net pnl. Not perfect but not too shabby either. Phase4: this is where I am currently am. Read tick data from .scid binary file and build hft performant algos. It did! All phases required approx 20-30 interactions (back and forth), some caused frustrations mostly because I was burning tokens and dealing with limits reached etc adding delays to my journey but having started my AI journey on 3/1 and considering where i am and what I have accomplished for just $20/month each for OpenAI and Anthropic, I am hooked. Tons of ideas flooding my brain cells atm but I will stop here. With a bit of creativity and curiosity, anything is possible for a “normal” non-technical user eliminating 90% of the need for ASCIL developers and system sellers for simple stuff. Disclaimer: I did ran into some funky business with one ASCIL study which crashed my laptop once so use AI with care and at your own risk. I also would not use AI on my general purpose/private/personal laptop due to security risks. |