2020 | This idea was already there just after my graduation. COVID wasn’t exactly the best time to find a graduate job. While hunting for opportunity, I spent my free time experimenting with new ideas I never had time during Uni.
Pazusoba was born during this phrase, aloneside maintaining WoWs Info and probably a few other apps, while learning native Swift and Kotlin development. By August 2020, I landed my current role and started my engineering journey.
Aug 2020 - Aug 2022 | Those first two years of work was extremely challenging. Everything was done by hands, without the assistant of LLMs. The steep learning curve left me with little energy for side projects. However, I did created this entire project somewhere in 2022, but back then, I figured I wouldn’t be able to do much with it until 2027 or 2028.
Late 2022 | ChatGPT arrived. The game changed forever.
Early 2023 | I entered the US market, and my trading journey officially began.
The time difference between Sydney and New York is honestly the worse. Staying up all night to trade live wasn’t sustainable, so automation became a necessity rather than a choice. You could say living in Sydney forced this project.
June 2023 | I hit my major success with options trading. Options taught me so much about market mechanics. The concept felt magical, limited loss & infinite profit. You can still see that influence in my original system design, asymmetry is the core philosophy.
August 2023 | Everything fell apart.
Deep reflection forced me to realise my approach was flawed. How many NVIDIA level stocks could I spot? As a solo retail trader, my reach is limited. There was no way I could hunt down a new NVIDIA every year to maintain consistent profits.
I have been very serious about trading since day one.I never wanted a single lucky lottery ticket, but a consistent compounding engine. For a retail trader, the standard 10 - 20% return is too slow for the goals I am pursuing. I needed something different.
2023 - 2025 | From late 2023 through June 2025, I was on a relentless experimental journey to find a path toward my ultimate dream. With the help from ChatGPT and GitHub Copilot, I iterated through countless failures, but in a new domain, any feedback, either positive or negative, is data.
Ultimately, I abandoned deep learning for now. I fell back on hand crafting features using pure, traditional machine learning. Old school? Maybe. But they do work, and they are explainable, not black boxes.
June 2025 - October 2025 | The first version training was over, and the initial results were mixed. I threw the algorithm into a paper trading account, but it was all over the place. Sometimes, it was legendary, but then a miserable failure follows. It required a massive amount of debugging and manual tuning.
November 2025 | After my paper account randomly got deleted, I moved straight to live trading. I didn’t want to wait anymore. The performance below could tell you all the story.
March 2026 | It took almost four months before the system finally started printing some profits. I was quite glad to see the hard work paying off, but I was NAIVE. The real challenge just started.
May 2026 | This month, the performance decayed significantly. I messed around with the parameters, and a series of heavy drawdowns had me genuinely worried.
Then, last Saturday, something clicked. It was funny how the brain works… I suddenly remembered a specific tool I had heard about back in 2020 while building Pazusoba. I don’t even know how the name stayed so clear in my memory, but it was the missing piece of the puzzle.
Originally, I had tried reinforcement learning, but it gave me unstable, mixed results. The black box nature of deep learning is a massive liability. My latest framework remains firmly algorithm focused and machine learning based. It doesn’t predict the future, it simply optimises my core trading strategy.
The new backtests look amazing, but then again, every backtest looks amazing. How does it work in live though? Nobody knows. History says it will probably break. This is the same for me, so I can really only wait until the end of this year to see how good it goes.
Back in the day, the project was originally called NAIVE. Looking back, thi was honestly a pergect reflection of where I was at. As the system evolved, I shifted to Funny Signal as a temporary placeholder. At the time, it was indeed just a simple signal. In the end, I dropped the “Signal” part and just called it FUNNY. It really fits well, and it is also how I see everything now, absurd and funny.
This past year has honestly felt like a Tragic Hero Arc. It was packed with challnges, failures, mistakes, and setbacks that made me question everything beyond just system trading. Like what I wrote in 《悲情英雄者》, the greatest enermy is always deeply within.
Anyway, my number one goal is finally done, I really need to focus on other things now.