Posts
Each post comes from a dated note I wrote at the time, not something put together afterwards. If a later post contradicts an earlier one, both stay up.
- My best backtest result was a bug, and fixing it reversed the answer
A limit I thought was holding my strategy back turned out to be the only thing protecting it. Someone found the bug by reading my code, and the corrected result pointed the other way.
- I set the pass mark first, then failed it twice in one day
Why writing down what counts as success before you look at the result is the only defence against your own reasoning, and what it costs to run tests you are allowed to fail.
- How I built survivorship-free NSE data from free bhavcopy files
Why most Indian stock datasets quietly flatter every backtest, how to rebuild ten years of history that does not, and the eight-month hole my own tests found in my own archive.
- My detector worked perfectly and was wrong for four days
My tick feed died at 15:15 for four days running. I blamed the broker and built a monitor. The market had changed shape underneath me, and the proof was already sitting in my own data.
- Eleven intraday ideas, and the cost floor none of them could clear
One of them was a real effect I still cannot trade. The best of them got to 0.97 of my cheapest cost. Here is what eleven intraday tests measured, and why more money would not have helped.
- One idea finally passed. Then I spent a week trying to break it.
After twenty-two failures something cleared the pass mark. Passing turned out to create more work than failing does, and I still would not call it good news yet.