26 · Backtesting the Soul

There are certain indicators, certain signals, certain cues you learn to recognise. As you see new maths questions, you train yourself to think of certain "realisations" — the next blocks that come after the present conditions. They're not certain (that's the probabilistic aspect), but there are a few possible progressions from any position. Just like the alpha that initiates certain buys and sells. Your focus should be knowing the cues (data collection), which continuation is most probable (the statistical tails, the Monte Carlo simulations in pragmatic application), and how to make the buy and sell decisions at the best time.

This is just like mathematical problem solving. You react by writing certain formulas when you see a certain expression — that's the metacognitive thought-process analysis, almost an enlightenment, that I had during my gap year after high school. It genuinely felt like my head was cracking open — a spiritual eureka phase where my mathematical skill took a leap over two weeks, along with emotional euphoria. Quantum leaps usually come after potential energy, latent energy, piles up.

You do this self-deconstruction and rebuilding all on your own. Even as you collect other people's semantics, heuristics and mnemonics, you practice their preaching until it corroborates into yourself. This is the psychoanalytical process necessary for exponential mastery in a radically shorter time period, in any endeavour. The counter to it is emotional burnout without creating anything — just hoping time does something for you. Seeing this play out in the market: volume spike plus price consolidation plus a news event means breakout probability increased. Certain conditions fusing into a certain probable "realisation" of that stock. Intuition comes from exposure to more data, but being able to surgically decompose and engineer your own thought process is something magnificent.

I talk about this a lot, but the ascent to mastery in mathematics is so similar to trading. In maths, you look at the thought process that led to your decision, then at the model answer's thought process. If you want to be great at competitive maths — the Korean CSAT, A-levels, olympiads, anywhere there's a time attack and pressure to perform — it's hard to be creative in the moment. But when you backtest your mathematical thought process — if you truly master metacognition — you can recall every link in the chain of thought that led to an outcome. Compare it with the model answer's chain, see which thought process should be etched in, then build heuristics and mnemonics to reinforce that reaction circuit. This is literally trading. Emulate that iterative process — machine learning leveraged into algorithmic trading — boom, done. The ability to self-study abstract subjects like maths and physics isn't just about formulas; the tacit thought process itself is the asset. Malleable metacognition. Not many people discuss this skillset. One of my personal alphas.

Decision logs: a trading decision log, or trading journal, is a detailed record of a trader's activity used to analyse performance, enforce discipline, and refine strategy. It tracks not only the technical details — entry price, profit and loss — but the rationale, the emotions, and the market context behind each decision, everything that culminated in that moment. What every master does is simple — but that simplicity has massive work ethic behind it.

Using the algorithmic thinking frameworks I've tacitly developed through maths — beyond deep thinking and scenario manifestation — this, along with backtesting, is literally how you train models. The model answer is the backtest. What decisions would have maximised the desired outcome, what context manifested, and what reaction cues should have been caught. Think harder and rarer: draw connections between fields so absurd that if you saw someone else leverage them, you'd think they were crazy. I've seen tarot traders make millions a month. Use what you're interested in — quantum computing, algebraic topology, colour coordination in fashion, entomology. There are model answers for fashion; there has to be a way to interpolate an algorithm that generates the output. For qualitative data it's possible — why would it be different for the quantitative data of trading?