It's his mother AI era, and quantitative trading is no longer an institutional privilege. Ordinary people like to eat meat! I have been spending most of my time on Vibe coding recently. My biggest gain is not how many useful tools or products I have made, nor how many ways I have found to make a fortune. The real gain was that this process allowed me to quickly understand and master many skills that I had never dared to imagine before, even to the point where many things could be quickly replicated. In other words, AI has given each of us an extra pair of writing wheel eyes. Taking the cryptocurrency market that we have been deeply cultivating for many years as an example, CZ was able to build the leading CEX in 2017, which was an extremely scarce ability at that time. Nowadays, it is not difficult for ordinary people to work with AI tools to achieve success. Not only that, but also the most profitable sectors in the cryptocurrency industry, including exchanges, stablecoins, contracts, quanta, and market makers. As long as you have a real idea, you can go out and do it. What we're going to talk about today is quantification. If this thing is done well, it can really make a lot of money. DeepSeek, the first AI in China, also earned its startup capital quantitatively. So the question is no longer 'can ordinary people touch quantification'. The question is: Do you know where to start. 1. First, turn your brain around: Quantify first and learn to calculate odds When many people hear about quantification, their first thoughts are still high numbers, code, institutions, and geniuses. Actually, the bottom layer is not that mysterious. First, understand four things: Probability. What percentage of things will happen. Expected value EV. Is it profitable to repeat this transaction for a long time. Variance. Will your account be thrown away before you wait for 'long-term profits'. Kelly's formula. How large should the position be. Many people die not in the wrong direction. It died due to a large position. The previous judgment was correct many times, and the last time I withdrew, I vomited everything back. One step forward is Bayesian. Speaking of human language, it means new news has arrived. How do you update your original judgment. The most valuable thing in the market is never who has an opinion. Is it possible for you to change probabilities faster and more accurately than others. So quantification is the first lesson, not the model. First, replace the gambler's mind. Don't keep asking whether it's going up or not. First, let's ask if the odds are correct, where is my advantage, and how should I place my position. 2. The second step is the most crucial: first learn to kick out the noise The most common mistake beginners make is to mistake luck for ability. After backtesting, I immediately felt like I had found the Holy Grail. The truth is usually hard to hear. If you test 100 strategies, there will always be a few that look particularly beautiful. This is not called alpha. This is called a sample that perfectly matches you. So the second layer must be supplemented with statistics. Its biggest function is to specifically slap the face. You should at least understand these things: Hypothesis testing. Is this result something real or purely luck. Multiple comparison issues. The more you measure, the more false signals there will be. Many people believe that their top 10 strategies are invincible, but they are completely useless once they are implemented. Regression analysis. Used to split the source of income. Is the money you earn really due to excess returns or is the overall market rising and you just drifting along. Maximum likelihood estimation (MLE). Essentially, it is a reverse inference. Seeing prices and fluctuations, infer the most likely parameters behind them. After completing this level, you will be severely awakened. Many previously thought 'abilities' are actually just sample period rewards. But this step must be taken. If the noise is not kicked off, learning anything later will be in vain. 3. Truly starting to act like a broad-based customer: Don't just focus on one transaction anymore Quantification really widens the gap, it's when you start looking at the system. You can't just focus on one coin, one candlestick, or one position. You need to start looking at combinations, correlations, and overall risks. Linear algebra will be used here. The most important thing is called the covariance matrix. Simply put, it's a relationship diagram of 'who and whom often move together'. You do 10 markets, not 10 independent risks. They will rise together, fall together, and fuck you hard together. Further ahead is PCA, principal component analysis. What it does is very practical: Compress a bunch of chaotic fluctuations into a few real driving forces. You think you're looking at 300 coins. Many times, there are only a few factors that truly determine the market trend: overall market risk appetite, liquidity, interest rates, and the popularity of a certain track. Next up is optimization. The meaning is not complicated either. Here are the risks, returns, costs, and position limitations, how to find the most reasonable combination. At this point, the way you conduct transactions will completely change. Previously relied on intuition. It's like doing engineering behind. 4. How do ordinary people start: Don't touch the most beautiful one as soon as you come up AI has lowered the threshold for tools. But the advantage threshold is still there. Code can be written by AI. There are a bunch of backtesting frameworks online. The data interface is also cheaper than before. There are still three things that are truly valuable: Unique data Others cannot obtain it, or even if they obtain it, they will not use it. Unique model You know exactly what you are estimating, where the boundaries lie, and where the mistakes lie. Unique Execution You are faster, more stable, and better able to handle complex situations than others. So the best route for ordinary people is not to immediately engage in reinforcement learning, complex agents, or neural networks. First, get the order right: Phase 1: Probability Theory Conditional probability, Bayesian, expected value, variance, Kelly. Phase 2: Statistics Significance, regression, testing MLE、 Overfitting. Phase 3: Linear Algebra+Optimization Matrix, covariance PCA、 Combination configuration. Stage Four: Stochastic Calculus and Derivatives Brownian motion, It ô lemma Black-Scholes、Greeks。 It ô Lemma can be understood as a chain rule in a random world. Greece is about breaking down option risk into how price, time, and volatility affect positions. Stage 5: Engineering Practice Python、 Data cleaning, backtesting, risk control, execution, logging, monitoring. Strategies can run, mistakes can be identified, and positions can be managed, all of which are more valuable than inspiration.