Here are the three questions I think this dataset is best placed to answer. The profile queries I ran show what it can support. **What the profile shows:** 422,729 daily rows covering 152 companies, each with one ticker, from 2012-01-02 to 2025-08-04. The median close is about $35.57 and the average daily volume is about 8.47M shares. The average daily high-low range is about 3.7% of price. 99,984 rows (about 24%) fall on or after ChatGPT's launch, which gives a large post-AI-surge sample. 4,888 rows have zero volume, so those need filtering or separate treatment in any volume analysis. ## 1. Did the AI surge change who wins in semiconductors, or did it lift everything? - **Question:** Compare each company's cumulative return and price level before and after late 2022, and check whether the gains are concentrated in a few names or spread across the sector. - **Why it's interesting:** It separates real AI beneficiaries from sector beta. - **Decision it serves:** Which stocks to overweight or underweight, and what belongs in a semiconductor basket. - **How to answer:** Use `company_name`, `date` and `close`, with the gold tables by company and year or month for the trend. ## 2. Has risk changed, and where is volatility concentrated? - **Question:** Has the daily high-low range (`(high-low)/close`, averaging about 3.7% overall) widened since the AI surge? Which companies carry disproportionate risk? - **Why it's interesting:** Strong returns often come with higher volatility, and that matters for position sizing. - **Decision it serves:** Risk-adjusted sizing and comparing companies on volatility. - **How to answer:** Compute the range ratio by company and period, and rank the companies before and after the surge. ## 3. When does unusual trading volume signal a real move? - **Question:** Do volume spikes relative to each stock's own rolling average (`volume`) predict subsequent price moves or volatility? Is that relationship different in the AI era? - **Why it's interesting:** It tests whether volume is actually a usable timing signal. - **Decision it serves:** Entry and exit timing. - **Data caveat:** The 4,888 zero-volume rows could produce false "dry-up" signals, so exclude them or check whether they cluster in specific tickers or dates. **Limits to keep in mind:** This is observational price data only. It has no fundamentals, sector labels or exchange field, so comparisons by exchange or sub-sector would need outside data. It also can't show that the AI surge caused any change, only that a change coincided with it. I can run any of these next, starting with question 1 using the gold tables by company and year.