Here are three questions I think this dataset can answer well. The figures below come from the profiling queries I ran, which covered 422,729 daily rows, 152 companies and tickers, and dates from 2012-01-02 to 2025-08-04. ## 1. Did the AI surge change the price level and risk profile of semiconductor stocks? - **Why it fits:** About 99,984 rows (23.7% of the data) fall after ChatGPT's launch. That leaves a long pre-AI baseline and a substantial post-AI window to compare. - **How to answer it:** Compare each company's growth in `close` and its average daily range `(high − low) / close` before and after late 2022. Then see which names moved most. - **Baseline:** Across the whole history the average daily range is about 3.69% of price. Post-launch ranges can be measured against that figure. - **Decision it serves:** Which stocks to overweight or underweight, and whether the AI re-rating is broad or concentrated in a few leaders. ## 2. Is the AI rally broad across the sector, or driven by a handful of companies? - **Why it fits:** With 152 companies over 13+ years, you can measure concentration directly. Examples are the share of sector gains from the top 10 names, and the dispersion of returns across companies before versus after 2023. - **Extra context:** Median close is about $35.57, so the universe spans many price levels, not just mega-caps. Returns should be normalised rather than compared in dollars. - **Decision it serves:** Whether to hold a whole-sector basket or pick leaders, and which companies belong in a benchmark. ## 3. When does trading volume signal unusual interest, and does it lead price moves? - **Why it fits:** Volume is extremely skewed. The average is about 8.47M shares, the median is only 898,500, and the maximum is about 3.69 billion. Raw volume is therefore a poor signal, and a per-company rolling baseline (z-score or ratio to median) is needed. - **Data caveat:** 4,888 rows have zero volume. These are likely halted days or illiquid listings and should be filtered or flagged before any analysis. - **How to answer it:** Flag days where volume is a multiple of a company's trailing median. Then test whether those days precede larger returns or ranges over the next few days. - **Decision it serves:** Timing entries and exits. The 4,888 zero-volume rows and the heavy skew in volume mean both the volume question (#3) and the price-range comparisons in #1 need cleaning and per-company normalisation before they can be trusted.