Here are the three questions I think this dataset can answer best. The figures below come from the profiling queries I ran. I haven't yet run the analyses themselves. **What the data looks like** - It has 422,729 daily rows covering 152 tickers, from 2012-01-02 to 2025-08-04. - 99,984 rows (23.7%) fall in the post-ChatGPT period, so the AI-era window has enough depth to compare against the roughly 10 years before it. - The average daily high-low range is 3.69% of price. - Volume is heavily skewed. The median is 898,500 shares, the mean is about 8.47M, and the maximum is about 3.69B. - 4,888 rows have zero volume, which points to thinly traded or delisted-style tickers. ### 1. Which semiconductor stocks re-rated most after the AI surge, and was it broad or concentrated? - **Method:** compare each company's price level and return before and after the ChatGPT date. Rank the companies and measure how much of the sector's gain comes from the top few names. - **Why it's interesting:** the data can show whether "the AI trade" lifted the whole sector or a handful of leaders. That bears directly on overweight/underweight decisions and on what belongs in a sector basket. - **Data needed:** `close`, `date` and `company_name`. The gold tables `semi_conductor_se_by_company_name_date__yyyy_mm` and `semi_conductor_se_by_company_name_date__yyyy` should cover most of it. ### 2. Did the AI era change risk and trading activity, and did it change it equally for every company? - **Method:** compare the daily range `(high − low) / close` and volume per company before and after late 2022. Then test whether the change is a sector-wide regime shift or driven by specific tickers. - **Why it's interesting:** a 3.69% average range is the baseline for risk sizing. If the post-2022 range or volume is materially different, position sizing built on the full 2012–2025 history would be mis-calibrated. - **Caveat:** use medians or per-company normalisation, because the volume skew (median 0.9M against mean 8.5M) would otherwise let a few mega-volume tickers dominate. ### 3. When does unusual volume carry information? - **Method:** flag days where volume is far above a company's own rolling baseline. Exclude the 4,888 zero-volume rows. Then test whether those spikes precede larger moves or a change in the price trend over the following days or weeks. - **Why it's interesting:** it tests the "volume signals unusual interest" idea for entry and exit timing on real data. It also shows whether the signal behaves differently before and after the AI surge. - **Caveat:** each stock needs its own baseline, because volume scales differ so widely across the 152 tickers. **Data notes** - The card lists about 149 distinct `company_name` values, but profiling shows 152 companies and 152 tickers. Check for name variants before any per-company comparison. - The data is observational. It can show association and regime changes, but not causal effects of AI. I can run any of these next. Question 1 is the quickest, since it is covered by the precomputed gold tables.