LessonThis is a lesson restated in new words. Any figure in it is an example from the source, not this desk's record.
Quantitative investing, in this lecture, is not a high-speed trading book. It is a way to hold equities so that firm-specific risk and industry risk are diversified away and market risk remains. Market risk is the piece you cannot remove without also removing the reason to own stocks instead of a Treasury bill. In the simulation, 30 names each had a 7 percent drift and 20 percent volatility. An equal-weight portfolio realized about that 7 percent. A second draw in which the market drift was negative realized about a 7 percent loss. Diversification changed which risks you held. It did not decide the sign of the market.
Why it matters
The lecture sorts equity risk into three buckets. Idiosyncratic risk is the firm: a new technology, a single product, a regulation aimed at one company. Industry risk hits a whole sector, such as health care, technology, or utilities. Market risk is the common piece, and the lecture points at GDP, inflation, and interest rates as the climate around it. One stock, the example is Nvidia, holds all three, and not in equal shares. A good year can be large. A bad year for that firm or that sector is also large. Stock picking is the game the lecture says it is not playing.
If every risk were hedged away, the finance argument the lecture uses is that you would earn a riskless rate, the Treasury case. The strategy keeps the market on purpose.
The idea in plain words
Thirty price paths start at 100. The average drift is 7 percent, so a naive endpoint is 107, but individual paths finish above and below because volatility, here standing in for all three risks plus a correlation the lecture does not dwell on, shocks each name. The equal-weight book is the average of those paths. With 30,000 dollars and 30 names, the lecture's intended split is an equal dollar amount in each. One badly garbled sentence says 11,000 dollars. Thirty thousand divided by 30 is 1,000, and "equal weight" is the instruction to trust. The worst name does not drag the book the way it would if it were the whole account. The best name does not lift it that way either. Run the same engine with a negative market drift and some paths still finish up, on the order of 20 percent in the lecture's glance at a few colored lines. The equal-weight result is still about minus 7 percent. The green path's good year was idiosyncratic or sector luck, and the portfolio was built to give that luck away. Correlation is the selection problem hiding under "just buy 30 stocks." The lecture wants low or negative pairwise correlation so names do not move as one cluster, while still sharing the market drift. A simulated pair frozen at correlation 0.5 moves together at a damped scale. A time-varying pair averages near 0.7, spikes toward 0.8, and dips toward 0.6. Another pair's correlation trends even while the prices diverge. The coefficient depends on whether you use daily, monthly, or annual returns. Public charts in the lecture make that concrete. Johnson & Johnson against Chipotle, picked as unrelated businesses, showed an annual correlation of about 0.01, a 12-month rolling average near 0.13, and a daily-window average near 0.17.
How to use it
Pick names across sectors, look at average pairwise correlation and at beta to the market, and know which window you used. A daily correlation and an annual correlation answered different questions for the same two tickers. Rebalance when one winner has silently raised your firm-specific bet. Keep the market exposure if the goal is the market drift. Move to Treasuries if the goal is to stop taking that exposure. The lecture says a long history of positive U.S. equity drift is why people show up for the market piece. It also says the diversified book loses when that piece is negative. Historical interest is not a forecast.
What usually goes wrong
Treating a single-name success as the strategy's success. The simulation's job was to throw that success away. Treating a correlation of 0.01 from annual returns as a permanent fact, after the 60-day window printed 0.735. Forgetting to rebalance after one name doubles. And reading "market risk only" as "the portfolio will be up." The minus 7 percent draw is the counterexample, produced by the same equal-weight rule.
Questions people ask
What did the 30-stock simulation assume?
Each name started at 100, with a 7 percent drift and 20 percent annualized volatility. The equal-weight average landed near that drift, up in one scenario and about 7 percent down in the other.
Which real pair looked uncorrelated?
Johnson & Johnson and Chipotle, at about 0.01 on annual returns. Shorter windows were higher, and a 60-day stretch reached roughly 0.735. Apple and Amazon were about 0.52 annually and 0.31 monthly.
Is beta required?
The lecture mentions portfolio beta as another look at market exposure and postpones the capital-asset pricing model. Pairwise correlation and rebalancing are the tools it actually uses.