Ten correlated tickers are about two bets

The measurement

Across the tickers our scanner sold puts on, the mean pairwise correlation of daily returns was 0.46. Some pairs were effectively one asset: TQQQ and QLD correlated 1.00 with QQQ, UPRO 0.89, NVDL 0.72, TNA 0.66.

A standard way to turn a correlation into a count is n divided by (1 plus (n minus 1) times the average correlation). With that correlation, a week of positions works out to about 2.0 independent bets. A later read on a 14-ticker book, using 90-day correlations of 0.58, came out at about 1.6, and a 9-ticker watchlist at 0.56 was also about 1.6.

Why it matters for a short-premium book

When positions are this correlated, a loss week is the whole book. Every put you sold that week is drawing on the same market move, so the 15 rows on your screen behave like two trades with fifteen times the paperwork. The variance of the outcome is set by the number of independent draws, not the number of tickers.

This is why a second position in QQQ after TQQQ is not a different trade. A different ticker on the same factor adds size to the same bet. It also explains why piling conditions onto entry rules did not help in our tests: every filtered subset was still drawing from the same Nasdaq factor.

What actually adds a bet

A position adds an independent bet only if it moves differently. In our 2016 to 2026 put credit spread replay, GLD correlated 0.19 with QQQ by monthly return, USO 0.11, TLT 0.26, while SPY was 0.87 and IWM 0.63. Low correlation is necessary but not enough: the second leg still has to earn something on its own, and of those only gold improved a combined book in our test. USO and TLT made it worse, and IWM was too correlated to help.

The combined QQQ plus GLD spread book had a Sharpe ratio of 1.22 against 0.99 for QQQ alone, and at equal drawdown odds it earned about 30% more. Gold's own interval still included zero, and four combinations were tried, so treat that as a lead rather than a finding.

What it changes in practice

Before calling a book diversified, compute its correlation and its effective bet count. Group positions by the factor they actually trade (a Nasdaq fund and its 2x and 3x versions are one factor) and cap the loss per factor, not per ticker. Our lesson on drawdown odds gives the fractions of equity at risk that kept a 20% drawdown unlikely; those apply per factor, because a factor is what gaps.