CPI drops 8:30 AM tomorrow. FinTwit is already positioned. I ran 273 CPI releases over 23 years.
Main finding: CPI day is statistically indistinguishable from any other day. Return difference vs non-CPI days: +0.04% (p=0.64). Volatility difference: none (p=0.74). All that positioning is for a random Tuesday.
What actually matters: the direction at the open predicts the close 75% of the time. By 10:30 AM it rises to 81%. By 2:30 PM it reaches 98.5%. The market prices CPI in the first hour. Everything after that is noise. Almost no intraday reversal.
Hot CPI days drift lower all day and close -0.96% on average. Cool CPI days drift higher and close +0.88%. The paths diverge at the open and rarely converge.
The one edge that exists: when the market sells off on CPI, $SPY is higher 20 days later 67% of the time, with an average gain of +1.00% (p=0.015). That includes the seven releases since April. Adding them made the signal slightly stronger — it was +0.88% with a 66% win rate before.
If tomorrow's print is hot and the market sells off, the data favors buying the weakness for a 20-day hold.
Three rules for tomorrow:
1. Before 9:30: wait for the print
2. 9:30 to 10:00: direction is largely set
3. After 10:30: very little new information arrives
$SPY $QQQ $IWM
Main finding: CPI day is statistically indistinguishable from any other day. Return difference vs non-CPI days: +0.04% (p=0.64). Volatility difference: none (p=0.74). All that positioning is for a random Tuesday.
What actually matters: the direction at the open predicts the close 75% of the time. By 10:30 AM it rises to 81%. By 2:30 PM it reaches 98.5%. The market prices CPI in the first hour. Everything after that is noise. Almost no intraday reversal.
Hot CPI days drift lower all day and close -0.96% on average. Cool CPI days drift higher and close +0.88%. The paths diverge at the open and rarely converge.
The one edge that exists: when the market sells off on CPI, $SPY is higher 20 days later 67% of the time, with an average gain of +1.00% (p=0.015). That includes the seven releases since April. Adding them made the signal slightly stronger — it was +0.88% with a 66% win rate before.
If tomorrow's print is hot and the market sells off, the data favors buying the weakness for a 20-day hold.
Three rules for tomorrow:
1. Before 9:30: wait for the print
2. 9:30 to 10:00: direction is largely set
3. After 10:30: very little new information arrives
$SPY $QQQ $IWM