Zero Fills, One Lesson: How Long to Trust a Sharpe Ratio

Nothing moved. The swing account holds thin slices of three ETFs bought on past signals - EWY (South Korea), MTUM (big US momentum names), SPY (the whole US market) - and today it didn't buy or sell a single share. Zero fills, zero dollars traded. The account still ticked up two dollars anyway, because that's what holding things during a green day does.

The reason it's fully invested instead of sitting in cash: SPY is trading well above its own 200-day average price, which the bot reads as 'the tide is coming in.' When that's true, the strategy stays close to fully invested. When the tide goes out - SPY drops below that long-run average - it's supposed to pull back toward cash instead of riding a falling market down. Today's read: tide's in, so it stayed in.

The honest number: the book sits at $979, funded with $1,000 back in June, so it's down $21 since it started - and on a risk-adjusted basis it's currently trailing SPY by about 3.8 percentage points of alpha over its short real-money life so far. That's not a good look and we're not dressing it up.

Reddit posted revenue up 61% year over year and profit that nearly tripled - and the stock dropped more than 20% anyway. The reason: the CEO said search referrals got 'choppy,' because Google's AI now answers people's questions right on the results page instead of sending them to click through to Reddit. The kicker - Reddit is one of the companies selling Google the data that trains those very answers. Good quarter, worse story.

Meanwhile Apple hit a fresh all-time high market cap while spending just 1.8% of its revenue building AI infrastructure. Alphabet is spending 37.5% of its revenue on the same thing and posted negative free cash flow doing it. Right now the market is paying up for the company renting AI capability instead of the one building the factory - a strange incentive if you're the one holding the shovel.

And because markets love an exclamation point: South Korea's Kospi index just had its best single day on record, up 16.5%, with Samsung and SK Hynix both up more than 24% - three days after a 17% selloff. Same underlying AI story, whiplashed in both directions inside a single week.

A Sharpe ratio measures return per unit of risk taken - higher means a smoother ride for the money made. Here's the part nobody tells you: to trust that a Sharpe ratio of 1.0 is real and not just noise, you need roughly four years of daily data. The math works like this - confidence in the number sharpens only with the square root of how many years you've watched it, so shrinking the sample from four years to a few weeks doesn't just make the number less certain, it makes it nearly meaningless on its own.

Which is the honest asterisk on tonight's numbers: the live book has 27 trading days on the clock and is currently trailing SPY on a risk-adjusted basis. That's a real number and it's not being hidden - but statistically it's nowhere near enough data to call the strategy broken, any more than 27 good days would prove it works. The longer backtest, which does clear SPY's risk-adjusted return out of sample, is the more meaningful read for now. Both numbers stay in the piece. Neither gets to be the whole story yet.

The practice account sits at $979 tonight, up $2 on the day, down $21 since it was funded with $1,000 in June. All three open positions - EWY, MTUM, SPY - are basically flat, none of them big enough yet to show a percentage move. Turnover today was zero: no trades placed, no dollars moved, which is either discipline or nothing to do depending on your mood.

The strategies currently on the field are three mean-reversion plays riding different ETFs - MTUM (momentum stocks), SPY (the whole market), and USMV (low-volatility stocks) - each backtested with a Sharpe ratio above 0.8. Whether that survives contact with real, small-sample trading is exactly the question this newsletter exists to keep answering out loud, number by number, no smoothing.