No cherry-picking
The numbers
Each daily pick is bought at the open and sold at the close of the same session, with 0.1% simulated trading costs. No leverage, no hindsight: the 43 sessions below are every session that was ever published.
Sessions measured
43
Closed up
47%
Beat the S&P 500
44%
Median day
-0.40%
Average day
-0.25%
Edge over the S&P 500
-0.19%
average per session
Best day
+8.10%
Worst day
-6.91%
Simulated capital against the market
-11.36% cumulative against -2.32% for the S&P 500 over the same sessions.
Does our conviction mean anything?
We label every pick before the session opens, so this split cannot be rigged after the fact. If the high conviction row does not end up ahead of the others over time, our conviction label is worthless and we will say so here.
Month by month
The same sessions, grouped by calendar month. The total is compounded rather than added up, so a day at +10% followed by one at −10% shows the loss it really was.
| Month | Sessions | Total | Average day | Volatility | S&P 500 |
|---|---|---|---|---|---|
| August 2026(launch month) | 12 | -5.01% | -0.39% | 2.78% | -1.59% |
| September 2026 | 21 | -7.94% | -0.35% | 3.04% | -1.03% |
| October 2026 | 10 | +1.36% | +0.15% | 1.84% | +0.29% |
Volatility here is the plain standard deviation of the daily returns inside the month, left un-annualised on purpose: scaling seven sessions up to a yearly figure would put a financial-looking number on a sample that cannot carry one.
August 2026 is marked as the launch month, and it should be read down rather than up. It is a partial month — the record opens on August 14, not on the first trading day — and every session in it ran while the system was still being brought into service. Two defects found on 25 August were in force throughout: the stored quantitative score did not match the one the ranking used, and the earnings-proximity guard never fired because the date it needed was never fetched. Both are fixed, neither was fixed in time for these sessions.
How to read this honestly
- The share of days that closed up tells you how often we were right, not how much you made. One bad day can wipe out several good ones.
- Beating the S&P 500 is the number that matters. Gaining 1% on a day the whole market gained 1.5% is not a win, it is a loss dressed up as one.
- Below roughly twenty sessions, everything on this page is still mostly luck. Treat it as such until the sample grows.
- These results are simulated on official open and close prices. A real order also pays a spread, which is only approximated here.
The session-by-session detail lives on the track record page.