Accountability first. Every session this desk puts direction reads on the board, and every read gets scored once the move has actually happened. No quiet deletions, no marking our own homework after the fact. Here is how the week landed.
The score
Across the week the desk logged 392 directional reads. 238 of them played out in the direction called. That is a 61% week. It sits below our longer-run numbers on purpose shown here so you can see the range, not a cherry-picked high: directional accuracy runs 78% measured over 30 days and 80% over 90 days. A single week is noisier than a quarter, and this was a choppier tape where short-horizon direction is harder to read. We would rather show you the honest 61% than dress it up.
What we called that worked
The clearest theme was a bearish lean on big-cap technology into the pullback. Reads that the crowd-favourite names had further to give back, NVIDIA, Apple, Tesla, Microsoft and Meta among them, largely played out as the tape rolled over. The same read carried through the index level, where our bearish calls on the NAS100 and the S&P tracked the move down. It was not uniform bearishness: a bullish read on Amazon also worked, which is the point of reading each name on its own evidence rather than painting the whole tape one colour.
What we called that missed
154 reads, 39% of the week, went the other way. That is the honest half of a scorecard, and it is the half most desks hide. A 61% week means nearly four in ten reads were wrong, and in a chop like this the misses clustered where a name looked set to continue and instead snapped back. We log those with the same timestamp as the wins, because a track record you can only see when it flatters you is not a track record.
How to read this
These are directional reads, scored after the move against what actually happened. They measure the quality of the read, not a live trading account, and nothing here is advice to buy or sell anything. The value is the discipline: a number that updates whether it helps us or not, week after week, so the method is visible and the claim is checkable.
Education only. No investment advice. Direction accuracy measured after the fact; past read accuracy does not predict future results.
