Turn time was hiding the real pace — a 31-32s swim, dressed up as 35-38s

1. "25m split" is a protocol term, not a measurement. Ask whether the timer stops at the hand touch (swim-only) or at the feet-on-wall (wall-to-wall including turn). They can differ by 4–7 seconds per 25, which is the entire range of variation in a fitness-day set. 2. If the bracketed split includes the turn, you're looking at turn efficiency, not swim speed. Ranking reps by 25m-wall-to-wall is ranking reps by turn, not by swimmer condition. Use it for turn work; don't read it as fitness signal. 3. If the swim-only split is stable when the wall-to-wall split varies, the set is telling you about the turn, not the swimmer. That's a turn-coaching rep, dressed as a fitness rep. Re-label the set before you read it.

That correction collapsed the whole analysis. The 35-38s bracket wasn't swim-pace — it was wall-to-wall, and the turn was eating 3–7 seconds per rep on top of the actual swimming. Once you subtract the turn, the swim-only 25m sits at a flat ~31-32s across every single rep of the nine. The variation in the wall-to-wall column was almost entirely turn efficiency, not swim speed. The "breakdown" in the middle of the set wasn't a breakdown at all — it was a turn that ran long.

This is the same move the 2026 05 28 Turn Testing Protocol 90daeb protocol made in May: isolate the variable you actually want to measure. That page argues for testing turns with a flags→5m→flags window so you're comparing turn type, not entry stroke. Today's session makes the inverse point: when you're already timing a set wall-to-wall, the swim-pace signal is underneath the turn tax, and you'll misread the set if you don't peel the layer off.