# Email to Martin — teaching the commentary to think like a sailor

**Subject:** Next step on the broadcast — and a way you can steer the commentary

Hi Martin,

First, thank you for the last round of notes — they were exactly the kind of detail we
needed. We've worked through all of them. The boats now carry a full crew of four, all
hiking on the windward rail and crossing over on each tack; the sails are fuller and take
a proper twist; the water reads with the breeze; and the marks and track rings sit down on
the surface instead of floating. On the tactical side, the calls now follow the team-racing
playbook far more closely — mark traps, splits, covers, the turn-back on the reach, and so on.

But I want to tell you about the bigger thing we're doing now, because this is where your
eye matters most.

**We're teaching the commentary to sound like a real sailing commentator — and we want to
use *your* judgement as the standard it learns from.**

Here's the idea, in plain terms. For every race, we've written out every single spoken line
into a simple spreadsheet — one line per row — along with the race clock, who's leading, and
which boats the line is about, so you can judge it without even watching the video. For each
line, all we'd ask is three things:

1. **Is it right?** Mark it *ok*, *wrong*, *dull*, or *irrelevant*.
2. **What would a good commentator have said?** Where a line is weak or wrong, just type the
   better line in your own words.
3. *(Optional)* a one-word reason — a fact error, a repeat, bad timing, wrong tone, wrong name.

One thing that helps most: don't only flag the lines that are *wrong* — flag the ones that are
technically correct but **flat or dull**, and write how you'd bring them to life. Those are exactly
the lines where a real commentator adds the most, and where we most need your ear.

That's it. Your "better" lines become the gold standard — the benchmark the system is measured
against and learns from. In other words, we're not asking you to learn our tools; we're asking
you to be the commentator whose calls the machine tries to live up to.

**How this ties back to the playbook.** The foundation document we sent you — the one built from
the GSYS team-racing clinic and the racing rules — is the rulebook the commentary works from.
Every spoken call is meant to describe a moment straight out of that playbook: a mark trap, a
split, a cover coming off the start, a turn-back on the reach. So when you mark a line and write
a better one, you're really doing two things at once — checking that the commentary read the
*playbook moment* correctly, and showing us how a sailor would actually say it out loud. Over
time your gradings and the playbook pull in the same direction: the calls stay honest to the
tactics, and they start to sound like they're coming from the rail rather than a rulebook.

The mistake we caught is a good example of exactly that. On a split — when one team sends a boat
up each side of the beat — the playbook is clear that it's the *covering* team that's caught out,
because they can't sit on both. Our commentary had it backwards and named the splitting team as
the one who "can't cover both." Your kind of note is precisely what surfaces that. It's now fixed,
and it's saved as a worked example the system will learn from.

I'm attaching the spreadsheets for the four races. There's no rush and no need to do all four —
even one race marked up would sharpen things a great deal. Mark a handful of lines, write a
couple of better ones, and send it back however's easiest.

Freshly rendered videos with the corrections are coming your way shortly — nothing needed from
you there.

Thanks again for lending your eye to this. It genuinely makes the difference between commentary
that's technically correct and commentary that sounds like it's coming from someone who's
actually sailed the racing.

Best,
Ulf