A movement with a method

People and models, on one counted problem.

An open chart for deaths in and on the water. Sea, coast, river, pond, pool, bath, and the spans above them. Not a club, and not a rescue service. A count from one water is never added to a count from another.

• 24,000 to 100,000

  Fisher deaths a year, depending on which serious estimate you trust. The distance is the reason the movement exists.

• 449

  Seafarer deaths sixty-six countries actually filed for 2024. Not the world.

• 35,642

  Recorded dead or missing in the Mediterranean since 2014, through 5 October 2026. A minimum. The count says so, and it will move.

A different water

These are not inside the figures above, and they are not a sum of each other.

• 4,345 a year

  U.S. unintentional drowning, every setting, 2018–2021. Pool, bath, river, and boat are inside it. A later window says over 4,500. They are not averaged.

• 376 a year

  Pool or spa deaths of children under 15, reported to CPSC for 2021–2023. Not the 4,345.

• 236,000 and 300,000

  WHO, two estimates of all drowning. The older figure is still on the page. The 2021 status report says 300,000. Do not average them.

Do not add these

A movement that sums unlike counts will design the wrong rescue. The ledger keeps the pairs apart.

• 24,000 a year and ~100,000 a year

  Same occupation. Two estimates. A quarter century apart. Averaging them invents a third number nobody counted.

• 449 and 24,000 a year

  449 is a filed merchant-fleet register. 24,000 is an old global fishing estimate. They are not the same people and they are not a sum.

• 556 deaths and 1,055 deaths

  556 is one year of U.S. recreational boating. 1,055 is nineteen years of U.S. commercial fishing. A harbor that mixes them will brief the wrong boat.

• 35,642 recorded and 86,170 recorded

  The Mediterranean figure is already inside the global migration figure. Adding them counts the same dead twice.

• 4,345 a year and over 4,500 a year

  4,345 a year is the 2018–2021 average. Over 4,500 a year is 2020–2022. The windows overlap. Averaging them invents a third United States.

• 4,345 a year and 376 a year

  4,345 is every setting and every age. 376 is pool and spa deaths of children under 15, reported to a different agency. Subtracting one from the other pretends they were counted the same way.

• 461 and 492

  461 and 492 are both ages 1–4, and they are different years. They are not a trend you can draw from two dots, and neither one is a pool count.

• 376 a year and nearly 90 children a year

  Nearly 90 children a year inside the home, mostly the bath, is not the pool file. Adding them builds a child-drowning total the reports refused to publish as one number.

• 4,345 a year and 556 deaths

  556 is pleasure boats. 4,345 is every unintentional drowning. The boats are inside the larger count only in the loose sense that some of the dead were boating. The years do not match. Do not subtract.

• 4,345 a year and ~236,000 a year

  One country, counted. The world, estimated. A lab that divides 4,345 by 236,000 and calls it the American share has invented a percentage.

• ~236,000 a year and 300,000 in 2021

  236,000 is still printed on WHO’s health-topics page. The 2021 status report says 300,000, and the fact sheet of 1 May 2026 says around 300,000 a year. Same agency, two estimates. Do not average them, and do not add either to a fishing total.

• 4,509 and over 4,500 a year

  4,509 is the count for 2022 in one paper. Over 4,500 a year is the 2020–2022 window. The year sits inside the window. Do not average them.

• 840 and 376 a year

  840 is all ages, death certificates, swimming pools, 2022. 376 a year is children under 15, pool or spa, reported to CPSC for 2021–2023. The 2022 CPSC count is 359, not 840. Do not average them.

• 341 and 556 deaths

  341 is watercraft drownings on 2022 death certificates. 556 is the Coast Guard’s recreational boating census for 2024, and not all of those deaths were drownings. Different years, different systems.

• 85,258 rescues and 32

  A rescue is not a death. 85,258 and 32 are two columns of one report from the agencies that filed. Do not subtract them and call the difference lives saved.

• 32 and 118

  32 and 118 are the guarded and unguarded columns. The association did not publish 150. Beaches that did not report are in neither column.

What it refuses

• A survival percentage. The forecast is a band, or it says the table does not apply.
• Invented hours for a suit or a raft. Those take a person off the immersion table. They do not come with a multiplier.
• A sum of unlike counts. The pairs above stay apart, including a life saved and a person aided.
• A dispatch. Distress is VHF channel 16, or the coast guard.

The compact

The rules. They are the movement. Break one and you are running a different project.

1. 01

   People decide

   The model may propose. A proposal no person has edited is not a proposal. The edit is the collaboration, and it stays visible.

2. 02

   Nothing here dispatches

   Distress is VHF channel 16, or the coast guard. A draft, a clock, or a clever branch does not start a boat and does not tell anyone to wait.

3. 03

   The disagreement stays

   24,000 and 100,000 are not averaged into a number that feels finished. Where the sources disagree, the chart shows both.

4. 04

   The dead are not a metric

   A range is allowed. An invented name, an invented toll, or an invented rescue is not. If it is not in the ledger, say so.

5. 05

   A change needs an adopter

   The work is real when a named service could take it up this year without a new agency. A movement that only produces a movement has not saved anyone.

6. 06

   Open, not silently writable

   Anyone may propose. No one may overwrite the canon from a browser. You export the text. A person merges it. That is how an open chart stays safe enough to use.

7. 07

   The branch, not the person

   An idea is judged only by what it does to a branch: the break it names, the change it asks, the service that could adopt it, and how you would know it failed. Office, fame, fluency, and whether a person or a model said it are not inputs. People overweight whoever spoke last and whoever holds rank. A model overweights whatever it is rewarded for saying. A leaderboard would be won by whichever side can produce more text. Take both out of the judgment.

8. 08

   Honor the bringing

   Name the person for bringing the idea, including the part that was wrong. Do not score them, and do not rank them against a model. A model can draft. It cannot be the author. The edit is the contribution. A patch is the same shape as a fence: kept if the chart gets truer or a branch gets a named adopter. Being corrected in public has to cost less than staying silent, or the next person will keep the bad number and skip the letter.

Brought, not ranked

The person is named for carrying an idea in. The tree judges the idea without them. There is no score, and a model is not on this list.

• A fence on the high-level bridge

  John, harbormaster, Pepperell Cove · Before the water

  The barrier stays on the branch. Twelve a year does not. The ledger toll is “No annual census.”

  Named for the ask, made for years. Not for the number.

Contributors

No one yet beyond the people already named in the chart. Bring a branch.