Agile Battleships

The board is interactive, so it does not print. Play it at michaeloreilly392.github.io/personal-site/battleships.html — or use the rules below on a sheet of squared paper, with someone else holding the answers, which is how the exercise is usually run in a room.

The board is eight squares by eight. Hidden on it are seven ships — one Carrier of five squares, one Battleship of four, one Cruiser of three, two Destroyers of two, and two Submarines of one — eighteen squares of ship in all. Every round gives you thirty shots and five minutes, and a fresh layout. Only the feedback changes.

1. Round one — the plan

   Place all thirty shots before firing any of them. Change your mind as often as you like; click a placed shot to take it back. When you press Fire, every shot resolves at once and you find out how you did. Until then you are told nothing at all. This is a year of work released on the last day.

2. Round two — the salvo

   Fire in fives. After every fifth shot you are told how many of those five hit, and how many ships have gone down so far — never which square, and never which ship. You can steer, but only every fifth shot, and only by inference. This is a quarterly release.

3. Round three — continuous

   Every shot answers immediately. Hit or miss, on the square, and the fleet list crosses each ship off as it sinks. The round ends the moment the last ship goes down, usually with shots to spare.

Nobody plays round three better because they have become better at Battleships in the last quarter of an hour. The skill is identical, the board is no easier, and the ammunition has not changed. What changed is how long a wrong assumption was allowed to survive — thirty shots, then five, then one — and that alone is worth two or three times the score.

Each round is a way of working, and the scores are the argument:

• Round one is the big-bang release

  Everything is decided up front by people who are reasoning carefully and cannot yet be told they are wrong. The plan is not lazy — it is often the most thought-through artefact anyone produces all year. It is just built entirely out of assumptions, and it finds out whether they held on the day it ships, which is the last day anything can be done about it.

• Round two is the quarterly release

  Feedback exists, but it arrives in batches and it arrives blurred. You are told that two of your last five shots hit, and you cannot tell which two, so you steer by inference and you carry your wrong ideas four shots further than you needed to. Most teams that describe themselves as agile are somewhere in this round.

• Round three is continuous delivery

  Every decision answers immediately, so a bad assumption costs one shot rather than thirty. Nothing here is more clever than round one — it is the same person making the same kind of guess. The difference is that the guess is cheap, and being wrong stops being expensive enough to be worth avoiding.

The uncomfortable half is that round one is not incompetence. It is a perfectly reasonable plan, made carefully, by someone who was simply not allowed to learn anything until it was too late to act on it. If a team is scoring like round one, the useful question is almost never whether they are working hard enough. It is how long they have to wait to find out whether they are working on the right thing.

Adapted from Agile Battleships by James Scrimshire, CC BY 4.0. Built from the published rules — the ship sizes, the thirty shots and the five-minute clock match the browser version by Box UK, which is where I first played it. None of its code, markup or artwork is used here.