Monday, April 30, 2012

Zendo: Testing in game form

I love Zendo. Its fans will often describe it as the best game, or gaming concept, ever made. It is deeply strategic without being overly competitive, in my opinion, and is also a lot of fun, especially with good players.

Here it is on Wikipedia:
http://en.wikipedia.org/wiki/Zendo_(game)

I suggest getting a few playing pyramids and a copy of Playing With Pyramids from Looney Labs.

Took me far too long (plus the time to watch a James Bach lecture, plus the time for it to sink in) to see that it contains a lot of the same thought processes as testing. My prediction: a good exploratory tester would kick arse at Zendo.

Here's the idea behind Zendo (icehouse piece version).

First you have a set of "pyramids" (icehouse pieces, treehouse pieces, or whatever they're calling them these days). They are often transparent, coloured, stackable pieces that come in sets of 15 (5 each of small, medium and large, or 1-point, 2-point and 3-point pieces if you prefer to call them that).

You also need a pile of black stones and a pile of white stones.

The "Master" player thinks up a rule, creates two "Koans" (small constructions made out of the pieces), one of which obeys the rule and one which does not. A rule could be "A koan has the Buddha nature if and only if it contains a blue piece" or "A koan has the Buddha nature if and only if it contains two medium pieces" or "...contains a blue piece pointing at a red piece" or "the value of the pieces is a square number"...etc.

Basically the idea for the other players is to guess this rule by building koans and having them marked as having the rule (white stone) or not (black stone) until they can determine enough about the system to guess at the rule.

If the master cannot build a koan that disproves the guess (build one that has the Buddha nature but disobeys the guessed rule, or build one that does not have the Buddah nature and obeys the guessed rule) then that player wins.

In order to do this you need to evaluate what the white koans have that the black koans do not and vice versa. Each guess will form part of the rest of your test plan as the resulting information is useful feedback. You need to make "quick attack" guesses that halve the problem or find useful information quickly rather than slow, plodding ones or wild guesses that make little advance to your current knowledge. You must not repeat tests (build exact copies of other koans) and reduce partial repetition to a minimum when you have enough knowledge to know better (building near-exact copies of other koans, such as the same koan with different colours after you have ruled out colour being important in the rule).

If it's not a good similie for exploratory testing I fail to see how.

Just a quick thought, I might formulate it into a more formal assessment later.

1 comment: