tegwick
b590e7fd59
Some checks failed
ci / check (push) Failing after 4s
Eight profiles on a common schema, each marking what was checked against a source this session and what is background recollection. Three are marked unverified in full — Machinations, the play substrates, most of RBG — and say so rather than reading as evaluations. Written straight after three review rounds whose entire yield was claims outrunning what had been checked, so the confidence rule is the first thing in the README. The survey changed the plan, which is what a survey is for. Track C was described in Positioning as open ground. It is not: Browne published 57 criteria for game quality, and Ai Ai already computes designer-facing measures — drama, lead changes, branching factor, completion, duration — from played games. The track becomes adopt, credit and find the gap. The gap looks real: those measures presume a leader, and SHARED GROUND has none — Modes.csv gives its tiebreak as "Not applicable". Track B probably adopts rather than builds. OpenSpiel implements CFR, best-response and exploitability over games that are simultaneous-move, imperfect-information and co-operative, which is all four of GROUND's awkward properties. "Does ATTACK ever pay" is a best-response question, and we spent three review rounds refining a two-policy sweep for it. The first Track B task is now one question — is exploitability meaningful for a co-operative game with a shared threshold — not a build. The cost of not surveying earlier is therefore measurable, and is recorded rather than glossed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
53 lines
1.6 KiB
Markdown
53 lines
1.6 KiB
Markdown
# GDL / General Game Playing
|
||
|
||
Stanford Logic Group · `unverified` for current status · `checked` for the
|
||
Ludii comparison
|
||
|
||
## What it optimises for
|
||
|
||
**Agent generality**: an agent that plays a game it has never seen, given
|
||
only the rules in a logical description language.
|
||
|
||
## How a game is defined
|
||
|
||
**GDL** — a declarative, logic-programming description of legal moves,
|
||
state update and terminal conditions. Verbose relative to Ludii.
|
||
`checked` (verbosity claim from the Ludii comparison)
|
||
|
||
## What it can analyse
|
||
|
||
Little by design. GGP is a competition framework for agents; the game is
|
||
input, not subject.
|
||
|
||
## What it does not do
|
||
|
||
Anything designer-facing. **Ludii is measured as materially faster** —
|
||
more than six times for Gomoku 15×15, more than twenty for Amazons and
|
||
Hex, and over two hundred for Chess. `checked`
|
||
|
||
## Relevance to clay-borg
|
||
|
||
**Track A, as the cautionary case.** GDL is the purest form of "define
|
||
any game formally" and its cost is speed and readability. It is the
|
||
argument for *not* reaching for maximal generality first.
|
||
|
||
## What to steal
|
||
|
||
The separation of **rules-as-data** from the agent, which our kernel has
|
||
in a weaker form (rules are Rust; only the edition is data).
|
||
|
||
## What to avoid
|
||
|
||
Its verbosity, and its framing of the game as an input to an agent rather
|
||
than as an artifact under study.
|
||
|
||
## Open questions
|
||
|
||
- Whether GGP is still active, and where the community went.
|
||
|
||
## Sources
|
||
|
||
- *A note on the empirical comparison of RBG and Ludii*
|
||
(https://arxiv.org/pdf/1910.00309)
|
||
- *Game Description Languages: the Good, the Bad, and the Ugly*, Björnsson
|
||
(https://ludii.games/TalksWorkshop/Bjornsson.pdf)
|