clay-borg/games/ground/examples/attack-value.rs
tegwick a928b5925c
Some checks failed
ci / check (push) Failing after 3s
CB-WP-0038: variant selection, H1 implemented, and H1 measured
ground-game packages hypotheses as selectable rules variants — a catalog,
a rules_delta.yaml, and prose — and their note is explicit that CSV text
alone is not executable here. So the kernel gains a Variant in game state:
in the state, therefore in the hash, therefore in the recording, because a
scenario replayed under a different variant would diverge silently.

Baseline is bit-for-bit what it was, asserted across seat counts and
seeds. A variant system that perturbs the baseline invalidates every
measurement this repo has.

H1-A and H1-B implemented from rules_delta.yaml and mutation-proven on
their own defects: "unclaimed" misread as face-up-and-unsolved, and the
attacker's Stress read after the attack's effects. Their `unchanged:` list
is asserted rather than trusted — that list is their claim about their own
experiment.

Measured, and three of their four criteria fail. DARVO arm rate is still
0 under greedy; ATTACK selection does not rise and falls for the rank-75
policy; group success collapses from 165/190/200 to 0 at 3/4/6 seats.
The mechanism is not the assumed one: greedy answers the pressure by
regulating, Stress plateaus at 3, so it never reaches the gate at 4 or the
arm at 5 — H1-A acts as a solve-rate tax and H1-B is unreachable under
competent play.

A harness defect was caught before the claim: sweep discarded refused
games silently and never reported its count, so "nobody won" and "nothing
played" printed identically. Reporting H1 as unwinnable on that basis
would have been the ADR-0018 family aimed at another repo's design. All
200 games ran in every cell; the zeros are real.

Chaos d8 = 8 — the window's first override, redrew L against a structural
L, so it changed nothing. Window 3 recorded in ChaosRollHistory.

NOT REVIEWED: tier L owes a separate-agent adversarial review, and no H1
result may reach ground-game until it has run.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-08 00:50:08 +02:00

233 lines
8.5 KiB
Rust

//! F17's reproduction: what is ATTACK worth, in each scoring mode?
//!
//! The maintainer reported *"there is no incentive to play attacks as long
//! as I have positive cards"*. F17 sat as a note because nothing
//! demonstrated it. This is the artifact.
//!
//! **`GreedyPolicy` plays ATTACK zero times, and that measures our bot.**
//! `bot.rs` ranks `Action::Attack => 10`, below everything else. Reporting
//! "the game gives no incentive" from a policy programmed to rank attack
//! last would be CB-WP-0025's C4 error again — one policy's behaviour
//! presented as the game's.
//!
//! So this varies **exactly one number**: ATTACK's rank in an otherwise
//! identical policy. 10 is greedy's. 75 puts it above SUPPORT and below
//! INVESTIGATE — *attack when convenient*. 95 puts it above SOLVE —
//! *attack whenever legal*.
//!
//! And it asks the question in **all three scoring modes**, because Blame
//! costs *personal* score: ATTACK may be worthless in the co-op mode and
//! earn its place in GR-E03 or GR-E04.
use cb_game_runtime::{ScenarioGame, Setup};
use cb_kernel::PlayerId;
use games_ground::bot::{play, Choice, GreedyPolicy, Policy};
use games_ground::{Action, GroundCommand, GroundState, ScoringMode};
/// Greedy's ordering with ATTACK's rank as a parameter.
///
/// A copy of the ranking rather than a call into it: `GreedyPolicy::rank`
/// is private, and the point is to vary one number while holding every
/// other preference identical.
struct Attacker(i32);
impl Policy for Attacker {
fn name(&self) -> &'static str {
"attacker"
}
fn choose(
&mut self,
state: &GroundState,
seat: PlayerId,
legal: &[GroundCommand],
_may_pass: bool,
) -> Choice {
let gated = state
.players
.get(&seat)
.is_some_and(|p| p.stress >= 4 && !p.freedom_gate_lifted);
let rank = |c: &GroundCommand| -> i32 {
match c {
GroundCommand::SelectAction {
action, problem, ..
} => match action {
Action::Ground if gated => 100,
Action::Attack => self.0, // the one varying number
Action::Solve
if problem
.and_then(|p| state.problems.get(&p))
.is_some_and(|p| p.claimed_by.is_some()) =>
{
5
}
Action::Solve => 90,
Action::Investigate => 80,
Action::Support => 70,
Action::Ground => 60,
},
GroundCommand::SpendFreedom => {
if gated {
96
} else {
0
}
}
GroundCommand::Reveal | GroundCommand::Resolve | GroundCommand::EndRound => -1,
_ => 50,
}
};
let mut best = 0usize;
for (i, c) in legal.iter().enumerate() {
if rank(c) > rank(&legal[best]) {
best = i;
}
}
Choice::Command(best)
}
}
/// `(games, good outcomes, attacks, DARVO armings)` over 200 seeds.
///
/// **`games` is reported, not assumed to be 200.** A rules variant can
/// make the engine refuse a position, and a harness that skips those
/// silently reports "nobody won" for a run in which nobody played.
fn sweep(
mode: ScoringMode,
variant: games_ground::Variant,
players: u8,
mk: &dyn Fn(u8) -> Vec<Box<dyn Policy>>,
) -> (u32, u32, u32, u32) {
let (mut games, mut won, mut atk, mut darvo) = (0, 0, 0, 0);
let mut errs: Vec<String> = Vec::new();
let mut setup_fails = 0u32;
for seed in 0..200u64 {
let Ok(mut st) = GroundState::setup(
&Setup {
players,
preset: format!("standard-{players}p"),
patch: Default::default(),
},
seed,
) else {
setup_fails += 1;
continue;
};
st.mode = mode;
// CB-WP-0038 T03: the SAME instrument measures both rule sets in
// the same run. Two harnesses would compare harnesses.
st.variant = variant;
let mut ps = mk(players);
// CB-WP-0038 T03: a game the engine REFUSES is not a game the
// table lost, and `else { continue }` made the two identical.
// H1's first run showed won/atk/darvo all exactly 0 at 3+ seats,
// which is the signature of every game being discarded rather
// than played -- so the count is now reported, not swallowed.
let g = match play(st, &mut ps) {
Ok(g) => g,
Err(e) => {
errs.push(format!("{e:?}"));
continue;
}
};
games += 1;
// In the co-op mode the table wins together. In the other two the
// question is whether SEAT 0 is among the winners — because that
// is what an individual player's incentive turns on, and ATTACK is
// an individual's choice.
let good = match mode {
ScoringMode::SharedGround => g.state.outcome.as_ref().is_some_and(|o| o.group_success),
_ => g
.state
.outcome
.as_ref()
.is_some_and(|o| o.winners.contains(&PlayerId(0))),
};
if good {
won += 1;
}
for (_, c) in &g.steps {
if let GroundCommand::SelectAction {
action: Action::Attack,
..
} = c
{
atk += 1;
}
}
for e in &g.events {
if matches!(e, games_ground::GroundEvent::DarvoTriggered { .. }) {
darvo += 1;
}
}
}
if setup_fails > 0 {
eprintln!(" !! {setup_fails} of 200 SETUPS failed ({players}p)");
}
if games != 200 {
eprintln!(" !! only {games} of 200 games ran ({players}p, {mode:?})");
}
if !errs.is_empty() {
eprintln!(
" !! {} of 200 games did not run ({}p): first = {}",
errs.len(),
players,
errs[0]
);
}
(games, won, atk, darvo)
}
fn main() {
println!("F17 — what is ATTACK worth, in each scoring mode?\n");
println!("CB-WP-0038 T03: run for BOTH rule sets. ground-game's H1 asks");
println!("whether problem pressure plus a high-Stress self-soothe make");
println!("ATTACK and DARVO relevant without making always-ATTACK optimal.\n");
println!("`won` is group success in SHARED GROUND, and \"seat 0 is among the");
println!("winners\" in the other two, because ATTACK is an individual's choice");
println!("and that is what an individual's incentive turns on.\n");
for (vlabel, variant) in [
("BASELINE ground-darvo-r0", games_ground::Variant::Baseline),
(
"H1 h1-problem-stress",
games_ground::Variant::H1ProblemStress,
),
] {
println!("================ {vlabel} ================\n");
for (label, mode) in [
("SHARED GROUND (GR-E02, co-op)", ScoringMode::SharedGround),
(
"COMMON PROBLEM (GR-E03, semi-co-op)",
ScoringMode::CommonProblem,
),
("BONDED COALITIONS (GR-E04)", ScoringMode::BondedCoalitions),
] {
println!("{label}");
println!(" rank=10 (greedy) rank=75 (sometimes) rank=95 (always)");
println!("seats won atk darvo won atk darvo won atk darvo");
for players in [2u8, 3, 4, 6] {
let (_, gw, ga, gd) = sweep(mode, variant, players, &|n| {
(0..n)
.map(|_| Box::new(GreedyPolicy) as Box<dyn Policy>)
.collect()
});
let cell = |r: i32| {
sweep(mode, variant, players, &move |n| {
(0..n)
.map(|_| Box::new(Attacker(r)) as Box<dyn Policy>)
.collect()
})
};
let (_, mw, ma, md) = cell(75);
let (_, aw, aa, ad) = cell(95);
println!(
" {players}p {gw:>4} {ga:>4} {gd:>5} {mw:>4} {ma:>4} {md:>5} \
{aw:>4} {aa:>4} {ad:>5}"
);
}
println!();
}
}
println!("(200 games per cell)");
}