Files
gogobee/cmd/expedition-sim/main.go
prosolis f2c2d774d4 J2: sim picker casts+consumes; T5 raid-content warning
The post-J1 sweep had the casters clustered at 19–22% L12 clear, vs
martials at 70–80%. A per-round trace across 240 boss-room fights showed
why: autoResolveCombat dispatched !attack only — zero spell_casts, zero
mid-fight consumable uses across every caster class. The entire "caster
cliff" was the sim measuring a strawman where casters couldn't cast.

J2a teaches the sim's autoResolveCombat to mirror a competent prod
player: heal at low HP if an inventory consumable is available, otherwise
cast the highest-EV damage spell (slot or cantrip), otherwise swing.
BuildCharacter now seeds the known-spell list via ensureSpellsForCharacter
so the synthetic spellbook is populated. A -trace flag on the cmd
attaches the raw CombatEvent stream to the last combat of each run for
post-hoc diagnostics.

A first re-baseline (n=100, all 10 classes) showed Ranger regressed
-35.8pp — the picker was burning L3 slots on lightning_arrow when
Ranger's weapon chassis (Hunter's Mark + Extra Attack) was the better
play. Added simMartialFirstClass to gate the picker off for Ranger and
Paladin (whose default kit is also weapon-first / no damage spells).
J2c experimented with widening the picker to control + heal spells;
heal-spell preempt cost druid 10pp (slot heals are 10HP vs 40HP
consumables) and control-spell scoring at 22 cost warlock 6.6pp. Both
reverted. Corpora retained under baseline_j2c*.jsonl for the post-mortem
in sim_results/j2b_findings.md.

Post-J2 L12 leaderboard (baseline_j2a_v2_all10.jsonl, n=100):
  fighter 80.0, ranger 80.0, paladin 78.4, rogue 76.8,
  druid 61.6, mage 53.4, sorcerer 50.6, warlock 48.2,
  bard 40.4, cleric 39.0.

The caster cluster is dissolved; martials are within ±5pp of J1 (sweep
noise). Bard/cleric still trail, but it's no longer a sim artifact —
their defaultKnownSpells damage rosters cap at L2 and the picker can't
pick spells they don't have. That's a prod-level fix, deferred.

J3 trace (sim_results/j3_findings.md): T5 dragons_lair walls every solo
class at 0% (Infernax 546 HP vs solo player HP 110–175; ~25% boss HP
eaten before TPK across all classes). Per the J3 plan menu, this is
party-shaped content the engine doesn't yet have parties for. Surface
a TwinBee-voiced heads-up in handleDnDExpeditionCmd's start path and a
matching tag in !expedition list — players see "raid-shaped — solo
runs not yet survivable" before they spend outfitting coin. No combat
or class balance changes.

Files: cmd/expedition-sim/main.go +trace flag; expedition_sim.go picker
+ SimCombatSummary.Events + spellbook seed; dnd_expedition_cmd.go
raidContentWarning + list tag. All baselines + traces + findings
checked in under sim_results/.
2026-05-17 15:43:41 -07:00

182 lines
5.4 KiB
Go
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
// expedition-sim drives synthetic players through expeditions for batch
// analysis. Re-uses production plugin paths against a fresh sqlite DB so
// outcomes mirror what live players hit.
//
// Single run:
// expedition-sim [-class fighter] [-level 5] [-zone goblin_warrens]
// [-bank 1000] [-cap 50] [-log] [-data DIR]
//
// Matrix mode (cartesian sweep over classes × levels × zones × N runs,
// one JSON object per stdout line, log suppressed by default):
// expedition-sim -matrix -classes fighter,mage -levels 5,10 \
// -zones goblin_warrens,wolf_den -runs 20
package main
import (
"encoding/json"
"flag"
"fmt"
"os"
"strconv"
"strings"
"gogobee/internal/plugin"
"maunium.net/go/mautrix/id"
)
func main() {
var (
class = flag.String("class", "fighter", "DnD class id (single-run mode)")
level = flag.Int("level", 5, "character level (single-run mode)")
zone = flag.String("zone", "goblin_warrens", "zone id (single-run mode)")
bank = flag.Float64("bank", 1000, "starting coin balance — must cover outfitting")
cap = flag.Int("cap", 50, "max autopilot bursts per expedition (each = up to autopilotRoomCap rooms)")
dataDir = flag.String("data", "", "data dir for the temp sqlite db (default: OS tempdir; ignored in matrix mode)")
userTag = flag.String("user", "@sim:expedition", "synthetic user id (single-run mode)")
logFlag = flag.Bool("log", true, "include per-row expedition log in output (single-run default true; matrix default false)")
matrix = flag.Bool("matrix", false, "matrix mode — sweep over classes × levels × zones × runs")
classes = flag.String("classes", "", "comma-separated class ids (matrix mode)")
levels = flag.String("levels", "", "comma-separated levels (matrix mode)")
zones = flag.String("zones", "", "comma-separated zone ids (matrix mode)")
runs = flag.Int("runs", 1, "replicates per (class,level,zone) cell (matrix mode)")
trace = flag.Bool("trace", false, "include raw per-round CombatEvent stream on the LAST combat of each expedition (boss room) — for J2 diagnostic sweeps")
)
flag.Parse()
plugin.SetSimIncludeTrace(*trace)
if *matrix {
// Matrix default: drop log to keep stdout manageable; explicit
// -log=true overrides.
includeLog := false
// Flag.Lookup tells us whether the user explicitly set -log.
flag.Visit(func(f *flag.Flag) {
if f.Name == "log" {
includeLog = *logFlag
}
})
runMatrix(*classes, *levels, *zones, *runs, *bank, *cap, includeLog)
return
}
runSingle(*class, *level, *zone, *userTag, *dataDir, *bank, *cap, *logFlag)
}
func runSingle(class string, level int, zone, userTag, dataDir string, bank float64, cap int, includeLog bool) {
dir := dataDir
if dir == "" {
var err error
dir, err = os.MkdirTemp("", "expedition-sim-")
if err != nil {
fail("mkdir temp:", err)
}
defer os.RemoveAll(dir)
}
res, err := runOne(dir, id.UserID(userTag), plugin.DnDClass(class), level, plugin.ZoneID(zone), bank, cap)
if err != nil {
if res != nil {
if !includeLog {
res.Log = nil
}
emitIndented(res)
}
fail("run:", err)
}
if !includeLog {
res.Log = nil
}
emitIndented(res)
}
func runMatrix(classes, levels, zones string, runs int, bank float64, cap int, includeLog bool) {
cs := splitNonEmpty(classes)
ls := parseLevels(levels)
zs := splitNonEmpty(zones)
if len(cs) == 0 || len(ls) == 0 || len(zs) == 0 || runs <= 0 {
fail("matrix mode requires non-empty -classes, -levels, -zones and runs > 0")
}
enc := json.NewEncoder(os.Stdout)
for _, c := range cs {
for _, lv := range ls {
for _, z := range zs {
for r := 0; r < runs; r++ {
dir, err := os.MkdirTemp("", "expedition-sim-")
if err != nil {
fail("mkdir temp:", err)
}
uid := id.UserID(fmt.Sprintf("@sim:%s-l%d-%s-%d", c, lv, z, r))
res, runErr := runOne(dir, uid, plugin.DnDClass(c), lv, plugin.ZoneID(z), bank, cap)
if res != nil && !includeLog {
res.Log = nil
}
if runErr != nil && res == nil {
// Synthesize a row so the corpus has one line per
// cell regardless of init failures.
res = &plugin.SimResult{
UserID: string(uid),
Class: c,
Level: lv,
Zone: z,
Outcome: "halted",
}
}
_ = enc.Encode(res)
_ = os.RemoveAll(dir)
}
}
}
}
}
func runOne(dataDir string, uid id.UserID, class plugin.DnDClass, level int, zone plugin.ZoneID, bank float64, cap int) (*plugin.SimResult, error) {
runner, err := plugin.NewSimRunner(dataDir)
if err != nil {
return nil, fmt.Errorf("init runner: %w", err)
}
defer runner.Close()
if _, err := runner.BuildCharacter(uid, class, level); err != nil {
return nil, fmt.Errorf("build character: %w", err)
}
runner.Euro.Credit(uid, bank, "expedition-sim bankroll")
return runner.RunExpedition(uid, zone, cap)
}
func emitIndented(res *plugin.SimResult) {
enc := json.NewEncoder(os.Stdout)
enc.SetIndent("", " ")
_ = enc.Encode(res)
}
func splitNonEmpty(s string) []string {
parts := strings.Split(s, ",")
out := parts[:0]
for _, p := range parts {
if p = strings.TrimSpace(p); p != "" {
out = append(out, p)
}
}
return out
}
func parseLevels(s string) []int {
var out []int
for _, p := range splitNonEmpty(s) {
n, err := strconv.Atoi(p)
if err != nil {
fail("bad level:", p)
}
out = append(out, n)
}
return out
}
func fail(args ...interface{}) {
fmt.Fprintln(os.Stderr, args...)
os.Exit(1)
}