# Gun campaign — ledger

**Goal (owner's mandate, 2026-09-26 overnight):** the movement campaign produced
a replicated champion (`TR_MOVEMENT=strafe`, `docs/movement_campaign.md`) which a
parallel job is shipping. *"Continue until you found an amazing movement. When
found do the same over for a gun."* This file is the GUN campaign's single source
of truth: later jobs **append** a `## Batch N` section and never edit an earlier
one (a wrong earlier number gets a correction line, not a rewrite). This file is
the gun successor to `docs/movement_campaign.md`; read that first, then this.

---

## OUTCOME (read this first)

**The shipped gun is UNCHANGED: `onlyPattern`** — the rack admits `Pattern` only
(`common_libs/gun_harness/selector.nim`, commit `e0666a5`). **Nothing beat it.**
The gun design space explored here is **CLOSED on evidence**, not on belief.

**Headline (MEASURED).** Four frozen-binary tournament sessions on the frozen
**15- and 33-opponent panels**: **1,233 battles** (phase 1: 270 + 297; phase 2:
270 + 396), **0 failed, 0 never-started, 0 liveness exclusions**. Each session is
ONE binary built from `git archive HEAD`; arms differ only by env; movement is
pinned `TR_MOVEMENT=strafe` in every arm, so every delta is a pure gun delta.
With the earlier 16-gun selector campaign (≈660 battles) the gun ledger now
stands on **≈1,900 measured battles**. What was **detectably worse**: `knn`
(−39.3 dmg/run, 0/15 opponents; −0.62 wins/run, p=0.012) and, on the wider
33-opponent panel, `TMHorizon` (−10.4 dmg/run, sign p=0.035). What was
**not distinguishable** (inside the MDE, no verdict): `bitbrain`,
`tmhorizon`@15, `rack_pk`, `rack_pt`, `len6`, `len16`, `depth100`, the two radial
controls.

### CLOSED axes (one-line reason each)

| axis | reason it is closed |
|---|---|
| other single guns | `knn` detectably worse; `bitbrain` a wash measured 3× (−3.6 / +10.3 / +1.8 dmg/run, all inside their MDEs); `TMHorizon` worse on 33 (`docs/gauntlet_bitbrain_vs_pattern.md`, Batches 1–2) |
| 2-gun selector racks | `rack_pk` nominally below `pattern` on wins; `rack_pt` == `tmhorizon`-alone → the selector just picks the corrector (`## Batch 1`) |
| selector mechanics | the virtual-fitness selector is negative value at every rack size tested (16, lean8, lean6, pairs); it never manufactured a win (`docs/selector_negative_value.md`) |
| lead amplitude | a full gain sweep (1.0/1.5/2.0/3.0) makes Pattern strictly worse at every range band — the lever is lead *information*, not amplitude (`docs/bitbrain_campaign.md` §0.3.2) |
| Pattern's own match-length / history params | `len6`'s single-session +0.49 wins (p=0.039 on 15) reversed to a wash on 33 (−0.04, p=1); `len16`/`depth100` flat; two batches, no replication (`## Phase 2`) |
| radial knobs (`TR_PATTERN_RAD_*`) | bearing-invariant **by construction** (job j99 proved `bmPath` is a structural no-op and scale/offset cannot change the aim bearing); in Batch 2 `rad_offset` is nominally *negative* |

### THE ONE OPEN AXIS

**A genuinely new source of lead INFORMATION** — not another knob on
BitBrain/TMHorizon (both measured neutral→negative), not amplitude (dead), not
radial (bearing-invariant). The standing mechanism is `docs/bitbrain_campaign.md`
§0.3.3: at 450+ px Pattern's own lead correlation with the required lead is only
**0.165**, so it is adding variance to a nearly uninformative signal.

**Honest caveat (MEASURED, `docs/state_window_gate.md`).** The best surviving
idea from the state line — a *single* wave-relative state at Q=4 — **does carry
signal at coarse resolution**: it predicts the miss-offset *bin* on held-out
battles at 0.4094 (vs 0.2348 majority), but those bins are **36/42/60 px wide =
4.58/5.34/7.63° at 450 px**, and its implied hit probability is **still the base
rate (~0.09)** — i.e. it is informative about **which side** the miss falls on,
not about **whether** the shot hits. The live hit window is
`atan(18/450) = 2.29°` half-width; **the measured signal is at ~4.6–7.6°, not at
the ~2° the 450 px hit window needs.** A temporal window does not fix this (the
window gate is negative; long contexts never recur). So the open axis is a *new
information source that resolves the arrival offset to better than ~2°*, which
does not exist yet.

**Cost estimate for trying it (INFERRED).** Build and *offline-gate* the new
information source first — require it to beat Pattern's 0.165 lead correlation
and, ideally, predict the arrival offset to <2.29° on held-out shots (days of
work; the offline instruments exist, e.g. `common_libs/tests/state_window_gate.py`).
Only then spend a live batch: one frozen binary, 6 env-only arms × 15 opponents ×
3 runs × 3 rounds = **270 battles ≈ 1–2 h wall** at conc 6 with `--wait-arena`
(exact command in *How to run a batch*). Do **not** open with a live batch.

### What would change our mind

A named new information source that (1) offline predicts the arrival offset to
better than the **~2.29° hit half-window at 450 px** on held-out battles **and**
(2) in a live frozen-panel batch beats `pattern` on **wins/run** with the winning
metric's CI excluding 0 and sign-flip p<0.05 while **damage is not detectably
down**. A damage-only gain without wins is the `ring`-mover trap and does **not**
count.

---

## What changed tonight (2026-09-26)

* **SHIPPED:** nothing in the gun. The rack is still `onlyPattern`.
* **NOT shipped, and why:** every candidate (other gun, small rack, selector
  mechanic, match-length/history parameter, radial knob) failed to beat `pattern`
  beyond the MDE on the frozen panels. Phase 2's one signal (`len6`, +0.49
  wins/run on 15 opponents) **did not replicate** on 33 opponents. The only
  shipped change of the night is the **movement** default (`tfil -> strafe`, see
  `docs/movement_campaign.md`); the gun was left untouched.
* **Revert:** no gun revert is needed (no gun default changed). Movement:
  `TR_MOVEMENT=tfil`.
* **Reproduce the key gun evidence (one command + the analyzer)** — phase 2,
  Batch 2, the session that killed `len6`:
  ```sh
  TOURNAMENT_NIMCACHE=/tmp/nc_j123 \
  tools/ab/tournament_run.sh \
      --arms tools/ab/arms_gun_b4.txt \
      --panel tools/ab/panel_gun_b2.txt \
      --runs 3 --rounds 3 --conc 6 --wait-arena 45 \
      --reference pattern \
      --outdir /tmp/ab/j123_b2
  python3 tools/ab/tournament_analyze.py /tmp/ab/j123_b2 --reference pattern
  ```

---

## 0. The question

The shipped rack admits **`Pattern` only** (`onlyPattern`). That decision was
taken because the virtual-fitness selector measured **negative value** at every
rack size tested, and Pattern is the best single gun by the DrussGT hit-rate
table (`docs/selector_negative_value.md`, `docs/gun_rack_analysis.md`, commit
`e0666a5`). But every one of those measurements — like every pre-campaign
movement claim — is **DrussGT-heavy**, and the movement campaign's standing
lesson is that a one-opponent result is not a result.

So Batch 1 asks the direct question, on a frozen 15-opponent panel:

> **Does any rack gun, or any small gun configuration, beat the shipped
> `Pattern` on damage/run AND round wins across 15 opponents — or is
> `onlyPattern` CONFIRMED rather than merely assumed?**

A **null is a successful outcome here**: it upgrades `onlyPattern` from an
assumption to a measured verdict across the panel, which is itself valuable.

---

## 1. Protocol (identical to the movement campaign; the instrument is reused verbatim)

`tools/ab/tournament_run.sh` + `tools/ab/tournament_analyze.py` +
`tools/ab/panel_movement.txt` (frozen panel), reused **unchanged**. The unit of
evidence is the **number of opponents**, not the number of runs.

| Element | Rule |
|---|---|
| Subject | ONE frozen binary built from `git archive HEAD` (`tournament_run.sh` records commit + binary sha256 in `session.json`) |
| Arms | env dicts only — **no per-arm rebuild, ever**; the arm file is a committed file (`tools/ab/arms_gun_b1.txt`) |
| Panel | the **frozen** `tools/ab/panel_movement.txt` (15 opponents). Adding/removing an opponent starts a new batch number |
| Pairing | per opponent: average the arm's runs, subtract the reference's average for that same opponent → one delta per opponent; then aggregate |
| Isolation | per-run bot dir + classic data dir, ephemeral ports, own process group; cleanup only by this session's outdir |
| Serialization | **one battle fleet at a time** via `--wait-arena`; never a broad `pkill robocode_shim` |
| Liveness | every declared env token must appear verbatim in OUR bot's own raw-env report, else the run is excluded and named |
| Never shipped | this is a measurement campaign: `git status` clean, shipped defaults untouched, `.gitignore` untouched |

> **Movement-default flip note (MEASURED).** During this job a parallel job
> (j120, commit `3fd6db9`) flipped the shipped `TR_MOVEMENT` default from `tfil`
> to `strafe`. Batch 1 built before the flip, Batch 2 after. **Every arm in both
> batches declares `TR_MOVEMENT=strafe`**, so the effective movement is identical
> and the two batches are directly comparable — the pin did its job. The reference
> rack lines in the boot reports read `PATTERN` for `pattern`, and the active 1v1
> rack reads `BITBRAIN` / `TMHORIZON` / `KNN` for the overridden arms.

### Movement is PINNED in every arm: `TR_MOVEMENT=strafe`

**Reason (hard rule).** A movement-default flip is landing from job j120 during
this run, so the default engine can change underneath us. Every arm therefore
sets **`TR_MOVEMENT=strafe` explicitly**. This (a) removes movement as a
confound, (b) makes all arms share exactly one movement, so every delta is a pure
**gun** delta, and (c) satisfies the analyzer's liveness rule for every arm
(each declared token must appear verbatim; an *undeclared* leaked `TR_MOVEMENT`
is fatal, but here every arm declares it). The reference arm is therefore the
shipped **gun** rack under the pinned movement, not under the (mutable) default.

---

## 2. Pre-registered decision rules (fixed BEFORE Batch 1 ran)

1. **Primary metrics:** **damage/run** and **ROUND WINS**. Secondary/explanation
   only: damage taken/run, incoming hit rate, mean distance. **Never hit rate
   alone** — that trap has inverted six verdicts in this project.
2. **BETTER than the reference** iff one primary metric is up with a
   cross-opponent **sign test p < 0.05** while the other does **not** go down;
   the mirror image for **WORSE**. Anything else is **NOT DISTINGUISHABLE** (a
   real answer, not a failure). Both the strict reading (the other metric's mean
   delta `>= 0`) and the substantive reading (not *detectably* down: not
   significant **and** smaller than that metric's MDE) are printed.
3. **A verdict must survive the between-opponent spread:** the pooled mean delta
   is reported with the SD across opponents, its SE, a 95% CI, and the **MDE**
   (α=0.05 two-sided, 80% power). An effect smaller than the MDE is reported as
   *not detectable* — never as *absent*, never as a win.
4. **Somewhere to stop:** if no arm beats the shipped `pattern` by rule 2 in
   Batch 1 **and** no arm shows a `>= +MDE` damage gain with p<0.10, then the
   gun stage's first phase is closed with *"the shipped `onlyPattern` rack is the
   best gun configuration we have measured across the panel"* — that is a
   **successful** outcome, and the campaign moves to a named next axis rather
   than inventing more arms. See *What would make us stop*.
5. **No promotion off a single metric, a single opponent, or a single run.**
   A change that wins damage by losing wins (or vice-versa) is not a win.
6. Every batch is shot with a **pre-registered prediction** stated in its section
   *before* the battles finish; a prediction that turns out wrong is recorded as
   wrong.

---

## 3. Stage 0 — what we already know (given, not re-derived)

| fact | value | source |
|---|---|---|
| shipped rack | `onlyPattern` (id 5), all other guns `off` | `common_libs/gun_harness/selector.nim` |
| why | selector measured negative value at every rack size (16, lean8, lean6, pairPC/PK/PL) and 10/10 adversaries | `docs/selector_negative_value.md`, `docs/gun_rack_analysis.md` |
| Pattern hit rate vs DrussGT | ~10.8% | given |
| KNN / Linear / Circular / WallBounce / GF vs DrussGT | 5.6 / 3.0 / 2.9 / 2.7 / 2.1 % | given |
| BitBrain when idle | == Pattern (30 runs/arm: 97/210 vs 97/210 wins) | `docs/bitbrain_vs_tmhorizon_ab.md` |
| BitBrain across 32 legacy opponents | neutral: **-3.6 dmg/run**, +0.02 wins/run, p=0.53/0.91 | `docs/gauntlet_bitbrain_vs_pattern.md` |
| lead **amplitude** | dead axis — full gain sweep {0, 0.25-1.0, 1.0, 1.25, 1.5, 2.0} run, nothing beats Pattern | `docs/bitbrain_campaign.md` |
| offline prediction checks | **veto-only** — may kill a design, never select one | `docs/offline_harness_trust.md` |
| spinner-specific claim | **UNTESTED**: no constant-turn spinner exists in the legacy roster (a note for a later job, not built here unless nearly free) | `docs/gauntlet_bitbrain_vs_pattern.md` |

**Movement context (why the pin):** the movement champion is `TR_MOVEMENT=strafe`
at its current defaults (range 325, tol 25, tilt 15/0.10); it replicated its win
over the shipped `tfil` in four independent sessions (Batch 4: 52.6% vs 38.5%
round-win rate). Movement is held constant at that champion for every gun arm.

---

## Batch 1 — does anything beat the shipped `Pattern`?

**Design.** One frozen binary, six env-only arms, one frozen panel
(`tools/ab/panel_movement.txt`, 15 opponents: 5 dodger, 3 pattern, 1
wall-follower, 1 corner-camper, 1 spinner, 1 rammer, 1 brawler, 2 aggressive
megas), **3 runs × 3 rounds per (opponent, arm) = 270 battles**. Arm file:
`tools/ab/arms_gun_b1.txt`. Reference: `pattern`.

| # | arm | env | what it isolates |
|---|---|---|---|
| 1 | `pattern` | `TR_MOVEMENT=strafe` | the arm to beat (shipped `onlyPattern` rack) |
| 2 | `bitbrain` | `+ TR_RACK_PATTERN=off TR_RACK_BITBRAIN=both TR_BITBRAIN_GAINS=1.0,1.25,1.5,2.0 TR_BITBRAIN_MEM=decay` | BitBrain-only, learned gain (panel replication of the 32-opp gauntlet at the pinned-movement standard) |
| 3 | `tmhorizon` | `+ TR_RACK_PATTERN=off TR_RACK_TMHORIZON=both` | TMHorizon-only corrector (predicted to lose; never panel-tested) |
| 4 | `knn` | `+ TR_RACK_PATTERN=off TR_RACK_KNN=both` | best non-Pattern single gun (never panel-tested) |
| 5 | `rack_pk` | `+ TR_RACK_KNN=both` | Pattern + KNN, **selector ON** (different-family hedge) |
| 6 | `rack_pt` | `+ TR_RACK_TMHORIZON=both` | Pattern + TMHorizon, **selector ON** (corrector-family hedge) |

*(`+` = the pinned `TR_MOVEMENT=strafe` is present in every arm; see §1.)*

**Pre-registered prediction (written BEFORE the battles finished):**
1. **No arm beats `pattern` on BOTH primaries**, and the point estimates sit
   inside the MDE. `onlyPattern` is confirmed across the panel.
2. `bitbrain` is a **wash** vs `pattern` (replicating the 32-opponent gauntlet:
   ±3.6 dmg/run, ≈0 wins) — the DrussGT-only gain-config penalty does not carry.
3. `tmhorizon` is **WORSE** than `pattern` (its own offline work predicts it
   needs ~80% side accuracy and can reach ~60%); no damage win, no win win.
4. `knn` is **WORSE** than `pattern` on damage (its DrussGT hit rate is roughly
   half Pattern's) and not better on wins.
5. The two small racks **do not beat `pattern`**; `rack_pk` lands closer to
   Pattern than `knn`-alone does (the selector at least partially hedges back),
   but the selector's poor ranking keeps it below the reference.
6. If any surprise exists, it is a **rack arm on damage without wins** — the
   `ring`-mover mirror-image trap — and it will not be read as a win.

### Outcome — direct answer

**MEASURED.** Session `/tmp/ab/j121_g1`, commit `1d8143a15e4a038c39dfc2153ff518fa003b8612`
(the pre-registration commit), frozen binary sha256 `aa49a45fec20…`, 15 opponents
× 6 arms × 3 runs × 3 rounds = **270 battles, 0 failed, 0 never started, 0
liveness exclusions**. Every arm declared `TR_MOVEMENT=strafe`, verified in each
bot's own raw-env report; the reference rack line reads `PATTERN` for `pattern`
and the overridden rack reads `BITBRAIN` / `TMHORIZON` / `KNN` for the others.

> **DIRECT ANSWER (MEASURED, n=15 opponents): NO — nothing beats the shipped
> `Pattern` across the panel on damage/run AND round wins.** Every arm except
> `knn` is **NOT DISTINGUISHABLE** from `pattern` under both the strict and the
> substantive reading of the pre-registered rule. The best nominal challenger,
> `tmhorizon`, is **+0.13 wins/run** (MDE 0.36) and **+8.7 dmg/run** (MDE 13.8)
> — an effect ~3× smaller than the design can detect — and `bitbrain` is
> `+0.11` wins / `+10.3` dmg, also inside the MDE. `knn` is **detectably WORSE**
> on both primaries. The shipped `onlyPattern` rack is therefore **CONFIRMED**
> rather than merely assumed, at the resolution of this panel. That is a real
> result, not a failure.

#### Pooled dashboard (all valid runs — explanation only, NOT the verdict)

| arm | runs | dmg/run | dmg taken/run | wins/run | round wins | win rate | incoming hit rate | mean distance |
|---|---:|---:|---:|---:|---:|---:|---:|---:|
| `pattern` (REF) | 45 | 99.6 | 155.9 | 1.51 | 68/135 | 50.4% | 12.84% | 435 |
| `bitbrain` | 45 | 109.9 | 151.2 | 1.62 | 73/135 | 54.1% | 12.87% | 427 |
| `tmhorizon` | 45 | 108.3 | 153.7 | **1.64** | **74/135** | **54.8%** | **12.53%** | 428 |
| `knn` | 45 | 60.3 | 175.7 | 0.89 | 40/135 | 29.6% | 13.31% | 430 |
| `rack_pk` | 45 | 107.3 | 161.3 | 1.42 | 64/135 | 47.4% | 13.65% | 432 |
| `rack_pt` | 45 | 105.3 | 149.9 | 1.64 | 74/135 | 54.8% | 12.90% | 424 |

#### Cross-opponent aggregation (the verdict layer, damage + wins)

| arm | metric | mean Δ | spread (SD) | 95% CI | sign test (wins/n) | p(sign) | p(sign-flip) | Wilcoxon p | MDE |
|---|---|---:|---:|---|---:|---:|---:|---:|---:|
| `bitbrain` | damage | +10.34 | 16.65 | [+1.12, +19.57] | 11/15 | 0.1185 | 0.03247 | 0.05006 | 12.05 |
| `bitbrain` | wins | +0.11 | 0.48 | [-0.16, +0.38] | 6/9 | 0.5078 | 0.4922 | 0.4764 | 0.35 |
| `tmhorizon` | damage | +8.71 | 19.10 | [-1.86, +19.29] | 11/15 | 0.1185 | 0.09918 | 0.1183 | 13.82 |
| `tmhorizon` | wins | +0.13 | 0.50 | [-0.14, +0.41] | 5/9 | 1.0 | 0.4062 | 0.3118 | 0.36 |
| `knn` | damage | −39.33 | 23.15 | [−52.15, −26.51] | 0/15 | 6.1e-5 | 6.1e-5 | 0.0007 | 16.75 |
| `knn` | wins | −0.62 | 0.59 | [−0.95, −0.30] | 1/11 | 0.01172 | 0.001953 | 0.004948 | 0.43 |
| `rack_pk` | damage | +7.68 | 32.09 | [−10.10, +25.45] | 9/15 | 0.6072 | 0.423 | 0.5895 | 23.21 |
| `rack_pk` | wins | −0.09 | 0.60 | [−0.42, +0.24] | 6/11 | 1.0 | 0.6738 | 0.5932 | 0.43 |
| `rack_pt` | damage | +5.69 | 16.50 | [−3.45, +14.83] | 10/15 | 0.3018 | 0.2026 | 0.2681 | 11.93 |
| `rack_pt` | wins | +0.13 | 0.41 | [−0.10, +0.36] | 6/10 | 0.7539 | 0.3242 | 0.1997 | 0.30 |

#### The pre-registered verdict (verbatim)

| rank | arm | Δwins/run | Δdmg/run | sign test wins | sign test dmg | verdict (strict) | verdict (substantive) |
|---:|---|---:|---:|---|---|---|---|
| 1 | `tmhorizon` | +0.13 | +8.7 | 5/9 p=1 | 11/15 p=0.1185 | **not distinguishable** | **not distinguishable** |
| 2 | `rack_pt` | +0.13 | +5.7 | 6/10 p=0.7539 | 10/15 p=0.3018 | **not distinguishable** | **not distinguishable** |
| 3 | `bitbrain` | +0.11 | +10.3 | 6/9 p=0.5078 | 11/15 p=0.1185 | **not distinguishable** | **not distinguishable** |
| 4 | `rack_pk` | −0.09 | +7.7 | 6/11 p=1 | 9/15 p=0.6072 | **not distinguishable** | **not distinguishable** |
| 5 | `knn` | −0.62 | −39.3 | 1/11 p=0.01172 | 0/15 p=6.104e-05 | **WORSE** | **not distinguishable** |

Reference `pattern`: 99.6 dmg/run, 1.51 wins/run, 12.84% incoming, 435 px.

#### Reading (MEASURED, with the mechanism)

* **`knn` is a clean, large loss and is dropped.** It deals **−39 dmg/run** and
  loses **−0.62 wins/run**, positive on **0/15** opponents on damage and 1/11 on
  wins (p=6e-5 / p=0.012). Its incoming hit rate is *no worse* (13.31% vs
  12.84%) — it simply cannot aim: the same movement, a worse gun.
* **The two small racks do not beat `pattern`.** `rack_pk` (Pattern+KNN,
  selector ON) is *not* the average of Pattern and KNN: it recovers most of KNN's
  damage loss (+7.7 vs KNN-alone's −39.3) but still lands **nominally below**
  Pattern on wins (−0.09). `rack_pt` (Pattern+TMHorizon) is statistically
  indistinguishable from `tmhorizon`-alone — adding Pattern to the rack changes
  nothing, i.e. the selector picks the corrector essentially always. **The
  selector's 16-gun negative verdict reproduces at a 2-gun rack**: it does not
  manufacture a win it did not have.
* **The only signal is a damage-side hint on the two Pattern-lead correctors**
  (`bitbrain` +10.3, `tmhorizon` +8.7, both 11/15 opponents). It is *below the
  MDE* (12.0 / 13.8) and *not significant by the pre-registered sign test*
  (p=0.1185), so it is **not a win and not a null**. Note the sign-flip test on
  the *mean* is nominally significant for `bitbrain` (p=0.032) while the sign
  test is not — the positive deltas are larger than the negatives — but with 5
  arms compared this is not compelling after multiplicity, and the effect is
  under the MDE. **This is exactly the "damage without wins" pattern this
  project keeps paying for** (the `ring` mover: +31 dmg/run and *fewer* wins);
  here the win deltas (+0.11/+0.13) are themselves sub-MDE.
* **The reference is stable:** `pattern` under `TR_MOVEMENT=strafe` measures
  99.6 dmg/run and 50.4% round wins here, matching the movement campaign's
  `strafe` reference in Batch 4 (103.5 dmg/run, 52.6%) — same movement, same
  panel, different job.

#### Pre-registered predictions — scorecard (an honest count)

| # | prediction | outcome |
|---|---|---|
| 1 | no arm beats `pattern` on both primaries, point estimates inside the MDE; `onlyPattern` confirmed | **CORRECT** |
| 2 | `bitbrain` is a wash vs `pattern` (gauntlet: −3.6 dmg, ≈0 wins) | **CORRECT on the verdict, WRONG on the point estimate**: +10.3 dmg here vs −3.6 in the 32-opponent gauntlet; both inside their MDEs. Direction differs, verdict (wash) holds |
| 3 | `tmhorizon` is WORSE than `pattern` (its own docs predict it loses) | **WRONG**: it is nominally *better* on both primaries (+8.7 dmg, +0.13 wins), though not distinguishable |
| 4 | `knn` is WORSE than `pattern` on damage, not better on wins | **CORRECT** (detectably on both) |
| 5 | the small racks do not beat `pattern`; `rack_pk` lands closer to Pattern than `knn`-alone | **CORRECT** |
| 6 | any surprise is a rack arm on damage without wins | **PARTLY CORRECT**: the nominal damage edge is on the single-gun correctors and the racks; every one of them is win-flat |

---

## Batch 2 — wider-panel calibration of the damage-side hint

The Batch-1 answer is decisive on the primary question (nothing beats `pattern`)
but the two corrector arms carry a sub-MDE damage hint that rule 3 forbids
calling either way. Resolving it needs **more opponents, not more runs**, so this
batch re-runs exactly those two arms plus the reference on a wider panel.

**Design.** One frozen binary, three env-only arms, panel
`tools/ab/panel_gun_b2.txt` = the j117 32-opponent legacy roster **minus Aurora**
(inert in j117: one fire in 6 battles — it cannot reveal a gun regression)
**plus SpinBot** = **33 opponents** (a superset of Batch 1's 15). **3 runs × 3
rounds = 297 battles**, `--conc 6`, `--wait-arena`. Arm file:
`tools/ab/arms_gun_b2.txt`. Reference: `pattern`. Movement pinned
`TR_MOVEMENT=strafe` in every arm (same rationale as Batch 1).

| # | arm | env | role |
|---|---|---|---|
| 1 | `pattern` | `TR_MOVEMENT=strafe` | reference |
| 2 | `bitbrain` | `+ TR_RACK_PATTERN=off TR_RACK_BITBRAIN=both TR_BITBRAIN_GAINS=1.0,1.25,1.5,2.0 TR_BITBRAIN_MEM=decay` | largest Batch-1 damage edge (+10.3) |
| 3 | `tmhorizon` | `+ TR_RACK_PATTERN=off TR_RACK_TMHORIZON=both` | second Batch-1 damage edge (+8.7) |

**Pre-registered prediction (written BEFORE the battles finished):**
1. The `+10` dmg/run hint **does not survive** the wider panel: `bitbrain` and
   `tmhorizon` damage deltas fall toward zero and stay inside the new (smaller)
   MDE; the per-opponent sign test remains non-significant. Rationale: both arms
   come from the same Pattern-lead-correction base, and the prior 32-opponent
   gauntlet measured `bitbrain` at **−3.6 dmg/run** on a largely overlapping
   roster; +10 on 15 opponents is plausibly a small-panel fluctuation.
2. Round wins remain **flat** for both arms — if anything they regress toward
   zero.
3. If instead the damage hint **survives** at a significant cross-opponent sign
   test with wins not down, it is recorded as the first gun to **beat** `pattern`
   by rule 2 — and the campaign immediately looks for what the corrector is
   exploiting. That would be a genuine up-set, and it would **not** be spun as a
   null.

### Outcome — Batch 2

**MEASURED.** Session `/tmp/ab/j121_g2`, commit
`c343c00aaf800bc05ce9bf1b277054c11f5f577c`, frozen binary sha256 `d267ab78f6ce…`,
**33 opponents × 3 arms × 3 runs × 3 rounds = 297 battles, 0 failed, 0 never
started, 0 liveness exclusions**.

**A movement-default flip landed between the two batches** (commit `3fd6db9`,
`TR_MOVEMENT` default `tfil` → `strafe`; the only source change). Because
**every** arm in **both** batches pins `TR_MOVEMENT=strafe`, the effective
movement is identical in Batch 1 and Batch 2 — the batches are directly
comparable and the gun comparison is unconfounded. This is exactly what the pin
was for.

> **DIRECT ANSWER (MEASURED, n=33 opponents): the Batch-1 damage hint did NOT
> survive.** `bitbrain` collapses to **+1.8 dmg/run** (19/33 opponents, sign test
> p=0.49) with **+0.11 wins/run** — a wash. `tmhorizon` **reverses sign and is
> detectably WORSE on damage**: **−10.4 dmg/run** (10/33 opponents, sign test
> p=0.035, sign-flip p=0.0047, 95% CI [−17.2, −3.6]) at unchanged wins (−0.03).
> The only open signal in Batch 1 was a **small-panel fluctuation**, and the
> wider panel resolves it **against** the correctors. Combined with Batch 1:
> **nothing beats the shipped `Pattern` on damage/run AND round wins**;
> `onlyPattern` is CONFIRMED on 15 opponents and again on 33.

#### Pooled dashboard (all valid runs — explanation only, NOT the verdict)

| arm | runs | dmg/run | dmg taken/run | wins/run | round wins | win rate | incoming hit rate | mean distance |
|---|---:|---:|---:|---:|---:|---:|---:|---:|
| `pattern` (REF) | 99 | 131.2 | 126.4 | 1.90 | 188/297 | 63.3% | 12.93% | 417 |
| `bitbrain` | 99 | 133.0 | 126.7 | **2.01** | **199/297** | **67.0%** | 12.91% | 417 |
| `tmhorizon` | 99 | 120.8 | 133.2 | 1.87 | 185/297 | 62.3% | 13.04% | 422 |

#### Cross-opponent aggregation (the verdict layer, damage + wins)

| arm | metric | mean Δ | spread (SD) | 95% CI | sign test (wins/n) | p(sign) | p(sign-flip) | Wilcoxon p | MDE |
|---|---|---:|---:|---|---:|---:|---:|---:|---:|
| `bitbrain` | damage | +1.80 | 18.95 | [−4.67, +8.26] | 19/33 | 0.4869 | 0.5914 | 0.5675 | 9.24 |
| `bitbrain` | wins | +0.11 | 0.48 | [−0.05, +0.28] | 12/19 | 0.3593 | 0.2436 | 0.5716 | 0.24 |
| `tmhorizon` | damage | **−10.40** | 20.01 | **[−17.22, −3.57]** | 10/33 | **0.03508** | **0.00472** | **0.006975** | 9.76 |
| `tmhorizon` | wins | −0.03 | 0.59 | [−0.23, +0.17] | 9/19 | 1.0 | 0.8459 | 0.4804 | 0.29 |

#### The pre-registered verdict (verbatim)

| rank | arm | Δwins/run | Δdmg/run | sign test wins | sign test dmg | verdict (strict) | verdict (substantive) |
|---:|---|---:|---:|---|---|---|---|
| 1 | `bitbrain` | +0.11 | +1.8 | 12/19 p=0.3593 | 19/33 p=0.4869 | **not distinguishable** | **not distinguishable** |
| 2 | `tmhorizon` | −0.03 | −10.4 | 9/19 p=1 | 10/33 p=0.03508 | **WORSE** | **WORSE** |

Reference `pattern`: 131.2 dmg/run, 1.90 wins/run, 12.93% incoming, 417 px.

#### Reading

* **BitBrain is a wash — now measured three times.** vs the shipped Pattern:
  −3.6 dmg/run on the 32-opponent legacy gauntlet (with `tfil` movement,
  `docs/gauntlet_bitbrain_vs_pattern.md`), **+10.3** on the 15-opponent Batch-1
  panel, **+1.8** here on 33 opponents. All three sit inside their own MDEs; the
  pooled evidence is a zero. The Batch-1 `+10` was a 15-opponent fluctuation.
* **TMHorizon is the one arm the wider panel separates — and it loses.** Its
  Batch-1 sign (+8.7) does not merely fail to replicate, it **inverts**
  (−10.4, 10/33, p=0.035). The damage loss is broad (24/33 opponents negative,
  worst `Aristocles` −61, `Jen` −54, `WallAvoider` −41), so it is not one
  outlier. Since TMHorizon was already predicted to lose by its own design docs,
  this is the first *confirming* live panel measurement of that prediction.
* **Wins are flat for both arms** (+0.11 / −0.03, MDEs 0.24 / 0.29): the
  correctors change damage at most, never survival. In this harness round wins
  are survival wins, so neither is a win.
* **The corrector family is closed** as a source of a `pattern`-beating gun at
  the resolutions tested.

#### Pre-registered predictions — scorecard

| # | prediction | outcome |
|---|---|---|
| 1 | the +10 dmg hint does not survive the wider panel; deltas fall inside the new MDE; sign test non-significant | **CORRECT** (`bitbrain` +1.8, p=0.49; `tmhorizon` even reversed to −10.4, p=0.035) |
| 2 | wins remain flat | **CORRECT** (+0.11 / −0.03) |
| 3 | if the hint survived at p<0.05 with wins not down, it is the first gun to beat `pattern` | **not invoked** |

**Final direct answer for the phase (MEASURED).** Across two frozen panels
(15 and 33 opponents), a single frozen-binary-per-session design, 567 battles,
six distinct gun configurations (five challengers + the reference), and movement
pinned identically in every arm:
**no rack gun and no small rack beats the shipped `Pattern` on damage/run AND
round wins.** Every candidate is either not distinguishable (`bitbrain`,
`tmhorizon`, `rack_pk`, `rack_pt` at n=15), a detectable loss
(`knn`; `tmhorizon` at n=33), or a sub-MDE damage-only wobble
(`bitbrain`). The shipped **`onlyPattern` rack is CONFIRMED**, not merely
assumed — the gun axis's first phase ends in a measured optimum of the tested
design space.


---

## What to try next (rewritten AFTER Batches 1–2 — recommendations, not results)

**The primary question is answered:** nothing beats the shipped `pattern` across
the frozen panel, and `onlyPattern` is CONFIRMED. Ranked by value per battle:

1. **The damage-side hint is now CLOSED.** Batch 2 (33 opponents) killed it:
   `bitbrain` is a wash (+1.8 dmg/run, p=0.49) and `tmhorizon` is detectably
   **worse** (−10.4 dmg/run, p=0.035). Do **not** spend another batch on the
   Pattern-lead correctors; three independent measurements of `bitbrain` vs
   Pattern now average ~0 (−3.6 / +10.3 / +1.8 dmg/run, all inside their MDEs).
2. **Lead INFORMATION remains the only named gun axis** (given: amplitude is
   dead). But the two live implementations of it (BitBrain's gain, TMHorizon's
   shift) are both measured negative/neutral across panels. A future axis would
   need a *different* information source, not another knob on these two — e.g. a
   new base prediction, or a gun-agnostic ensemble. Do not re-open the correctors.
3. **The selector is confirmed negative at a 2-gun rack** (`rack_pt` ==
   `tmhorizon`; `rack_pk` below `pattern` on wins). Do not re-open it with a
   larger rack: the 16-gun negative already stands and Batch 1 shows the
   mechanism does not manufacture a win.
4. **Drop `knn`** (detectably worse on both primaries, 0/15 on damage).
5. **Never chase damage without wins.** This harness's round wins are survival
   wins, so the gun's job is to *kill*; a damage gain that does not raise the win
   rate is the `ring`-mover trap in a new costume (and `tmhorizon`'s n=33 result
   is a *negative* damage signal with flat wins — also not a win).
6. **Spinner-specific claim (owner's) remains UNTESTED.** The legacy roster has
   **no constant-turn spinner**; SpinBot is a periodic circle-mover and is in
   both panels (Batch 1: `bitbrain` +25.3 dmg / +0 wins, `tmhorizon` +15.2 / +0
   wins; on the 33-panel SpinBot's deltas are in the per-opponent tables).
   Building a true constant-turn spinner opponent is a nearly-free follow-up and
   is the one part of the owner's claim this campaign has not touched.

## What would make us stop

**PHASE STATUS (updated after Batch 2): STOPPED — the condition fired.** Across
Batches 1–2 no arm beats `pattern` beyond the MDE, and no arm shows a
`>= +MDE` damage gain with p<0.10 (`bitbrain` +1.8 vs MDE 9.24; `tmhorizon`
actually negative). The gun stage's first phase is therefore **closed** with
*"the shipped `onlyPattern` rack is the best gun configuration we have measured
across the panels (15 and 33 opponents)"* — a successful outcome, not a failure.
Further gun work must open a **named new axis** (see *What to try next*), not
another arm on the same axis.

* **Stop the first phase** once a batch's best arm cannot beat `pattern` beyond
  the MDE, or when a gun arm's damage gain is bought with a detectable win loss.
  At that point *"`onlyPattern` is the measured optimum of this rack"* is the
  conclusion, not a failure (§2 rule 4). **This fired after Batch 2.**
* **Stop a single batch early** only for a contract violation (arena not free,
  liveness FAIL, non-zero exit rate) — never because the numbers look boring.
* **Do not invent more arms on the same axis** once two consecutive batches fail
  to improve on `pattern` beyond the MDE; move to a *named* new axis instead
  (candidate list in "What to try next").

---

## How to run a batch (exact commands)

```sh
# 0. wait for the arena (other campaign jobs may be fighting)
TOURNAMENT_NIMCACHE=/tmp/nc_j121 \
tools/ab/tournament_run.sh \
    --arms    tools/ab/arms_gun_b1.txt \
    --panel   tools/ab/panel_movement.txt \
    --runs 3 --rounds 3 --conc 6 --wait-arena 45 \
    --reference pattern \
    --outdir /tmp/ab/j121_g1

# Batch 2 (the wider-panel calibration) was the same command with
#   --arms tools/ab/arms_gun_b2.txt --panel tools/ab/panel_gun_b2.txt \
#   --outdir /tmp/ab/j121_g2 --wait-arena 50

# 1. the paired per-opponent table, sign tests, MDE and the pre-registered verdict
python3 tools/ab/tournament_analyze.py /tmp/ab/j121_g1 --reference pattern
```

`--reference` may be ANY arm of the session: re-analyzing an old session with a
different reference is a free pairwise comparison with no battles.

## Session log (outdirs are in `/tmp` and are NOT committed)

| session | commit | battles | arms | verdict |
|---|---|---:|---|---|
| `/tmp/ab/j121_g1` | `1d8143a` | 270 (0 failed, 0 never started, 0 excluded) | pattern, bitbrain, tmhorizon, knn, rack_pk, rack_pt | **nothing beats the shipped `pattern`**; all arms not distinguishable except `knn` = WORSE; `onlyPattern` CONFIRMED |
| `/tmp/ab/j121_g2` | `c343c00` | 297 (0 failed, 0 never started, 0 excluded) | pattern, bitbrain, tmhorizon | **the Batch-1 damage hint does not survive**: `bitbrain` +1.8 dmg/run p=0.49 (wash); `tmhorizon` −10.4 dmg/run p=0.035 (WORSE); wins flat |

---

# Phase 2: tuning the incumbent

**Owner's mandate (phase 2):** the phase-1 campaign closed the "other gun / other
rack" design space (nothing beats the shipped `Pattern`; the correctors are a
wash or a loss). What remains OPEN is the **incumbent's own tuning**: Pattern's
match-length / history parameters had NEVER been swept. Phase 2 asks the direct
question:

> **Can the shipped `Pattern` be improved by tuning its own match parameters —
> and if so, by how much on damage/run and round wins?**

The lead-amplitude axis is already known dead (`docs/bitbrain_campaign.md`), and
the radial knobs are known non-winners (job j99: `bmPath` structural no-op; live
+0.28 pp p=0.62 for scale 0.98, −0.42 pp p=0.46 for offset −20). Those are
therefore **controls** here, not candidates.

## Task A — exposing Pattern's match-shape parameters (what and why)

**MEASURED (code read).** `common_libs/guns/pattern_matcher.nim` had exactly two
tunable knobs, both RADIAL (`TR_PATTERN_RAD_SCALE`, `TR_PATTERN_RAD_OFFSET`), and
neither can change the lead bearing (job j99 proved the bearing is untouched), so
neither was ever the "lead information" axis. The parameters that actually
control **how the pattern is matched** were compile-time constants:

| parameter | code | what it controls | exposed as |
|---|---|---|---|
| match-key length | `PatternLen = 10` | the length of the movement segment compared (the search key); also how far after the match the replay starts | `TR_PATTERN_LEN` (int, default 10) |
| search depth | implicit `HistorySize = 500` | how far back the best-match scan may reach (`scanEnd` was always `count−PatternLen−1`) | `TR_PATTERN_DEPTH` (int, default 500 = full buffer) |
| similarity/search radius | **does not exist** | the search always takes the single lowest-cost match; there is no acceptance threshold or radius to expose | **not exposed — nothing to expose** |
| history buffer capacity | `HistorySize = 500` | the fixed `array[HistorySize]` backing store | runtime depth limit only; the buffer ceiling cannot be raised at runtime (see below) |

**Why env and not `-d:`.** The phase-1 instrument (`tools/ab/tournament_run.sh`)
builds ONE frozen binary from `git archive HEAD` and every arm differs only by its
env dict. A `{.intdefine.}` knob would need one binary per arm, which the
instrument forbids. Both new knobs are therefore resolved lazily from the
environment on the first `predict`, exactly like the radial knobs, and default to
the pre-knob constants. `setMatchParams(patternLen, histDepth)` is the explicit
offline/unit-test twin that writes the same fields.

**What could NOT be exposed cheaply (MEASURED).** `HistorySize` sizes four fixed
`array[HistorySize(+1)]` fields in the gun object. Raising it above 500 at
runtime is impossible without a heap buffer; the runtime `TR_PATTERN_DEPTH` knob
therefore *lowers* the effective search depth within the existing 500-entry
buffer. A depth above 500 is clamped to 500, and a non-positive or unparsable
value falls back to the shipped default. `PatternLen` is clamped to
`1..HistorySize`.

## Task A — default parity (MEASURED, byte-for-byte)

* **Baseline-vs-new parity dump.** A throwaway harness replayed 800 ticks of
  `tools/fixtures/drussgt_vs_crazy.jsonl` through one `PatternMatcherGun` at all
  four power bins (3200 predictions) and printed every point at full precision.
  Compiled once against a `HEAD` worktree (`/tmp/j123_base`, before the change)
  and once against the modified tree: **the two dumps are byte-identical**
  (`diff -q` clean). The shipped default path is unchanged.
* **The knobs are live and self-falling-back.** `TR_PATTERN_LEN=6/16` and
  `TR_PATTERN_DEPTH=100/20` each change the dump; `TR_PATTERN_LEN=banana`
  reproduces the default dump exactly.
* **Existing guards pass** (no new guard tests, per the "cut ceremony" rule):
  `common_libs/tests/test_pattern_radial_offset.nim` (6 checks) and
  `common_libs/tests/test_gun_harness.nim` (all checks) pass; the env-report
  guard `test_env_report.nim` passes with `TR_PATTERN_LEN` / `TR_PATTERN_DEPTH`
  registered in `knownEnvNames()` and the effective-values report.
* **Clean-archive compile** is exercised by `tournament_run.sh` itself, which
  builds the frozen binary from `git archive HEAD`.

## Task B — Batch 1 (pre-registered BEFORE any battle)

**Design.** One frozen binary, six env-only arms, the FROZEN 15-opponent panel
`tools/ab/panel_movement.txt`, **3 runs × 3 rounds per (opponent, arm) = 270
battles**, `--conc 6`, `--wait-arena`. Arm file: `tools/ab/arms_gun_b3.txt`.
Reference: `pattern`. Movement pinned `TR_MOVEMENT=strafe` in every arm.

| # | arm | env over the pin | what it isolates |
|---|---|---|---|
| 1 | `pattern` | (none) | the arm to beat (shipped `onlyPattern` rack) |
| 2 | `len6` | `TR_PATTERN_LEN=6` | shorter movement segment compared |
| 3 | `len16` | `TR_PATTERN_LEN=16` | longer movement segment compared |
| 4 | `depth100` | `TR_PATTERN_DEPTH=100` | shallower history / match search |
| 5 | `rad_offset` | `TR_PATTERN_RAD_OFFSET=-20` | control: aim 20 px short |
| 6 | `rad_scale` | `TR_PATTERN_RAD_SCALE=0.95` | control: scale the aim distance |

**Pre-registered prediction (written BEFORE the battles finished):**
1. **No arm beats `pattern` on both primaries.** The incumbent is already tuned —
   the phase-2 null. The point estimates sit inside the MDE.
2. `len16` is **WORSE or flat** on damage: a 16-tick key matches rarely in a
   ~500-tick buffer, so the gun falls back to the linear forecast (the weaker
   base) more often; wins flat.
3. `len6` is **flat** on damage (more matches but a noisier replay) and flat on
   wins; possibly a sub-MDE wobble in either direction.
4. `depth100` is **not distinguishable** from `pattern` — the full buffer already
   contains the useful candidates; a sub-MDE damage wobble is possible.
5. The two radial controls are **not distinguishable** and, per job j99, cannot
   change the real aim bearing; any damage delta is the fire-gate channel only.
6. If any surprise exists, it is a **damage-only wobble without wins** — the
   `ring`-mover trap — and will not be read as a win.

**Pre-registered Batch-2 trigger (task rule).** A Batch-1 arm is promoted to a
higher-power Batch 2 **only if** it beats the reference `pattern` by the
campaign rule 2 (one primary up at cross-opponent sign-test p<0.05 while the
other does not go down) **AND** its damage CI excludes 0 **AND** the sign-flip
permutation test gives p<0.05. Otherwise Batch 1 is the answer.


### Outcome — Batch 1

**MEASURED.** Session `/tmp/ab/j123_b1`, commit
`2a98aba91b0ec023f580cd2485cafba2404d275d`, frozen binary sha256 `287d292fb7d1…`,
**15 opponents × 6 arms × 3 runs × 3 rounds = 270 battles, 0 failed, 0 never
started, 0 liveness exclusions**. Arena serialization held: the parallel job
j124 (`/tmp/ab/j124_spinner.run.log`) detected this session's processes at
04:08–04:18 and **waited**; no foreign battle ran concurrently. Every arm
declared `TR_MOVEMENT=strafe`; the new `TR_PATTERN_LEN` / `TR_PATTERN_DEPTH`
tokens appear verbatim in each overriding arm's own raw-env report and the
reference arm shows the shipped defaults 10 / 500.

#### Pooled dashboard (all valid runs — explanation only, NOT the verdict)

| arm | runs | dmg/run | dmg taken/run | wins/run | round wins | win rate | incoming hit rate | mean distance |
|---|---:|---:|---:|---:|---:|---:|---:|---:|
| `pattern` (REF) | 45 | 101.1 | 161.0 | 1.27 | 57/135 | 42.2% | 13.27% | 428 |
| `len6` | 45 | 110.8 | 151.2 | **1.76** | **79/135** | **58.5%** | **12.28%** | 424 |
| `len16` | 45 | 103.3 | 158.8 | 1.53 | 69/135 | 51.1% | 13.07% | 427 |
| `depth100` | 45 | 99.9 | 154.5 | 1.53 | 69/135 | 51.1% | 12.77% | 429 |
| `rad_offset` | 45 | 107.9 | 157.2 | 1.69 | 76/135 | 56.3% | 13.21% | 431 |
| `rad_scale` | 45 | 105.9 | 157.1 | 1.38 | 62/135 | 45.9% | 12.62% | 429 |

#### Per-opponent paired deltas (arm − `pattern`), damage and wins

| opponent | style | Δdmg_len6 | Δwins_len6 | Δdmg_len16 | Δwins_len16 | Δdmg_depth100 | Δwins_depth100 | Δdmg_rad_offset | Δwins_rad_offset | Δdmg_rad_scale | Δwins_rad_scale |
|---|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|
| DrussGT | dodger | +8.4 | +0.67 | -21.7 | -0.33 | +3.6 | +0.67 | +10.5 | +1.33 | +18.7 | +0.67 |
| Diamond | dodger | -22.3 | +0.00 | -18.1 | -0.33 | -54.7 | -0.33 | -30.9 | +0.00 | -17.0 | -0.33 |
| Dookious | dodger | -6.6 | +1.33 | +7.6 | +1.00 | -0.8 | +0.67 | +2.6 | +0.67 | -15.6 | -0.33 |
| GresSuffurd | dodger | +32.3 | +0.00 | +31.9 | +0.00 | -6.6 | +0.00 | +28.4 | +0.67 | -0.9 | +0.00 |
| CassiusClay | dodger | +15.0 | +0.67 | +6.8 | +1.00 | +10.0 | +0.33 | +12.5 | +0.67 | +40.4 | +1.00 |
| RetroGirl | pattern | +48.0 | +1.33 | +20.3 | +0.67 | +59.9 | +1.00 | +4.1 | +0.00 | +55.5 | +0.33 |
| TripHammer | pattern | +6.1 | +0.00 | +0.1 | +0.33 | -21.6 | +0.00 | -9.3 | +0.00 | +4.7 | +0.33 |
| Coriantumr | pattern | +1.9 | -0.33 | -3.9 | +0.33 | +2.0 | +1.00 | -5.3 | -0.33 | -10.3 | -0.33 |
| WallAvoider | wallfollower | +28.5 | +1.00 | -23.5 | +0.00 | +41.0 | +0.00 | +23.6 | +1.00 | -24.0 | +0.00 |
| HawkOnFire | cornercamper | -8.9 | +0.67 | -9.5 | +0.00 | -9.3 | -0.33 | -12.0 | +0.33 | -5.0 | +0.33 |
| SpinBot | spinner | +5.7 | +0.00 | +13.3 | +0.00 | -13.5 | +0.00 | -1.7 | +0.00 | -0.3 | +0.00 |
| DiamondStealer | rammer | -16.1 | +0.00 | -9.8 | +0.00 | -14.6 | +0.00 | +31.8 | +1.33 | +1.1 | -0.67 |
| BlitzBat | brawler | +9.1 | +0.00 | +17.8 | +0.33 | -6.3 | +0.33 | +9.6 | +0.33 | +9.2 | -0.67 |
| YersiniaPestis | aggressive | +21.0 | +1.33 | +2.6 | +0.67 | -18.6 | +0.33 | +27.4 | +0.67 | +3.0 | +0.67 |
| Ascendant | aggressive | +22.4 | +0.67 | +18.7 | +0.33 | +10.9 | +0.33 | +9.3 | -0.33 | +12.6 | +0.67 |

#### Cross-opponent aggregation (the verdict layer, damage + wins)

| arm | metric | mean Δ | spread (SD) | SE | 95% CI | sign test (wins/n) | p(sign) | p(sign-flip) | Wilcoxon p | MDE |
|---|---|---:|---:|---:|---|---:|---:|---:|---:|---:|
| `len6` | damage | +9.63 | 18.97 | 4.90 | [−0.88, +20.13] | 11/15 | 0.1185 | 0.06976 | 0.09384 | 13.72 |
| `len6` | wins | **+0.49** | 0.58 | 0.15 | **[+0.17, +0.81]** | 8/9 | **0.03906** | **0.007812** | **0.01269** | 0.42 |
| `len16` | damage | +2.18 | 16.62 | 4.29 | [−7.02, +11.38] | 9/15 | 0.6072 | 0.6142 | 0.712 | 12.02 |
| `len16` | wins | +0.27 | 0.42 | 0.11 | [+0.03, +0.50] | 8/10 | 0.1094 | 0.04688 | 0.01852 | 0.30 |
| `depth100` | damage | −1.24 | 26.52 | 6.85 | [−15.93, +13.45] | 6/15 | 0.6072 | 0.8677 | 0.5137 | 19.19 |
| `depth100` | wins | +0.27 | 0.42 | 0.11 | [+0.03, +0.50] | 8/10 | 0.1094 | 0.04688 | 0.046 | 0.30 |
| `rad_offset` | damage | +6.71 | 17.18 | 4.44 | [−2.80, +16.23] | 10/15 | 0.3018 | 0.1522 | 0.1323 | 12.43 |
| `rad_offset` | wins | +0.42 | 0.54 | 0.14 | [+0.12, +0.72] | 9/11 | 0.06543 | 0.01465 | 0.01424 | 0.39 |
| `rad_scale` | damage | +4.80 | 21.10 | 5.45 | [−6.89, +16.49] | 8/15 | 1 | 0.4095 | 0.6293 | 15.27 |
| `rad_scale` | wins | +0.11 | 0.51 | 0.13 | [−0.17, +0.40] | 7/12 | 0.7744 | 0.5107 | 0.289 | 0.37 |

#### The pre-registered verdict (verbatim)

| rank | arm | Δwins/run | Δdmg/run | sign test wins | sign test dmg | verdict (strict) | verdict (substantive) |
|---:|---|---:|---:|---|---|---|---|
| 1 | `len6` | +0.49 | +9.6 | 8/9 p=0.03906 | 11/15 p=0.1185 | **BETTER** | **BETTER** |
| 2 | `rad_offset` | +0.42 | +6.7 | 9/11 p=0.06543 | 10/15 p=0.3018 | **not distinguishable** | **not distinguishable** |
| 3 | `len16` | +0.27 | +2.2 | 8/10 p=0.1094 | 9/15 p=0.6072 | **not distinguishable** | **not distinguishable** |
| 4 | `depth100` | +0.27 | −1.2 | 8/10 p=0.1094 | 6/15 p=0.6072 | **not distinguishable** | **not distinguishable** |
| 5 | `rad_scale` | +0.11 | +4.8 | 7/12 p=0.7744 | 8/15 p=1 | **not distinguishable** | **not distinguishable** |

Reference `pattern`: 101.1 dmg/run, 1.27 wins/run, 13.27% incoming, 428 px.

#### Reading and the single red flag

* **`len6` is the only arm that beats `pattern` by the campaign rule** — on the
  WINS primary: +0.49 wins/run, sign test 8/9 p=0.0391, sign-flip p=0.0078,
  wins CI [+0.17, +0.81] excluding 0 — while damage is nominally up (+9.6, but
  inside the 13.7 MDE and its CI includes 0). A 6-tick match key is broadly
  positive (10/15 opponents non-negative, only `Coriantumr` −0.33 negative).
* **RED FLAG (MEASURED): all five arms are nominally positive on wins**
  (+0.49 / +0.27 / +0.27 / +0.42 / +0.11), including `rad_offset`, which by job
  j99's exact structural proof **cannot change the aim bearing** and whose live
  effect was measured as a tie (+0.28 pp, p=0.62). The reference `pattern` also
  measures **42.2% wins here vs 50.4% in j121 on the same 15-panel** (its damage
  is unchanged at 101–99.6), so the leading explanation for the all-arms-positive
  pattern is a **session-level low reference**, not five independent gun wins.
  The within-session paired deltas remain valid, but `len6`'s win is exactly the
  kind of single-session effect that must replicate on a second, wider panel.
* **The two shape knobs behave plausibly.** `len16` (fewer, more specific
  matches → more linear fallback) is damage-flat; `depth100` (shallower search)
  is damage-flat at a larger MDE. Neither is a detectable win.

#### Pre-registered predictions — scorecard

| # | prediction | outcome |
|---|---|---|
| 1 | no arm beats `pattern` on BOTH primaries | **CORRECT** (`len6` beats on wins only; damage CI includes 0) |
| 2 | `len16` flat/worse on damage, wins flat | **PARTLY WRONG**: damage flat (+2.2), wins nominally +0.27 |
| 3 | `len6` flat on damage and wins | **WRONG**: wins +0.49 significant; damage +9.6 sub-MDE |
| 4 | `depth100` not distinguishable | **CORRECT** (damage −1.2), though wins nominally +0.27 |
| 5 | radial controls not distinguishable / bearing-invariant | **CORRECT**: `rad_scale` flat; `rad_offset` nominal +0.42 wins but sign p=0.065 (not significant) |
| 6 | any surprise is damage-only without wins | **WRONG in direction**: the surprise is wins-with-nominal-damage |

## Batch 2 — higher-power confirmation of `len6` (+ the structural control)

**Why we run it (task trigger, with one recorded deviation).** The task's Batch-2
trigger is: an arm beats the reference with **the winning primary's CI excluding
0 AND sign-flip p<0.05**. `len6` satisfies it on wins (CI [+0.17, +0.81];
p(sign-flip)=0.0078). *Deviation, recorded:* the Batch-1 pre-registration added a
stricter clause — "its **damage** CI excludes 0" — which `len6`'s damage
([−0.88, +20.13]) does not meet. We run Batch 2 anyway because (a) the task rule
keys on the **winning** primary and (b) the all-arms-positive red flag means a
replication is the only way to tell a real `len6` effect from a session-level
reference anomaly. This is a deviation from the letter of my own pre-registration,
not from the task's criterion.

**Design.** One frozen binary, four env-only arms, the wider panel
`tools/ab/panel_gun_b2.txt` (**33 opponents**, superset of Batch 1's 15),
**3 runs × 3 rounds = 396 battles**, `--conc 6`, `--wait-arena`. Arm file:
`tools/ab/arms_gun_b4.txt`. Reference: `pattern`. Movement pinned
`TR_MOVEMENT=strafe` in every arm.

| # | arm | env over the pin | role |
|---|---|---|---|
| 1 | `pattern` | (none) | reference |
| 2 | `len6` | `TR_PATTERN_LEN=6` | the candidate |
| 3 | `len16` | `TR_PATTERN_LEN=16` | dose-response on key length |
| 4 | `rad_offset` | `TR_PATTERN_RAD_OFFSET=-20` | **structural positive control** (bearing-invariant) |

**Pre-registered prediction (written BEFORE the battles finished):**
1. **`len6`'s +0.49 wins does NOT replicate.** On 33 opponents its wins delta
   collapses toward the MDE and the sign test (or the sign-flip test) fails.
   Rationale: all five Batch-1 arms were positive on wins including the
   bearing-invariant `rad_offset`, and the reference was 8 points low on win rate
   versus j121's same-panel measurement — a single-session low reference, not a
   `len6` effect.
2. **`rad_offset` is not a detectable win**; if it, too, replicates a wins gain,
   the wins axis is being moved by something other than the aim bearing and the
   whole Batch-1 wins signal is an artifact.
3. `len16` stays within its MDE on both primaries.
4. Damage stays within the MDE for every arm.
5. **If `len6` DOES replicate** (winning-metric CI excludes 0, sign-flip
   p<0.05, damage not detectably down), it is recorded as the campaign's
   **champion candidate** — and the shipped-default flip remains a separate,
   explicit decision that is **NOT** taken in this job.

### Outcome — Batch 2

**MEASURED.** Session `/tmp/ab/j123_b2`, commit
`1b59b6581075cf4019773dd74e6db9eecf2fcb02`, frozen binary sha256 `8155c39fbc6a…`,
**33 opponents × 4 arms × 3 runs × 3 rounds = 396 battles, 0 failed, 0 never
started, 0 liveness exclusions**. Arena serialization held: the job waited at
the start of the run for j124's spinner fleet (`/tmp/ab/j124_spinner`) and only
started once it was gone; no foreign battle ran concurrently with the measurement.
Every arm declared `TR_MOVEMENT=strafe`; `TR_PATTERN_LEN` / `TR_PATTERN_RAD_OFFSET`
appear verbatim in each overriding arm's raw-env report.

#### Pooled dashboard (all valid runs — explanation only, NOT the verdict)

| arm | runs | dmg/run | dmg taken/run | wins/run | round wins | win rate | incoming hit rate | mean distance |
|---|---:|---:|---:|---:|---:|---:|---:|---:|
| `pattern` (REF) | 99 | 129.5 | 124.4 | 2.11 | 209/297 | 70.4% | 12.54% | 428 |
| `len6` | 99 | 131.4 | 121.0 | 2.07 | 205/297 | 69.0% | 12.04% | 423 |
| `len16` | 99 | 125.8 | 121.4 | 2.04 | 202/297 | 68.0% | 12.36% | 426 |
| `rad_offset` | 99 | 126.9 | 132.4 | 1.91 | 189/297 | 63.6% | 13.03% | 425 |

#### Per-opponent paired deltas (arm − `pattern`), damage and wins

| opponent | style | Δdmg_len6 | Δwins_len6 | Δdmg_len16 | Δwins_len16 | Δdmg_rad_offset | Δwins_rad_offset |
|---|---|---:|---:|---:|---:|---:|---:|
| Aristocles | dodger | +0.7 | +0.00 | -12.7 | +0.00 | -5.3 | +0.00 |
| Ascendant | other | +13.8 | -0.67 | -23.1 | -1.33 | +0.7 | -1.00 |
| BlitzBat | regular | -2.6 | +0.00 | +4.4 | +0.33 | -12.0 | +0.33 |
| BrokenSword | other | -17.7 | +0.00 | +6.0 | +0.33 | -15.9 | +0.33 |
| CassiusClay | dodger | +3.7 | +0.67 | -7.5 | +0.00 | +32.4 | -0.33 |
| Cigaret | dodger | -26.3 | -2.00 | -5.1 | -0.33 | -29.1 | -1.67 |
| CigaretBH | dodger | +17.6 | +0.33 | +1.1 | +0.33 | +18.2 | +0.00 |
| Coriantumr | other | -3.9 | +0.00 | -14.1 | -0.33 | +3.9 | +0.00 |
| Diamond | dodger | -5.2 | -0.33 | -8.8 | -0.67 | -13.7 | -0.67 |
| DiamondHawk | other | +3.2 | -0.33 | -3.9 | +0.00 | -6.6 | -0.33 |
| DiamondStealer | regular | +29.5 | +1.00 | +27.4 | +0.33 | +16.5 | +0.33 |
| Dookious | dodger | +16.6 | +0.33 | +14.5 | +0.33 | -2.1 | +0.00 |
| DrussGT | other | -13.2 | +0.00 | +20.7 | +0.33 | -15.9 | -0.33 |
| FloodMini | regular | +1.0 | +0.00 | +12.0 | +0.00 | +20.7 | +0.00 |
| GresSuffurd | dodger | +10.8 | +0.33 | -11.1 | +0.00 | +4.8 | +0.00 |
| HawkOnFire | regular | +24.9 | +0.67 | +26.3 | -0.33 | -2.6 | +0.67 |
| Jen | dodger | -36.3 | +0.00 | -27.0 | +0.00 | -16.1 | -0.33 |
| Komarious | dodger | +3.3 | +0.33 | +1.2 | +0.00 | -11.4 | +0.33 |
| KurtWaveSurfer | dodger | -18.9 | -0.67 | -35.9 | +0.00 | -7.8 | +0.00 |
| LightningBug | other | -15.3 | +0.00 | -10.7 | +0.00 | -20.7 | +0.00 |
| LionWWSVMvoid | dodger | +19.1 | +0.00 | -1.3 | +0.00 | -1.7 | +0.00 |
| Lukious | dodger | +17.6 | +0.00 | +1.8 | +0.33 | +19.6 | +0.33 |
| PatternRobot | regular | +21.3 | +0.00 | +6.7 | +0.00 | -1.3 | +0.00 |
| Phoenix | other | -8.2 | +0.33 | -29.5 | -0.33 | -4.4 | -0.67 |
| RetroGirl | other | -29.6 | -0.33 | -6.7 | +0.00 | -17.0 | -1.67 |
| RougeDC | dodger | +7.5 | +0.00 | -1.9 | -0.33 | +2.3 | +0.00 |
| Shadow | other | +3.1 | -0.67 | -10.9 | -0.67 | -10.4 | -0.67 |
| SpinBot | spinner | -14.7 | +0.00 | -9.5 | +0.00 | -3.5 | +0.00 |
| TripHammer | other | +0.9 | -0.33 | -7.6 | -0.33 | -16.8 | -1.00 |
| WallAvoider | regular | +35.7 | -0.33 | +3.3 | +0.33 | -5.6 | -1.00 |
| WaveSurferGF | dodger | +12.5 | +0.33 | -16.5 | -0.67 | +26.3 | +0.33 |
| WaveSurferPG | dodger | -1.2 | -0.67 | +0.3 | +0.33 | -22.7 | +0.00 |
| YersiniaPestis | other | +10.8 | +0.67 | -4.7 | +0.00 | +8.6 | +0.33 |

#### Cross-opponent aggregation (the verdict layer, damage + wins)

| arm | metric | mean Δ | spread (SD) | SE | 95% CI | sign test (wins/n) | p(sign) | p(sign-flip) | Wilcoxon p | MDE |
|---|---|---:|---:|---:|---|---:|---:|---:|---:|---:|
| `len6` | damage | +1.83 | 17.09 | 2.98 | [−4.00, +7.67] | 20/33 | 0.2962 | 0.5417 | 0.4859 | 8.34 |
| `len6` | wins | −0.04 | 0.54 | 0.09 | [−0.22, +0.14] | 10/20 | 1 | 0.7631 | 0.8806 | 0.26 |
| `len16` | damage | −3.72 | 14.45 | 2.52 | [−8.65, +1.22] | 13/33 | 0.2962 | 0.1494 | 0.09657 | 7.05 |
| `len16` | wins | −0.07 | 0.38 | 0.07 | [−0.20, +0.06] | 9/19 | 1 | 0.379 | 0.5303 | 0.19 |
| `rad_offset` | damage | −2.69 | 14.72 | 2.56 | [−7.72, +2.33] | 11/33 | 0.08014 | 0.2997 | 0.1952 | 7.18 |
| `rad_offset` | wins | −0.20 | 0.56 | 0.10 | [−0.39, −0.01] | 8/20 | 0.5034 | 0.05926 | 0.05875 | 0.28 |

#### The pre-registered verdict (verbatim)

| rank | arm | Δwins/run | Δdmg/run | sign test wins | sign test dmg | verdict (strict) | verdict (substantive) |
|---:|---|---:|---:|---|---|---|---|
| 1 | `len6` | −0.04 | +1.8 | 10/20 p=1 | 20/33 p=0.2962 | **not distinguishable** | **not distinguishable** |
| 2 | `len16` | −0.07 | −3.7 | 9/19 p=1 | 13/33 p=0.2962 | **not distinguishable** | **not distinguishable** |
| 3 | `rad_offset` | −0.20 | −2.7 | 8/20 p=0.5034 | 11/33 p=0.08014 | **not distinguishable** | **not distinguishable** |

Reference `pattern`: 129.5 dmg/run, 2.11 wins/run, 12.54% incoming, 428 px.

#### Reading

* **`len6`'s Batch-1 wins win did not replicate.** On 33 opponents it collapses
  from +0.49 wins/run to **−0.04 wins/run** (10/20, p=1) and from +9.6 to
  **+1.8 dmg/run** (20/33, p=0.30, CI [−4.00, +7.67], MDE 8.34). Both primaries
  are squarely inside the MDE. Verdict: **NOT DISTINGUISHABLE**.
* **The Batch-1 red flag is explained.** All five Batch-1 arms were wins-positive
  including the bearing-invariant `rad_offset`; here `rad_offset` turns
  **negative** on both primaries (−0.20 wins, −2.7 dmg) and the other two arms
  are wins-flat. The Batch-1 effect was a **session-level low reference**:
  `pattern` measured 42.2% win rate on 15 opponents in Batch 1, versus **70.4%**
  here on its 33-opponent superset and 63.3% in j121 Batch 2 on the same 33-panel.
  A single-session paired comparison is internally valid, but a wins-only signal
  in one session is not evidence until it replicates — which is exactly what
  Batch 2 tested.
* **The dose-response is flat and slightly negative for longer keys.**
  `len16` is −3.7 dmg/run and −0.07 wins/run (damage sign test p=0.30; if
  anything it drifts toward the linear fallback as predicted in Batch 1).
* **No shipped default was touched**; this job is a measurement.

#### Pre-registered predictions — scorecard

| # | prediction | outcome |
|---|---|---|
| 1 | `len6`'s +0.49 wins does not replicate; delta collapses toward the MDE | **CORRECT** (+0.49 → −0.04 wins, p=1; damage +9.6 → +1.8) |
| 2 | `rad_offset` is not a detectable win | **CORRECT** (it is nominally *negative*: −0.20 wins, −2.7 dmg) |
| 3 | `len16` stays within its MDE on both primaries | **CORRECT** (−3.7 dmg vs MDE 7.05; −0.07 wins) |
| 4 | damage stays within the MDE for every arm | **CORRECT** |
| 5 | if `len6` replicates it is the champion candidate | **not invoked** |

## Phase 2 — final direct answer

> **MEASURED (two sessions, 15 then 33 opponents, 666 battles, 6 frozen-binary
> arms + the reference): NO — the shipped `Pattern` cannot be improved by tuning
> its own match-length or history parameters at the resolution we can measure.**
> The incumbent is already tuned on this axis.

* The one Batch-1 signal (`len6`, +0.49 wins/run on 15 opponents, sign p=0.039,
  sign-flip p=0.008) **reversed to a wash** on 33 opponents (−0.04 wins/run,
  p=1; +1.8 dmg/run, p=0.30). It was a session-level low reference, not a gun
  effect.
* On damage/run and round wins, the best Batch-2 challenger is **inside its MDE
  on both primaries** (len6: +1.8 dmg/run vs MDE 8.34, −0.04 wins/run vs MDE
  0.26). Longer keys (`len16`) and a shallower search (`depth100`) do not help.
* The two pre-existing radial knobs remain what j99 measured: bearing-invariant
  and not winners; in Batch 2 `rad_offset` is nominally *negative*.
* **No shipped default changed.** Exposing the knobs (`TR_PATTERN_LEN`,
  `TR_PATTERN_DEPTH`) does not change the default path — proven byte-identical
  before any battle.

*(MEASURED vs INFERRED. The knob exposure, the byte-identity parity dump, the
270- and 396-battle sessions, every per-opponent delta, the CIs, MDEs, sign
tests and sign-flip tests above are all MEASURED. The "incumbent is already
tuned" reading is a MEASURED null at these resolutions; any claim that no
setting whatsoever could ever help would be INFERRED and is deliberately not
made.)*

## What to try next (rewritten AFTER Phase 2)

The phase-1 conclusion stands and is now joined by the phase-2 one:

1. **The match-shape axis is CLOSED.** Match-key lengths 6 and 16 and a
   shallow (100-tick) history search do not beat the shipped defaults (10,
   full 500). Do not spend another batch on `TR_PATTERN_LEN` / `TR_PATTERN_DEPTH`
   unless a specific mechanism is named; the MDE is ~8 dmg/run and ~0.26
   wins/run, so only a very large effect could even be seen.
2. **`TR_MOVEMENT=strafe` + default `Pattern` is the measured optimum of the
   tested design space**, and Pattern is now measured-tuned on its own
   parameters as well.
3. **The named open axes remain the ones phase 1 named**: a *different*
   information source for the lead (not another knob on BitBrain/TMHorizon,
   both measured negative/neutral), and the owner's **spinner-specific claim**
   (j124 is now on it). The lead-amplitude axis and the radial axis stay dead.
4. **Session-level reference variance is large and must be respected.** The same
   frozen reference on the same movement measured 42.2% wins (Batch 1, 15
   opponents) and 70.4% wins (Batch 2, 33 opponents, superset). Any wins-only
   signal from one session must be replicated before it is believed — this
   job's Batch-1 `len6` is the second campaign example (after j121's `bitbrain`
   damage hint) of a single-session wobble dying on replication.

## Session log (Phase 2 rows appended)

| session | commit | battles | arms | verdict |
|---|---|---:|---|---|
| `/tmp/ab/j123_b1` | `2a98aba` | 270 (0 failed, 0 never started, 0 excluded) | pattern, len6, len16, depth100, rad_offset, rad_scale | `len6` the only rule-2 win (wins +0.49, sign p=0.039); **red flag: all 5 arms wins-positive incl. the bearing-invariant `rad_offset`** |
| `/tmp/ab/j123_b2` | `1b59b65` | 396 (0 failed, 0 never started, 0 excluded) | pattern, len6, len16, rad_offset | **`len6` does not replicate** (−0.04 wins, p=1; +1.8 dmg, p=0.30); all arms not distinguishable; the incumbent is already tuned |