From 36c4334487961e89fb04e5adf44d3e2c4bb4f849 Mon Sep 17 00:00:00 2001 From: Davide Cappellini Date: Sat, 26 Sep 2026 14:48:23 +0200 Subject: [PATCH] bitbrain gun README: the gain warning was wrong and contradicted its own verdict section (it recommended the direction measured decisively harmful) --- common_libs/guns/bitbrain_gun.README.md | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/common_libs/guns/bitbrain_gun.README.md b/common_libs/guns/bitbrain_gun.README.md index 46eca8d..fe595bf 100644 --- a/common_libs/guns/bitbrain_gun.README.md +++ b/common_libs/guns/bitbrain_gun.README.md @@ -79,12 +79,17 @@ Default is **off**; the shipped rack never calls it. Minimal `.env`: ``` TR_RACK_BITBRAIN=both TR_RACK_PATTERN=off -TR_BITBRAIN_GAINS=1.0,1.25,1.5,2.0 +TR_BITBRAIN_GAINS=1.0 # 1.0 = identity = aims EXACTLY like Pattern TR_BITBRAIN_MEM=decay TR_BITBRAIN_LOG=1 ``` -⚠️ **Warning:** the **default** candidate set `{0, 0.25, 0.5, 0.75, 1.0}` can only reach 1.0 — it can only **shrink** the lead (toward HeadOn), which was measured **harmful** at long range. The useful set is the one that allows gains **above 1.0**: `1.0,1.25,1.5,2.0`. +⚠️ **`TR_BITBRAIN_GAINS=1.0` is the identity** — with one candidate at 1.0 the gun aims **exactly like Pattern**. +**Do not read the candidate list as a recommendation.** Measured live: fixed gains **above** 1.0 are +*decisively harmful* (1.5 → −93.8 dmg/run, p=0.0006; 1.25 → −34.2 dmg/run), a learner restricted to **≤ 1.0** +is a **wash** (−2.5 dmg/run, p=0.87), and zero lead (= HeadOn) is **catastrophic** (14 vs 279 dmg/run). +So the default set `{0, 0.25, 0.5, 0.75, 1.0}` is **not** a recommendation either — it merely *allows* the gun +to shrink the lead toward HeadOn. Multi-candidate lists are for running the experiment, not for playing. ## MEASURED VERDICT