diff --git a/crates/sds-core/src/features/firing.rs b/crates/sds-core/src/features/firing.rs
index 42d97b4..1ff4aa9 100644
--- a/crates/sds-core/src/features/firing.rs
+++ b/crates/sds-core/src/features/firing.rs
@@ -197,8 +197,13 @@ impl Feature for PKill {
     type Norm = Bounded;
     type Input<'a> = Volley<'a>;
     const NAME: &'static str = "p_kill";
+    // Names the locations, on purpose. The old sentence - "enough damage in one
+    // place to destroy the target outright" - described the computation this
+    // replaced *and* the one that replaced it, so the basis fingerprint, which
+    // hashes descriptions, could not tell them apart. A sentence that survives a
+    // change of meaning was too vague to be load-bearing.
     const DESCRIPTION: &'static str =
-        "The chance this volley does enough damage in one place to destroy the target outright.";
+        "The chance the head or centre torso is destroyed, counting only the damage that lands there.";
 
     fn measure(volley: &Volley<'_>) -> Score<Bounded> {
         if !volley.has_locations() {
@@ -216,7 +221,7 @@ impl Feature for PMissionKill {
     type Input<'a> = Volley<'a>;
     const NAME: &'static str = "p_mission_kill";
     const DESCRIPTION: &'static str =
-        "The chance this volley takes a leg and leaves the target unable to fight on.";
+        "The chance a leg is destroyed, or the target dies, counting only the damage that lands there.";
 
     fn measure(volley: &Volley<'_>) -> Score<Bounded> {
         if !volley.has_locations() {
@@ -309,7 +314,7 @@ impl Feature for Overkill {
     type Input<'a> = Volley<'a>;
     const NAME: &'static str = "overkill";
     const DESCRIPTION: &'static str =
-        "How much of this volley's damage is expected to land on a target that is already dead.";
+        "How much of this volley's damage is expected to overflow the locations it lands on.";
 
     fn measure(volley: &Volley<'_>) -> Score<Bounded> {
         let mean = volley.damage.mean();
diff --git a/plan/training.md b/plan/training.md
index bb398f2..7debfee 100644
--- a/plan/training.md
+++ b/plan/training.md
@@ -526,13 +526,30 @@ label comparison's `p_kill +0.225` at 91% agreement. Nothing in a corpus
 recorded before this change says which `p_kill` wrote it, and nothing would
 object.
 
-- [ ] Give the feature basis a version the manifest records, bumped when a
-      feature's meaning changes rather than when the set does - the same rule
-      `sds/epoch.py` applies to a match, applied to the basis. A name list
-      answers "are these the same columns"; this answers "do the columns mean
-      the same thing", and only the second question would have caught this
-      change. The epoch itself is the wrong place: it versions what a match is,
-      and the bot's own scoring is not that
+The basis fingerprint hashes each feature's name, normalisation and
+one-sentence description, which is a better question than the name list and
+still not the right one. Tested against this change, it reported the basis had
+moved - but only because six features were added. Its `REWORDED` list was
+empty: `p_kill`'s sentence was byte-identical across a change that moved its
+answer by three orders of magnitude, because "enough damage in one place to
+destroy the target outright" describes the old computation and the new one
+equally well. Had this change touched `p_kill` alone and added nothing, the
+fingerprint would have called the basis unchanged.
+
+The descriptions of `p_kill`, `p_mission_kill` and `overkill` now name the
+quantity rather than the intent, which closes this instance. It does not close
+the hole: a description is a proxy for a meaning, and a proxy that a person can
+forget to update is one they will.
+
+- [ ] Hash a `BASIS` constant in `sds-core` that a person bumps when they change
+      what a feature computes, alongside the description. The same shape as
+      `sds/epoch.py`, and for the same reason: enumerating what might have
+      changed does not work, so one number is declared instead. The epoch itself
+      is the wrong home - it versions what a match is, and the bot's own scoring
+      is not that.
+- [ ] Make a vague description fail rather than pass. A sentence that still fits
+      after the computation changed was too vague to be load-bearing, and that
+      is a property worth a review rule if not a test
 
 ## Leaving the map: not yet, and why