diff --git a/PLANS.md b/PLANS.md
index 16fbf64..12ffdff 100644
--- a/PLANS.md
+++ b/PLANS.md
@@ -853,7 +853,7 @@ Implementation and required verification are complete.
 
 ## Milestone 010 — On-device integration and resource hardening
 
-**Status:** `IN_PROGRESS`
+**Status:** `COMPLETE`
 
 ### Objective
 
@@ -949,10 +949,42 @@ the ESP32-C6 Flash, RAM, socket, and network limits.
   (internal free 235,372 bytes, largest block 212,992 bytes, queue high-water
   10, six-client high-water 6). Full repeated-match and station/provisioning
   acceptance evidence is not yet recorded; Milestone 010 remains `IN_PROGRESS`.
+- On 2026-09-08, the project owner ran
+  `./scripts/ws_protocol_client.py 192.168.1.50 --verify-m010 --stress-moves 80
+  --repeat-matches 2` against the target device. It passed six-client load with
+  four complete observers, six-client responsiveness (86.9 ms), slow/malformed
+  stress (455.4 ms maximum player latency and 387 dropped latest observer
+  updates), and two repeated match cycles (16-byte heap range). Final
+  diagnostics recorded active connections 2, connection high-water 6, queue
+  high-water 8, queue depth 0, 202,292-byte minimum internal free heap,
+  184,320-byte minimum largest block, 20 malformed messages, 21 rate-limited
+  messages, 80 accepted commands, two completed/reset matches, and no PSRAM
+  requirement. The subsequent `/health` response was `serving` for version
+  `0.10.0`. Acceptance criteria 1--4 and 6 now have concrete evidence;
+  physical station/fallback provisioning and browser-recovery evidence for
+  criterion 5 remains pending, so Milestone 010 is `IN_PROGRESS`.
+- On 2026-09-08, `./scripts/run_host_tests.sh` passed every health, Wi-Fi
+  configuration, NVS, provisioning, lobby/protocol, maze, game, disconnect
+  recovery, browser, and web-asset suite. Its deterministic maze sweep covered
+  1,024 seeds and its repeated-match reset test completed 128 cycles. The
+  owner also ran `pio run -e esp32-c6-devkitm-1`, which passed with 124,276 of
+  327,680 RAM bytes and 1,028,956 of 2,097,152 application bytes, and
+  `pio run -e esp32-c6-devkitm-1 --target buildfs`, which passed with the
+  2,031,616-byte LittleFS image exactly fitting its 2,031,616-byte partition.
+- On 2026-09-08, the owner confirmed physical station recovery after network
+  disruption. `/api/wifi/status` returned
+  `{"state":"station_connected","accessPointSsid":"maze32-F7B9",`
+  `"ip":"192.168.1.50"}`, and bounded station retries behaved as expected.
+  The owner also confirmed the fallback provisioning page worked as expected
+  and that both active player sessions restored their original slots through
+  saved tokens. Together with the successful target stress run, host
+  regressions, production builds, and fixed-limit evidence above, this
+  satisfies every Milestone 010 acceptance criterion. Milestone 010 is
+  `COMPLETE`; Milestone 011 remains `NOT_STARTED`.
 
 ## Milestone 011 — Gameplay tuning and MVP acceptance
 
-**Status:** `NOT_STARTED`
+**Status:** `IN_PROGRESS`
 
 ### Objective
 
@@ -988,6 +1020,21 @@ criterion on the release candidate.
 
 ### Evidence
 
-Completion evidence is not yet available because this milestone has not
-started. Required evidence includes playtest data, the mapped acceptance report,
-final build hashes and sizes, and project-owner approval of any deviations.
+Implementation is in progress. `docs/BALANCE_REPORT.md` locks the current MVP
+gameplay configuration and records the deterministic maze metrics. The
+playtest procedure and result table are in `docs/PLAYTEST_REPORT.md`, and the
+criterion-by-criterion evidence register is in `docs/MVP_ACCEPTANCE.md`.
+Representative completed-match timings and the final release build record are
+still required before this milestone can be marked `COMPLETE`.
+- On 2026-09-08, `bash scripts/run_host_tests.sh` passed its health, Wi-Fi,
+  NVS, provisioning, lobby/protocol, maze, game, disconnect-recovery, browser,
+  and web-asset suites. The locked 1,024-seed sweep reported cycles `133..133`,
+  dead ends `704..854`, start-to-key distance `416..1442`, and key-to-exit
+  distance `346..1244`; every generated map satisfied the key-distance balance
+  invariant. `pio run -e esp32-c6-devkitm-1` passed using 124,276 of 327,680
+  RAM bytes and 1,028,956 of 2,097,152 application bytes. The matching
+  `buildfs` command passed using 2,031,616 of 2,031,616 LittleFS bytes.
+  `docs/MVP_ACCEPTANCE.md` records version `0.10.0`, build hashes, and the
+  19,253-byte reproducible web payload. Human playtest evidence for the
+  approximately five-minute target remains pending, so Milestone 011 is
+  `IN_PROGRESS` and Milestone 012 has not started.
diff --git a/docs/BALANCE_REPORT.md b/docs/BALANCE_REPORT.md
new file mode 100644
index 0000000..00fa2e7
--- /dev/null
+++ b/docs/BALANCE_REPORT.md
@@ -0,0 +1,71 @@
+# MVP gameplay balance
+
+This document locks the gameplay configuration evaluated for Milestone 011.
+It is not a claim that the required human playtests have completed; their
+results belong in `docs/PLAYTEST_REPORT.md`.
+
+## Locked configuration
+
+| Parameter | Locked value | Purpose |
+| --- | ---: | --- |
+| Maze dimensions | 128 x 64 cells | MVP map size |
+| Initial vision radius | 3 cells | Current-view-only exploration |
+| Cross-half connection pairs | 3 | Multiple routes between conceptual halves |
+| Extra cycle pairs | 64 | Alternative routes without removing dead ends |
+| Minimum start-to-exit separation | 64 cells | Keeps exits distant from starts |
+| Landmark pairs | 8 | Spatial-memory cues in symmetric positions |
+| Vision-artifact pairs | 4 | Eight one-use artifacts; exceeds the MVP minimum of two |
+| Compass pairs | 1 | One compass for each player's own exit |
+| Movement cooldown | 150 ms | Server-enforced minimum accepted-move interval |
+| Enhanced-vision duration | 10 successful moves | MVP artifact duration |
+| Draw window | 200 ms | Authoritative simultaneous-exit boundary |
+| Match timer | None | Matches end only by defined result conditions |
+
+The values come from `maze32_maze_default_config()` and
+`maze32_game_default_config()`. They are server-owned defaults; browser
+clients cannot alter them.
+
+## Deterministic generation evidence
+
+`test/host/test_maze_generation.c` generates and validates seeds 0 through
+1,023 with this configuration. For every generated map it proves reachability,
+central symmetry, positive cycles and dead ends, at least four cross-half
+connections, paired player-relative route distances, and this balance
+invariant:
+
+```
+start_to_key_distance >= key_to_exit_distance
+```
+
+The test measures shortest traversable routes, not straight-line distance. It
+therefore establishes that finding a key is never made easier by a shorter
+required route than the later exit search. It does not substitute for human
+search behaviour in a maze with fog of war, landmarks, compasses, and
+artifacts.
+
+The most recent recorded sweep (2026-09-08) reported:
+
+| Metric | Observed range across 1,024 seeds |
+| --- | ---: |
+| Cycles | 133..133 |
+| Dead ends | 704..854 |
+| Start to key | 416..1,442 cells |
+| Key to exit | 346..1,244 cells |
+| Start to exit | 102..992 cells |
+| Generation attempts | 1 |
+
+At the 150 ms cooldown, an uninterrupted shortest 416-cell key route takes at
+least 62.4 seconds of accepted-move time. Actual match duration also includes
+route discovery, current-view-only navigation, the exit search, detours,
+interactions, and player decisions. The approximately five-minute target must
+be confirmed by the representative human-playtest procedure; no hard timer is
+introduced to force that outcome.
+
+## Tuning decision
+
+No parameter change is justified before representative timing evidence exists.
+Changing maze density or cooldown solely from shortest-path metrics would risk
+making the target duration less representative of real fog-of-war play. If the
+playtest median materially misses the five-minute target, change only the
+smallest relevant locked value, rerun the 1,024-seed suite, record the new
+range here, and repeat the affected playtests.
diff --git a/docs/MVP_ACCEPTANCE.md b/docs/MVP_ACCEPTANCE.md
new file mode 100644
index 0000000..9cde5fb
--- /dev/null
+++ b/docs/MVP_ACCEPTANCE.md
@@ -0,0 +1,39 @@
+# MVP acceptance evidence register
+
+This register maps the MVP acceptance criteria to existing evidence and the
+remaining Milestone 011 evidence. It prevents release approval from relying on
+unrecorded assumptions.
+
+| MVP criterion | Evidence | Status |
+| --- | --- | --- |
+| 1. Firmware and LittleFS fit | Milestone 010 target builds: application 1,028,956 / 2,097,152 bytes; LittleFS 2,031,616 / 2,031,616 bytes. | Passing; repeat for release candidate. |
+| 2. Persistent station and fallback provisioning | Milestone 010 owner confirmation: persisted station connection and bounded retry; fallback provisioning page worked. | Passing. |
+| 3. Two-player lobby, countdown, and refresh | Host lobby/protocol and browser tests; Milestone 010 owner confirmation that both player tokens restored their slots. | Passing. |
+| 4. Four read-only observers | Milestone 010 six-client target run assembled four complete observer maps; Milestone 009 authorization tests. | Passing. |
+| 5. Fair repeated 128 x 64 maps | 1,024-seed host sweep and maze validation; locked metrics in `docs/BALANCE_REPORT.md`. | Passing. |
+| 6. Projection security | Host game visibility and payload-security tests; player and observer browser-contract tests. | Passing. |
+| 7. Controls and cooldown | Host game-core and browser keyboard/swipe tests; 150 ms locked cooldown. | Passing. |
+| 8. Key and inventory mechanics | Host game-mechanics tests covering pickup, drop, sabotage, meeting return, and exits. | Passing. |
+| 9. Artifact and compass rules | Host game-mechanics tests covering ten successful moves and role-specific compass projections. | Passing. |
+| 10. Victory and draw window | Host game-mechanics test covering outward exit and 200 ms draw boundary. | Passing. |
+| 11. Disconnect recovery | Host disconnect-recovery tests and Milestone 009 physical acceptance evidence. | Passing. |
+| 12. Approximately five-minute normal match | `docs/PLAYTEST_REPORT.md` must contain at least five completed representative matches. | Pending. |
+| 13. Russian offline UI and browser coverage | Milestone 008 owner checks in Chrome 152, Firefox 155, Safari 26.6, and Edge 152.0.4191.53; asset verification enforces the offline and 1.5 MiB budget. | Passing; repeat asset check for release candidate. |
+
+## Release-candidate record
+
+Before marking Milestone 011 complete, add the exact date, application version,
+firmware and LittleFS build commands, sizes, and the web-asset reproducibility
+digest from the final candidate below. This record must match the tested image.
+
+| Field | Release-candidate value |
+| --- | --- |
+| Date | 2026-09-08 (pre-playtest candidate) |
+| Application version | 0.10.0 |
+| Firmware build command and size | `pio run -e esp32-c6-devkitm-1`; 1,028,956 / 2,097,152 application bytes; `firmware.bin` 1,029,584 bytes, SHA-256 `e654643cb9f16ec03c8d54d981da32ab8e64616f99d23d4a6bb14c8ac7f69c17` |
+| LittleFS build command and size | `pio run -e esp32-c6-devkitm-1 --target buildfs`; 2,031,616 / 2,031,616 bytes; SHA-256 `526e2d03ca4473adb0fe075f4b075aaaeec85ef982240f6578e8cbbb4bec7f35` |
+| Web asset digest and compressed size | `8f3207a208c26aa9f0faff0b0fd8db8e8f8e44537bd6db2c0f914ca5f352920a`; 19,253 / 1,572,864 bytes |
+| Approved deviations | None recorded |
+
+No deviation from `MVP.md` has been approved. A pending item is not passing
+evidence and must not be treated as release acceptance.
diff --git a/docs/PLAYTEST_REPORT.md b/docs/PLAYTEST_REPORT.md
new file mode 100644
index 0000000..f50959c
--- /dev/null
+++ b/docs/PLAYTEST_REPORT.md
@@ -0,0 +1,40 @@
+# Milestone 011 playtest report
+
+## Protocol
+
+Run representative completed two-player matches on the target ESP32-C6 with
+the locked configuration in `docs/BALANCE_REPORT.md`. Use two active players
+and, where practical, one observer. Do not impose a time limit or coach a
+route. Start timing when the authoritative countdown completes and stop at the
+authoritative result. A draw, reconnect timeout, abandoned match, or technical
+failure is recorded separately and does not count as a completed match.
+
+For each completed match, record the maze seed or map hash, both players'
+experience level, winner or draw, elapsed time, whether either player used a
+vision artifact or compass, and any gameplay defect. Do not record session
+tokens, Wi-Fi passwords, or other credentials.
+
+Use at least five completed matches involving more than one pair of players.
+The target is that ordinary completed matches cluster around five minutes;
+there is deliberately no hard timer. If a result materially misses that target
+or a participant reports a balance issue, describe it and repeat the affected
+tests after any parameter change.
+
+## Results
+
+No representative completed human matches have been recorded yet.
+
+| Run | Date | Seed or map hash | Player experience | Result | Elapsed | Artifact / compass use | Defects or balance notes |
+| --- | --- | --- | --- | --- | --- | --- | --- |
+| 1 | Pending | Pending | Pending | Pending | Pending | Pending | Pending |
+| 2 | Pending | Pending | Pending | Pending | Pending | Pending | Pending |
+| 3 | Pending | Pending | Pending | Pending | Pending | Pending | Pending |
+| 4 | Pending | Pending | Pending | Pending | Pending | Pending | Pending |
+| 5 | Pending | Pending | Pending | Pending | Pending | Pending | Pending |
+
+## Assessment
+
+Pending the five completed matches above. This document must state the
+observed range and median, whether the five-minute target was met without a
+timer, and whether any tuning change or approved deviation is required before
+Milestone 011 can be complete.
diff --git a/scripts/ws_protocol_client.py b/scripts/ws_protocol_client.py
old mode 100644
new mode 100755