Tundra Arms League, 2026-11-17: recordings, tables and report
==============================================================

WHAT THIS IS
The spectator view of 5 league matches, recorded from start to finish, with the tables and the report built
from them. A spectator is sent every march, speed-up, teleport, building change and score change of a match as
it happens. The recordings are that stream, written down line by line with the time each update arrived.
Recording is read-only: a spectator cannot act in a match.

In all: 425 player appearances, 27,314 marches sent, 46,282 speed-ups, 2,084 teleports, 127 crumb waves.

MATCHES
match          start UTC         result                       recorders  recording
1250 v 1230    2026-11-17 12:00  WBI 156499 : SYN 349553      1          20261117_g19_m61_rec1.jsonl.xz
396 v 398      2026-11-17 12:00  KKO 291368 : ATM 138058      1          20261117_g7_m62_rec1.jsonl.xz
608 v 667      2026-11-17 12:00  BUD 378872 : PpP 128710      1          20261117_g10_m62_rec1.jsonl.xz
394 v 437      2026-11-17 14:00  ExE 152209 : 2VT 250561      2          20261117_g7_m61_merged.jsonl.xz
567 v 559      2026-11-17 19:00  ROD 344804 : VIS 136528      3          20261117_g9_m61_merged.jsonl.xz

A match is named by the two states, e.g. "567 v 559". Matches of the day that were recorded only in part are not
in the package; the report lists them.

FILES
README.txt                 this file
MANIFEST.csv               every file with its size and SHA-256
report/                    the report (PDF)
tables/matches.csv         one row per match
tables/players.csv         one row per player and match: counts, each timing measure with the place among all
                           players. "uid" is the game id used in the events and recordings;
                           "player key" is a short hash of it.
tables/measures.csv        each timing measure across all players: how many players, fastest, 1%, 5%, 10%, 25%, median
tables/waves.csv           one row per player and crumb wave they picked up from: when the first pick-up left
                           after the wave appeared, the gaps between pick-ups, teleports next to the wave
tables/events_<match>.csv.xz  every update of the match as one row (the recording, decoded and labelled)
tables/event_1945.csv      the pick-ups at 14:37 left on the clock in 567 v 559 (19:45:22 UTC), one row per march
tables/tal_findings.csv    every event counted in the report, one row per event
recordings/                the recordings, xz, one JSON object per line
                           The single recordings of matches that also have a merged one are in a second
                           this zip too.

RECORDINGS
<day>_<group>_<match>_merged.jsonl.xz   a match watched by two or three recorders, put together (see below)
<day>_<group>_<match>_rec<N>.jsonl.xz   what one recorder received

match         file                                  updates  first to last update, UTC
1250 v 1230   20261117_g19_m61_rec1.jsonl.xz          30460  12:00:03 to 12:59:59
396 v 398     20261117_g7_m62_rec1.jsonl.xz           47625  12:00:03 to 12:59:59
608 v 667     20261117_g10_m62_rec1.jsonl.xz          37633  12:00:03 to 12:59:59
394 v 437     20261117_g7_m61_merged.jsonl.xz         58038  14:00:02 to 15:00:00
394 v 437     20261117_g7_m61_rec1.jsonl.xz           58045  14:00:02 to 15:00:00
394 v 437     20261117_g7_m61_rec2.jsonl.xz           58038  14:00:04 to 15:00:00
567 v 559     20261117_g9_m61_merged.jsonl.xz         33191  19:00:04 to 20:00:00
567 v 559     20261117_g9_m61_rec1.jsonl.xz           28997  19:10:03 to 20:00:00
567 v 559     20261117_g9_m61_rec2.jsonl.xz           23261  19:20:03 to 20:00:01
567 v 559     20261117_g9_m61_rec3.jsonl.xz           33191  19:00:04 to 20:00:00

Line kinds ("k"):
  push   an update the server sent. "n" its name, "o" the decoded content, "raw" the bytes as received (base64),
         "t" the arrival time on the recorder's clock (unix seconds).
  resp   the server's answer to a question: the two alliances with start and end time of the match, and the score
         board (asked for every 30 s).
  note   one header line naming the match.

Times. "t" is when an update reached the recorder, about 0.6 s after the server acted, the same delay for every
player. The server's own stamps inside an update (starttime, endtime) are whole seconds. Every spectator is sent
the same updates, so two recordings of one match differ only in arrival times and in how long each recorder was
connected (a recorder that joined later, or reconnected, has fewer lines). In a merged recording "t" is the
earliest arrival among the recorders after taking out the difference between their clocks, and "t_base" is the
arrival in the recording it was built on, where that was later.

What was taken out of the recordings: the recorder's own log lines, the questions it sent, and its sign-in. They
say nothing about the match. Every push line is kept exactly as recorded.

EVENTS TABLE
Columns: match, utc, t, event, action, uid, nick, tag, side, march_id, march_type, march_kind, target_type,
  target, target_uid, sx, sy, ex, ey, starttime, endtime, state, troops, battleresult, extra.
"action" is what happened: dispatch (a march sent), speedup, rally launch, battle, arrive, return, teleport,
object appears / object gone (crumbs: target "war crumb", position in sx/sy, points in extra), building update.
"utc" and "t" are the arrival time; "starttime"/"endtime" are the server's whole seconds for the march.

HOW TO READ THE NUMBERS
A timing value is a measurement of what the server did and when. It does not show how an order was given: a
second device, a second person on the account or a practised routine can each explain a fast action. The tables
put each value next to what every other player did in the same matches.
A gap under 0.25 s between two updates in a match with one recorder depends on how the two updates were
delivered and is not used on its own. Places are among all players of these matches with enough values on
that measure.

CHECKING THE PACKAGE
MANIFEST.csv lists the SHA-256 of every file. tables/events_<match>.csv.xz can be rebuilt from the recording of that
match; tables/waves.csv and tables/players.csv are computed from the events.
