Tune the grading
Change how fast a reaction must be for Perfect, how the 0-100 score works and what players see and hear on every pop.
Every re-equip is timed from the hit to the moment a totem is back in a hand, and graded against
thresholds in performance. A session's grades and fails become a score from 0 to 100, which is what
the rating leaderboard ranks by, weighted by the speed it was set at. What each grade looks and sounds
like is feedback, in messages.yml.
Before you start#
Thresholds live in plugins/ExyliaTotemTrainer/config.yml, the feedback in
plugins/ExyliaTotemTrainer/lang/<language>/messages.yml. After a change run /totemtraineradmin reload;
no restart is needed. Colour tokens such as {success} are explained on
Colours.
The grades#
| Grade | Default threshold | Points |
|---|---|---|
PERFECT | 0.25 s | 100 |
EXCELLENT | 0.40 s | 85 |
GOOD | 0.60 s | 70 |
OK | 0.90 s | 50 |
SLOW | anything slower than ok | 25 |
A threshold is the slowest reaction, in seconds, that still earns the grade, inclusive: a
reaction of exactly 0.25 s is PERFECT.
performance:
defaults:
perfect: 0.25
excellent: 0.40
good: 0.60
ok: 0.90
by-ticks: {}
by-mode: {}
points:
perfect: 100
excellent: 85
good: 70
ok: 50
slow: 25
fail-penalty: 15
full-credit-pops: 20
speed-weights:
"25": 0.8
"20": 1.0
"15": 1.5
"10": 2.5
"5": 5.0
reference-ticks: 20Thresholds per speed and per mode#
The thresholds a session uses are resolved once when it starts, most specific first:
by-ticks.<ticks>— the tick interval, written as a string key such as"10".by-mode.<modeId>— the mode id.defaults.
Both override maps ship empty: a reaction is graded against one standard at every speed, and how hard
a speed is lives in speed-weights instead (see the rating). Grading a fast interval harder
here as well would charge a player twice for picking it. A mode entry is checked only when no entry
matches the speed, so a by-ticks block always wins over a by-mode block.
The score#
The score of a session, a round or a whole duel is:
- The mean of the points of every pop's grade.
- While the session has fewer pops than
full-credit-pops, the mean is scaled bypops / full-credit-pops. A full session is untouched;full-credit-pops: 0turns the scaling off. - Rounded, minus
fail-penaltyper fail, clamped to 0..100. No pops at all is 0.
Five pops, all GOOD, no fail, with the defaults: a mean of 70, scaled by 5/20 to 17.5, rounded to
18. The same five pops ending in a miss: 18 − 15 = 3. Twenty pops all GOOD ending in a
miss: 70 − 15 = 55.
A mean on its own rates one lucky pop as highly as a long clean run: a single GOOD pop would sit at
70 above a twenty-two pop session on a leaderboard sorted by score. Scaling below full-credit-pops
makes a short session score like a short session.
In a duel, each round is scored on its own and the match score is the average of the round scores. Pops, fails and the best reaction are summed across rounds.
Rating#
The score says how cleanly a session was played and is blind to speed, which is what a duel round is decided on. A leaderboard asks something else: a clean run at 10 ticks is harder than a clean run at 20. The rating is the score times the weight of the interval it was played at:
rating = round(score × weight(ticks))weight(ticks) is the value under performance.speed-weights for that interval, written as text such
as "10". An interval with no weight of its own weighs reference-ticks / ticks, so with the default
20 a 40-tick interval would weigh 0.5. A score of 80 is a rating of 200 at 10 ticks and 80 at
20 ticks.
| Key | Default | What it does |
|---|---|---|
speed-weights | 25: 0.8, 20: 1.0, 15: 1.5, 10: 2.5, 5: 5.0 | What one point of score is worth at each interval. The shipped weights rise a little faster than the plain ratio: the last few ticks are the hard ones. |
reference-ticks | 20 | The interval that weighs 1.0 when it has no weight of its own. |
The rating is what the rating leaderboard sorts by and what the session summary screen shows. It is
computed from the stored score each time a board is read, so retuning the weights re-ranks every board
without touching a record. See Statistics.
Feedback#
Everything a player sees and hears is an ExyliaLib effect under feedback in messages.yml, so each
language folder carries its own: a title, an action bar,
a sound, and optionally a boss bar, particles or a firework. Remove a channel to keep it quiet; remove
a whole section and nothing plays. The full shape of an effect block is documented in the
library's effects page.
Per grade#
Each grade has an effect and a chat line, empty for none. Both know four placeholders:
| Placeholder | Value |
|---|---|
%grade% | The grade name, PERFECT to SLOW. |
%reaction% | The reaction in seconds, two decimals, no unit. |
%pops% | Pops so far in this session or round. |
%streak% | Pops in a row since the last fail. |
By default a grade shows no title, plays ENTITY_EXPERIENCE_ORB_PICKUP at volume 0.8 and a pitch
that drops with the grade, and writes the action bar:
| Grade | Action bar | Pitch |
|---|---|---|
perfect | {success}⚡ PERFECT {muted}· {letters}%reaction%s | 1.8 |
excellent | {success_light}✦ EXCELLENT {muted}· {letters}%reaction%s | 1.5 |
good | {info}✔ GOOD {muted}· {letters}%reaction%s | 1.2 |
ok | {warning}• OK {muted}· {letters}%reaction%s | 1.0 |
slow | {error}▼ SLOW {muted}· {letters}%reaction%s | 0.7 |
Only perfect has a chat line by default: %prefix% {success}Perfect pop! {highlight}%reaction%s.
feedback:
perfect:
effect:
action-bar:
text: "{success}⚡ PERFECT {muted}· {letters}%reaction%s"
duration: 2.0
sound:
name: ENTITY_EXPERIENCE_ORB_PICKUP
volume: 0.8
pitch: 1.8
category: MASTER
chat: "%prefix% {success}Perfect pop! {highlight}%reaction%s"The other moments#
| Section | When | Default |
|---|---|---|
fail | No totem in hand when a hit landed | Title {error}&lPOPPED OUT / {letters}You did not swap in time, sound ENTITY_ITEM_BREAK 1.0 · 0.8 |
countdown | Before the first hit of a session or round | Title {primary}&l%time% / {letters}Press {highlight}E {letters}and swap totems from your inventory, action bar {secondary}Starting in {highlight}%time_precise% ⌚, sound BLOCK_NOTE_BLOCK_HAT 1.0 · 1.4. %time% is the seconds left, %time_precise% the same with its unit. |
match-found | Both duellists arrived in the arena | Title {primary}&lMATCH FOUND / {letters}Teleporting…, sound ENTITY_EXPERIENCE_ORB_PICKUP 1.0 · 1.2 |
round-won | A round went to you | Title {success}&lROUND WON, sound ENTITY_PLAYER_LEVELUP 1.0 · 1.4 |
round-lost | A round went to your opponent | Title {error}&lROUND LOST, sound BLOCK_NOTE_BLOCK_BASS 1.0 · 0.7 |
round-draw | Both dropped the totem in the same tick | Title {warning}&lDRAW / {letters}Both of you dropped it — replaying, action bar {warning}Round replayed, sound BLOCK_NOTE_BLOCK_SNARE 1.0 · 0.8 |
match-won | The series is yours | Title {success}&lVICTORY / {letters}Match won, sound UI_TOAST_CHALLENGE_COMPLETE |
match-lost | The series is theirs | Title {error}&lDEFEAT / {letters}Match lost, sound ENTITY_WITHER_HURT 0.6 · 0.8 |
training-complete | A full-inventory mode without refill ran out | Title {success}&lCOMPLETE / {letters}Every totem used, sound UI_TOAST_CHALLENGE_COMPLETE |
Sound values are volume · pitch. The countdown's title and action bar count the same seconds and
stop together, whether the timer ran out or a punch cut it short.
The chat line accepts the same placeholders as the effect. A server that wants a quieter screen can
empty the action bars and write one chat line per grade instead; a server that wants none of it
removes the chat values and keeps the sounds.
Check that it works#
Run /totemtraineradmin reload, start /totem train normal, and let a few totems pop. Each one should
show the action bar of its grade with the reaction in seconds; a very fast swap reads PERFECT. Miss a
totem and the "POPPED OUT" title appears, then the summary screen shows the score and the rating.
Going further#
- Pick the modes the grades are applied to: Choose and tune the modes.
- See where the rating ends up: Read stats and leaderboards.
Something missing on this page? Tell us on Discord