Every fragment tea-bot has ever served, ranked by the one difficulty measure that is not circular: its position in the bot’s own prompt list. What players did with it comes second, and only once there is enough of it to mean anything.
Two rules, both of which the previous version of this page broke
syll is the objective difficultytea-bot draws prompts from prompts.txt, a file of 9,217 three-letter fragments ordered easiest to hardest. The line number is the difficulty: ing is line 1. That number is fixed by the bot before anyone plays, so it cannot be bent by who happened to draw the prompt.
Lines 9,166 and up exist in the file but sit past the bot’s selection ceiling — 52 prompts it can never choose.
An observed solve rate is printed only once a prompt has at least 20 attempts in the current scope. Below that you get the raw count and the words “not enough data” — never a percentage, and never a zero.
On this data that is a brutal filter: 140 of 10,540 prompts clear it in multiplayer. The old site applied no floor at all, then bucketed prompts by array index, and so labelled 2,729 of its 2,750 prompts “Easy”.
tea-bot writes a turn to the log only when the turn ends. Rejected and invalid guesses are never recorded at all, so this rate measures turns survived inside the timer, not first-guess accuracy.
A player who typed nine wrong words and then a right one appears here identically to one who got it first try. That is why solve rate sits in the nineties nearly everywhere, and why solve time is the measure worth reading.
Median solve time and sample size, over ten bands of the syll index
Multiplayer · prompts grouped into ten bands of the objective syll index. Hollow points are bands with fewer than 100 timed solves behind them.
Attempts behind each band
Across the 8 bands with at least 100 timed solves, median solve time runs 3.6 s to 6.2 s — a 1.7× spread, while the solve rate stays between 87.3% and 97.8% throughout. Almost every prompt eventually gets answered, so how often one was survived separates them barely at all. Ranking by that is how the old site put 2,729 of 2,750 prompts in a single “Easy” bucket.
Provisional bands (fewer than 100 timed solves): 8–9k (n=92), 9k+ (n=4).
| syll band | Prompts | Attempts | Solved | Solve rate | Median solve |
|---|---|---|---|---|---|
| 1–1,000 | 1,000 | 14,507 | 14,189 | 97.8% | 3.7 s (n=14,189) |
| 1,001–2,000 | 906 | 6,127 | 5,928 | 96.8% | 3.8 s (n=5,928) |
| 2,001–3,000 | 799 | 1,627 | 1,534 | 94.3% | 3.6 s (n=1,534) |
| 3,001–4,000 | 792 | 1,567 | 1,410 | 90.0% | 3.8 s (n=1,410) |
| 4,001–5,000 | 932 | 2,725 | 2,471 | 90.7% | 3.7 s (n=2,471) |
| 5,001–6,000 | 692 | 1,120 | 993 | 88.7% | 4.0 s (n=993) |
| 6,001–7,000 | 623 | 1,081 | 979 | 90.6% | 4.6 s (n=979) |
| 7,001–8,000 | 256 | 361 | 315 | 87.3% | 6.2 s (n=315) |
| 8,001–9,000 | 105 | 117 | 92 | 78.6% | 7.0 s (n=92) |
| 9,001–9,217 | 6 | 6 | 4 | 66.7% | 6.4 s (n=4) |
1,683 matching · sorted by attempts
| Prompt | syll | Attempts | Solve rate | Median solve |
|---|---|---|---|---|
| gaa | 6,764Very HardVery Hard | 1 | not enough datan = 1 | 9.5 sn = 1 |
| gak | 5,620Very HardVery Hard | 1 | not enough datan = 1 | 6.3 sn = 1 |
| gbu | 5,215Very HardVery Hard | 1 | not enough datan = 1 | 2.3 sn = 1 |
| gch | 7,170Very HardVery Hard | 1 | not enough datan = 1 | 12.0 sn = 1 |
| gde | 6,706Very HardVery Hard | 1 | not enough datan = 1 | 10.0 sn = 1 |
| gek | 5,860Very HardVery Hard | 1 | not enough datan = 1 | 3.8 sn = 1 |
| gfe | 6,236Very HardVery Hard | 1 | not enough datan = 1 | 3.2 sn = 1 |
| ggb | 7,014Very HardVery Hard | 1 | not enough datan = 1 | 6.1 sn = 1 |
| ggc | 7,066Very HardVery Hard | 1 | not enough datan = 1 | 1.8 sn = 1 |
| ggd | 8,369Very HardVery Hard | 1 | not enough datan = 1 | 14.1 sn = 1 |
| ggn | 7,397Very HardVery Hard | 1 | not enough datan = 1 | 2.7 sn = 1 |
| ggp | 7,761Very HardVery Hard | 1 | not enough datan = 1 | 13.4 sn = 1 |
| ghc | 6,501Very HardVery Hard | 1 | not enough datan = 1 | 13.1 sn = 1 |
| ghg | 5,912Very HardVery Hard | 1 | not enough datan = 1 | 6.1 sn = 1 |
| ghj | 6,503Very HardVery Hard | 1 | not enough datan = 1 | 5.3 sn = 1 |
| ghk | 7,490Very HardVery Hard | 1 | not enough datan = 1 | 9.5 sn = 1 |
| ghn | 5,020Very HardVery Hard | 1 | not enough datan = 1 | 1.2 sn = 1 |
| ghp | 6,306Very HardVery Hard | 1 | not enough datan = 1 | 2.8 sn = 1 |
| gja | 6,206Very HardVery Hard | 1 | not enough datan = 1 | 11.0 sn = 1 |
| gje | 7,411Very HardVery Hard | 1 | not enough datan = 1 | 11.9 sn = 1 |
| gka | 5,760Very HardVery Hard | 1 | not enough datan = 1 | 7.1 sn = 1 |
| gke | 8,181Very HardVery Hard | 1 | not enough datan = 1 | 9.6 sn = 1 |
| gkh | 8,701Very HardVery Hard | 1 | not enough datan = 1 | 4.3 sn = 1 |
| glb | 9,164Very HardVery Hard | 1 | not enough datan = 1 | 1.3 sn = 1 |
| gmu | 6,371Very HardVery Hard | 1 | not enough datan = 1 | 2.3 sn = 1 |
| gnl | 5,275Very HardVery Hard | 1 | not enough datan = 1 | 6.3 sn = 1 |
| gnn | 6,478Very HardVery Hard | 1 | not enough datan = 1 | 16.0 sn = 1 |
| gny | 6,768Very HardVery Hard | 1 | not enough datan = 1 | 10.8 sn = 1 |
| gok | 7,156Very HardVery Hard | 1 | not enough datan = 1 | 18.0 sn = 1 |
| gpa | 6,331Very HardVery Hard | 1 | not enough datan = 1 | 11.4 sn = 1 |
| gpe | 6,330Very HardVery Hard | 1 | not enough datan = 1 | 4.5 sn = 1 |
| gpi | 5,179Very HardVery Hard | 1 | not enough datan = 1 | 2.1 sn = 1 |
| gpl | 6,226Very HardVery Hard | 1 | not enough datan = 1 | — |
| gsa | 5,379Very HardVery Hard | 1 | not enough datan = 1 | 14.4 sn = 1 |
| gse | 6,112Very HardVery Hard | 1 | not enough datan = 1 | — |
| gsi | 5,611Very HardVery Hard | 1 | not enough datan = 1 | 17.0 sn = 1 |
| gsu | 6,640Very HardVery Hard | 1 | not enough datan = 1 | 4.8 sn = 1 |
| gte | 6,355Very HardVery Hard | 1 | not enough datan = 1 | — |
| gtr | 5,474Very HardVery Hard | 1 | not enough datan = 1 | 2.6 sn = 1 |
| gtu | 7,163Very HardVery Hard | 1 | not enough datan = 1 | 3.6 sn = 1 |
| gtz | 8,654Very HardVery Hard | 1 | not enough datan = 1 | — |
| guf | 5,329Very HardVery Hard | 1 | not enough datan = 1 | 1.4 sn = 1 |
| guv | 7,010Very HardVery Hard | 1 | not enough datan = 1 | 3.6 sn = 1 |
| guz | 5,489Very HardVery Hard | 1 | not enough datan = 1 | 4.2 sn = 1 |
| gyd | 7,052Very HardVery Hard | 1 | not enough datan = 1 | — |
| gyg | 6,739Very HardVery Hard | 1 | not enough datan = 1 | 4.7 sn = 1 |
| gys | 6,892Very HardVery Hard | 1 | not enough datan = 1 | 6.2 sn = 1 |
| haa | 5,157Very HardVery Hard | 1 | not enough datan = 1 | 2.3 sn = 1 |
| haj | 5,240Very HardVery Hard | 1 | not enough datan = 1 | 5.3 sn = 1 |
| hax | 7,287Very HardVery Hard | 1 | not enough datan = 1 | 4.9 sn = 1 |
10,400 of 10,540 prompts sit below the 20-attempt floor and show no rate. Among the 140 that clear it, solve rates average 98.0% across a 85%–100% range — a narrow band, which is why the syll index and not the rate is the primary sort. Sorting by solve rate or median solve time ranks only prompts at or above the floor; below it the numbers are still printed with their n, but a one-sample median is not a ranking.
4,429 prompts are Unrated — not easy, just absent from prompts.txt. 4,422 of them are the wrong length for the list, which only covers 3-letter fragments; custom games can set any prompt length. The other 7 are 3-letter oddities: Cyrillic homoglyph spoofs that look like Latin letters, blanks-mode prompts carrying a wildcard character, and a little junk. Show them
Three modes are missing on purpose. Red, Green and Yellow Tea score whole rounds rather than turns, so tea-bot logs them as round submissions: no response time, and — unlike turns — rejected words are kept. Their “solved” means first-guess validity, not surviving the timer. Folding two different definitions into one rate would be worse than leaving them out, so they are left out.