bullet
@reminegs
By win rate · min 10 games
Accuracy means turns survived. tea-bot never logs a rejected or invalid word — only correct, timeout, quit and nothing_works reach the database — so this is not first-guess correctness.
Activity
All games turns per day
View as table
| Month | Turns | Correct | Games |
|---|---|---|---|
| August 2026 | 287 | 281 | 3 |
Speed and placement
Response times are recorded for correct answers only — timeouts and quits carry no time at all
View as table
| Response time | Answers | Share |
|---|---|---|
| 0–2s | 109 | 38.8% |
| 2–4s | 107 | 38.1% |
| 4–6s | 38 | 13.5% |
| 6–8s | 13 | 4.6% |
| 8–10s | 8 | 2.8% |
| 10–12s | 1 | 0.4% |
| 12–14s | 4 | 1.4% |
| 14–16s | 1 | 0.4% |
| 16–18s | 0 | 0.0% |
| 18–20s | 0 | 0.0% |
| 20s+ | 0 | 0.0% |
Outcome mix
What happens on a turn, by game mode
View as table
| Mode | Turns | Correct | Timed out | Other |
|---|---|---|---|---|
| Black Tea | 287 | 281 · 97.9% | 6 | 0 |
Games
3 all games games · newest first
| Game | When | Result | Players | Turns | Correct | Length |
|---|---|---|---|---|---|---|
| Black Tea#1353 | Solo practice | 169 | ||||
| Black Tea#1352 | Solo practice | 21 | ||||
| Black Tea#1351 | Solo practice | 97 |
Head to head
Multiplayer opponents with at least 3 shared games · top 10 by games together
No repeat opponents yet
bullet has not shared 3 or more multiplayer games with anyone under these filters.
Personal records
Bests within all games, under the current filters
Prompt difficulty
Difficulty is objective: the bot's prompt list runs easiest to hardest, and a prompt's line number is its rank
| Tier | Attempts | Solve rate | Median time |
|---|---|---|---|
| Common | 193 | 97.9% | 2.2 s |
| Familiar | 94 | 97.9% | 2.4 s |
Tiers come from the bot’s own prompt list, which is ordered easiest to hardest, so the line number is the difficulty: Common lines 1–1,000, Familiar lines 1,001–2,500, Uncommon lines 2,501–4,500, Rare lines 4,501–7,000, Brutal lines 7,001+. Not rated covers prompts outside that list — custom games use four-letter fragments, which the index does not cover. They are kept rather than dropped, because they are also the hardest prompts in the data and removing them would flatter the rates above. Median time is a median, not a mean: response times are right-skewed against a hard timeout, so an average sits well above a typical turn.
