Platform numbers

Registered-account totals. Last computed 25 Oct 2026, 04:58 UTC.

Trainees
18

Registered accounts with training progress

Finished sessions
171

Across universal and legacy modules

Observations
570

Judged evaluations plus recorded card guesses

Curated targets
3,232

Catalog targets with an image

Results by scoring method

Each row keeps one module, method, version and hint group together. Different measures are not averaged together. A dash means no established chance baseline for that score.

Results by scoring method
Module and methodMean result
Symbol CardsNo recorded sketch hints465 recorded guesses
Method and version
Method:
five-symbol-cards
Rubric:
recorded-guesses-v1
Model:
not-applicable
Score mapping:
hits/recorded-guesses
22.2%hit rateChance per guess: 20.0%
Free Response *No recorded sketch hints22 evaluations
Method and version
Method:
llm
Rubric:
unknown
Model:
google/gemini-2.5-flash
Score mapping:
benevolent-curve-v1:1
22.4%AI similarityChance per guess: —
Intuition Forecast *No recorded sketch hints16 evaluations
Method and version
Method:
llm
Rubric:
unknown
Model:
google/gemini-2.5-flash
Score mapping:
default-linear-v1:1
21.4%AI similarityChance per guess: —
Intuition Forecast *No recorded sketch hints14 evaluations
Method and version
Method:
llm
Rubric:
unknown
Model:
google/gemini-2.5-flash
Score mapping:
benevolent-curve-v1:1
15.5%AI similarityChance per guess: —
Free Response *Assisted: recorded sketch hints12 evaluations
Method and version
Method:
llm
Rubric:
unknown
Model:
google/gemini-2.5-flash
Score mapping:
benevolent-curve-v1:1
33.2%AI similarityChance per guess: —
Free Response *No recorded sketch hints12 evaluations
Method and version
Method:
llm
Rubric:
unknown
Model:
google/gemini-2.5-flash
Score mapping:
default-linear-v1:1
13.1%AI similarityChance per guess: —
Free Response *No recorded sketch hints7 evaluations
Method and version
Method:
llm
Rubric:
target-grounded-v2
Model:
google/gemini-2.5-flash
Score mapping:
benevolent-curve-v1:1
23.1%AI similarityChance per guess: —
Free Response *No recorded sketch hints7 evaluations
Method and version
Method:
keyword-fallback
Rubric:
unknown
Model:
signal-similarity-v1
Score mapping:
default-linear-v1:1
25.2%stored scoreChance per guess: —
Free Response *Assisted: recorded sketch hints5 evaluations
Method and version
Method:
llm
Rubric:
target-grounded-v2
Model:
google/gemini-2.5-flash
Score mapping:
benevolent-curve-v1:1
26.0%AI similarityChance per guess: —
Location *No recorded sketch hints5 evaluations
Method and version
Method:
location-practice-engine
Rubric:
unknown
Model:
canvas-proximity-v1
Score mapping:
benevolent-curve-v1:1
77.4%stored scoreChance per guess: —
Location *No recorded sketch hints3 evaluations
Method and version
Method:
location-practice-engine
Rubric:
unknown
Model:
canvas-proximity-v1
Score mapping:
default-linear-v1:1
41.7%stored scoreChance per guess: —
Free Response *No recorded sketch hints1 evaluations
Method and version
Method:
llm
Rubric:
unknown
Model:
gpt-4o-mini
Score mapping:
default-linear-v1:1
28.6%AI similarityChance per guess: —
Intuition Forecast *Hint history unknown1 evaluations
Method and version
Method:
llm
Rubric:
target-grounded-v2
Model:
google/gemini-2.5-flash
Score mapping:
benevolent-curve-v1:1
0.0%AI similarityChance per guess: —

* Fewer than 30 observations. These descriptive results may change substantially with more data.

How to read these numbers

  • AI similarity is a model’s assessment of a submission against the known target. A percentage is not a probability of being correct or evidence of performance beyond chance. Other stored scores and keyword fallbacks are labeled separately.
  • Assisted rows contain sessions with recorded sketch hints or a hint bonus. Their stored scores may include that bonus. “No recorded sketch hints” describes the available records; it does not establish a controlled, unassisted condition.
  • Rows separate recorded method, rubric, model and score mapping. Unknown legacy versions remain explicitly unknown. Each judged evaluation has equal weight within its row; repeated evaluations are not independent participants.
  • Only classic five-symbol card guesses have a displayed mathematical baseline: 20% per independent uniform guess. The 25% constant used for Free Response progression is a game rule, not a measured chance baseline. Existing scores, belts and rewards are unchanged.

Card completion and missing guesses

Card hit rate is total correct recorded guesses divided by all recorded guesses, including unfinished sessions. Scheduled cards without a recorded guess and timed-out cards are excluded from that denominator. An unfinished session is not necessarily abandoned. Mobile arcade games are excluded.

Legacy card sessions: 71 marked finished; 17 unfinished. 465 recorded guesses; 0 timeouts; 157 scheduled cards without a usable outcome.

Limits of these results

People choose when and what to practice. Target pools, repeated users and selective completion can affect these figures. These descriptive aggregates are not a controlled study, significance test or validated measure of an ability.

Guest and synthetic test accounts are excluded. 0 failed or invalid score records are excluded from means. Results refresh every ten minutes.

Try a practice session

Sketch, reveal the target and review the similarities. Your first evaluation is free.

Try one round — no account needed
v1.1.35