Base Difficulty

Base difficulty: Level 3 edit

Level distribution for verb (directional_movement): 3: 2 13: 2 23: 3 24: 1 26: 5 28: 2 120: 1 135: 2 145: 5 210: 1 254: 1 300: 1 330: 2 415: 1 460: 1 470: 9 1040: 1 1050: 2 1120: 1

Per-Language Difficulty Overrides

No difficulty overrides set. The base difficulty level will be used for all languages.

Add Difficulty Override
-1 to exclude, 1-1299 for level
Frequency and Tier Signals
Combined Frequency Rank: — (run python -m storage.admin --calc-ranks to populate)
Tier Signals
Source Tier Rank used
cambridge_yle Starters (1 of 3) 325

Rank used = synthetic rank this tier contributes to the combined frequency rank (lower = more common).

Corpus Frequency
Corpus Freq (per M) Best form Share Zipf rank Forms
19th_books 618.60 357 100% 163 4
20th_books 681.78 336 100% 153 4
wiki_math 99.24 1922 94% 946 3
wiki_geography 458.16 627 94% 271 4
wiki_biology 164.18 1517 50% 836 3
wiki_modern_life 535.89 377 94% 226 4
wiki_arts 327.76 834 94% 397 4
wiki_society 143.05 1286 94% 920 3
wiki_linguistics 58.68 2454 94% 1818 2
wiki_physical_science 70.09 2453 94% 1895 2
wiki_history 264.82 1024 94% 474 4
early_modern_science 248.67 852 94% 439 4
religious_translated 195.99 1121 94% 496 4
cooking 106.22 1189 94% 989 2
legal_scotus 177.81 1101 94% 646 4
eu_parliament_debates 117.80 1396 94% 870 3
openstax_science 179.22 1341 94% 750 3
openstax_society 477.88 369 94% 268 4
wpa_life_histories 1025.18 233 94% 113 4

Best form = the single best-ranked form in the corpus, share-blind — how the spelling itself ranked. Share = how much of that spelling belongs to this lemma, when several senses compete for it. Zipf rank = all the forms combined (variant spellings included), each scaled by its share, and what feeds the combined frequency rank. Combining forms improves on the best form, but a share below 100% pushes the other way and usually wins: a sense holding a small slice of a common spelling is a rare word, so its rank lands downrated below the best form — and once it falls past what the corpus can support it is capped at the rank an unlisted word would get.