Field guide · Tree fairness review · 10 October 2026
Which crowds actually need help?
Policy update: the current rule accepts five candidates and starts free support at six. This page preserves the earlier four-candidate evaluation. Read the current assessment →
The earlier crowded-proposal list was not a list of uniquely problematic dinosaurs. We now test each whole candidate crowd, including every competitor, against useful next guesses.
17 comparisons · 35 distinct five-plus states · all 1,308 legal source guesses tested. The live sample remains 36 genera; these are research decisions for expansion.
Largest state in each comparison
An informative probe must distinguish between wrong answers. Merely winning or removing the guessed genus does not count as a tree split. These choices use only the remaining state, never the hidden answer.
Abrosaurus · 5 candidates · blind siblings
Anchor: Shaximiao Formation · CN
Every remaining candidate: Daanosaurus, Qijianglong, Omeisaurus, Abrosaurus, Zigongosaurus.
blind crowd; distinguishing evidence needed.
Standard-fact gaps: Zigongosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved.
Exact worst-case additional guesses to solve on this source tree: 5, including the final correct guess, without bonus help.
Bisticeratops, Navajoceratops, Terminocavus · 6 candidates · useful branches
Anchor: Kirtland Formation · US
Every remaining candidate: Titanoceratops, Pentaceratops, Torosaurus, Navajoceratops, Terminocavus, Bisticeratops.
one informative guess guarantees fewer than five.
Useful next guess: Titanoceratops. Every wrong response leaves at most 4 candidates.
- Shared Triceratopsini: Torosaurus (1).
- Shared Chasmosaurinae: Pentaceratops, Navajoceratops, Terminocavus, Bisticeratops (4).
Standard-fact gaps: Titanoceratops: no_genus_linked_occurrence, period_unresolved, region_unresolved.
Exact worst-case additional guesses to solve on this source tree: 5, including the final correct guess, without bonus help.
Yeneen · 7 candidates · useful branches
Anchor: Bajo de la Carpa Formation · AR
Every remaining candidate: Traukutitan, Overosaurus, Inawentu, Yeneen, Bonitasaura, Neuquensaurus, Rinconsaurus.
one informative guess guarantees fewer than five.
Useful next guess: Overosaurus. Every wrong response leaves at most 4 candidates.
- Shared Rinconsauria: Rinconsaurus (1).
- Shared Colossosauria: Traukutitan (1).
- Shared Titanosauria: Inawentu, Yeneen, Bonitasaura, Neuquensaurus (4).
Exact worst-case additional guesses to solve on this source tree: 4, including the final correct guess, without bonus help.
Ceratonykus, Jaculinykus, Khulsanurus, Ondogurvel, Parvicursor · 5 candidates · blind siblings
Anchor: Baruungoyot Formation · MN
Every remaining candidate: Ceratonykus, Ondogurvel, Khulsanurus, Jaculinykus, Parvicursor.
blind crowd; distinguishing evidence needed.
Exact worst-case additional guesses to solve on this source tree: 5, including the final correct guess, without bonus help.
Daliansaurus, Sinusonasus · 6 candidates · useful branches
Anchor: Yixian Formation · CN
Every remaining candidate: Daliansaurus, Jianianhualong, Liaoningvenator, Sinovenator, Mei, Sinusonasus.
one informative guess guarantees fewer than five.
Useful next guess: Jianianhualong. Every wrong response leaves at most 3 candidates.
- Shared Sinovenatorinae: Sinovenator, Mei (2).
- Shared Troodontidae: Daliansaurus, Liaoningvenator, Sinusonasus (3).
Exact worst-case additional guesses to solve on this source tree: 4, including the final correct guess, without bonus help.
Wendiceratops · 6 candidates · useful branches
Anchor: Oldman Formation · CA
Every remaining candidate: Spinops, Coronosaurus, Wendiceratops, Centrosaurus, Monoclonius, Avaceratops.
one informative guess guarantees fewer than five.
Useful next guess: Centrosaurus. Every wrong response leaves at most 3 candidates.
- Shared Centrosaurini: Spinops, Coronosaurus (2).
- Shared Centrosaurinae: Wendiceratops, Monoclonius, Avaceratops (3).
Exact worst-case additional guesses to solve on this source tree: 4, including the final correct guess, without bonus help.
Damalasaurus · 34 candidates · useful branches
Anchor: Daye Group · CN
Every remaining candidate: Daanosaurus, Eomamenchisaurus, Tonganosaurus, Dashanpusaurus, Nebulasaurus, Xinjiangtitan, Huangshanlong, Damalasaurus, Oshanosaurus, Dachongosaurus, Qijianglong, Lingwulong, Anhuilong, Shunosaurus, Omeisaurus, Tienshanosaurus, Fushanosaurus, Analong, Rhomaleopakhus, Jingiella, Tongnanlong, Huashanosaurus, Jinchuanloong, Datousaurus, Bellusaurus, Abrosaurus, Hudiesaurus, Sanpasaurus, Protognathosaurus, Zigongosaurus, Klamelisaurus, Chinshakiangosaurus, Chuanjiesaurus, Yuanmousaurus.
useful branches remain; further navigation or evidence needed.
Useful next guess: Abrosaurus. Every wrong response leaves at most 19 candidates.
- Shared Sauropoda: Damalasaurus, Oshanosaurus, Dachongosaurus, Sanpasaurus, Chinshakiangosaurus (5).
- Shared Mamenchisauridae: Daanosaurus, Eomamenchisaurus, Tonganosaurus, Xinjiangtitan, Huangshanlong, Qijianglong, Anhuilong, Omeisaurus, Tienshanosaurus, Analong, Rhomaleopakhus, Jingiella, Tongnanlong, Bellusaurus, Hudiesaurus, Zigongosaurus, Klamelisaurus, Chuanjiesaurus, Yuanmousaurus (19).
- Shared Eusauropoda: Dashanpusaurus, Nebulasaurus, Lingwulong, Shunosaurus, Fushanosaurus, Huashanosaurus, Jinchuanloong, Datousaurus, Protognathosaurus (9).
Standard-fact gaps: Oshanosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved; Dachongosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved; Zigongosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved.
Exact solve cost not computed: the bounded search covers at most ten candidates.
Microdontosaurus · 34 candidates · useful branches
Anchor: Chaya Group · CN
Every remaining candidate: Daanosaurus, Eomamenchisaurus, Tonganosaurus, Dashanpusaurus, Nebulasaurus, Xinjiangtitan, Huangshanlong, Oshanosaurus, Dachongosaurus, Qijianglong, Lingwulong, Anhuilong, Shunosaurus, Omeisaurus, Tienshanosaurus, Fushanosaurus, Analong, Rhomaleopakhus, Jingiella, Tongnanlong, Huashanosaurus, Jinchuanloong, Datousaurus, Bellusaurus, Abrosaurus, Hudiesaurus, Sanpasaurus, Microdontosaurus, Protognathosaurus, Zigongosaurus, Klamelisaurus, Chinshakiangosaurus, Chuanjiesaurus, Yuanmousaurus.
useful branches remain; further navigation or evidence needed.
Useful next guess: Abrosaurus. Every wrong response leaves at most 19 candidates.
- Shared Sauropoda: Oshanosaurus, Dachongosaurus, Sanpasaurus, Microdontosaurus, Chinshakiangosaurus (5).
- Shared Mamenchisauridae: Daanosaurus, Eomamenchisaurus, Tonganosaurus, Xinjiangtitan, Huangshanlong, Qijianglong, Anhuilong, Omeisaurus, Tienshanosaurus, Analong, Rhomaleopakhus, Jingiella, Tongnanlong, Bellusaurus, Hudiesaurus, Zigongosaurus, Klamelisaurus, Chuanjiesaurus, Yuanmousaurus (19).
- Shared Eusauropoda: Dashanpusaurus, Nebulasaurus, Lingwulong, Shunosaurus, Fushanosaurus, Huashanosaurus, Jinchuanloong, Datousaurus, Protognathosaurus (9).
Standard-fact gaps: Oshanosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved; Dachongosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved; Zigongosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved.
Exact solve cost not computed: the bounded search covers at most ten candidates.
Datousaurus, Shunosaurus · 10 candidates · useful branches
Anchor: Shaximiao Formation · CN
Every remaining candidate: Daanosaurus, Dashanpusaurus, Qijianglong, Shunosaurus, Omeisaurus, Yuzhoulong, Datousaurus, Abrosaurus, Protognathosaurus, Zigongosaurus.
useful branches remain; further navigation or evidence needed.
Useful next guess: Abrosaurus. Every wrong response leaves at most 5 candidates.
- Shared Mamenchisauridae: Daanosaurus, Qijianglong, Omeisaurus, Zigongosaurus (4).
- Shared Eusauropoda: Dashanpusaurus, Shunosaurus, Yuzhoulong, Datousaurus, Protognathosaurus (5).
Standard-fact gaps: Zigongosaurus: no_genus_linked_occurrence, period_unresolved, region_unresolved.
Exact worst-case additional guesses to solve on this source tree: 5, including the final correct guess, without bonus help.
All immediately blind states
Four of the 35 starting states have no informative next tree guess. Only the Baruungoyot state has complete provisional standard facts. The others remain evidence gaps, not approved difficult puzzles.
crowd-1: Ceratonykus, Ondogurvel, Khulsanurus, Jaculinykus, Parvicursor.
Standard-fact gaps: none in the provisional screen.
crowd-17: Daanosaurus, Qijianglong, Omeisaurus, Abrosaurus, Zigongosaurus.
Standard-fact gaps: Zigongosaurus.
crowd-30: Damalasaurus, Oshanosaurus, Dachongosaurus, Sanpasaurus, Chinshakiangosaurus.
Standard-fact gaps: Oshanosaurus; Dachongosaurus.
crowd-31: Oshanosaurus, Dachongosaurus, Sanpasaurus, Microdontosaurus, Chinshakiangosaurus.
Standard-fact gaps: Oshanosaurus; Dachongosaurus.
The two basal-sauropod five-candidate states arise after a reaching guess in a more deeply nested sauropod branch. Complete their missing periods and regions before treating these as fair hidden answers.
Decisions
- Oldman: do not give Wendiceratops a special exception. All six competitors belong in the assessment. A Centrosaurus or Coronosaurus probe divides the worst state into at most three. Qualified referrals remain possible; this is not a claim that all six are securely known from one quarry.
- Baruungoyot: the five alvarezsaurids are a real blind crowd on this source tree after an outside reaching guess. If one of those five has already been guessed, only four remain. Develop free sourced help for the five-candidate state, rather than making a necessary clue cost an optional clue used.
- Abrosaurus locality state: five direct Mamenchisauridae siblings also lack a useful tree split. Their standard-fact and horizon gaps still need completion before approving answers or a new clue.
- Datousaurus / Shunosaurus broader state: do not treat all ten as automatically blind. Abrosaurus separates four remaining mamenchisaurids from five other sauropods; further useful probing can narrow the latter. The five-sibling block can arise after a different guess, so help belongs to the actual remaining state.
- Kirtland / Ojo Alamo, Yixian and Anacleto: keep their useful tree splits. A formation count of five or more alone is not sufficient to grant help. Their separate specimen, standard-fact and horizon questions still govern hidden-answer eligibility.
- Damalasaurus / Microdontosaurus: keep pending hidden-answer approval. The 34-candidate diagnostic contains metadata gaps and broad overlaps; it is not 34 confirmed dinosaurs at a Tibetan dig site. Useful initial branches do not repair missing evidence.
Discovery archive
A referenced discovery or collection event for the name-bearing type fossil, with the event type stated. This is not the earliest fossil ever attributed to the genus, and it is not the naming, publication, acquisition or preparation year.
A broad date range is enough for the five Baruungoyot alvarezsaurids. Exact dates and the event type can stay in the reference evidence. Jaculinykus’s 2016 date is documented collection, not a separately verified first-discovery year.
| Genus | Found / collected | Evidence |
|---|---|---|
| Khulsanurus | 1987 | explicit discovery year |
| Parvicursor | By 1996; exact year unresolved | inferred upper bound, not an exact discovery date |
| Ondogurvel | 1999 | explicit collection year in firsthand author account |
| Ceratonykus | 2003 | explicit discovery year |
| Jaculinykus | 2016 | explicit expedition collection year |
Parvicursor’s bound is an inference from the type fossil’s original 1996 description. We have not verified its exact discovery year in the primary sources accessed.
Local palaeontologists checked the expedition archive: the recorded discovery or collection of its type fossil was before 1997.
Leaves Khulsanurus, Parvicursor (2).
Local palaeontologists checked the expedition archive: the recorded discovery or collection of its type fossil was in 1997 or later.
Leaves Ondogurvel, Ceratonykus, Jaculinykus (3).
This is proposed free support for a genuinely blind crowd. It is not activated. Complete specimen indexing and expanded-tree/profile approval first; unknown dates must remain possible.
Discovery evidence and remaining gaps
- Averianov and Lopatin. The second taxon of alvarezsaurid theropod dinosaurs from the Late Cretaceous Khulsan locality in Gobi Desert, Mongolia. Published online 2021.
Full original article in author-uploaded copy; Introduction and Materials and methods. The introduction dates Alifanov's discovery of the partial skeleton to 1987; materials identifies PIN 4487/27.
- Averianov and Lopatin. A re-appraisal of Parvicursor remotus from the Late Cretaceous of Mongolia. Published online 2022.
Full redescription in author-uploaded copy; Abstract, Introduction and systematic account. Original 1996 description not accessed. The redescription identifies the holotype and its original description in 1996. We infer that this particular fossil had already been found by 1996; we do not use 1996 as its discovery year. Gap: The often repeated exact date 1992 was not established in the primary sources accessed. Exact-year and decade clues remain unapproved.
- Averianov and Lopatin. Ondogurvel, a new dinosaur from the Late Cretaceous of Mongolia. Retelling Paleontology, Borissiak Paleontological Institute, 2022, pp.120–121.
Full Russian institutional account by the describing authors, in author-uploaded copy. Formal original article accessed at abstract level only. The authors date Alifanov's finding and excavation of the described skeleton to 1999. The cited DOI is the associated formal description, not this institutional chapter. Gap: Confirm the catalogue number against the formal systematic account before incorporating a specimen-indexed gameplay field.
- Alifanov and Barsbold. Ceratonykus oculatus gen. et sp. nov., a new dinosaur from the Late Cretaceous of Mongolia. Paleontological Journal 43, 2009.
Full original article in author-uploaded copy; Introduction p.96 and systematic account. The paper dates the described material's discovery by the joint Russian–Mongolian expedition at Khermiin Tsav to 2003.
- Kubo et al. A new alvarezsaurid dinosaur from the Upper Cretaceous Baruungoyot Formation of Mongolia. PLOS ONE, 2023.
Full original open-access article; systematic account and Acknowledgments. Acknowledgments identifies the Nemegt Educational Expedition 2016 as the crew that collected the studied specimen.
Method and limits
- Reconstruct every reaching wrong guess for all 17 current five-plus pool comparisons, using their chosen reviewed locality anchors. Evaluate every remaining answer, including competitors not named as proposal targets.
- Test all 1,308 selected source genera as subsequent guesses. A useful tree guess produces at least two different WRONG relationship responses. Simply winning or removing the guessed genus is not a branch split.
- The next guess is selected from the remaining candidate state alone: minimise the largest wrong-response bucket, then the uniform-prior diagnostic average, prefer a possible answer, then alphabetical order. No hidden-answer knowledge is used.
- The navigation plan repeatedly chooses the best informative probe and stops below five or at a blind crowd. This is a reproducible witness strategy, not an optimal hint-use strategy.
- For states of ten or fewer, exact minimax searches all distinct guess partitions to minimise worst-case additional guesses to solve, including the final successful genus guess. It assumes no future bonus help; the earlier reaching guess is excluded. Larger states have no exact cost claim.
- Five candidates in a flat source-tree sibling block need help before random elimination. Five in a broader clade may have useful probes. Branch counts alone do not determine fairness.
- These states use the frozen PBDB source tree and all three provisional standard clues. Their structure does not establish the most recent phylogeny or the approved future Expert tree.
- The starting states reproduce a single reaching wrong guess and the selected locality anchor; this is not an exhaustive simulation of every game history, anchor choice or optional-clue order. Subsequent witness plans preserve all prior observations.
- Known non-primary overlaps, qualified referrals and missing relevant locality coverage are retained. Presence of a named genus in a crowd is not confirmation of a shared physical dig site or secure contemporary assemblage.
- Unknown standard facts are retained conservatively. Where listed, they must be researched before claiming that a complete clue set makes the answers indistinguishable.
- Only the discovery-date research proposal is added. No taxonomy, locality exclusions, live answer roster, counters or automatic-help runtime are changed.
Reproduce with python evaluate-tree-probes.py. The JSON contains every reaching context, all five-plus states and their subsequent navigation plans.