Sample and Expert · PBDB snapshot 9 October 2026
Expert is available for daily and practice expeditions: 1,281 legal guesses and 1,246 eligible answers. The original sample retains separate puzzles. Both rosters share one discovery journal covering the full curated roster. Read Expert profiles and approved decisions →
How this working tree was built.
This is a 36-genus interaction prototype. Its data are a transparent working starting point, not the promised comprehensive release dataset or a definitive consensus phylogeny.
Roster and provenance
The 36 familiar genus records and their ancestor records were retrieved from the Paleobiology Database on 9 October 2026. Each entry links to its stable PBDB taxon identifier. The raw responses, retrieval URLs and retrieval time are preserved with the versioned source code. The API queries use taxa/list.json, named genera, parent information and the all_parents relationship.
This named sample does not establish a complete accepted-genus extraction or resolve every synonym. The full version still needs published inclusion rules, a dated snapshot, explicit non-avian boundaries and a completeness audit. Recognised prototype input aliases include T. rex and Cryptovolans; they are name matches rather than recommendations.
Curator’s Notes
PBDB supplies the full guess-roster baseline within our published non-avian scope. Explicit, sourced exceptions exclude duplicate identities, unavailable informal names and exceptionally dubious material; all are listed prominently in Curator’s Notes. Other information—including relationships, diet, age and locality—may follow the latest valid, relevant literature as needed, with evidential support and uncertainty reviewed. Material departures are cited and explained in Curator’s Notes.
Inspect every current tree amendment, clue choice and genus lineage →
The working hierarchy
PBDB ancestry is a taxonomic classification that may skip intermediate clades or reconcile different published opinions. A direct export gives some familiar dinosaurs inconsistent depths. The prototype therefore uses an explicit simplified working hierarchy, with the amendments below. Genus endpoints are retained, while many intermediate names are omitted from the dataset. Within that dataset, feedback uses the deepest shared node exactly. Compact display further hides some linear intermediate labels without changing feedback.
| Amendment | Reason and reference |
|---|---|
| Place Theropoda and Sauropodomorpha under Saurischia. | Use the conventional basal arrangement for the prototype. This is a selected hypothesis, not a claim that basal dinosaur relationships are settled. Museum dinosaur collection; discussion of alternatives. |
| Connect Gallimimus through Coelurosauria and Maniraptoriformes; connect Maniraptora through Maniraptoriformes. | Raw PBDB ancestry places the sample ornithomimosaur directly under Theropoda. The working hierarchy retains shared coelurosaur and maniraptoriform ancestry. Chinzorig et al. 2017, ornithomimosaur analysis; Pelecanimimus study and maniraptoriform relationships. |
| Connect Neornithischia through Genasauria. | Keep compatible shared ancestry for the represented ornithischian branches. Madzia et al. 2021, ornithischian phylogenetic nomenclature. |
| Connect Ceratopsidae through Neoceratopsia and Coronosauria. | Restore shared ancestry with the represented protoceratopsid branch. Farke et al. 2014, ceratopsian phylogeny; Madzia et al. 2021. |
Sibling branches in this hierarchy are not claims of equal evolutionary distance. No percentage closeness, divergence date or branch-length score is calculated. Unresolved or omitted intermediate relationships need further review before release.
Group descriptions and further reading
49 group overviews are short original paraphrases from the American Museum of Natural History, the Natural History Museum, and palaeontologist Thomas R. Holtz Jr.’s University of Maryland courses on theropods, bird relatives, ornithischians and sauropodomorphs. Each overview links its source. The remaining 23 finer groups use a clearly labelled classification summary, with no anatomical traits inferred from their parent.
All group records include PBDB links. Maniraptoriformes is explicitly mapped to taxon 57429, verified with the PBDB API on 9 October 2026. Wikipedia links in group and genus profiles offer optional accessible overviews and pictures; their content is not imported or used to generate clues.
Clues and journal filters
Broad diet and period categories are based on individual Natural History Museum Dino Directory entries. Discovery regions group modern fossil locations into the Americas, Europe, Asia and Africa. These are discovery regions, not reconstructed prehistoric continents. A clue lists all regions recorded in this prototype; it does not imply that a genus was found exclusively in one region. The dataset is not an exhaustive occurrence catalogue.
Diet categories for Oviraptor, Gallimimus and Pachycephalosaurus are explicitly marked “likely”, reflecting the uncertainty of dietary interpretations. Each clue uses exactly the category available in the journal filter. Coelophysis is listed under both Triassic and Jurassic to cover the broad genus-level range used here; choosing either period retains it.
Genus-level regional corrections include Europe for Allosaurus and Stegosaurus, beyond the museum's abbreviated USA profiles. References: Portuguese Allosaurus study; European Stegosaurus occurrence. Brontosaurus uses the museum's reinstatement article and its related sauropod references.
Optional clues are presented as field support: analysis kits and locality records brought in to reveal evidence. Support requests are recorded as clue use. Results and journal records show guesses and clues used, with no performance stars, points or budget meter.
Automatic field reports
Once a player reaches a shared direct-parent branch with at least six unguessed, plausible answers, the game checks whether their tree positions and all three standard clues cannot separate the remaining candidates. If so, a field report brings extra referenced evidence without adding to clues used. The evidence must narrow the crowd to at most five. Useful tree splits or unrequested standard clues prevent the report from arriving prematurely. Five or fewer indistinguishable candidates are left for the player to test. Reports stay in the expedition history after completion.
Local palaeontologists, archive bulletins, field dispatches and a budget surplus provide alternative fictional expedition wording. The chosen wording stays the same within an expedition; it changes neither the evidence nor result counts. No real new scientific discovery is implied.
The current 36-genus roster has no qualifying crowd. Its largest identical sibling group has two members before guesses. The mechanic is tested using fictional larger-branch fixtures, which are not included in the published roster. Twenty-seven journal entries now show and allow searches of the country listing in their preserved museum profile, with a source and an explicit non-exhaustive qualification. These listings are not precise dig sites or complete geographic ranges. The nine hand-entered profiles receive no inferred locality listing. Larger groups need reviewed extra evidence, such as a verified specimen formation or locality, before they become hidden answers.
Balance audit
For the 36-genus sample, broad-group plus diet, one listed period and one listed region produces intersections of 1–8 genera. Seven answers become unique in that check: Baryonyx, Coelophysis, Iguanodon, Kentrosaurus, Plateosaurus, Spinosaurus and Stegosaurus. Multi-region clue combinations can be more revealing still. The prototype has fixed clue wording; adaptive clue broadening is not implemented.
These results are a limitation of the small interaction sample. They do not establish whether the three clues are suitably broad in the comprehensive roster. The production audit must examine the remaining possibilities given actual guess feedback and all prior clues, including clue redundancy and sparse groups.
Daily puzzles and records
A deterministic date-based schedule chooses one shared answer from this fixed dataset each UTC day. Repeats are allowed. Practice chooses a random genus and can also repeat. The prototype allows 15 guesses; that number is provisional.
Only a solved mystery earns a journal discovery. Practice discoveries count towards the collection. Blue and gold card borders mark Sample and Expert discoveries respectively; an Expert solve permanently upgrades the border. A successful Daily expedition earns a dated ink stamp, regardless of clue use. Cards prioritise the best Daily result, with practice bests and roster-specific results inside the entry. Repeated successful Daily expeditions add dated log entries. Sample and Expert results contribute to the same genus entry, with roster provenance preserved. Best records prioritise fewer clues, then fewer guesses; repeat solves increment the encounter record. Failed expeditions and Give up reveal the answer without awarding a discovery or Daily stamp. The finished map highlights the answer’s complete lineage in this working dataset and retains previous wrong guess branches, faded. Pinch zoom and the Fit control let players inspect larger maps. A surrendered daily expedition stays closed after reloading.
Progress is saved in this browser on this device. There is no cross-device account or synchronisation. The dataset and answers are client-side because this is a private prototype; it has no anti-cheat or competitive verification service. You can export a JSON journal record from Sources & methodology in the app.
Dataset audit
The 9 October 2026 audit captures a comprehensive PBDB subtree and records classification questions, missing evidence and clue-balance findings. Its candidate roster is a review queue. The original sample uses the same 36-genus dataset; Expert uses the curated release described below.
Read the earlier eight-case inclusion checkpoint → · Historical v11 source-coverage register →
Read the dataset audit and decisions for review →
Branch Survey · find the expanded-roster pools needing additional evidence →
Locality Survey · inspect all large groups and every overlapping occurrence →
Future found-in locality clues will compare complete reviewed occurrence footprints, retaining non-primary places, credible uncertain identifications and unresolved coverage. A preferred type-specimen locality chooses an anchor, not an exclusive genus bucket. Collection records, physical dig sites, geological groups, formations and members are distinct scopes. The locality screen creates review leads; it does not expand the playable roster or approve hidden answers.
Current Expert reference
The 11 October 2026 curation selects 1,281 legal guesses, with 1,246 eligible profiles and 35 conservative answer holds. Individually sourced exclusions, working-tree changes and qualified diet inferences are explicit. All eligible profiles pass the documented blind-sibling crowding check; 442 use free locality support. A separate runtime audit demonstrates one deterministic route from the root for all 1,246 answers with three requested clues: all finish within 11 guesses. This is not an optimality proof or a guarantee after arbitrary guesses. Expert daily and practice games use these eligible profiles. The original sample remains available separately.
Search current profiles, sources and approved decisions →
What to review next
Playtest the Expert roster, tree navigation and clue presentation, then select the smaller Explorer roster. Browser visual QA was unavailable in this build environment, so desktop and phone layout review remains part of the first playtest.