Findings
Everything on this page was written by an AI model from the sources each variant’s page lists. It is a reading of them, not a source: follow a variant through to its sources to check it.
What the sources say when compared, variant by variant: where they agree, where they disagree, and where they say nothing.
Where sources disagree
Entries where the sources contradict each other.
Bashi-Bashi
Sloperama documents two incompatible games under this name: version A with 144 tiles and 8 red fives, version B with 112 tiles and 6 red fives, differing also on scoring and payout. The chart carries one undifferentiated node. Sloperama also says Bashi-Bashi is based on Japanese modern AND Filipino rules, while the chart descends it from Filipino alone.
Beijing style
A second source settles what the first could not, and contradicts it. Sloperama assumed 144 tiles and recorded flower handling as unknown; the blog it cites states 136 tiles and that the eight flower and season tiles are not used, and its 17-wide two-high wall confirms 136 arithmetically. Prefer 136 and no flowers. The wild-tile mechanism the chart summarises as 'drawn wild tile' is corroborated and much more specific than either summary: the indicator is face up, and the wild tiles are the four one greater in number, wrapping from nine to one.
Fuzhou style
The detailed entry gives 132 tiles with one honour removed and no flowers or jokers. The abbreviated list on the same page describes Fuzhou as having honours treated like flowers and a dora-type joker, which is the detailed Fujian entry's description, not Fuzhou's. One of the two is contaminated, and there is no second source to break the tie.
Pusser's Bones
A complete ruleset is on record and the search for a second source is over, and none was found. Every trail leads back to one man.
The numbers are stable across the two forms of his account: the points, doubles, fixed-score, bonus and penalty tables are word-for-word identical between the website and the book, as is the scoring procedure — points times two, doubled once per double, rounded to the nearest hundred, divided by a hundred, and that many cents from each loser. 98 sentences are identical. They are one document in two forms.
One real discrepancy, and it is a scoring one. On holding four identical bones, the website covers only the hand-sorting case: declare the set, keep it CONCEALED, place it face down. The book adds a case the website lacks — if a set of three has already been declared and the fourth is then plucked, it joins the declared, open set — and then contradicts itself, a later passage saying a fourth bone plucked into a set goes face down. The difference is worth double: a set of four minors scores 8 open against 16 concealed, majors 16 against 32.
The points table settles it. It lists open and concealed rows separately for sets of four, so if every four-of-a-kind were concealed the two open rows would be unreachable. The book's All-Of (declared) rule is operative, the website is incomplete rather than wrong, and the book's later face-down sentence is the error. Build from the book; keep the contradiction on the record.
Why there is still one source. The BoardGameGeek thread, the last open lead, turns out to be a single post linking to the Gun Plot site: a link, not a description of the game. The site itself is now gone — the domain resolves but serves a certificate for an unrelated host — so the Wayback copies are its surviving form. Wikipedia's sentence on the variant and an RAN veterans' slang glossary both stand outside the author's work, and together they establish that the game and its name are genuine Royal Australian Navy culture: the glossary, unconnected to him, glosses 'Bones' as 'Mah-jongg tiles' and carries his register — grey funnel line, roughers, gash. Neither corroborates a single rule. The tradition clearly existed. What is missing is a second person's account of how it was played, which the author expected, writing that the origins rest with no one individual and that the rules were never officially documented.
Copyright: the book is © 2023 Hot Rod Handbooks, all rights reserved, and sells commercially; the Gun Plot site separately restricted reproduction of substantial parts. Mechanics are recorded here as facts, as tile counts are for every other variant. Neither text is reproduced and neither file is filed in this repository.
Shenzhen style
The source contradicts itself on the same page: the abbreviated list says the game is played with kongs and no pungs or chows, the detailed entry says the winning hand must be pungs only, hedged with 'if I recall correctly'. The chart follows the kongs wording without flagging the conflict. Tile count, hand size, flowers, scoring and payout are all recorded as unknown.
Tui Dao Hu
The name may not denote a variant at all. Wikipedia gives 推倒胡 as the ordinary Mandarin term for a scoreless or chicken hand, paired with Cantonese 雞糊, in a discussion of variants that allow such hands to win. Sloperama, working from the same maque.games page, guessed the meaning as 'push down hu' and recorded a variant with twelve ways of going out and multiple winners. Both can be true, since a ruleset can be named after the scoreless hand it permits, but the chart presents as a distinct regional style what at least one source treats as a common noun. Everything else about the node remains unknown: no tile count, no hand size, no scoring.
Where the sources don't back the claim
The tree claims something its own source doesn't say.
Modern inventions
No source groups these as a category, and the chart says as much in its label. Re-reading shows the group mixes three kinds of thing. There are new games that borrow mahjong's material: Mhing, a card game, and now Rummikub and Mahjong solitaire, added here. There is a tournament ruleset built from two existing traditions, Mahjong Masters Millions. And there is a competition format layered on an existing ruleset, Duplicate. Duplicate in particular is closer to the MCR line than 'not descendants' implies, but the chart's placement is fixed and is recorded as a caveat on that node. Two variants Sloper lists without a chart node, the 12-tile game and Taoist Mahjong, belong to other families and are left for those passes.
Nanyang variants
Sloperama catalogues each Southeast Asian variant separately and never groups them or asserts a common parent; the grouping term does not appear. This is consistent with the chart drawing the link dashed, and confirms the grouping is just a way of organising the tree, with no source behind the lineage.
Where the sources partly agree
They agree on some key facts but disagree or say nothing on others.
AMJA card
Two sources confirm the AMJA card exists as the second of the two annual American cards and that the NMJL and AMJA differ in minor scoring details. The association's own site could not be read: amja.net serves an expired certificate, and it was not fetched over an unverified connection. So the node is attested but not primary-sourced, and no source read here describes how the two cards differ in practice.
Chinese casino style
The sources agree that casino mahjong is whatever the casino's rules say. It doesn't descend from any one game. Every one of them describes money play whose rules are set by the venue or its regulator, and in which the house takes a cut of each won hand: Macau's regulation fixes the commission by stake band, the banked tile game beside it takes 5 per cent, and the British casinos that run mahjong say outright that the rules vary from table to table. Where they divide, they divide because they are describing different houses, which is not a disagreement: Macau's law gives 136 tiles, no flowers, chows allowed but claimable only by the player to the discarder's right, and the winner alone paid; the British pages describe a 144-tile game with chows; and this node describes a 144-tile table with no chows where everyone scores. All three can be true at once, and the first two are firmly sourced. What stays unsupported is this node's own table. It rests on one report from June 2009 that can no longer be read by anyone: reachmahjong.com serves a hosting placeholder and the archive's captures of the board answer with a permissions wall. Nothing places it either — the entry never says which country the casino was in, so 'Chinese' comes only from the chart's label, and no published casino ruleset found here matches its no-chow, everyone-scores shape. As with Malaysian 3-player, the node names one of several documented forms and the chart carries it as though it were the whole category. The Macau law does settle a neighbour outright: its simplified form is Macau simplificado, tile for tile as catalogued.
Hong Kong Taiwanese
No longer unsupported. Chinese Wikipedia carries a full article on 港式臺灣麻雀, the 16-tile game of Hong Kong and Macau, giving 144 tiles with eight flowers, a 16-tile hand, additive and unlimited fan, a ready declaration called ding and a carry-over stake called la, and citing a Hong Kong Mahjong Association league rules document. Two cautions. The article itself carries a single-source reliability warning, so it is one account rather than a consensus. And it contradicts the tree's shape: it describes the game as a blend of Hong Kong, authentic Taiwanese, Japanese and MCR rules, where the chart makes it simply a child of Taiwanese style. Its additive scoring is also the feature the article uses to contrast it with Guangdong, which the chart places on a different branch entirely.
Hua ma que, North China flower sparrow
The game is real and described in writing. Liu Yishu's 1941 manual gives its tile list, its wild tiles and its 16-tile hand, as translated by Stanwick and Xu. The Nakamura set photographed at Dalian in 1924 matches that list. What is argued rather than shown is the descent. Stanwick and Xu say the 1870s kings and jokers are early hua ma que flowers, and Ebashi separately puts the pre-dragon Himly branch in North China. Both call it a hypothesis, and neither has a 19th-century rule text. The sources agree that the game existed, where it was played and what its flowers were. They are silent on how it was scored.
Macau official casino mahjong
The rules are as firmly established as anything in this compendium — they are law, published in the official gazette, reproduced across two instruments two years apart and still named by the regulator today. What is not established is everything around them. No source outside the Macau government describes this game being played, so there is no witness to whether tables follow the text or treat it as a floor; the one independent account of mahjong in Macau casinos, the Wizard of Odds page, documents Mah Jong Pai Gow, which is a pai gow game dealt from mahjong tiles and not this at all. The descent is this compendium's reading and nothing more: the regulation cites no ancestry, and the parent here is inferred from rule shape, a 136-tile wall with fan scoring, no minimum, winner-only payment and a Cantonese fan list. One structural point belongs on the record. This node is the conventional form, article (a) throughout the 1991 text; article (b) of the same regulation is the 112-tile game the chart files separately as Macau simplificado under Mainland regionals. They are two forms of one law, and a reader who finds one should be sent to the other.
Malaysian 3-player
The 84-tile one-suit game is confirmed with a full component breakdown. But Sloperama also records a second three-player Malaysian game of 108 tiles, which the chart does not carry at all, so the node covers only one of two documented forms.
Modern HKOS
The ruleset itself is now well documented. The Hong Kong Mahjong Association's 2023 rulebook fixes it: a 3-faan threshold, a 10-faan cap, 3 faan for semi-flush and all triplets and 7 for a flush (the 三三制), and the discarder paying the whole hand. Wikipedia, the Mahjong Wiki, Chinese Wikipedia and Sloper all describe the same game, with the minimum still a table convention: Sloper's correspondents play from 1 fan in person. The date is the weak point. Only the Mahjong Wiki dates the change, now putting the 3-3 system, with its 3-fan minimum, in the 1970s. Chinese Wikipedia gives the mechanism (flowers and the 三三制 overwhelming the old four-tai cap, then minimums of one, two and three tai) without dates. Kwan, writing in 2009, calls the 3-fan minimum mainstream only 'in recent years'. The chart's '1970s' therefore rests on one source, and a later date is possible for the minimum becoming the norm rather than an option.
Siamese Mah Jongg
Sloperama gives the substance: Gladys Grad, 2015, two players each playing two hands on two-tiered racks, with the concept applicable to any variant. The chart's own canonical source for the node, mahjongg.org, does not corroborate it: its archived homepage is a tournament and Master Points site that does not mention Siamese Mah Jongg. The variant is attested, but by the other source, not the cited one.
Wang ma que, Changsha kings sparrow
The rules survive only through one text, Liu's 1941 manual, as Stanwick and Xu translate it. Chinese Wikipedia repeats the wild-tile functions but gives no independent source for the Changsha game. The case for putting it here is structural. Its eight kings are the Himly set's jokers expanded, and they match the Vietnamese eight-king game. That makes wang ma que the likeliest bridge between the 1870s proto sets and Vietnamese play. The chart cannot draw that link, because it runs Vietnamese play through the disputed Nanyang node.
Western / British Empire style
The rules are well attested by Sloperama and the descent from Babcock is independently confirmed. The relationship to Wright-Patterson is not: Wikipedia treats Western classical as largely meaning the Wright-Patterson rules today, while the chart makes Wright-Patterson one child among five. Sloperama's own naming discussion, preferring British Empire style precisely to separate it from American variants, supports the chart keeping them apart, but the sources do differ.
Zung Jung
The rules are now primary-sourced: Alan Kwan's own site gives 136 tiles, a 13-tile hand, 44 additive patterns with no basic points, and winner-only payment, all matching Sloperama. That is the strongest evidence any node in this compendium has, because it comes from the system's designer. The date and the parentage remain open: the official site gives no creation date, so the chart's 2000s against the wiki's 1990s is still unresolved, as is whether Zung Jung descends from Guangdong Style or sits beside it.
Where there is only one source
One source describes it, and nothing else confirms it.
Fujian style
124 tiles, winds as flowers and a wild tile as joker match the chart exactly. But it rests on one contributor who says openly that he does not understand the scoring, and the Fuzhou entry on the same page appears to have borrowed this description. Treat the tile count as the only firm part.
Israeli Mah-Jongg Association
Still only one source, and a second one has proved hard to find. Sloperama gives 144 tiles or kards, a 13-tile hand, flowers melded on receipt, all players paid, and scoring whose numbers vary greatly from other forms — the last being the only real hint at what distinguishes it. A targeted search in English and Hebrew found nothing about the association's own rules; the results were swamped by American mahjong and general Jewish-mahjong material, which is a different subject and should not be mistaken for this node. Anyone with Hebrew and a route to the association itself would likely resolve this quickly; web search alone does not.
Mahjong solitaire (tiles only, not mahjong play)
Only Wikipedia was read. The node is placed and labelled so that a reader can see what it is not: a mahjong variant. It shares the tile set and the name, and nothing of the play. The 'centuries-old Chinese game called the Turtle' is Activision's marketing, and Wikipedia notes that Lockard never claimed it. It is a modern invention in the plainest sense, 1981, and descends from no ruleset in this tree.
Rummikub
Only Wikipedia was read, so the verdict is single-source. The node records a borrowing, not a descent: Rummikub takes mahjong's tiles-on-a-rack form into a rummy game with numbered tiles in four colours and two jokers. It is here so the tree shows where the best-known mahjong-inspired tile game sits, in the chart's own group for games that do not descend from the mahjong line.
Sichuan 72-tile game
72 tiles, bams and dots only, a 7-tile hand winning on 8. Sloperama attributes it solely to Dragon Chang and itself flags that it differs from the other Sichuan account on the same page. The chart is right to carry it as a separate node rather than merging it.
Where independent sources agree
Two or more sources, read separately, agree on the defining facts.
American mah jongg
Five sources. Tiles, hand and the annual card are agreed; Wikipedia adds the features that separate this branch structurally from every other node in the tree, melds of five or more and no chow at all. The joker date is consistent across sources at 1961 or the 1960-61 season. Nothing here records what is printed on any card, which is deliberate.
Babcock rules
Four sources agree on the 1920 red book and its role in introducing the game to the West, and Wikipedia adds the man, his dates and a 1923 second edition. Sloperama gives the rules themselves, including that kongs and flowers were dropped from Babcock's simplified system, and Wikipedia's scoring article independently derives the Babcock system from Chinese Classical, which supports the chart's descent. The 136-or-144 ambiguity traces to Babcock's own opening sentence describing a set of 136 pieces.
Chinese archaic
The date is well supported, and there is more of a ruleset behind it than this entry used to say. An earlier pass here said no written 1890 ruleset lies behind the node. That was wrong. Wilkinson's 1890 manuscript describes the play of chungfa, and his 1925 memorandum quotes it directly: four players named for the compass points, 15 cards each 'as in Khanhoo', scoring 'tricks' by a tariff agreed before play, and sequences that fill a hand but 'score nothing'. Stanwick quotes the passage, Sloper's catalogue summarises it, and the Depaulis-Tsui chart scores it. They agree, but they are not independent: all three rest on the one copy Depaulis found.
The objects agree with the text. Wilkinson's own 1893 catalogue, read here in Culin's reprint, describes set No. 43 from Ningbo: 108 suit tiles, four each of the winds, 中 and 發, and eight blanks, 140 tablets in all. His 1889 Ningbo set in the British Museum has the same count and a box stamped 中發. Set beside its parent, the 1870s sets, the change is clean. The kings, the rulers and the seasons are gone, and 發 has arrived, so the 136-tile core of the later game exists by 1889.
This is not yet the classical game. The hand is 15 tiles, not 14. There is a mixed-suit triplet of aces or nines, which mahjong does not have. Sequences score nothing, and the chart finds concealed sets not rewarded more highly, the opposite of the Late Qing novels. Four other points of the Late Qing game are unknown for 1890. The 'sparrow' name also belonged elsewhere: in 1890 ma que was the name of the paper money deck around Hankou, and the tiles were chung fa.
One gap remains. Nobody here has read the 1925 memorandum itself. Wilkinson marked it 'Not for Publication', and Depaulis's 2002 newsgroup thread that transcribes it could not be reached: Google Groups refuses automated requests and the Wayback Machine has no capture. Every rule above is therefore second-hand, and comes through one discoverer. The objects, and Wilkinson's 1893 catalogue, are what corroborate it independently.
Chinese classical
Four sources agree on the scoring structure, and Wikipedia independently derives the Babcock system from Chinese Classical, supporting the chart's placement of Babcock rules under this node. The node now also rests on three 1920s primary texts. They agree on the core: 136 tiles, with flowers absent or optional; 13 in hand; points for sets, 10 or 20 for going out, then doublings; small bonuses for a single wait and for a self-draw; a limit of about 300; and the same set of limit hands, including thirteen orphans. They disagree on the payout that the other sources treat as defining. Babcock (1923) and Tchou (1924) have the losers settle among themselves, so a loser can outscore the winner. The Shanghai manual of 1924 (first printed 1914) says the other hands are not reckoned at all when someone wins, as in the Late Qing novels. Dealer doubling was also in flux. All three texts describe a 'new method' (Babcock's New Method, Tchou's nouveau mode, the manual's 近時新法) in which the self-drawn winner is paid double by all, and in the first two the discarder pays double too. That is the discard-and-self-draw payout of the child node Shanghai / Hong Kong new style, already current by 1923-24. Babcock's claim that no Chinese rules were in print is wrong: the Shanghai manual was first published in 1914. Two more 1920s texts by Chinese authors, Tow (1924) and Wei (1925), agree with Babcock and Tchou that the losers settle among themselves and that East pays and receives double. Kwan (2009) traces the 'early rules' to Hatsuhara Shigeki's reconstruction of 1952, built from five Chinese books of the 1910s-20s, and describes a late classical stage with more hands and a higher base and limit.
Chinese pre-classical
This node now rests on its primary source. Mauger's 1915 article, read here in the Chine ancienne text edition with Persée's page numbers, gives the whole game. It has a 136-tile set, a wall of four lines, 13 held and 14 to go out, and draw-and-discard with the last discard claimable. There is a points table, five worked hands and a payment example. It is recognisably the Late Qing game: 10 for going out, three of a kind scored, sequences worth nothing, doublings for honour sets and for one suit, all three losers paying the winner.
Three things set it apart, and they are what the node stands for. First, a triplet's value turns on whether it was dealt or drawn (8 against 4 for honours and terminals), not on whether it was exposed. Second, the wind tiles count only for the player whose seat they name, and East's own winds double his winnings, but East is not otherwise paid or charged double. Third, losers settle their differences among themselves after paying the winner, which the Late Qing novels do not show. Sloper's 1915 column and the Depaulis-Tsui chart both record the last two differences, and the text bears them out.
The tile set is its own thread. Mauger's winds may be court titles (duke, marquis, general, minister), his dragons may be 龍 and 鳳, and his figure shows 品 for the myriads suit. That northern pattern is attested earlier in the Sheng Xuanhuai set (c. 1884-92) and in the Noguchi set in Chiba. Stanwick and Ebashi both read it as a North China form that spread south. Where Mauger's game itself was played is less clear: he names Beijing, Shanghai, Suzhou and Hankou, and his informant's Mandarin points to Hankou. One small conflict is recorded: the article runs pp. 238-281 in Persée, not the 228-81 the Depaulis-Tsui chart gives.
Duplicate mahjong
The rules are now read at the source, not only confirmed to exist, and they change what the node is. MIL's current page defines duplicate play as the 'Duplicate Format of MCR'. It is the MCR ruleset played with 136 tiles and no flowers, pre-arranged and serial-numbered walls, and bridge-style IMP scoring against the average score on the same deal. So it is a competition format for an existing ruleset, not an invention outside the mahjong line. By descent it belongs beside MCR, and the chart's placement under 'modern inventions (not descendants)' is best read as 'a modern invention in how the game is contested'. The chart's parent link is fixed, so this is recorded here rather than redrawn. MIL itself dates from 2015 and has been an IMSA observer since 2016.
Filipino style
Three sources agree on the 144-tile set and the 16-tile hand winning on 17. Wikipedia also confirms the flower handling, which is how the chart tells Filipino from Taiwanese: dragons and winds are grouped with the flowers. The one disagreement is jokers, which Sloperama hedges and Wikipedia denies; prefer Wikipedia, since Sloperama admits it cannot state the rule.
Guangdong MCR / GMCR
The 2020s date is confirmed by two wiki pages, and the entry now has real content: GMCR was assembled to make Guangdong Style competitive, drawing only on rules already present in historic Guangdong rulesets and stripping out pure-luck wins like Blessing of Heaven and the self-drawn win. Its own page also states the descent chain in words — GMCR from Guangdong Style, Guangdong from HKOS, HKOS adopted as a backlash to Hong Kong New Style, and that from Chinese Classical — which matches the chart's shape through this whole branch, including the intermediate new-style node.
Guangdong style
Two wiki pages agree that Guangdong Style is a 13-tile game built on HKOS with New Style winning conditions and additive scoring, born in the 1980s, matching the chart. Both also independently support the chart's caveat about influence running two ways: one calls it popular in South China and Taiwan, the other says it has more in common with Hong Kong, and with Taiwan post-1990s, than with its mainland neighbours.
Hong Kong old style
Tiles, hand, scoring and payout agree across the sources: 136 tiles with flowers optional, 13 in hand, fan only with fu abolished, and only the winner paid. The date conflict that made this node 'conflicting' is gone at its source. Re-read on 2026-09-25, the Mahjong Wiki's HKOS and Guangdong pages both date the Clear Chapter simplification to the 1950s and give the 1970s to the later 3-3 system, which the chart's '1950s, fu removed' now matches. The earliest dated fan-only ruleset is Nguyên Xuân Mai's (Haiphong, 1950), as charted by Depaulis and Tsui. The descent is disputed. Kwan derives HKOS directly from the late classical game, with New Chapter added afterwards, and has no pattern-heavy stage before it. The chart and the Mahjong Wiki put a new-style stage between. What the 1920s texts do show is Kwan's payout step: Babcock's New Method and Tchou's nouveau mode are his 出銃么二制, and Sloper's early-forms table scores HKOS at 4 against the 1920s New Method and 0 against 1920s Chinese Classical. The node's detail is otherwise sound. The later stages (二二制, 三三制, the rising minimum, discarder pays all) belong to its descendants and are recorded here only for context.
Houchi rules
Two independent sources agree that the 1952 Houchi/Hochi publication is the origin point of modern riichi. The Chombo Club gives the exact serialisation window, 11 November to 5 December 1952, and Amano Daizou as author; riichi.wiki, written by a different community, independently makes the same ruleset the hinge of the modern game. This is now one of the better-supported nodes in the Japanese branch.
India / Mumbai style
The rotating round rules that distinguish this node are confirmed in detail, each round with different conditions. The author also notes that he corrected his earlier belief that it was the same as British Western, after visiting Mumbai in 2009.
Japanese classical / arushiiaru
Sloperama and the wiki agree on 136 tiles, a 13-tile hand and point-and-double scoring, and the Krakow Chombo Club history independently dates the first Japanese mahjong organisation to 1929. Sloperama's early-forms table adds real structural evidence: Japanese Classical scores identically with the 1920s New Method and only 4 against standard Chinese Classical.
Korean traditional
104 tiles and a 13-tile hand are confirmed by two sources, and Sloperama's suit list independently confirms bamboo's absence, which is the chart's distinguishing detail. Wikipedia adds two rules that mark the branch: season tiles used but called flowers, and no melded chows. Note that the Korean three-player game described by Wikipedia keeps bamboo 1 and 9, so it is a different reduction from this node's 104-tile game. None of this establishes the parent, which is why the chart draws the link dashed.
Late Qing style
This node now rests on its primary source, not only on summaries of it. Li Boyuan's novel, read here in the Wikisource text, shows the game as a pastime of Jiangnan officials. In chapter 31 it plays out and scores one complete hand. Exposed sets count half of concealed ones: an exposed honour pung scores 4 and a concealed honour pung 8. Going out adds 10, and the pair adds nothing. The total is then doubled, and one-suit is a doubling. That is the count-then-double structure Sloper's catalogue and the Depaulis-Tsui chart assign to this node, and it is the structure of the later Chinese classical game. Sessions ran in eight rounds with a reseat by dice after four, and points were kept in counters.
Han Bangqing's 海上花列傳 (1892), also read at the source, carries the same game eleven years earlier and settles two points Li leaves open. First, the hand: its chapter 26 winner holds 14, discards one and waits on 13, so the 13-tile hand is in the text by 1892. Second, the payment: 'the three others settle their counters' on a discard win, so in Shanghai in 1892 all three losers paid, not just the discarder. Its bystander's reckoning (22 × 8 = 176) shows the doublings multiply by two. The 1892 text also has seats drawn by lot, the dealer set by dice, eight-round sessions, a pre-agreed cap on losses, and the named hand 對對和.
What the texts do not settle. Neither novel names the doublings in its scored hand. Sloper's 'only the winner is paid' means losers do not settle among themselves. That fits both novels: the losers pay the winner in 1892, and the dealer loses 200-odd dollars on one hand in 1903. But neither novel states the rule. Sloper's 1903 table and the Depaulis-Tsui 1892-1911 chart describe much the same game. The chart pools six novels, though, and a feature it scores may not come from Li. Li's first half was serialised from April 1903, which dates the node, but chapters after 48 may be Ouyang Juyuan's.
Macau simplificado
112 tiles and a 4-tile hand winning on 5 are now confirmed from the law that created them. Sloperama, working from a 2006 newsgroup post, catalogued the game correctly: Macau's official casino regulation, Portaria 135/91/M, gives 112 tiles — the three suits and the green dragon — a hand of 4 winning on 5, no flowers, and payment to the winner only. It adds what the catalogue could not: this is not a regional style that grew up in Macau but a second, faster form written into the casino rules in 1991, where a discard may be claimed only to win and never for a chow or a pung, and where the House takes a commission on every won game. Its extreme hand size is a casino's doing, not a region's, and the regulator still names those rules today.
Mahjong Masters Millions
Two sources, and the second turns a bare label into a dated, attributed ruleset: created for the 1997 Mahjong Masters Millions tournament by Zuwei Thomas Feng and Hiroyuki Takashaki, as Internet Mahjong Server rules — a deliberate hybrid of modern Chinese and Japanese play. Pattern selection follows the Japanese side, including the sacred-discard and missed-discard rules and better scoring for concealed hands, while the Japanese low-scoring patterns are dropped and a self-drawn win instead takes auxiliary points in the Chinese modern style.
It also decodes Sloperama. Its entry says to score the sets, then 'determine the aux value, then multiply' without saying what aux is; the preset documentation shows auxiliary points in use, ten of them for a self-drawn win. The one disagreement is flowers: Sloperama calls them optional and gives 136 or 144 tiles, while this source says the ruleset does not use flowers and seasons at all, which resolves the count to 136.
The chart's placement under modern inventions is well supported by this. A 1997 tournament ruleset assembled deliberately from two traditions is an invention, not a descendant.
The payment side, re-read, is as hybrid as the scoring. Only the winner is paid. A discarder pays alone, and a self-draw is split three ways. East winning takes ×1.5 but East losing is not penalised. Points pass through a conversion table rather than being doubled. Sloper's timeline dates the Internet Mahjong Server to 1996, a year before the tournament, which fits the ruleset being an IMS product.
Mainland regionals
The wiki independently confirms both of the chart's caveats about this group: over 100 variants today, and a 1990s divergence in which some built on Old Style and some on an additive New Style. It also dates the Sichuan/Chengdu Bloody rules to the 1990s, matching the chart's separate node.
Malaysian 4-player
164 tiles and 24 flowers match exactly, from two Sloperama pages. The winning-hand restriction to All Chow, All Pung or a flush is a substantive rule the chart omits.
MCR / Chinese official
Sloperama and the wiki agree on 144 tiles, a 13-tile hand and pattern scoring; the wiki adds the 8-point minimum. Wikipedia independently dates the sport certification to January 1998 and the rules to that September, matching the chart. One wrinkle: the wiki's infobox is built on a 2006 rulebook, not the 1998 edition the chart cites.
MCR scoring reference
Primary-sourced. The rulebook itself was fetched and read: 144 tiles of which 108 are suit tiles, a 13-tile hand, eight flowers replaced from the back of the wall, and an 8-point minimum enforced by a 10-point forfeit to each opponent for declaring below it. That matches the wiki's infobox and Sloperama's summary, and puts a number on the penalty neither of them gives.
Mhing
Well attributed and now described in play. Suntex International of Easton, Pennsylvania published it in 1982, designed by Robert Sun, and Spear's licensed it in the UK in 1984. The card-count caveat resolves. Owl Works gives the playing deck as 150 cards: 108 suit cards, 28 honours, 8 flowers and 6 wilds. Spear's 158 is that deck plus six scoring cards and two blanks. The play is a close card transposition of the mahjong structure, with 13 dealt, chow and pung on discards, pung over chow, and four sets and a pair to go out. It adds wild cards, a 2-6 player count and a published hand chart. That is why it sits among the inventions: a new product built on the game rather than a ruleset descended from one.
Modern riichi / yonma
Three sources now agree on 136 tiles, a 13-tile hand and no flowers in play, and riichi.wiki adds the han-and-fu structure and the one-yaku minimum. The chart's 1970 date for red fives is still uncorroborated: riichi.wiki confirms red five dora exist and are optional but dates the modern variant only to the 1960s, with the popularity surge after 1969.
Modern rules
Two independent sources agree that the 1967 rules introduced bazoro, oka and ura dora, exactly as the chart says. There is an apparent conflict with the 1957 Tokyo rules, which already carry a zoroba, and it resolves cleanly: riichi.wiki specifies that 1967 introduced TWO HAN bazoro, the fixed form still built into modern scoring tables, where 1957 had a variable bonus rolled on the dice that did not count toward the minimum. The chart's attribution is correct once bazoro is read as the fixed form.
Nanjing style
Two independent sources agree, and the second is far richer than the 2006 newsgroup report the node rested on. Both give 144 tiles and dragons treated as flowers; Baidu makes the arithmetic explicit, twenty flower tiles being the twelve dragons plus the eight seasons and plants. Both make flower count the scoring engine. Baidu adds the structure the chart lacks: minor and major hands, four hard flowers to qualify for a minor, a flower kong alongside exposed and concealed kongs, and a major hand retaining the dealership. The local name is Jinling Mahjong.
Nepalese
144 tiles with a 10-tile hand winning on 11 confirms the chart's 11-tile winning hand precisely, and the source identifies it as Western-like but with fewer tiles in hand, which supports its placement under the Western branch.
NMJL card
The conflict is settled. Four secondary sources date the founding to 1937 in New York City, two of them naming the Essex House meeting, and the League's own website now confirms it from the inside: in 2026 it celebrates 89 years, which counts back to 1937. Wikipedia's statement that the League published Official American Rules in 1935 cannot be reconciled with that and should be treated as the outlier; whatever the 1935 publication was, the organisation dates itself from 1937.
Pong
120 tiles and the Dobree 1955 attribution match. The source adds the defining structure: four threesomes and no pair, pungs and knitted pungs only, an 11-tile hand winning on 12. That is a far more distinctive game than the chart's one-line entry suggests.
Proto-mahjong
Well dated from surviving objects, and now traced to where each origin story starts. The chart's c. 1870s rests on four tile sets and their collectors' own records. The two Glover sets are in the American Museum of Natural History and the Brooklyn Museum; Glover was at Fuzhou in 1872-73, and his 1875 cover note names the game after the sparrow and counts its 148 pieces. Himly's set from Ningbo is lost, but he described it tile by tile in 1889 and 1901. Culin's 1893 report, read here in scan, lists a Glover set piece by piece. They agree on what the proto set was: the three money suits that carried over, the winds, eight blanks and four seasons, plus a layer of 'king' tiles that did not survive. The Himly set adds suit jokers. There are no flowers, and one or both of the later dragons are missing.
'Rules lost' also holds up, and the one source that seems to contradict it falls apart when checked. Wikipedia says the consul F.E.B. Harvey wrote down the rules Chen Yumen taught him in 1861. Wikipedia's own footnote calls the story probably forged. The essay it cites takes the story from a 2008 CCTV documentary and does not mention rules. The museum the documentary said holds the diary told Sloper and three other researchers that it does not. The closest thing to proto-mahjong rules is Wilkinson's 1890 note on chungfa, and that describes the later, c. 1890 form, not the 1870s one.
Two things stay open. The first is who made the game. Chen Yumen (1817-1878) is a real, documented official of Ningbo, but the earliest texts that credit him with the game are Powell's 1923 magazine piece and Chen Ailu's 1925 one, and no late-Qing source names him. Late-Qing and early-Republican writers name Ningbo as the place, and Du Yaquan's 1933 account puts it among Fujian and Guangdong seafarers first. The second is what the kings did. Xu Ke's Taiping account matches the Himly set's tiles, and Zhang Deyi's 1889 diary still lists them as optional flowers. Stanwick and Xu's inner-and-outer-flower reading is a reasoned hypothesis, not a recovered rule.
Ebashi's Japanese study, read on 2026-09-24, gives the period a frame the English sources lack. The game before Ningbo's reform was 花麻雀 'flower sparrow', and the pruned 136-tile game after it was 清麻雀 'plain sparrow'. On that reading this node is the flower-sparrow stage, and Chinese archaic is the plain-sparrow stage that followed. Two branches that kept the proto extras and survived into the 20th century are now in the tree as this compendium's own additions: North China's hua ma que and Changsha's wang ma que.
Regional US house rules
Three sources support a distinct pre-NMJL layer: a 1924 American Code of Laws produced by a committee, the proliferation of incompatible rule sets that prompted it, and Babcock's own membership of that committee. This entry is a real stage in the history, not a placeholder. No source gives it tiles or a hand size because it is a period of time, not a single ruleset.
Sanma
Now properly sourced. Wikipedia gives the exact tile reduction and three rules of its own, and states explicitly that the three-player game is not a mere adaptation but has its own idiosyncrasies, which strengthens the chart's decision to give it a node. That cuts against riichi.wiki's framing of three-player play as an accommodation for a missing fourth. Both are describing real practice; the chart's node is better supported than the previous pass concluded.
Shanghai / Hong Kong new style
Other sources now confirm this entry, not just the chart. Sloperama's Hong Kong New Style entry matches on tiles, hand and payout, and its 'many many special hands' against Old Style's 'few' is the substantive difference the branch depends on. The wiki's mainland page describes later pattern additions to the classical game as Shanghai New Style, and its GMCR page states outright that HKOS was adopted as a backlash to HKNS, which had evolved from Chinese Classical. Three sources therefore place a new-style stage between Chinese Classical and Hong Kong Old Style. The 1920s date now has contemporary support, for the payout at least. Babcock (1923), Tchou (1924) and the Shanghai manual (1924) each describe a 'new method' paying the self-drawn winner double and, in the first two, making the discarder pay double. That is the self-draw and discard payout the wiki gives New Style, in an early form: the discarder pays double, not all. Tow (1924) shows the Shanghai game already differing from the Cantonese in walls, base score and special-hand values. What no 1920s text shows is the defining feature, the large number of added winning patterns; the manuals of 1914-24 list only ten special hands. The node also bundles two stages the wiki separates, Shanghai New Style (1920s-40s) and Hong Kong New Style (1950s-70s). Depaulis and Tsui date their Shanghai New Style column 1950-60 and place it in Hong Kong. The names conflict. English Wikipedia calls today's standard Hong Kong faan game 'Hong Kong New Style'; the Mahjong Wiki reserves that name for the pre-HKOS stage and calls the modern game New Chapter; Chinese Wikipedia runs 老章, 清章, 新章. Racster puts the Shanghai base at 20 against Hong Kong's 10, while Tow calls a base of 10 'quite common in Shanghai'. Kwan's history (2009) challenges the node itself: he derives HKOS directly from the late classical game, and his only pre-HKOS growth in hands is the late classical inflation. On his reading, this node is the late classical game plus the new-method payout, not a separate pattern-heavy stage.
Sichuan / Hunan / Tibetan
108 suit-only tiles match. The source is more precise than the chart in a useful way: Hunan and Tibetan/Sichuan are given different winning conditions, so this single node covers at least two distinct games.
Sichuan bloody rules
The best-evidenced node in the mainland group: five sources, four of them independent of the chart. All agree on 108 tiles with honours and flowers removed, the one-missing-suit requirement and the continuation of play after the first win. The wiki independently dates the Chengdu adoption to the 1990s, matching the chart. Two sources also qualify the no-chow rule: chows cannot be called from a discard except to complete a winning hand, which is narrower than a flat prohibition.
Singapore style
Two independent sources agree on the animal tiles that define the variant, and Wikipedia names all four: cat, mouse, cockerel and centipede. Sloperama's 148 total and requirement of all twelve flowers is consistent with that. Wikipedia's mid-hand payout rule is a structural feature neither the chart nor Sloperama records.
Taiwanese style
Four sources agree on the 144-tile set and the 16-tile hand, and two independently describe tai scoring as linear, which the wiki's infobox also shows. Wikipedia adds two structural features the chart omits and which distinguish the branch sharply from Hong Kong play: recurring dealerships, and multiple players winning from one discard. This is among the best-supported nodes in the tree.
Tokyo rules
Two independent sources now agree on the 1957 date, Amano's authorship and the one-han minimum that gives the node its name, and riichi.wiki supplies the Japanese titles, 東京ルール and 全一飜縛り麻雀. The page also settles a question the next node raises: the 1957 rules have no kandora and no uradora, so the chart is right to place ura dora at 1967. Its dice-driven han bonus is the early, variable ancestor of bazoro, and it explicitly does not count toward the minimum.
Twenty Point Mah Jong
1924, R.F. Foster, and the 20-point minimum that gives the game its name all confirm the chart. Sloperama adds the motive and the verdict: Foster's attempt satisfied nobody but paved the way for American mah-jongg in the 1930s, which supports the chart's placement of it beside the American line rather than in it.
Vietnamese classical
Two independent sources give 160 tiles built from eight jokers and eight extra flowers, matching the chart exactly. Sloperama adds the eight kings and queens and the nineteen ways of going out, which is the feature that ties the classical and modern forms together.
Vietnamese modern
Two independent sources agree, and Wikipedia is the more precise: triplicating or quadruplicating the eight jokers gives exactly 176 or 184 tiles, where Sloperama says 176 or more and the chart says 176+.
WMPA / WMF
136 tiles, no flowers, and the distinctive squared-faan scoring. Sloperama states independently that WMPA is Korean and based on Japanese Modern, which is exactly the placement the chart gives it, under the Japanese branch but coloured Korean.
Wright-Patterson
Sloperama and The Mahjong Line agree on the Wright Field, Ohio origin among Air Force officers' spouses. Sloperama adds 144 tiles, 70 to 80 hands and a Charleston, which places it closer to Western than to NMJL American.