G: I found something useful almost immediately by shifting the search toward Japanese SG-1000/Othello Multivision sources: the commonly repeated 1983 date for the Japanese SG-1000 version appears to be wrong. The Japanese release was almost certainly 1984, and the best-supported surviving date is December 1984. I could not, however, find a defensible day of the month—and there is a good historical reason that may simply never have been recorded cleanly.

The important wrinkle is that this wasn't really a Sega release in the normal sense. The Japanese cartridge was published by Tsukuda Original (ツクダオリジナル) as part of the software line for its Othello Multivision, an officially compatible SG-1000-derived machine. Sega itself confirms that Othello Multivision cartridges could be played on the SG-1000. citeturn922426search1turn634531search7 The Japanese cartridge's catalog number was OM-G008, price ¥4,300, and contemporary packaging identifies Tsukuda Original rather than Sega. citeturn521535search2turn256036search9

Release date

The strongest Japanese database I found that gives better than a year is ピコピコ大百科 (Pikopiko Encyclopedia), which lists:

007 ジェームズ・ボンド — December 1984 — Tsukuda Original — ¥4,300. citeturn256036search0

Its complete SG-1000 chronology likewise places 007 James Bond and Space Armor in December 1984, after the November releases and before Girl's Garden in January 1985. citeturn256036search1 Sega Retro independently gives 1984-12 for OM-G008. citeturn922426search3

There is one dissenting database: the French Sega archive Guardiana says November 1984. citeturn256036search11 But the Japanese sources I found either say December 1984, merely 1984, or explicitly admit that the exact date is unknown. One Japanese collector's SG database actually has:

発売日：不明
“Release date: unknown”

for OM-G008. citeturn521535search0

That last point matters because Sega itself warns that precise release dates were often not formally established for this period. On Sega's official SG-1000 software archive, it says that because clear release dates were not set at the time, Sega has had to classify games using the copyright dates printed on their packages. citeturn922426search0 So I would not invent December 15 or December 20 or anything similar just to fill your catalog field. I specifically searched Japanese combinations for OM-G008, the title, Tsukuda, December 1984, and likely date formats, and nothing credible produced a day.

For cataloging, I would therefore record it as:

Japan: December 1984 [day unknown]

with perhaps a note such as “Some sources give November 1984; Japanese retrospective databases generally give December.”

What the game actually is

There's another useful correction to the usual English-world description. This Japanese cartridge is essentially the Japanese SG-1000/Othello Multivision adaptation of the early Parker Brothers James Bond 007 game, rather than some obscure independent Japanese Bond game.

The original game was created for the early-1980s American market and appeared on Atari/Coleco/Commodore platforms. A Japanese retrospective from 4Gamer describes Parker Brothers as having launched it for the 1983 Christmas market, followed by Tsukuda Original's Japanese Othello Multivision release in 1984. citeturn634531search4 Modern preservation databases identify the underlying game as developed by On Time Software, with the SG-1000 version ported by Tsukuda Original and licensed from Parker Brothers. citeturn256036search14

That explains where the 1983 date contaminating some English references probably comes from: 1983 belongs to the origin of the Parker Brothers project, not the Japanese Tsukuda cartridge. The Japanese release followed the next year.

There's even surviving ROM evidence supporting a distinct Tsukuda conversion. MAME's software database records at least two Japanese revisions, v2.6 and v2.7, under OM-G008, with Tsukuda Original as publisher and a 1984 date. citeturn256036search12 That's a nice little cataloging detail I hadn't expected to find.

Gameplay and subject matter

It's a side-scrolling vehicle shooter/action game, rather than a conventional run-around-as-Bond game. Japanese retrospectives describe Bond primarily operating the Lotus Esprit, moving across land and underwater and using various weapons. citeturn634531search9

The game draws its four missions from four Bond movies:

Diamonds Are Forever
The Spy Who Loved Me
Moonraker
For Your Eyes Only

The Japanese ピコピコ大百科 specifically identifies those four movies as the stage inspirations. citeturn256036search2

So the connection is thematic rather than an attempt to adapt one film from beginning to end. The stages extract recognizable Bond action set pieces—especially the transformation/underwater capabilities associated with the Lotus Esprit—and turn them into scrolling arcade-style missions.

A Japanese player retrospective rather charmingly summarizes it as essentially: although the title says 007, apart from the familiar music at the beginning, the game itself is basically a horizontal shooter in which you control the Bond car; the reviewer remembers roughly four stages based on Bond films and says that, primitive as it was, wanting to see what came next made it exciting at the time. citeturn521535search0

That seems a pretty good period-appropriate summary.

Why it wound up on an SG-1000 at all

This is perhaps the most interesting historical context for your catalog.

Tsukuda Original produced the Othello Multivision, a licensed SG-1000-compatible machine with Othello built into it. It had a tiny dedicated cartridge library—only eight Tsukuda cartridges are generally recognized. 007 James Bond, catalog number OM-G008, was the last-numbered title in that line. citeturn634531search10turn634531search7

Because the machines were compatible, these cartridges also ran on Sega's SG-1000 and SC-3000, which is why databases now usually classify 007 as an SG-1000 game even though its actual Japanese commercial identity was an Othello Multivision cartridge.

That's a distinction I would preserve in a catalog:

Platform: Othello Multivision / SG-1000 compatible
Publisher: Tsukuda Original
Catalog: OM-G008
Japanese release: December 1984 (exact day presently undocumented)
Price: ¥4,300
Genre: horizontal scrolling action/shooter
Origin: Japanese conversion of Parker Brothers' James Bond 007
Japanese ROM revisions known: v2.6, v2.7

And therefore I would change an existing “1983 — SG-1000 — Japan” record rather than merely qualify it. The evidence against 1983 for this particular release is pretty strong.

One further thing I can try, if this is worth the effort for your cataloging experiment, is going a level more archaeological: search scans of 1984 Japanese computer/game magazines, Tsukuda catalogs and newspaper/toy-industry advertising from November–December 1984. That's probably the only realistic way we might tease an actual on-sale day out of this; the normal Japanese web databases seem to have reached their limit at “December 1984.”

G: [Non-text content: thoughts; content_type=thoughts]

G: What you’re noticing is real, and the SG-1000 example is unusually good because Sega itself effectively says, “we did not have exact release dates for this stuff.” That points to a structural change in what “release date” meant, not just to nerds occasionally losing trivia.

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: The basic answer is that older games often did not have a single event corresponding to our modern concept of “release date.” Sometimes they did, which is why certain old dates are extremely solid. Sometimes the product simply leaked into commerce through distribution, and decades later historians are being asked to assign a date to something that nobody at the time thought needed one.

That distinction explains the weird unevenness you’re seeing.

Today, when we say a game “released on October 17,” several systems reinforce that date simultaneously: marketing announces it; retailers are told not to sell before it; reviewers work against it; digital storefronts activate at a defined time; corporate databases record it; players discuss launch day online. The release date is itself an object being manufactured by the publisher.

For a lot of early console software, the underlying model was closer to:

finished → manufactured → distributor → wholesalers → individual stores → eventually appears on shelves.

There may never have been a meaningful moment where somebody said, “This game goes on sale nationally Tuesday, November 13.”

That is particularly clear with your SG-1000 case. Sega's own historical archive says:

当時は明確な発売日が設定されていなかったため…

roughly, “Because definite release dates were not established at the time…”

Sega therefore classifies some SG/SC software using the copyright year on the packaging. That isn't some amateur database admitting defeat. The company that released the games says the datum sometimes did not exist in the first place. citeturn502159search0turn502159search4

And amusingly, only a few years later Sega's records look completely different. Its Master System list gives dates such as Fantasy Zone — June 15, 1986, Hokuto no Ken — July 20, 1986, etc. citeturn502159search16

So you're looking right across the transition.

Why some ancient games nevertheless have beautiful exact dates

There are several different ways a date can survive, and people tend to collapse them all into the innocent-looking database field Released: YYYY-MM-DD.

A game can have an exact date because there was a genuinely coordinated launch. A major new console, flagship title, arcade conversion, heavily promoted licensed property, etc., was much more likely to have publicity saying essentially “ON SALE JULY 15.” The SG-1000 hardware itself is a perfect example: Sega confidently gives July 15, 1983 and later explicitly celebrated that anniversary. citeturn502159search6turn502159search7

Or the date can survive because somebody happened to preserve excellent evidence. An advertisement dated September 18 saying “NOW AVAILABLE”; a distributor price sheet; retailer purchasing records; a magazine's new-release column; correspondence between publisher and dealers; newspaper advertising. Historians can triangulate those.

That's substantially how researchers reconstruct early American dates now. Atari Archive's release-date project explicitly says early-game dating is difficult because it was poorly tracked, and reconstructs dates from newspaper ads, gaming periodicals, trade publications and a contemporary retailer-oriented newsletter. citeturn502159search9

So an apparently authoritative:

Adventure — June 1980

might really mean:

“June 1980 is the smallest interval into which all surviving evidence allows us to squeeze its commercial appearance.”

And an apparently authoritative:

Pac-Whatever — June 17, 1982

might mean:

“A dated retailer document says shipments were scheduled June 17.”

Both become an identical database field.

That's one reason your anticipated nerd wars over which date counts actually do exist. They tend to happen when there's enough evidence to fight over.

One person wants:

publisher-announced date

another:

first confirmed retail sale

another:

first shipment from distributor

another:

first advertisement saying ‘available now’

another:

copyright/database date

another:

national availability rather than one store breaking the intended date.

The paradox is that argument is a luxury created by evidence.

For an obscure game you instead get:

“Well, Game Machine mentioned it as forthcoming in October; this December catalog lists it normally; somebody remembers buying it around Christmas; box says ©1984.”

There aren't enough competing dates to have a fight.

And commerce itself makes this worse

This is the part I think most directly answers your intuition that surely somebody should have remembered.

Suppose Tsukuda manufactures James Bond cartridges during November 1984.

On November 22 it tells distributors they can order them.

A Tokyo wholesaler receives inventory December 3.

A large Tokyo toy store puts them out December 6.

A smaller Osaka shop gets them December 14.

Some regional stores don't receive them until January.

When was it released?

In 1984, everyone involved might have considered that an absurd question.

Tsukuda made it.

The distributor carried it.

The shop stocked it.

Customers bought it.

Nothing important happened at midnight between December 5 and December 6.

The historical question only becomes important thirty years later, once we invent giant databases in which every game has a field called release_date.

And now somebody discovers that reality never generated the value required by the schema.

That is the conceptual piece I think you were circling.

Why the problem isn't evenly distributed by age

It's not simply:

older = fuzzier.

It is much more like a probability function involving era × company × country × distribution system × prominence × surviving documentation.

A 1977 Atari game might have an astonishingly well-established date because Atari paperwork survived.

A 1984 Japanese cartridge from a small toy company might have almost nothing.

A 1985 Nintendo game might be extremely well documented because Nintendo had begun operating a much more controlled software ecosystem.

A 1989 shareware PC game might become fuzzy again because someone uploaded version 1.0 to a BBS and nobody can establish which BBS got it first.

An arcade game introduces another mess because you can distinguish announcement, trade-show debut, test location, first commercial installation, widespread rollout, and overseas release.

Computer software can be worse: completion date, duplication date, mail-order availability, dealer availability and first review may all differ.

Consequently, the history isn't progressing smoothly from “no records” to “records.” Different distribution models repeatedly make dates sharper or blurrier.

There's also a huge survivor-bias effect in what we think we know

This is probably contributing substantially to your feeling that the gaps are bizarre.

Gaming databases give the impression that we possess a gigantic systematic historical record.

We don't.

We possess a gigantic retrospective reconstruction whose entries have wildly different evidentiary pedigrees.

The Video Game History Foundation is still actively collecting retailer catalogs, magazines, publisher materials, development records and other ephemera precisely because these mundane documents frequently contain information that wasn't preserved anywhere else. Its retailer-catalog collection includes material going all the way back to 1973, and its archive explicitly describes itself as a continuing work in progress. citeturn502159search1turn502159search8

The spectacularly important games also undergo enormous amounts of historical archaeology. Twenty people investigate Super Mario Bros. Nobody spends three years determining whether OM-G008 007 James Bond first appeared in Japanese toy shops on December 7 or December 14.

So eventually a circular process emerges:

well-known game → more research → better metadata → easier to research → more confident databases

versus

obscure game → copied secondary date → little investigation → same date copied everywhere.

That can even create false certainty. Ten websites agreeing on “1983” isn't ten sources. It may be one unsourced database entry copied nine times.

Which is almost certainly something you want to watch for in the catalog you're constructing.

I think there are actually four useful categories for your project

Rather than imagining release-date quality as simply exact/unknown, I'd mentally classify dates something like:

1. Designed date — There really was an official national launch date. “On sale July 15.”

2. Documentary date — Contemporary evidence establishes a very narrow date, even if nobody necessarily orchestrated a launch.

3. Reconstructed window — Evidence supports “December 1984,” “late 1984,” or “between November and January.”

4. Conventional date — Databases give a date, but its original evidence is unclear or merely repeated.

That fourth category is important. A date with eight digits looks more precise than “Winter 1984,” but may actually contain less historical knowledge.

And there is a fifth state that databases hate:

There was no singular release date.

That's not “information lost.” It's a historical question imposing a modern category on a commercial process that didn't have one.

I think that's the part that resolves your original apparent paradox most cleanly.

The nerds didn't necessarily forget whether the game came out on December 12.

December 12 may never have been a fact.

What was factual might instead have been something like:

Tsukuda began distributing OM-G008 during December 1984.

And forty years later we're trying to force that perfectly sensible historical statement through a database column that only accepts YYYY-MM-DD.

That also gives us a better methodology for the next games you throw at me: rather than merely hunting for “the release date,” I can specifically look for the earliest recoverable commercial chronology—announcement, planned release, shipment, retail appearance, contemporary “now on sale” notices—and tell you what kind of date we've actually recovered rather than silently pretending all dates mean the same thing.

G: Yes. I’ll treat mid-1985 as the only starting bound, then work backward only when I find dated contemporary evidence—ads, catalogs, magazine listings, retailer material, scans, or personal records. I’ll keep a running “earliest confirmed by this source” trail and translate the relevant Japanese each time.

G: [Non-text content: thoughts; content_type=thoughts]

G: A useful wrinkle has appeared: Beep is giving us actual contemporary checkpoints, but its cover dates are misleading. The January 1985 issue was actually on sale December 8, 1984, and the March issue February 8, 1985. That gives us much finer “photographs of the market” than a retrospective database does.

G: [Non-text content: thoughts; content_type=thoughts]

G: Now we have a genuinely interesting backdoor hit. A collector digitized an original Tsukuda promotional leaflet, not a later game database. It advertises an Okamoto Ayako victory promotion running November 10, 1984–January 9, 1985, and the leaflet itself shows 007 ジェイムスボンド among the Othello Multivision cartridges at ¥4,300. That has already pushed the contemporary-evidence boundary back from February 1985 to at least November 10, 1984—and it makes the retrospective “December 1984” date look seriously questionable.

G: [Non-text content: thoughts; content_type=thoughts]

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: Yes. This was worth trying, because the side-road approach has already produced better evidence than the databases did. More importantly, it has produced evidence that makes me think “December 1984” is probably not merely imprecise but actually too late.

I deliberately started with your artificial premise: we know only that it existed by mid-1985. I then worked backward rather than searching for “007 release date.”

The running archaeological trail

1. By June 30, 1985 — definitely in the commercial library.
A contemporary issue of the Japanese amusement-industry newspaper Game Machine, dated July 15, 1985, surveys home video-game systems. It says of the compatible SG-1000 II/Othello Multivision systems:

上の2機種は互換性がありカートリッジはセガ社から40種類、ツクダから8種類、計48種類が6月末までに発売されている

My translation:

“The two machines above are compatible, and by the end of June, 40 cartridges had been released by Sega and eight by Tsukuda, for a total of 48.”

All eight Tsukuda Multivision cartridges therefore existed commercially by June 30. Since 007 is one of Tsukuda's final eight-title library, that gives our first hard contemporary boundary. citeturn435689search18turn542739search0

So:

≤ June 30, 1985

2. February 8, 1985 — 007 is explicitly sitting in a contemporary consumer game catalog.

This one is considerably better.

I found newly digitized 600-DPI scans of Beep, and its March 1985 issue was actually released February 8, 1985. The issue has a substantial feature called:

テレホビーゲーム大作戦
Tele-Hobby Game Grand Strategy

and explicitly covers:

ファミリーコンピュータ／セガSC・SGシリーズ／スーパーカセットビジョン／オセロマルチビジョン

“Family Computer / Sega SC-SG series / Super Cassette Vision / Othello Multivision.”

Then there is an actual テレホビーゲーム・カタログ — “Tele-Hobby Game Catalog”, including a section devoted to Tsukuda's Othello Multivision. citeturn516634view0

I went into the scan rather than relying on the OCR. And there, in the Multivision catalog, is:

007 ジェイムスボンド
ツクダ　4,300円

along with a screenshot and description. The text essentially says:

“A game based on the film Diamonds Are Forever. Avoid enemy attacks while heading out to destroy the oil base—an action game.”

The significant thing for our purposes isn't the description. It's that on February 8, 1985 Japanese consumers were being shown 007 as an ordinary ¥4,300 item in the current Othello Multivision catalog, alongside Q*bert, Guzzler, Three-Player Mahjong, Space Armor, etc. This isn't a later historian remembering it. It's a game magazine putting it in front of customers. The archived scan is genuinely contemporary material; Gaming Alexandria's copy is a 600-DPI scan of the original magazine. citeturn189488view1turn516634view0

New boundary:

≤ February 8, 1985

3. December 8, 1984 — interesting negative evidence.

Beep's first issue, nominally “January 1985,” actually hit stores December 8, 1984. I searched its digitized text and its new-software material. I could find Sega's contemporary titles—Champion Soccer, Champion Boxing, Lode Runner, etc.—but not 007 or Tsukuda's Multivision line.

That absolutely does not establish “not released yet.” Beep had only just launched; Tsukuda was a relatively marginal toy company in the videogame market, and absence from one magazine isn't evidence of nonexistence. But it becomes interesting once we compare it with what I found next. citeturn478218view0turn962293view0

So I scribbled this down as:

Dec. 8: no positive sighting; weak negative only.

I would not use it as a lower boundary.

4. And then the weird collector backdoor paid off: November 10, 1984

I searched for exactly the sort of obsessive-personal-archive material you suggested, and found a Japanese collector who has digitized catalogs he personally accumulated from childhood onward. His collection includes an original Tsukuda Original Othello Multivision sales leaflet. He isn't presenting release-date research; he's basically saying, “Here are my old catalogs.” Which is precisely why this is useful. citeturn383011view0turn840820view0

And the leaflet is datable.

Across the top:

岡本綾子優勝記念
'84 全英女子オープン

“Commemorating Ayako Okamoto's victory — '84 Women's British Open.”

Then:

〈実施期間〉59.11.10～60.1.9

That is Shōwa dating:

Promotion period: November 10, 1984 – January 9, 1985.

And underneath that promotion, Tsukuda displays the software currently associated with the Multivision.

Sitting right there in the bottom-left of the software lineup is:

007 ジェイムスボンド
4,300円

with its screenshot.

The leaflet's heading for this section is:

オセロマルチビジョン用ゲームカートリッジ・ソフトいろいろ

roughly:

“Various game-cartridge software for Othello Multivision.”

There is no “coming soon,” “scheduled for December,” “advance announcement,” or tentative price notation attached to 007. It is presented exactly like the other purchasable cartridges, with a normal retail price. citeturn210983view0

You can see the surviving leaflet itself here: urlTsukuda Othello Multivision promotional leaflet scanhttps://ratscats.client.jp/cs/omv.jpg

So our boundary just jumped to:

≤ November 10, 1984

And that is vastly more interesting than “December 1984.”

What I think that does to the December date

I would now reject December 1984 as a defensible release month unless somebody can produce primary evidence explaining the contradiction.

There is a theoretical escape hatch: perhaps Tsukuda printed a November 10–January 9 promotional flyer which included a game that wasn't going to reach stores until December. Companies certainly advertise forthcoming products.

But the leaflet gives us none of the usual signals for that. 007 is mixed straight into the existing software lineup at ¥4,300, with no qualifying mark. And this is specifically a purchase promotion:

期間中お買上げの方に限り……

“Limited to customers purchasing during the promotion period…”

That makes November 10 considerably stronger than merely “the game had been announced by November 10.”

I would phrase the conclusion conservatively as:

007 James Bond was being marketed by Tsukuda Original as part of the available Othello Multivision cartridge lineup no later than November 10, 1984.

That is something I think we can actually document.

It also suddenly makes the French Guardiana database's unexplained “November 1984” much more interesting. I wouldn't use Guardiana as the evidence—the whole point of this exercise is not to—but apparently whoever put November in there may have had access to information that the December-dating databases didn't. citeturn842526search5

And I found another useful artifact while wandering around

An old Yahoo Auctions listing preserved photographs of another Othello Multivision product catalog. It appears to be an earlier generation of the catalog because its software lineup is shorter. It has Q*bert, Guzzler, Space Mountain, Three-Player Mahjong, etc., but not 007. citeturn842526image0

Unfortunately, that particular flyer isn't visibly dated in the surviving auction material, so it can't yet give us the other half of the bracket. But conceptually this is exactly what we want:

Catalog A: 007 absent
↓
Catalog B: 007 present
↓
date both catalogs
↓
release/marketing window

If I can find a dated version of that earlier leaflet—or determine from its hardware/software contents when it must have been printed—we may be able to squeeze November substantially further.

And there are some clues for doing exactly that. The catalog generations contain things such as the revised FG-2000 Multivision, Okamoto's golf game, Three-Player Mahjong, Challenge Derby, etc. Their introduction sequence can potentially be used as internal fossils to date an otherwise undated piece of paper.

Where I'd put the pin right now

Not:

December 1984

and certainly not:

1983

but:

Japan — no later than 10 November 1984; exact release date currently unknown. Probably November 1984 or earlier.

I would give November 10 a different metadata status than an actual release date. Something like:

Earliest contemporary availability/marketing evidence: 1984-11-10

because that's precisely what we've proven.

The really enticing possibility now is that November 10 itself might turn out to be the launch date. The leaflet starts a major Christmas-season promotion that day, and 007 and Space Armor appear as the two final numbered Multivision cartridges. It would make commercial sense for Tsukuda to introduce new software alongside the promotional push. But that's a hypothesis, not something the leaflet proves.

So I think the next archaeological target has become much narrower than when you asked me the question:

September 1 – November 10, 1984, concentrating particularly on late October/early November toy-industry material, Christmas catalogs, Tsukuda advertising, and the transition between those two generations of Othello Multivision catalog.

That is now small enough that another lucky piece of ephemera could conceivably get us down to a week—or even produce the damned date.

And this is a lovely Japanese vocabulary case for the project: 発売日(はつばいび) is literally “release/sale-start date,” whereas what we've actually established here is closer to 販売中(はんばいちゅう) — “on sale/currently being sold.” The latter may ultimately be the historically knowable fact. memcite

G: That hypothesis about the November/December dates is plausible enough that I’ll treat this last pass differently: I’m going to look for company-history and employee-memory evidence, not ordinary game databases—Tsukuda corporate retrospectives, interviews, trade-history pieces, catalog scans, and any digitized internal/business records that mention the Multivision software line or OM-G008 specifically.

G: [Non-text content: thoughts; content_type=thoughts]

G: I found a promising lead that is much closer to what you had in mind than the web databases: there is a 1986 book by Tsukuda Original founder Mitsuo Tsukuda himself, 佃光雄その人と遺言 (“Mitsuo Tsukuda: The Man and His Testament”), 214 pages, published only about two years after the Multivision period. I’m checking whether any contents, excerpts, library metadata, or later quotations expose the videogame chapter—and I’m also tracing a 2007 long-form interview with former Tsukuda planning chief Takeshi Wakui in the Japanese toy trade press.

G: [Non-text content: thoughts; content_type=thoughts]

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: I did find evidence that the sort of source you were imagining exists for Tsukuda Original as a company, but I did not find a surviving digitized source that gets us closer than the November 10 evidence for 007. That negative result is fairly meaningful after this pass.

The strongest lead is a 1986 book written by Tsukuda Original founder Mitsuo Tsukuda (佃光雄), titled 佃光雄その人と遺言 — roughly Mitsuo Tsukuda: The Man and His Testament. It is 214 pages and was published in May 1986, so chronologically it is almost ideal: written essentially immediately after the 1983–84 Multivision experiment. Unfortunately, Japan Search/NDL explicitly says no digital content exists for the book, and the copies I could locate are physical used-book holdings rather than scans. citeturn246683search1turn246683search2

That book is probably the single most promising unexplored source I found. I can't establish from online metadata that it discusses the Multivision specifically, but if you were trying to squeeze the last few drops out of this question, that's the book I'd want somebody in Japan to physically open and check the index/contents for テレビゲーム, オセロマルチビジョン, セガ, or パーカーブラザーズ.

There is also a genuine oral-history trail around Tsukuda. Takeshi Wakui (和久井威), who ran product planning and was one of the major figures behind Tsukuda Original's successful products, gave a multi-part long interview to the Japanese toy trade journal 月刊トイジャーナル (Monthly Toy Journal). Part 2 appeared in the June 2007 issue, pp. 72–76. Later Japanese writing quotes it extensively for things like how Tsukuda acquired Othello and Slime, so it clearly contains detailed first-person memories of product acquisition and development. citeturn142709search0turn142709search1

But searches for the interview plus オセロマルチビジョン, テレビゲーム, 007, etc. did not turn up any quotation or indexed passage about the videogame operation. What survives online overwhelmingly concerns Othello, Slime, Rubik's Cube, and other successful toy products. So there may be a relevant paragraph buried in those physical Toy Journal issues, but I couldn't substantiate one from the Japanese web.

There is one other oral-history source that helps explain the corporate mechanics, though not our date. A Japanese government-supported media-arts oral-history project interviewed Sega veteran Shuji Sato. When asked specifically about the Othello Multivision, he says:

「オセロマルチビジョン」はツクダオリジナルね。
“Othello Multivision was Tsukuda Original's.”

And then, concerning these compatible/OEM machines:

OEMはどっちかというと、私が絡んでやってるんですよ。

Approximately:

“As for the OEM side of things, I was more or less the one involved with that.”

That tells us there was a concrete Sega-side OEM/business relationship involving identifiable personnel, rather than Tsukuda simply independently cloning the machine. citeturn384963search32

That matters because it suggests another category of lost document that should once have existed: Sega/Tsukuda OEM correspondence, licensing paperwork, manufacturing schedules, ROM handoff records, approval sheets, etc. I could not find any of that digitized.

And the cartridge manuals themselves point to something surprisingly formal inside Tsukuda: they give the responsible unit as

株式会社ツクダオリジナルテレビゲーム事業部

“Tsukuda Original Co., Ltd., Television Game Division.” citeturn246683search13

So this was not apparently some completely casual two-person side project. There was a named Television Game Division, which makes it very likely that contemporary internal paper once included things such as production schedules, product approval sheets, price lists, dealer notices, and correspondence with Sega/Parker Brothers.

The problem is what happened afterward. Tsukuda Original eventually went through corporate restructuring, became part of Bandai, transferred operations, became Palbox, and ultimately disappeared into the MegaHouse lineage. A 2003 Bandai announcement documents the transfer/reorganization directly. citeturn553918search31 That sort of corporate history is exactly how mundane 1984 videogame-division files tend not to make it into a nice public archive.

So after this search, I'd describe the surviving evidence pyramid like this:

Best actual date evidence

November 10, 1984: Tsukuda's own dated promotion already presents 007 James Bond as ordinary ¥4,300 Multivision software. This remains our earliest concrete contemporary sighting.

Potentially better sources which exist but aren't digitized

Mitsuo Tsukuda's 1986 autobiographical/company-history book.
Takeshi Wakui's Toy Journal long interview and probably other toy-industry interviews.
Physical back issues of Japanese toy-trade publications.
Possible Tsukuda/Sega corporate records that have not surfaced publicly.

Things I specifically failed to find

an internal memo stamped “OM-G008”;
a master production/shipping schedule;
a Parker Brothers licensing document with a Japanese launch date;
an interview in which someone says “007 went on sale on November X”;
a digitized Tsukuda company archive comparable to the Atari material you're thinking of.

There was one thing from the search that reinforces your interpretation of the database dates. A Japanese collector/research page devoted specifically to Multivision cartridges says, rather candidly:

1983年から、1984年位まで発売されていたようですが、ほとんど資料を所有してい…

Essentially, “They seem to have been released from 1983 through around 1984, but I possess almost no documentary material…” citeturn246683search7

That is a very different epistemic culture from a modern database confidently printing 1984-12-01. The people closest to the Japanese collecting scene are often openly admitting that the records simply aren't there.

So I think your proposed genealogy is now quite plausible, though still not demonstrable:

“November 1984” may descend from somebody encountering Tsukuda's November promotional material or equivalent contemporary evidence and recording the month without preserving the citation.

“December 1984” may descend from a December publication/catalog appearance being treated as a release datum, then copied repeatedly until it acquired the appearance of independent consensus.

I wouldn't claim we've proven those transmission paths. But after trying to get behind both dates, November is the one for which we now have an actual primary-source foundation, whereas December still appears to be floating on retrospective assertion.

For your catalog, I'd now be comfortable with something like:

Release: by 10 November 1984 (Japan). Exact date unknown.
Contemporary Tsukuda promotional material dated 10 November 1984–January 1985 lists OM-G008 as regular-priced available software. Later sources variously report November or December 1984; no primary evidence for the December date has been located.

And if this particular obscure cartridge ever bothers you enough to pursue it beyond internet archaeology, the 1986 Mitsuo Tsukuda book is where I'd put the next dollar. It is tantalizingly close to the event, written by the founder, and currently sitting in Japanese used-book shops for about ¥1,000–1,400 rather than behind some impossible institutional wall. citeturn246683search2turn246683search4

That may literally be one of those cases where the answer—or a clue to it—is still sitting on paper and simply never entered the searchable internet.

G: Your instinct about the “numbered corporate sequence should have left paperwork” problem is exactly why one more angle is worth trying. I’m going to check the book as a bibliographic object rather than a web page—NDL records, used-book catalogs, possible scans, OCR/transcription fragments, and any amateur quotation/translation trail—then I’ll give you the clean identifier data.

G: [Non-text content: thoughts; content_type=thoughts]

G: The last pass produced an unexpected answer: the book actually has been digitized by Japan’s National Diet Library. The catch is access, not existence—the scan is restricted to NDL/eligible transmission, so it doesn’t show up as a normal freely readable PDF. That gives us a much cleaner target than “maybe somebody scanned it.”

G: [Non-text content: thoughts; content_type=thoughts]

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: Yes — and the last check turned up one useful correction to what I told you before.

The book has been digitized by Japan’s National Diet Library. The NDL record says it was digitized in 2022 and is available through the NDL Digital Collections, but not as an unrestricted public PDF. Access is limited to the NDL itself or eligible library/individual transmission services; the NDL page specifically notes that individual transmission is unavailable to people residing outside Japan. citeturn106996view0

So the situation is:

Physical book exists: 佃光雄『佃光雄 : その人と遺言』, 勁草出版サービスセンター, May 1986, 214 pp. citeturn106996view0turn728530search2
Digital scan exists: NDL persistent ID 12262666, DOI 10.11501/12262666. citeturn106996view0
It is not freely open-access from the U.S.
I found no amateur English translation, Japanese transcription, OCR dump, or pirated/free scan indexed on either the Japanese or English web.
I also found no indexed excerpts mentioning オセロマルチビジョン or television games.

The bibliographic record gives us something else interesting: pages 203–214 are a “brief chronology of Mitsuo Tsukuda” (佃光雄略年譜). citeturn106996view0 That may actually be more promising than reading 200 pages of memoir. If the chronology lists major corporate developments by year/month, 1983–84 could conceivably include the Multivision venture.

About the ISBN

I can't find one, and the authoritative NDL bibliographic record does not list an ISBN at all. That is significant because the record is otherwise quite detailed. So I think the book probably was published without an ISBN, rather than the ISBN simply being missing from a random used-book listing.

For identification, use these instead:

Title: 佃光雄 : その人と遺言
Author: 佃光雄
Publisher: 勁草出版サービスセンター
Publication: May 1986
Pages: 214
NDL call number: GK141-56
NDL Bibliographic ID: 000001855626
Japanese National Bibliography number / 全国書誌番号: 87033106
NDL persistent ID: info:ndljp/pid/12262666
DOI: 10.11501/12262666 citeturn106996view0

The 全国書誌番号 87033106 is probably the closest thing to the “equivalent identifier” you were asking for if you're cataloging the physical book itself.

And I do have one slightly out-of-left-field thought about your larger mystery.

The fact that OM-G001 through OM-G008 are so orderly may actually be pointing us toward the wrong documentary ecosystem. Those numbers may not have originated with the development team at all. They may have been sales/inventory SKUs assigned by Tsukuda's product-management or distribution system.

If so, the documents most likely to preserve a precise introduction date wouldn't necessarily be programmer records or videogame magazines. They'd be things like:

卸売価格表 — wholesale price lists
商品台帳 — product ledgers/registers
商品コード表 — product-code lists
受注票 / 発注書 — order forms/purchase orders
玩具業界向け新製品案内 — new-product notices for the toy trade
年末商戦資料 — Christmas/year-end sales materials

That would actually reconcile the two things bothering you: Japan could be obsessively systematic enough to give it OM-G008, while modern game historians nevertheless can't find the release date, because the systematic paperwork belonged to the toy wholesale chain, not the videogame-information ecosystem that otaku later preserved.

And Tsukuda Original was fundamentally a toy company, not a computer company. That distinction may be the key reason our expected “computer nerd paper trail” isn't where we've been looking.

At this stage, though, the book is now a concrete lead rather than a hypothetical one. Somebody with NDL access could jump straight to pp. 203–214, then search the scan around 昭和58年 / 昭和59年 (1983/1984), ツクダオリジナル, セガ, テレビゲーム, and オセロマルチビジョン. If there's anything useful in the book, that's probably the fastest way to expose it. memcite

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: That works. For this conversation, I’ll use this hierarchy:

“Release date” = earliest date contemporary documentation reliably shows the game existed, unless a more specifically defined launch/on-sale date can be established.
I’ll distinguish hard evidence from inference: e.g. “advertised as available by Nov. 10” versus “probably released in November.”
For originally Japanese games, I’ll start with Japanese primary or near-primary material—magazines, catalogs, trade publications, company records, advertisements, interviews, collector scans of contemporary ephemera—rather than treating English-language database repetition as evidence.
English/Western sources can still be useful as leads, but I’ll trace their claims backward when possible instead of counting repeated dates as corroboration.
If the evidence bottoms out at a month, season, or bounded interval, I’ll give you that actual resolution rather than manufacturing a day.

So for our 007 James Bond example, under this standard the working catalog datum would currently be “documented as existing by 10 November 1984”, not “December 1984.”

G: I’ll use the hierarchy we just established: first look for an explicit official/on-sale date; if that fails, look for contemporary Japanese primary material that narrows the earliest reliable existence date beyond “1983.” I’ll treat English databases only as leads, not evidence.

G: [Non-text content: thoughts; content_type=thoughts]

G: I’ve found a useful complication: a June 15, 1983 Game Machine issue already lists Borderline among the software associated with Sega’s forthcoming home system. That proves the game existed publicly before the SG-1000 launched—but not that it was on sale. So I’m now trying to find the first contemporary source that changes its status from “announced/listed” to “released/available.”

G: [Non-text content: thoughts; content_type=thoughts]

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: This one turned out to be much more interesting than I expected, because “1983” appears to be wrong as the Japanese SG-1000 release year. There is fairly strong Japanese documentary evidence for March 1984.

The best source I found is not a modern database. It is the May 1, 1986 issue of the Japanese amusement-industry trade newspaper Game Machine. It published a retrospective catalog titled essentially “SG-1000 / SC-3000 / Mark III — Sega Home Software Has Grown to 70 Titles”, and the list supplies release months for the software. In that list the early cartridges are not ordered by product number; rather, their actual release chronology is given. Borderline G-1001 is listed as March 1984. citeturn140145search9turn901909search0

That produces a rather striking early chronology:

July 1983: N-SUB G-1003, Mahjong G-1004, Tsumeshogi G-1006, Congo Bongo G-1007, Yamato G-1008, Star Jacker G-1010
August 1983: Safari Hunting G-1002, Champion Tennis G-1009, Champion Baseball G-1011
…
March 1984: Borderline G-1001, Champion Golf G-1005, Zippy Race, Exerion. citeturn901909search0

That same March 1984 chronology has subsequently survived in some Japanese retro-game catalogs, which probably inherited it either from this source or the underlying Sega records. ピコピコ大百科, for example, gives Borderline as 1984年03月, not 1983. citeturn711225search0turn711225search1

The particularly useful contemporary clue

There is also a Game Machine issue from June 15, 1983, a month before the SG-1000 hardware launched, which already mentions ボーダーライン among Sega's planned home-game cartridge lineup. citeturn710071search11

That reconciles a lot.

By June 1983:

Borderline existed as an announced/planned SG/SC product.

But the later Sega chronology says:

It did not actually release until March 1984.

So this looks like a classic case where product number and announcement chronology got mistaken for release chronology. It was G-1001 because it was among the very first cartridges planned, not because it was the first cartridge sold.

And there is a rather nice independent piece of evidence supporting why Borderline was so early in the development queue. In an oral-history interview with Masamitsu “Moo” Niitani / Compile personnel, the account explicitly says that the first work Sega gave Compile consisted of three SG-1000 conversions:

SG-1000版の「N-SUB」、「ボーダーライン」、「サファリハンティング」の3つの開発が、セガから頂いた最初の仕事。

Translation:

“Development of the three SG-1000 versions of N-SUB, Borderline, and Safari Hunting was the first work we received from Sega.”

They recreated them while looking at the arcade originals. citeturn367081search37

So G-1001 makes perfect sense as an internal/project number, even with a March 1984 commercial release.

Why the familiar “1983” date probably arose

Sega's own current archive is partly responsible for the confusion. It lists Borderline under 1983, but Sega explicitly warns at the top of that very page:

当時は明確な発売日が設定されていなかったため、各タイトルのパッケージに記載されたコピーライトを参考に分類してあります。

In English:

“Because clear release dates were not established at the time, titles have been classified with reference to the copyright information printed on their packages.”

So Sega's “1983” is explicitly not claiming “released in 1983.” It is a classification based on package copyright dating. citeturn259580search1

That is exactly the sort of metadata laundering we were discussing: somebody sees Sega's table headed 発売日, sees “1983,” and copies it as a release year without reading Sega's warning that that isn't actually what the number means.

There are then English-derived sites that go a step further and turn it into July 15, 1983, apparently by assuming it was a launch title simply because the console launched that day. I found no Japanese primary support for that. The Japanese trade chronology actually contradicts it.

Where I'd put your catalog entry

Under the standard we've established, I would currently use:

Borderline (SG-1000/SC-3000)
Japan: March 1984
Exact day not presently established.

And I would regard that as significantly better than our 007 result because this isn't merely an “existing by this date” boundary. We have a Japanese industry source from only two years later specifically assigning a release month.

I wouldn't yet manufacture March 1, March 15, etc. I searched Japanese combinations around March advertisements, 「新発売」, 「好評発売中」, G-1001 and various March/April 1984 dates and didn't find a period document yielding a day.

So its evidentiary ladder currently looks like:

June 15, 1983 — publicly documented as a planned/announced Sega cartridge. citeturn710071search11
↓
1983 copyright — explains later databases classifying it as a 1983 title. citeturn259580search1
↓
March 1984 — Japanese trade-industry retrospective gives this as its actual release month. citeturn140145search9turn901909search0

So I think we actually improved the catalog datum from “1983” to “March 1984,” rather than merely adding precision to 1983.

And incidentally, this is a useful warning for the rest of your SG-1000 cleanup: G-1001 through G-1010 emphatically do not represent release order. There may be several other apparent “1983” titles whose accepted dates are copyright years rather than sales dates. memcite

G: [Non-text content: thoughts; content_type=thoughts]

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: Yes. The article is the May 1, 1986 Game Machine piece headed 「セガ社家庭用ソフト70種」 (“70 Sega Home-Use Software Titles”). It says explicitly that the list is arranged in release order and gives the release month rather than exact day. citeturn412927view1

Here’s the release-month list as given there, converted from Shōwa years to Western years. I’m separating the 70 Sega titles from the 8 Tsukuda Original titles that the article prints alongside them.

Sega — 70 titles

July 1983

N-Sub
Mahjong
Serizawa Hachidan no Tsumeshogi
Congo Bongo
Yamato
Star Jacker

August 1983

Safari Hunting
Champion Tennis
Champion Baseball

October 1983

Flipper
Pop Flamer

November 1983

Paca Paca? / “Packer” [the scan/OCR is poor here]
Galaga

December 1983

Monaco GP
Pachinko
Space Slalom

January 1984

Sindbad Mystery

March 1984

Borderline
Champion Golf
Zippy Race
Exerion

April 1984

Golgo 13
Pachinko II

May 1984

Orguss

July 1984

Safari Race

September 1984

Lode Runner

October 1984

Boxing

November 1984

Champion Soccer
Hustle Chumy

December 1984

Home Mahjong
Flicky
Uranai Angel Cutie

January 1985

Girl’s Garden

February 1985

Zaxxon

March 1985

Champion Pro Wrestling
GP World

April 1985

Hyper Sports
Shinnyushain Tooru-kun

May 1985

Star Force

June 1985

Space Invaders

July 1985

Dragon Wang — My Card
Zoom 909 — My Card
Pitfall II
Choplifter
Doki Doki Penguin Land

The July 1985 Game Machine article independently gives those five July releases and their months, so that portion is especially easy to confirm directly. citeturn120999search24

Then the May 1986 article continues with the later releases:

August 1985

Bank Panic
Chack’n Pop
Teddy Boy Blues
Hang-On

September 1985

Drol
Great Baseball
Great Soccer
Elevator Action

October 1985

Rock n’ Bolt
Hang-On II

November 1985

Astro Flash
Championship Lode Runner
Sokoban
Champion Ice Hockey

December 1985

Pitfall II — My Card/reissue
Choplifter — My Card/reissue
Doki Doki Penguin Land — My Card/reissue
Hero

January 1986

Satellite 7
Pit Pot
Bomber Jack
Seishun Scandal
F-16 Fighting Falcon

February 1986

Ninja Princess
See-Saw

March 1986

F-16 Fighting Falcon / later-format entry
Machine Gun Joe

April 1986

Champion Kendo
Ghost House

A caveat on that second half: the original page is extremely densely typeset, with ditto marks for repeated months, and the OCR is ugly. The month sequence itself is readable from the scan, but a few title/media pairings around the 1985–86 transition deserve a manual line-by-line verification before I’d use this as a canonical dataset. The source page itself is here and explicitly labels the column 発売月 (“release month”). citeturn479663view0

Tsukuda Original — 8 titles printed in the same article

This part is clearer:

October 1983 — Q*bert
October 1983 — Guzzler
May 1984 — Space Mountain
May 1984 — Sannin Mahjong
June 1984 — Challenger Bee
July 1984 — Ayako Okamoto no Match Play Golf
December 1984 — 007 James Bond
December 1984 — Space Armor

Those are particularly interesting because they show exactly where the December 1984 007 date came from: Game Machine itself was publishing that month retrospectively in 1986. And its earlier July 1985 survey gives the same December 1984 assignment. citeturn120999search24

So I need to slightly revise what I said earlier: December 1984 isn't merely some later database inference. It is a period Japanese trade-publication datum, repeated by Game Machine in both 1985 and 1986. Our November 10 Tsukuda leaflet still creates a genuine contradiction, but now it's a much more interesting one: primary Tsukuda promotional evidence versus near-contemporary trade-industry release-month reporting.

If you want, I can next do the useful version of this: extract the entire 70-title table into a clean title → YYYY-MM dataset and manually verify the questionable OCR entries directly against the scan, so you've got something reliable to use for the rest of your SG-1000 catalog work.

G: I’ll treat this as a hardware-family history rather than a list of trivia: how the machines relate electrically and commercially, what My Card actually was, where the Mark III fits, and why American “Master System” labels sometimes get retroactively attached to Japanese SG/SC-era software.

G: [Non-text content: thoughts; content_type=thoughts]

G: [Non-text content: reasoning_recap; content_type=reasoning_recap]

G: The cleanest way to think about this is that Sega did not have four unrelated 8-bit systems. It had one evolving hardware family that began as a computer/console pair, then grew into the Mark III/Master System line.

The family tree is basically:

SC-3000 / SG-1000 (1983)
→ SG-1000 II (1984)
→ Sega Mark III (1985)
→ Master System outside Japan (1986)
→ Japanese Master System (1987)

The confusing part is that Sega kept a surprising amount of backward compatibility while changing the branding and the software media.

1. SC-3000 and SG-1000 were siblings, not successors

The SC-3000 was conceived as Sega's home computer: built-in keyboard, BASIC/programming capability, printer/data-recorder support, and games as one use among several. Sega released it on July 15, 1983. citeturn161289search9turn161289search11

The SG-1000 appeared the same day. Internally it used essentially the same basic architecture, but Sega stripped away the built-in keyboard and sold it as a cheaper dedicated game console. Sega itself describes it as effectively the SC-3000 specialized for games. citeturn161289search7turn161289search11

So:

Machine	What it fundamentally was
SC-3000	computer version
SG-1000	game-console version

Their game cartridge ecosystem was shared. Sega's own documentation calls the cartridges “For SC/SG”, meaning both families. citeturn161289search0turn161289search2

That means if your catalog has early titles marked SC-3000/SG-1000, that's correct. They generally aren't separate versions of the game; they're the same software running on closely related machines.

The SG-1000 could even be turned partly back into an SC-3000 by adding the SK-1100 keyboard, which enabled SC software and peripherals. citeturn161289search0

2. SG-1000 II was mostly a redesign

The SG-1000 II was not the next major technical generation.

Think of it as:

SG-1000, redesigned and cleaned up

rather than:

SG-1000 → new incompatible successor.

The real generational break comes with the Mark III.

This is why early Sega histories sometimes make the numbering feel strange. Sega eventually called Mark III its “third-generation” home console, even though there had technically already been SG-1000 and SG-1000 II models. Sega treats SC-3000/SG-1000/II collectively as the preceding architecture. citeturn821259search3

3. “My Card” was a storage medium, not another console

This is probably the piece causing the most catalog confusion.

The Sega My Card was essentially a tiny ROM cartridge flattened into a credit-card-like form.

Instead of this:

[ chunky cartridge ]

Sega could manufacture:

[ thin card containing ROM ]

The normal My Card held 32 KB. citeturn161289search0

The original SG-1000 and SC-3000 did not have a card slot, so Sega sold the:

Card Catcher C-1000

which plugged into the ordinary cartridge slot and provided a My Card socket. citeturn161289search12

Thus:

SG-1000 + Card Catcher + My Card

was simply another way of loading SG-1000 software.

It wasn't emulation or a separate platform.

Example from your 70-title list

Games such as:

Dragon Wang
Zoom 909

appearing as My Card releases are still part of the SG-1000/SC-3000 software ecosystem.

That's probably why some catalog sources appear to shift classifications abruptly around 1985. The physical medium changed while the underlying target hardware initially did not.

4. Then Sega complicated matters by creating a second generation of My Card

In October 1985 Sega introduced the Sega Mark III.

This is the important hardware transition.

The Mark III kept enough of the old architecture that it could run SG-1000/SC-3000 game cartridges, but its graphics hardware was substantially enhanced. Sega explicitly designed it as an upward-compatible successor. citeturn821259search3turn821259search5

It also had a card slot built directly into the machine.

But now Sega produced:

My Card Mark III

These looked like the old My Cards and were still 32 KB, but they were games written for the new Mark III hardware.

Sega explicitly warns:

My Card Mark III cannot be used on the SC/SG series.

citeturn821259search4

So there are two things that look almost annoyingly similar:

Format	Intended hardware
Sega My Card	SG-1000 / SC-3000 generation
My Card Mark III	Mark III / Master System generation

And the Mark III can play both, because it is backward-compatible.

An SG-1000 can't.

That distinction is extremely useful for cataloging.

5. The Mark III is basically the Japanese ancestor of the Master System

This is where the American terminology enters.

The Sega Mark III launched in Japan in October 1985. It was Sega's substantially upgraded 8-bit machine: better graphics, more colors/sprites, improved scrolling, while retaining backward compatibility with the older SG/SC cartridge software. citeturn821259search3

Then Sega prepared it for overseas sale.

The machine released in North America and other markets as the:

Sega Master System

was essentially the internationalized Mark III platform, with a redesigned case, different regional cartridge connector/BIOS arrangements, and some other implementation differences.

So in broad software-platform terms:

Mark III = Japanese version of the hardware generation that the West calls Master System.

That is why game databases often put Mark III games under “Master System.”

That convention is reasonable, provided you're cataloging platforms by hardware/software family rather than exact Japanese retail branding.

6. But Japan itself eventually got something literally called “Master System”

This creates another layer of naming stupidity.

Japan did not immediately rename the Mark III when the international Master System appeared.

Instead:

1985: Japan gets Sega Mark III
1986: overseas markets get Master System
1987: Japan gets a redesigned/upgraded machine also called Master System

The Japanese Master System launched October 18, 1987. Sega describes it basically as a Mark III with several accessories/features integrated:

FM sound unit built in
rapid-fire function built in
direct 3-D Glasses connection

citeturn821259search8turn821259search3

So the Japanese history is:

Mark III → Japanese Master System

whereas the American history starts:

Master System

because America never commercially received the SG-1000 or Mark III under those names.

7. The cartridge lineage remained backward-compatible in Japan

This is one of the most elegant—and confusing—things Sega did.

Sega's official documentation says the old SC/SG cartridges could also be used on:

Sega Mark III
Japanese Master System
Othello Multivision
Pioneer SD-G5

and consequently were sometimes called:

全機種用 — “for all models.”

citeturn161289search0

So physically, in Japan, you could have this situation:

1983 SG-1000 cartridge
↓
works in SG-1000
↓
works in Mark III
↓
works in Japanese Master System

That is a real hardware lineage.

The Mark III retained the old cartridge interface deliberately. citeturn821259search3

8. The American Master System is where backward compatibility gets messy

The Western Master System isn't physically interchangeable with Japanese cartridges in the simple Famicom-style sense.

The Japanese SG/SC/Mark III/Master System cartridge connector lineage is related, but the Western Master System uses a different physical/region implementation. Technical documentation shows that the Japanese machines preserve the older SG-series cartridge interface much more directly. citeturn667072search0turn667072search1

Western Master Systems also use BIOS/header checks that prevent normal Japanese SG-1000 software from simply working even where an electrical adapter might otherwise make it possible. citeturn667072search2

So:

Mark III / Japanese Master System are directly backward-compatible with SG-1000.

but:

American/European Master System should not be treated as an ordinary SG-1000-compatible machine.

Architecturally they're descendants; commercially/physically they're a regionalized branch.

That distinction probably explains some contradictory database labels you've encountered.

9. “Sega Card” is the Western name for the later card concept

When Sega took the Mark III architecture overseas, My Card Mark III essentially became:

Sega Card

on the Master System.

The Western Master System had both:

cartridge slot
Sega Card slot

So an American owner could buy something like a Sega Card rather than a cartridge.

Technically the card format is closely related across generations, but there are software/region compatibility wrinkles. The important catalog distinction is simpler:

Japanese pre-Mark-III My Card
= SG-1000 generation

Japanese My Card Mark III
= Mark III generation

Western Sega Card
= Master System generation

citeturn667072search5turn821259search2

The card format then died because 32 KB became restrictive just as console games were getting much bigger.

10. Sega's “Gold Cartridge” marks another useful dividing line

In 1986 Sega began releasing its larger Mark III games as Gold Cartridges.

Sega says the first big transition came with 1-megabit Gold Cartridges in June 1986, allowing substantially larger games such as Fantasy Zone. citeturn821259search3turn821259search7

This effectively killed My Card Mark III.

So media evolution looks like:

SG/SC era

ordinary cartridge
↓
My Card added as cheaper 32 KB alternative

Mark III era

My Card Mark III initially emphasized
↓
games become bigger
↓
Gold Cartridge becomes dominant

Master System era

large cartridges dominate

This is why the 1985 part of that Game Machine list you asked about looks like Sega is bouncing strangely between cards and cartridges. You're seeing a media transition in progress.

11. The thing that may be confusing your catalog most

Suppose your catalog says something like:

Hang-On — Master System — Japan — 1985

You might reasonably think:

“Wait. The Japanese Master System didn't exist until 1987.”

Correct.

What the catalog usually means is:

software belonging to the Mark III/Master System platform family

because Hang-On was released as My Card Mark III on October 20, 1985. citeturn821259search0

Calling that a “Japanese Master System game” is technically understandable but historically imprecise.

The retail platform printed on the Japanese product at that time was:

Sega Mark III

not Master System.

Similarly:

Fantasy Zone — Master System, Japan, 1986

is again essentially Mark III software, even though databases lump Mark III and Master System together because the software runs on both. Sega itself does this to a degree: its current Master System software page says it lists Gold Cartridges usable on both Mark III and Master System. citeturn821259search7

So the database category Master System often really means:

Mark III/Master System software architecture

rather than:

“This exact machine bearing MASTER SYSTEM branding existed in Japan when the game appeared.”

That's almost certainly one of the anomalies you've noticed.

12. But SG-1000 games that later have American Master System versions are another category entirely

You also mentioned that some items in the SG list are connected in your catalog to American Master System versions.

Here you need to distinguish:

Same old software running through backward compatibility

Example:

Japanese SG-1000 game → Japanese Mark III runs it

That's one binary/product lineage.

versus:

Later Master System remake/port/reimplementation

Example conceptually:

SG-1000 version of Game X
and later
Master System version of Game X

Those can be different releases built for different hardware capabilities, despite sharing the same arcade source and name.

Given Sega's habit of repeatedly converting its arcade catalog, this happens quite a bit.

A 1983–85 SG-1000 Choplifter, Pitfall II, Hang-On-related release, etc. should therefore not automatically be conflated with a later American Master System release merely because the title matches.

You need to ask:

Is this the SG-1000 program being treated as backward-compatible software, or is there a distinct Mark III/Master System version?

For catalog purposes, that is an important distinction.

The family in one diagram

                         1983
                ┌──────────────────┐
                │     SC-3000      │
                │ home computer    │
                └────────┬─────────┘
                         │ shared architecture/software
                ┌────────┴─────────┐
                │     SG-1000      │
                │ game console     │
                └────────┬─────────┘
                         │
                         │ 1984 redesign
                         ▼
                ┌──────────────────┐
                │    SG-1000 II    │
                └────────┬─────────┘
                         │
        Original My Card │ + Card Catcher
        belongs here ────┘

                         │ backward-compatible evolution
                         ▼
                         1985
                ┌──────────────────┐
                │ Sega Mark III    │
                │ upgraded video   │
                └───────┬──────────┘
                        │
             My Card Mark III
             Gold Cartridge
                        │
           ┌────────────┴────────────┐
           │                         │
           ▼                         ▼
       1986 overseas              1987 Japan
   ┌────────────────┐       ┌──────────────────┐
   │ Master System  │       │ Master System    │
   │ US/EU/etc.     │       │ upgraded Mark III│
   └────────────────┘       └──────────────────┘

For your catalog, I'd use three platform buckets

If your purpose is historical precision rather than merely emulator compatibility, I think the cleanest classification would be:

SG-1000 / SC-3000 family
For the original cartridge and original My Card games.

Sega Mark III
For Japanese games released from October 1985 onward specifically for the upgraded hardware—including My Card Mark III and Gold Cartridge games—even though many databases call these “Master System.”

Master System
For the overseas versions beginning in 1986, plus Japanese releases from the actual Japanese Master System period if you want branding rather than architecture to drive the label.

Then add compatibility relationships separately rather than allowing them to redefine the platform.

That prevents precisely the weird result:

“Japanese Master System game — released two years before the Japanese Master System.”

The shortest possible version

SC-3000 = computer.
SG-1000 = cheaper console based on it.
My Card = alternate SG/SC software medium.
Mark III = much stronger backward-compatible successor.
My Card Mark III = games requiring that successor.
Master System = internationalized Mark III family; Japan later adopted the same name for an enhanced Mark III model.

So, yes: the Japanese-only things you're seeing labeled “Master System” are probably not SC-3000 software. They're much more likely Mark III software being categorized under the broader Master System family name. citeturn821259search3turn821259search7

A useful Japanese shorthand for the distinction is 上位互換(じょういごかん), “upward/backward compatibility”: the later machine can use the older software, but that does not make the older software native to the later platform.