Who Stole From Whom: 1,400+ Plagiarism Hits Across 9 Cabinet Manufacturers

After cracking 120,000+ trivia questions out of 1980s arcade ROMs, we ran a cross-cabinet similarity scan. The results read like a courtroom. Exidy and GEI share 227 verbatim questions. Bally Sente, who licensed Trivial Pursuit from Horn Abbot, shares 201 questions with GEI, questions GEI did not license. Status Games turns up everywhere. The bar-trivia industry was running on stolen content.

April 24, 2026 · analysis, plagiarism, 1980s-arcade, trivia

← Archive

Who Stole From Whom

A cross-cabinet similarity scan across 123,241 questions · 8 manufacturers · 1,414 verbatim cross-mfr matches

The Setup

By April 2026 we’d cracked the question banks of every major English-language bar-top trivia cabinet from the 1980s, Greyhound Electronics, Status Games, Bally Sente’s Trivial Pursuit machines, Exidy Fax, Merit Industries’ Trivia Whiz family, Kramer’s Trivia Madness, Enerdyne’s Super Trivia Master, Triumph’s MTV Rock-N-Roll, Techstar/Sunn’s Trivia Quest. 123,241 records total, all browseable side-by-side.

Sitting on that pile of clean, normalized question data, an obvious question: were the cabinet operators stealing from each other?

Answer, with the receipts in hand: yes. A lot.

The Method

Every question gets normalized (lowercase, strip punctuation, collapse whitespace), then sliced into 6-word shingles, overlapping six-word sequences hashed to 64-bit prefixes. Two questions that share a high fraction of their shingles are saying the same thing in near-identical phrasing.

To suppress noise, template shingles that appear in more than 5% of any single source’s records get filtered out. These are question stems like “what was the name of”, common phrasings that don’t indicate copied content. We then compute Jaccard similarity on the remaining content shingles for every cross-manufacturer record pair. Anything ≥ 0.5 counts as a strong match.

Cross-manufacturer is the critical filter. GEI re-released their own catalog dozens of times across series 1, 2, …, 18, a question appearing in gtsers14 and gtsers14a is just GEI being GEI. The interesting story is when Manufacturer X’s question shows up in Manufacturer Y’s ROM, byte-for-byte.

The full analysis is at scripts/analysis/cross_cabinet.py ; the report data is committed at scripts/analysis/cross_cabinet_similarity.json .

The Results

1,414 cross-manufacturer strong matches out of 123,241 records. That’s about 1% of the entire archive, but the distribution is heavily clustered:

matchesmanufacturer pair
268GEI ↔ Status Games
227Exidy ↔ GEI
201Bally Sente ↔ GEI
180GEI ↔ Merit
107Merit ↔ Status Games
64Exidy ↔ Status Games
56Bally Sente ↔ Status Games
43Bally Sente ↔ Exidy
35GEI ↔ Kramer
32Exidy ↔ Merit

GEI is in the top four pairs for one simple reason: GEI shipped the most ROMs, more than any other manufacturer in the database, so the absolute count of theft incidents involving them is highest. That doesn’t necessarily mean GEI was the perpetrator, they could be the victim. The matches alone don’t say which way the copying flowed.

But the chronology often tells you. And so do specific cases.

The Source Behind the Source: Fred Worth and the Columbo Trap

Before getting to the smoking guns in our archive, the relevant prior art. In October 1984, Fred L. Worth sued Selchow & Righter and Horn Abbot Ltd. for $300 million1, alleging that Trivial Pursuit had lifted questions wholesale from his 1974 Trivia Encyclopedia2. The proof: Worth had planted a deliberately false fact in his book, claiming Lt. Columbo’s first name was “Philip.” Horn Abbot’s Trivial Pursuit shipped a card with the same fictional answer.

Worth lost. The 9th Circuit affirmed the district ruling in 19873. Facts aren’t copyrightable. Even a question copied verbatim from one book to another isn’t infringement so long as it’s a fact (or even a fact-with-a-deliberately-planted-typo, in the Columbo case). The Supreme Court denied cert in March 19884.

That ruling set the legal floor for everyone in this archive, retroactively. The cabinet manufacturers shipping in 1984-1986 were already copying freely while the case was still in district court; the 1987 Ninth Circuit affirmation just confirmed nobody owed anyone anything for it. The cabinet operators we’re about to look at mostly didn’t even bother changing the wording.

The Smoking Gun: Bally Sente vs. GEI

Bally Sente licensed Trivial Pursuit from Horn Abbot Ltd. in 1984 to build the arcade adaptation. Their question banks aren’t homemade. they’re literally the questions from the Trivial Pursuit board game, paid for and contracted. Six cabinets shipped under the deal: Genus Edition5, Genus II6, All-Star Sports7, Baby Boomer8, Young Players9, and the Spanish-language Ataque Trivial (via Maibesa)10. Every record traces back to a Horn Abbot question card.

GEI did not license Trivial Pursuit.

But here are four random samples from the 201 verbatim matches between Bally Sente and GEI:

bally_triviayp: “What shape are playing cards in India?” → Round gei_gtsers9: “What shape are playing cards in india?” → Round

bally_triviayp: “How many days are there in a Leap Year?” → 366 gei_gtsers8: “How many days are there in a leap year?” → 366

bally_triviayp: “What does a Geiger counter measure?” → Radioactivity gei_gtsers12a: “What does a geiger counter measure?” → Radioactivity

bally_triviayp: “How many days are there in a Leap Year?” → 366 gei_gt103aa: “How many days are there in a leap year?” → 366

The case isn’t open-and-shut on any single question, “How many days in a leap year?” is something two writers might independently compose. But 201 of them, all sharing exact phrasing, all matching content GEI did not pay to use, and many of them dated after Bally’s 1984-1986 Trivial Pursuit run? That’s a pattern.

triviayp (Trivial Pursuit Young Players Edition) is a particularly good source for this, it’s a children’s question set with very specific phrasing choices. GEI shipping the same wording suggests the questions arrived not from independent research but from one bartender’s well-worn Trivial Pursuit cartridge.

Exidy ↔ GEI: The Top Pair

The biggest cross-mfr overlap is Exidy ↔ GEI at 227 matches. Exidy Fax shipped in 1983, the earliest cabinet in the entire archive. GEI’s Greyhound Electronics series 1 came in 1984. The arrow of copying almost certainly points GEI → from → Exidy, not the other way.

Sample matches:

exidy_fax2: “Whose nickname is The Iron Lady?” → Margaret Thatcher gei_gtsers10: “Whose nickname is ’the iron lady’?” → Margaret Thatcher

exidy_fax2: “What is Wolfman Jack’s real name?” → Bob Smith gei_gtsers1: “What is wolfman jack’s real name?” → Bob Smith

exidy_fax2: “What is Barbie doll’s boyfriend’s name?” → Ken gei_gtsers10: “What is barbie doll’s boyfriend’s name?” → Ken

The capitalization differences (“The Iron Lady” → “the iron lady”) are artifacts of the GEI extractor, not the original ROM, both cabinets displayed mixed-case in-game. The phrasing is identical.

Where the Catalog Started

When you watch the matches scroll by, a small set of questions keeps recurring, the same record appearing across three or four manufacturers. These look like they originated in one source document that multiple cabinet outfits got hold of and copied wholesale.

The most common multi-cabinet hits:

The Status Games sex-trivia set has its own constellation of overlaps with GEI’s adult-oriented sextriv1, sextriv2, sexappl ROMs. Some of these questions are vocabulary items that you really would not see two writers compose from scratch, and yet they sit identically in two different mfrs’ cabinets.

The Industry Picture

The 1980s bar-top trivia market wasn’t a market for questions. It was a market for boxes, cabinets, video controllers, ROMs, coin acceptors. The questions were the cheapest part of the build, and operators treated them like community property.

Some manufacturers (Bally Sente) paid for licensed content. Some (Exidy, the early ones) likely wrote their own. Most of the rest appear to have been pulling questions from board games, almanacs, and, most often, each other’s ROMs, dumped overnight in a warehouse and re-shipped under a new label.

The reason this archive can exist is the same reason the theft was possible: 1980s arcade ROMs were never built to keep secrets. The encryption schemes we cracked across this project, base-40 radix, OR-0x40, dual-LUT scramble, 5-bit packing, LZ phrase tables, were designed to save ROM space, not to keep questions proprietary. A motivated operator with a ROM dumper and an afternoon could do exactly what we just did.

The Limits

A few caveats worth flagging:

  1. Common-knowledge facts are not theft. “How many days in a leap year?” is the kind of question any writer could compose. The plagiarism signal is the phrasing pattern repeating, not any single question.

  2. Word-for-word matches across different ROMs don’t always mean copying. Some question writers worked freelance; they might have sold the same question twice. Some questions came from public reference works.

  3. Our extraction has bugs. Status Games question prefixes are sometimes mis-attached (we documented this in the Status Games crack post ). That noise ends up in the similarity scan too.

  4. Jaccard ≥ 0.5 is a high bar but not perfect. Two questions covering the same topic with different phrasings will miss the threshold; two with shared question stems can hit it coincidentally.

The 1,414 strong matches reported here are the floor of the actual plagiarism, not the ceiling.

What’s Next

The full report is committed to the repo , every manufacturer-pair count, every source-pair count, plus eight sample matches per mfr pair so you can read the receipts yourself. Run python3 -m scripts.analysis.cross_cabinet to regenerate it.

The next thing to add: a chronology overlay. If a question appears first in 1983 (Exidy Fax) and later in 1986 (a GEI series), the direction is obvious. Right now we have years for every cabinet but haven’t wired them into the similarity output. That’s a next-pass upgrade.

Also worth doing: inverse cluster, finding questions unique to a single manufacturer, which gives us a fingerprint of what they actually wrote themselves vs. what they imported.

But the headline finding stands: in a market built on cheap content, the questions belonged to whoever had the screwdriver.

Other analyses

Browse the cracks

References

  1. “The author of trivia books filed a $300 million…” UPI Archives, October 24, 1984  ↩︎

  2. Wikipedia, The Trivia Encyclopedia  ↩︎

  3. Worth v. Selchow & Righter, 827 F.2d 569 (9th Cir. 1987), Justia  ↩︎

  4. “Judicial Pursuit”, Washington Post, March 29, 1988  ↩︎

  5. Trivial Pursuit (Genus), Arcade-Museum / KLOV · MAME romset triviag1 ↩︎

  6. Trivial Pursuit (Genus II), Arcade-Museum · MAME romset triviag2 ↩︎

  7. Trivial Pursuit (All Star Sports), Arcade-Museum · MAME romset triviasp ↩︎

  8. Trivial Pursuit (Baby Boomer), Arcade-Museum · MAME romset triviabb ↩︎

  9. Trivial Pursuit (Young Players), Arcade-Museum · MAME romset triviayp ↩︎

  10. Ataque Trivial / TP en Español, Arcade-Museum · MAME romset triviaes2 ↩︎