Cracking MTV Rock-N-Roll Trivia Part 2 (1986)

A Pac-Man conversion kit that turned bar-top cabinets into MTV-licensed rock trivia. 3,283 questions across Heavy Metal, General Rock, Classic Rock, and Rock Videos, compressed with a 103-entry phrase-token dictionary baked into the program ROM.

April 24, 2026 · crack, triumph, rocktrv2, z80, pacman, mtv

← Archive

MTV Rock-N-Roll Trivia Part 2

Triumph Software · 1986 · Pac-Man conversion kit · 3,283 questions across 4 category chips

The Cabinet

By 1986 the bar-top trivia market was saturated and the original Pac-Man cabinets from 1980 were aging out of their arcade runs. Triumph Software spotted the gap: take a used Pac-Man cabinet, swap the ROMs, slap an MTV logo on the bezel, and resell it as an MTV-licensed rock trivia machine. The hardware stayed the same, Z80 @ 3 MHz, Namco WSG sound, the Pac-Man video generator, but the game was now a four-category rock music quiz.

MAME files it as rocktrv2 in pacman/pacman.cpp, sharing the driver with the original Pac-Man plus a pile of other conversion kits. The name, Part 2, implies there was a Part 1, but every catalog (Arcade-Museum , MAME source, archive.org, EmuParadise) only ever surfaces Part 2. There is no rocktrv parent set in MAME, no Part 1 ROM dump, no flyer, no second-pressing record anywhere. Triumph either planned a Part 1 and never shipped one, or used “Part 2” as question-set branding from the cabinet’s first release. The cabinet you would have actually played at a 1986 bar said “Part 2” on the marquee.

Four category chips:

The Data

All 3,283 records browse live at /browse?source=triumph_rocktrv2 , categorized by chip, searchable, with the full answer set and three decoys for every question. 30 Heavy Metal records were hand-verified end-to-end against real-world facts (Bad Company vocalist → Paul Rodgers, Queen’s country → U.K., Ted Nugent leaving the Amboy Dukes → FALSE); the rest inherit that trust.

How We Cracked It

The Bank Window

The Pac-Man conversion kit exposes a standard Z80 bank-select at $5FF0: the CPU writes a bank number there, and a 32 KB window at $8000-$FFFF maps the selected question ROM. No XOR, no bitswap, no address scramble. just a banked read. The 8 physical banks are arranged as 4 (Q, A) pairs, one pair per category.

category   Q-bank  A-bank
HM         0       4
GR         1       5
CR         2       6
RV         3       7

Each Q-bank is a stream of questions, separated by a literal > byte. Each A-bank is a matching stream of 4-slot answer groups, one per question, also >-separated.

The Phrase-Token Dictionary

Every question ROM is plain ASCII, until you hit a byte in the range 0x5B..0xC1. Those bytes are tokens that expand via a dictionary baked into the auxiliary program ROM (1.aux) between offsets $38FF..$3B09. The dictionary itself is >-delimited, one phrase per entry:

0x5B  " "
0x5C  ","
...
0x77  "THE"
0x7B  "WHO"
0x89  "LIKE"
...
0xB4  "HEAVY METAL"
0xB5  "PAUL McCARTNEY"
...

103 entries total. The most common tokens are short function words (THE, IS, OF, A); the tail carries long proper nouns (band names, album titles, MTV VJs) that would otherwise eat hundreds of bytes per ROM.

A few special glyphs that aren’t really tokens:

The Correct-Answer Nibble Table

Each question has 4 answer slots in the A-bank, but only one is correct. The correct-slot index lives in a packed 4-bit table, a nibble per question, tucked into whichever chip had free tail space. For HM, GR, and RV, the table sits at the tail of the A-high bank; for CR it’s at the tail of the Q-high bank.

Nibble values:

1..4   → slot index (which of the 4 answers is correct)
5      → TRUE   (the record is a true/false question, answer = TRUE)
6      → FALSE  (the record is a true/false question, answer = FALSE)
0      → mid-table marker, rare, appears in CR only

True/false records are encoded in the A-bank as a 4-slot group of [TRUE, "", FALSE, ""]; the nibble tells us which side to score.

Yield

HM  892    GR  907    CR  823    RV  661

total: 3,283 records (2 CR records skipped, Q/A stream counts
                      disagreed by one, almost certainly a cut
                      question during development)

What It Tells Us

Triumph’s trick is the business model, not the tech. The crack itself is mid-difficulty, a straight phrase-token dictionary, no encryption. but the conversion kit economy is fascinating. By 1986 Pac-Man cabinets were cheap and plentiful; anyone with a Z80 assembler and question data could ship a new game on used hardware. MTV licensing made this one commercial, but the same pipeline let dozens of smaller outfits mint new arcades out of old ones.

The dictionary-based compression is also a nice reminder that question cabinets were always ROM-constrained. 103 phrase tokens saves maybe 15-20% of the text mass, enough to fit one more category per chip, which at retail meant one more quarter per play cycle.

References

Cross-archive analyses

  1. MTV Rock-N-Roll Trivia (Part 2)1 (1986, Triumph Software, Inc.) · Arcade-Museum · MAME romset: rocktrv2 ↩︎