Cracking Status Games Hangman (1984)
The only Status Games title with no phrase table in the ROM, tokens inferred from their own question contexts. 894 questions recovered.
April 19, 2026 · crack, status-games, statriv2, hangman, 8085
Hangman
Status Game Corporation · 1984 · Intel 8085 @ 12.4 MHz · same statriv2 board · 894 records
The Game
Hangman1 (Status Game Corporation, 1984) took the bar-trivia formula and bolted it onto the old paper-and-pencil guessing game. Instead of four answer buttons to a multiple-choice prompt, the screen showed a category, a row of underscores, and a doomed stick figure. Players guessed letters; wrong ones added a limb. Get the phrase before the figure swings, keep your quarter.
It shipped into the same NYC-area bars as Status Games’ other 1984 cabinets, Triv Quiz and Status Trivia 2, on the same Intel 8085 board at 12.4 MHz. Status Games reused everything: hardware, monitor, coin door, answer buttons, boot sequence. The only things that changed per title were the program EPROM and the question ROMs.
Which is why Hangman looked like a trivially easy target. Same hardware,
same MAME driver (statriv2), same OR 0x40 encoding,
same phrase-token compression
as the
other seven Status Games titles. Open the aux ROMs, decode with the same
pipeline, done.
Except it wasn’t.
The Data
All 894 records browse at /browse?source=statriv2_hangman . full token coverage, two stubborn source-corrupt records left over (real ROM damage, not extraction).
- Extractor:
scripts/statriv2/extract.py(see thehangmanentry in_GAME_CONFIG) - Shared decoder:
scripts/statriv2/format.py
How We Cracked It
The Phrase Table That Wasn’t There
Every other Status Games title stores its phrase table as plain ASCII in
the program ROM, a run of 0xFF-separated English strings like
WHAT IS THE , WHO PLAYED , IN WHAT MOVIE . Index byte 0x80 picks
the first, 0x81 picks the second, and so on. The extractor walks the
program ROM, finds the longest run of 0xFF-delimited ASCII, and
auto-builds the token map. Works for every Status Games title.
Hangman’s program ROM has nothing. No run of ASCII phrases. Not encoded,
not ciphered, just absent. The main_hang_*.u36/u7/u8/u9 files
contain the 8085 game loop, the wheel-value tables for the letter guess
scoring, and the stick-figure bitmap, but no phrase dictionary.
Meanwhile the question ROMs (aux_2764.1 through aux_27c256.8) were
full of token bytes in the 0x80-0xBF range, waiting to be expanded.
~1,000 records with roughly a fifth of them referencing phrases that
lived nowhere on the board.
Inferring Tokens From Themselves
If the phrase table isn’t in the ROM, then in the running cabinet it must live somewhere volatile, or be computed. Disassembling the 8085 to chase that down would take days. But there was a cheaper path: the tokens expand into questions, so each unresolved byte has dozens of contextual usages. Group those up, read them, guess the phrase.
For each unmapped token, pull every record that contains it and print
the surrounding text with the token as [XX]:
=== 0xA5 (11 usages) ===
'how ' [A5] ' dick long direct'
'why ' [A5] 'god create women'
'why ' [A5] 'ginger rogers end her affair'
'how much ' [A5] 'mata hari claim to charge'
'what kind of work ' [A5] ' marilyn monroe do after being'
'during how filming ' [A5] 'elizabeth taylor and richard burton'
Every usage is a “how/why/when” question with a past-tense verb one
slot away. The answer is always a name or a reason. 0xA5 = DID .
The real prize was 0xAC (27 usages):
=== 0xAC ===
'greatest f' [AC] 'affecting sexual frequency'
'what famous ' [AC] 'and director appeared in a stag'
'what famous ' [AC] ' carried two green balls'
“Greatest f” + unknown + “affecting sexual frequency” → “greatest
fACTOR affecting sexual frequency”. Token 0xAC is the stored
phrase ACTOR . Verify with “what famous ACTOR and director appeared
in a stag movie” → Sylvester Stallone. Yes.
And then 0xBD and 0xBE, each only two or three usages:
=== 0xBD ===
'on ' [BD] ' what was arnold' → pig
'where was ' [BD] ' located' → Hooterville
=== 0xBE ===
'who was the doctor on ' [BE] → Doc Adams
'where did ' [BE] ' take place' → Dodge City
Arnold the pig. Hooterville. 0xBD = 'GREEN ACRES' . Doc Adams, Dodge
City. 0xBE = 'GUNSMOKE' . Both phrases match the corresponding token
slots in Status Trivia 2’s table, the same TV-show constant baked into
a different game’s code.
The Sports Bank
One more surprise. A handful of records used a second range of tokens in the 0xC0-0xD4 band, bytes not in any Status Games title’s normal 0x80-0xBF phrase map. They all clustered around sports questions:
[C4]gold medals [D1] carl lewis win male 1984 olympics → 4
[C4]substitutions [D3]allowed [CC] rugby [CD] → None
[C3]win a volleyball [CD] → 15
what was horse won [D4] 1985 kentucky derby? → Spend-a-buck
who invented [D4] t formation? → George Halas
A second, smaller phrase table stuck above the regular one, for the
sports category only. Same inference trick: 0xC4 = HOW MANY ,
0xD1 = DID , 0xD3 = ARE , 0xCC = IN , 0xCD = GAME ,
0xD4 = THE . A handful of records, but without the extended tokens
they were unreadable.
The Holdouts
Four tokens in the 0xC0-0xCA range (0xC0, 0xC1, 0xC2, 0xCA)
showed up only in two records and nowhere else, not enough context
to infer from hangman alone. The fix came from a sibling set: Super
Triv II’s auto-extracted phrase table (same Status Games board,
same 8085) has those exact bytes as FOOTBALL, BASEBALL, CAREER,
RECORD, a sports-heavy phrase group reused across the line.
Borrowing the four entries resolved the Obed Ariri record cleanly:
he hit 4 ... football career record does obed ariri hold
→ Field goals (Touchdowns / Interceptions)
That’s “What football career record does Obed Ariri hold?”, and
Ariri did hold the NCAA career field-goal record at Clemson. The
question text itself still carries typos (hoege, fiege) baked
into Status Games’s ROM burn, but the decoded phrase is now
meaningful and the record survives the [XX] filter. One token
(0xC1 = BASEBALL) still only appears in a secondary malformed
record that gets dropped on the “fewer than 2 answers” rule, but
all four tokens are now mapped.
The Duplication Trick
Final wrinkle: the same record would extract eight times. "what was horse won THE 1985 kentucky derby" appeared identically across
aux_2764.4, aux_2764.5, aux_2764.7 and three more chips. Not a parsing
bug, the cabinet’s chip-select hardware mapped several EPROMs to
overlapping address windows so the 8085 would land on a question no
matter which bank was selected. MAME loads them all; the concatenator
sees 8× copies.
Fix is trivial, dedupe on (question, answer) at write-time. Fix is
only trivial if you know that’s what you’re looking at, which took a
minute of hex diffing across the ROMs to confirm.
Result
894 unique questions, full token coverage, two stubborn source-corrupt records left over (real ROM damage, not extraction). The same cracking pipeline as the rest of the Status Games line . but none of the phrase data was in the place the pipeline expected.
The token table for Hangman lived only inside its own questions, waiting to be reconstructed from context.
References
Related
More cracks
Cross-archive analyses
Hangman (1984, Status Game Corporation) · Arcade-Museum · Flyer · MAME romset:
hangman↩︎