Identity and evidence precedence
This is the Bally/Midway physical product released January 1992, IPDB #20, WPC part-number family 16-20017 (Operations Manual 16-20017-101, Operator's Handbook 16-20017-103, WPC Schematic Manual 16-20017-102A). It covers the taf_* clone tree rooted at taf_l5: 20 driver rows spanning Bally production revisions L-1 through L-7 with their D-*/LED-Ghost-Fix pairings, prototype revisions P-2/P-3, the newest known H-4/I-4 revision, the German-distributor-only L-6/D-6, a home/competition L-5C patch, and a 2026 community ball-saver patch (taf_d7bs) of the L-7 prototype. Every one of these is a firmware revision for the same physical machine.
The 1994 Williams "Addams Family Special Collectors Edition / Gold" (tafg_*, IPDB #21) is a separate physical machine and is out of scope for this definition. Three independent facts establish this: the WPC part-number family changes entirely (WPC 50038 vs this machine's WPC 20017); pinned PinMAME's own taf.c carries a 2000-era maintainer comment reading /* 161000 TAFG is not a clone of TAF anymore */ immediately above a block that adds a dedicated Buy-In button and separate sound ROMs for tafg; and the retained manual set is physically distinct (Gold's manual is retained separately at manuals/by-machine/williams.the-addams-family-gold.1994/, and this definition's three PDFs were visually confirmed to never print "Gold" or "Special Collectors Edition" anywhere). machines/stubs/tafg_lx3.json is left untouched; it needs its own curation pass with Gold's own manual set as evidence, not a silent merge into this record.
Evidence precedence for this definition: the retained known-working VPX table ("The Addams Family (Bally1992) v2.3.2 (g5k)" by g5k, Sliderpoint and 3rdaxis, cGameName = "TAF_L7") is runtime and mechanism-causality ground truth; the Operations Manual (16-20017-101) and Operator's Handbook (16-20017-103) control physical construction, part numbers, polarity, quantities, and device presence; pinned PinMAME (taf.c) controls controller generation, public address topology, and mechanism-counter causality; the retained VPX geometry supplies normalized coordinates. All three manual PDFs carry an OCR text layer, but the OCR misreads digits and glyphs on the two-column tables, so every printed table used here was visually re-transcribed from rendered pages into external:pinmame-review-artifacts/the-addams-family-1992/manual-transcription.md rather than trusted from OCR text.
GEN_WPCFLIPTRON (PINMAME_HARDWARE_GEN_WPCFLIPTRON = 0x8, pinned PinMAME's own comment names this exact machine: "Fliptronic flippers, Addams Family 2/92 - Twilight Zone 5/93") with wpc_dispDMD. This definition reuses controllers/pinmame/wpc-fliptronic.json unchanged -- the same profile Twilight Zone introduced.
- Switches: dedicated coin-door 1-8, matrix 11-88 as drive column then return row, Fliptronic 111-118.
tafGameData declares no custom switch column (swCol = 0).
- Solenoids: standard drivers 1-28 (all fitted; unlike Monster Bash's unfitted 4/7, every one of TAF's 28 printed solenoid-table rows has a real device), Fliptronic upper-flipper circuits 33-36, Fliptronic lower-flipper circuits 45-48.
tafGameData declares no custom solenoids (custSol = 0), so nothing above 51 is claimed. Addresses 37-44 are simply unused address space on this generation (no LPDC board; see wpc-fliptronic.json's own note), not a duplicated WPC-95-style mirror pair.
- Lamps: full 8x8 matrix 11-88 except two genuinely unused positions (41, 76). No auxiliary lamp column (
lampCol = 0).
- GI: five strings on public addresses 0-4; address 3 is printed "Not Used" and has no playfield object.
Opto polarity -- a clean cross-check, no conflict
Sweeping the Operations Manual's Switch Locations parts list (printed 2-39) for the Non-negotiable #1 opto signature (blank Switch Number, populated paired opto-assembly part number, "Opto" in the description) finds exactly seven addresses: the four Bookcase optos (53-56, assembly A-15017/A-15018), the Bumper Lane opto (57, assembly A-14231/A-14232), and the two Thing position optos (84/85, assembly A-15285). tafGameData's inverted-switch mask --
{0x00,0x00,0x00,0x00,0x00,0x7c,0x00,0x00,0x18,0x00,0x00,0x00} -- covers exactly the same seven positions, column for column (column 5 bits 2-6 = 53-57) and row for row (column 8 bits 3-4 = 84-85). Full agreement. Unlike Monster Bash's Dracula-position optos or Centaur's lamp 113, there is no unresolved polarity conflict anywhere in this definition.
One address deserves a note despite not being a real opto: item 23 ("Ticket Opto.") has the word "Opto" in its printed description, but both its Switch Number and Switch Assy columns read blank/"Not Used" -- the manual's signature for "enumerated, never fitted", the same pattern this project uses for Monster Bash's knocker (solenoid 7) and Twilight Zone's third magnet.
Both upper Fliptronic flippers are real, fitted hardware
Non-negotiable #3 asks whether the Fliptronic block hides a repurposed position (Monster Bash puts a spinner on F7). Here the opposite question needed settling: are both upper flipper positions genuinely fitted, or is one a template row (Monster Bash's F5/F6/F8 print "NOT USED")? Three independent sources say both are real:
tafGameData.hw.flippers sets both FLIP_SW(FLIP_U) and FLIP_SOL(FLIP_U) -- unlike Monster Bash, where FLIP_SOL(FLIP_L) only is set despite FLIP_SW(FLIP_L|FLIP_U).- The Operations Manual prints a full illustrated, itemized parts breakdown for both
A-15205-R "Flipper Assembly - Upper Right" (EOS switch SW-1A-193, coil FL-11630 "Flipper Coil - Red", crank link, bumper plug, 20 line items) and A-15205-L-1 "Flipper Assembly - Upper Left" -- not blank template pages.
- The retained VPX script wires
SolCallback(sURFlipper) = "SolURFlipper" and SolCallback(sULFlipper) = "SolULFlipper", each rotating an independently animated table object (Flipper1 upper-right, Flipper2 upper-left, rendered with "Thing" hand art).
The two positions serve different gameplay roles. The upper-left flipper is "Thing's Mini-Flipper", marketed in the manual's own "'Thing Flips' Automatic Calibration" section as "an exclusive Williams/Bally pinball innovation": an AI-calibrated automatic flip enabled by the far-left flipper return lane, not a normal player-controlled flipper -- "the ball will be diverted to the upper left mini-flipper and the game ('Thing') will attempt to shoot the ball into the swamp... between 50-60% of the time." taf_stateDef's stTFlip case is the only place in the driver's own ball-routing logic that reads any upper-flipper solenoid (core_getSol(sULFlip)), confirming the asymmetry. The manual names this feature's own calibration inputs explicitly: the Bumper Lane Opto (57, printed "above the upper left mini-flipper"), the three Swamp Million targets (45/47/48), and the Swamp Lock Upper switch (71).
The upper-right flipper is never named in the rules text and never referenced in taf.c's ball-routing sim; it is documented here as ordinary Fliptronic hardware with its own EOS/button pair and coil (FL-11630), CPU/software-timed like the left one (FLIP_BUT(FLIP_L) gives only the lower positions the direct hardware button-to-coil bypass on this generation). This project has not independently confirmed from static sources alone whether the ROM ties its firing unconditionally to the right flip button or gates it further -- that nuance is not required to recreate the physical device correctly, since a table author just wires the coil the way the known-working retained script does.
Ball path, trough, and shooter
A ball drains through the Outhole (18) into a three-position trough -- Left Trough (15) nearest the drain, Center Trough (16), Right Trough (17) at the release position -- and solenoid 4 (Ball Release) kicks the ball at Right Trough into the shooter lane, where it rests on switch 27 (Ball Shooter) until the player pulls the manual plunger. TAF has no auto-plunger.
Bookcase (Vault) and Thing hand -- motor-position mechanisms, not discrete sensors
Two mechanisms are driven purely by internal motor-position counters inside taf_handleMech, with no discrete playfield sensor behind either pair of switches:
- Bookcase (A-14970 assembly, solenoid 27):
locals.bookPos is a 0-199 tick counter, incremented while solenoid 27 is asserted and wrapping at TAF_BOOKTICKS = 200. swBookOpen (81) asserts for bookPos < 15; swBookClose (82) asserts for 100 <= bookPos < 115. The retained script never calls Controller.Switch(81) or (82); it only reads Controller.GetMech to animate the visible vault_base/vault_plastic/vault_post/vault_screws/vault_upright rotation and swap the base's shadow image.
- Thing hand (A-14711 Hand Drive Assembly, solenoid 25):
locals.thingPos is a 0-399 tick counter, wrapping at TAF_THINGTICKS = 400. swThingDn (84) asserts for thingPos < 15; swThingUp (85) for 200 <= thingPos < 215. The retained script reads Controller.GetMech to drive Thing.RotY, handMAGNET.RotY, and a non-linear thingBox.Rotx lookup table (a 26-branch piecewise table in ThingMotor) so the visible box lid tracks the hand's rise smoothly.
Both pairs of switches are recorded here with a documented spatial projection onto their own mechanism's retained primitive (vault_base for 81/82, Thing for 84/85) rather than an invented sensor position, following the same allowance Monster Bash's Dracula-position optos used for Primitive Drac.
The Thing hand is a separate mechanism from the four Bookcase optos (53-56, "Bookcase Opto 1-4"): those four are fixed playfield sensors near the loop habitrail that calibrate ball detection around the bookcase area, not part of the rotation-position counter above.
Thing saucer -- capture, magnet grab, and kickout
A ball entering the Thing saucer closes switch 87 (Thing Eject Hole). Solenoid 6 (Thing Magnet) then energizes a magnet under the saucer so the Thing hand mechanism appears to grab the captured ball (ThingMagnet(Enabled) sets ThingBall = True and drives the hand-grab sound/animation); solenoid 26 (also literally named "Thing Eject Hole" on the printed solenoid table) kicks the ball back out. All three devices -- the capture switch, the magnet, and the eject kicker -- share one physical assembly (A-15368 Eject Assembly) and one retained script object, ThingSaucer. The separate Thing Kickout hole (switch 77, solenoid 7, assembly A-15200) is a distinct device elsewhere on the playfield.
Swamp lock
Balls lock in sequence down the Vault habitrail: Swamp Lock Upper (71), Center (72, which additionally drops Wall70 to hold the ball), Lower (73, which drops Wall69). Solenoid 28 (Swamp Release) releases locked balls into the Lockup Kickout hole (switch 74), and solenoid 8 (Lockup Kickout) kicks the released ball back to the playfield.
Jet bumpers -- resolving a label discrepancy at address 34
Five B-12030-2 jet bumpers on switches/solenoids 31-35. The Operator's Handbook's rendered Switch Matrix wiring page (printed p.9-10) happens to render the column-3/row-4 cell identically to row 3 ("Center Left Jet"), which would duplicate switch 33's label. Two independent sources disagree with that page and agree with each other: the Operations Manual's Switch Locations parts list (2-39) names item 34 "Center Right Jet", and the retained script's own comment on Bumper4_Hit (the handler that pulses switch 34 and fires solenoid 12) reads the same. Per this project's "prefer the parts list, resolve typos via the symmetric partner" rule, address 34 is Center Right Jet, matching the natural Upper-Left / Upper-Right / Center-Left / Center-Right / Lower naming of the five-bumper cluster.
Cousin It -- four targets, one address
Public switch 44 ("Cousin It") is shared by four standup targets on one figure -- SW44a1, SW44a2, SW44b1, SW44b2 in the retained script, each independently pulsing switch 44 -- matching the printed parts list's two item rows ("44a ... Cousin It (2)" / "44b ... (2)", four targets total). All four cluster tightly in the retained geometry (x in [0.4075, 0.4292], y in [0.3903, 0.4645]), consistent with one small creature figure rather than four separated targets.
Grave word features
Two independent five-letter word features exist on separate address ranges. The jet-bumper cluster's lamp column 3 spells G-R-E-E-D (lamps 31-35, one letter per jet). A different set of standup targets and lamps spell GRAVE: switches 41 ("Grave G"), 42 ("Grave R"), and 86 ("Grave A") are dedicated standup targets, while lamps 43 ("Grave G"), 44 ("Grave R"), 14 ("Grave A"), 17 ("Grave V"), and 48 ("Grave E") light the letters -- four different lamp-matrix columns, not a contiguous run, so "V" and "E" are lit by other shots (the Vault and Electric Chair features respectively) rather than by dedicated standup targets of their own.
Magnets
Three under-playfield magnets implemented through the retained script's shared cvpmMagnet helper class rather than a scripted SolCallback: Left Magnet (solenoid 16, LeftMagnet.InitMagnet LMagnet), Upper Magnet (solenoid 23, UpperMagnet.InitMagnet UMagnet), and Right Magnet (solenoid 24, RightMagnet.InitMagnet RMagnet). None has a dedicated printed sensor switch; they hold and redirect the ball generically rather than gating a specific shot the way the Thing magnet does.
Lower flippers
Two FL-15411 flippers on Fliptronic circuits 45/46 (right) and 47/48 (left). FLIP_BUT(FLIP_L) gives these two positions the direct hardware button-to-coil bypass circuit (minimum latency), dropping to the hold winding once the end-of-stroke leaf switch closes. The retained table runs Const UseSolenoids = 2 (fast flips).
General illumination
UpdateGI's Case 0 ("Left String") drives collection GILeftString (7 bulbs: GI11, GI41, GI38, LLightning1, GI14, GI28, GI15) and Case 4 ("Right String") drives GIRightString (4 bulbs: GI42, GI34, GI33, SwampLight). There is no Case 1, Case 2, or Case 3 block, matching the manual's own printed table exactly: GI02 "Insert House String" and GI03 "Insert People String" are backbox strings with no playfield object, and GI04 is printed "Not Used" outright.
A modeling quirk worth flagging for future curators
The seven "THING" name-chase header lamps (public 81-87, spelling the word across the top of the backbox insert/header area rather than the playfield proper) sit at a small negative normalized y in the retained table (-0.0288 to -0.0292), just beyond the table's own defined top=0 edge. This matches the manual's own Lamp Locations diagram, which draws these seven lamps in a row above the cabinet outline rectangle -- i.e. a real design detail (a header strip mounted above the play surface), not a table-authoring bug. This definition clamps their recorded y to the schema-valid boundary 0.000000 and discloses the raw offsets in vpx-geometry.txt; a future curator encountering the same pattern on another WPC game should look for the same explanation before assuming a bug.
Similarly, every lamp address on this table has a co-located "b"-suffixed Light object (L11/L11b, L12/L12b, ... L87b) at an offset under one bulb diameter -- a systematic brightness-doubling technique used uniformly across the whole table, not evidence of physical two-bulb inserts (the manual marks no lamp position with a (2) bulb-quantity note). One address, 45 ("The Mamushku"), has two candidate primary objects instead of the usual primary+"b" pair (L45old and L45bold, neither referenced anywhere in script.vbs, relying on VPinMAME's default lamp-matrix name binding); L45bold is treated as the current art revision here.