AUTOX, Father of the Computox

Last weekend the Curator of the Museum of Computing uncovered some images of the Computox and related material in an old PC backup. These included some pictures of the Computox from September 2010, and scans of nine pages from the Model Engineer Magazine.

Single player Computox game from 2010 (note the failed bulb on the right O)

The magazine scans are split across three articles from 1956 and describe the workings of another relay-based tic-tac-toe machine, the AUTOX. The name is an abbreviation of Automatic Noughts and Crosses. Unfortunately, scans of these issues cannot be found on the internet, hence why the machine did not appear in the earlier timeline of interesting machines. The articles are credited to ETP and I've included images of the relevant pages below. It's interesting to note that ETP references the work of both Davies and Haufe on the first page.

ETP's introduction to the AUTOX

The AUTOX is a single player machine housed in an old gramophone case. The design is based around a single uniselector and eleven telephone relays. As such, even without supporting two player games, its implementation is quite a bit simpler than the Computox. Whilst the AUTOX uses lamps to indicate the computer's moves, it doesn't do this for those of the player. Instead, the player makes their moves by closing a shutter over the desired square rather than pressing a button (see below). The act of rotating a shutter into place opens or closes up to eight switches, which act as latching relays, similar to the *XA relays in the Computox. At the end of a game, these shutters must be rolled back by the player before starting the next.

The AUTOX player shutter mechanism

The computer turn is initiated by pressing a button, which causes the uniselector to sweep out a full rotation, testing 24 possible game positions in preference order (1 - attempt to finish a computer two O line to win, 2 - block a player two X line, or 3 - chose a square from the list: E, A, J, C, B, H) and selecting the first available one found. Like the Computox, the AUTOX can play either first or second at the player's discretion, and will adjust its strategy slightly based on the choice made. There is also a third mode of play where the player starts and the AUTOX will play imperfectly, matching the Haufe design or "secret switches" of the Computox.

One of the interesting things about the article scans is that they are annotated in places, suggesting that they were used to check off progress in the construction of the Computox.

AUTOX winner schematic annotated with ticks

So, can we confirm that the Computox is a derivative of the AUTOX? The article provides two tables that detail the move selection mechanism, from the uniselector, through the relays to the relay coils that claim the computer move, and light the lamp.

AUTOX uniselector and relay wiring

A preliminary investigation suggests that the connections detailed in these tables are replicated between the larger uniselector and *OA, *OB, *XA and *XB relays in the Computox. Therefore, we can be pretty confident that the computer opponent in the Computox implements the same strategy as the AUTOX, and the Computox is a derivative / extension of it.

The details of the connections between the uniselector and relays have been added to the relay spreadsheet. They make a significant dent in reverse-engineering the remaining logic, which is a bit of a relief.