Were there ever plans published, or kits, or a machine made, for a noughts-and-crosses computer game which did not use electronics – using only wire, an electricity source, and light bulbs ?

  • . A Playing against the computer game .
  • . I'm looking for the most optimal solid-state designs
  • . Machines containing things like electromagnetic relays / solenoids, are not accepted as answers, since I'm looking for alternate computer logic .
  • . Machines containing multi-contact-switches would qualify, although, of course I'm also looking for the most optimal solid-state designs, which would decrease these to minimal amounts, perhaps at the cost of containing more wiring.
  • . Machines containing a rotating component, like How to build a working digital computer, are not preferred answers, but they would be interesting to know about, if they only use wire, an electricity source, and light bulbs ?


Why I think one may have existed, and how it could have functioned .

  • It would use memory-boards, to compare the situation on the playing-board, to match it to one of the memory-boards, so when a memory-board is matched, it lights up a new square on the playing-board as the computer's next move.

  • Each memory-board would just be a length of wire, with gaps at certain positions to represent noughts or crosses (it may require separate memory-boards, one for noughts, one for crosses), and both sides of the gap wired up to a square on the playing-board, so when the human plugs in a piece, it closes/completes that part of the wire to allow electricity to flow. (so each square on the playing-board, would be wired up to lots of memory-boards).
    - Maybe, to significantly reduce the amount of memory-boards, some sort of board rotation for matching), could be used, simply using wiring.
    - NOTE - An innovation ( a computing / logic / circuit innovation ) would be, that for the human player , each plug-in-piece, would be numbered from 1st move to last move, each one of these plug-in-pieces would be of a different length to reach down to activate a different move-number/platform/circuit of the game.
    ( Or, of course, the plug-in-piece could have multi-contact-switches, varying for the move number ).

  • I wonder if theories used in Multiway switching / Switching circuit theory may be helpful. - I don't know if such a machine could work, I have not fully thought through all the details.
    - Anyone trying to visualize if this would work, use a 2x2 board.

  • (Note – for a computer, where a person simply plays against another person, it would be very simple for a machine not using any electronics, just using wire, electricity-source, and light-bulbs, to simply detect a win, when an entire row has noughts or crosses. I assume the original computer battleships game worked like that, so that's not the question in this post.)

  • Also, my other question, which they erased / closed Could you make a calculator without electronics ( using only, wire, electricity-source, and light-bulbs )? , had a useful innovation regarding the 'carry' in a binary calculation, it could be relevant for this question also.
    ( Quote "An innovation, could be plug-in-pieces/electrical-contact-pieces ( acting like switches ) that would be of a different length ( depending on what part of a number they are ) to reach down to activate a different circuit of the calculation ( useful for the 'carry' of a binary calculation ). ( Or, of course, the plug-in-piece could have multi-contact-switches, depending on what part of a number they are )."
    - NOTE. - What if this innovation, detailed above, regarding the 'carry' of a binary calculation, could lead to making an arithmetic-logic-unit which contains no transistors ( or even to other integrated-circuits not containing transistors ), by erasing that question there is no proof that I may have invented it.

  • Comments are not for extended discussion; this conversation has been moved to chat.
    – Chenmunka
    Commented Mar 16, 2021 at 16:18
  • 3
    There are attempts to re-open this question using trivial edits. (Also its duplicate retrocomputing.stackexchange.com/q/18293/10260)
    – DrSheldon
    Commented Mar 21, 2021 at 17:21
  • I wonder if the the robot arm that solved mazes, built by 'Thomas Ross' in 1933, at the 'University of Washington', which used a type of electromechanical memory, if it could have been designed to use 'No Electronics Or Relays', you just plug in the wall-panels to form the maze, and the individual plugs of the panels activate switches which either make, or, disconnect electrical connections, so when you plug in the last panel, the robot arm would immediately know the correct path
    – infomtn
    Commented Dec 22, 2021 at 21:30
  • This computer meets all requirements, including being made of "solid state" components. Because, what could be more "solid state" than holding your machines "state" in solid matchboxes and beads? (Suitable for study by a condensed matter physicist.) Although very retro it also features a modern "reinforcement learning" algorithm.
    – davidbak
    Commented May 10, 2022 at 20:56

3 Answers 3



In 1978 Danny Hillis and Brian Silverman, two sophomores at MIT, built an entirely mechanical tic-tac-toe (noughts-and-crosses) playing machine. It contains absolutely no electronics. The design could easily be converted to a device that uses wires and electric lights instead of little wooden flags to display its results.

Hillis and Silverman actually built two editions of the TinkerToy tic-tac-toe player.

Version 1, constructed in 1978, uses three-state logic gates, and is a cube about one meter on a side. This went first to the Mid-America Museum of the Arkansas Department of Parks and Tourism, and then to the Computer Museum in Boston, where it was on display in 1989. It currently resides in the Computer History Museum in Mountain View, CA.

Version 2, made in 1980, uses serial logic, and looks like a demented clockwork wall tapestry. It was built because the original cube required constant adjustment of its strings and pulleys, and the Mid-America Museum wanted a more robust version for a demonstration display. This is the one described in the October 1989 Scientific American.

  • Good reference, but I definitely saw a switches-and-lights implementation when I was aged 11 or so, which was quite a bit before 1978. I'm pretty sure that the builder did not invent it, but he's dead so I cannot ask,
    – dave
    Commented Mar 14, 2021 at 4:10
  • 1
    The description at the link is hilarious! Commented Mar 14, 2021 at 17:14
  • I wonder if that's the same machine or a different one from the one I saw at the computer museum in/near Boston, Massachusetts some decades ago?
    – supercat
    Commented Mar 14, 2021 at 19:21
  • @A.I.Breveleri: Scientific American showed an alternative version (still using Tinker Toy brand materials) which was smaller and I think more interesting. The one I saw at the Computer Museum used lots of vanes and purely combinatorial logic, but the one in Scientific American used sequential logic, even though an acute reader noted that a piece of the program was in the wrong place, and where it should have been, and the designer of the machine confirmed the fix.
    – supercat
    Commented Mar 14, 2021 at 21:56
  • 1
    @supercat: Danny Hillis and Brian Silverman built two editions of the TinkerToy tic-tac-toe player. - Version 1, in 1978, uses three-state logic gates, and is a cube about one meter on a side. This went first to the Mid-America Museum of the Arkansas Department of Parks and Tourism, and then to the Computer Museum in Boston, where it was on display in 1989. - Version 2, in 1980, uses serial logic, and looks like a demented clockwork wall tapestry. It was built because the Mid-America Museum wanted a more robust version. This is the one described in the October 1989 Scientific American. Commented Mar 14, 2021 at 23:52

A 1949 film clip hints at an electromechanical implementaton. The information content is low, but a picture of a uniselector suggest it's switches-and-relays logic.


1949 was definitely in the stored-program computer age (EDSAC ran its first program in May 1949) but this clearly was not computer technology.


With only wires and lights: NO.

The straight answer is NO. It can not be done, as neither wires not 'light globes' are active elements, thus they can't be modified/rearranged to model the game. Any solution will at least need a combination of complex (read multi contact) switches that will be moved either manually or machine controlled (which in addition needs clocking, like due a motor).

But YES if we add switches

Noughts-and-crosses 'programming' follows an extreme simple and straight forward scheme, allowing to draw up a fixed tree of moves. As a result this can be implemented as a complex but terminal number of switches and connections inbetween.

The Logikus

An early device (and general available) that possible could be used to 'encode this' would be the 1968 Logicus by Kosmos:

enter image description here

(picture taken from here)

There is a German Wiki-entry, as well as a whole site dedicated to the Logikus. A bit more information (in English) can be gathered from a page about an emulator.

There were at least two licenced versions in the US and Canada by Logix Enterprises of Montreal and New York called O-600 and SF-5000, ca 1970, as well as kind of a simpler clone by Tandy as part of their "Science Fair" Series (as late as 1977). There's an English language video describing it - and another, showing more details, although she doesn't seem as exited :))

For this I's stay with the original.


It was an educational toy, first produced in 1968, consisting of 10 switches, 10 light bulbs and a button to power whatever is connected when pressed.

  • Each switch could be in one of two positions, up or down.
  • Each switch had 10 contacts, of them
    • half were closed
    • and half were open in either position
  • Each contact could be feed by up to three wires
  • Each contact could feed 3 wires.

Each light bulb could be feed by a single wire.

The button was connected to the batteries and all light bulbs.

To plug a 'program' wires would connect the button with some contacts of some switches and routed from there. Connecting two contacts

  • in series forms an AND relation
  • in parallel forms an OR relation

Doing either over 'open' contacts negates logic, introducing NOT

In combination any switch can be seen as a single bit input, representing two possible values, while each lamp was a single bit output as lit or dark.

Implementing the game

While a single Logikus has not enough switches and contacts to represent the game in a plain version, using prepared input (which in turn could come from one or more other Logikus') could allow to program the strategy on a single Logikus:

(just a quick write up, I bet I forgot some combination)

Lets have binary inputs for

  • Two own tokens and an empty field in a row
  • No token at center
  • One own token plus two empty in a row
  • No token in corner

These 4 conditions can easy be combined to 4 output bits to lit one of 4 commands to the player:

  • Place token in empty field and win
  • Place token at center
  • Place token in row with one own and two empty (in corner)
  • Place token in empty corner

All of these can be made with easy AND/OR/NOT connections.

For a detailed output which field is to be used this result can be used in conjunction with the input data (from previous stage) to form an output stage naming exactly which field the token is to be place usind another Logikus.


While this can be used to play the whole game - sufficient switches/contacts given - it would not play it automatic, as there are no sensors, actors and stepping. All it can do is to value a certain game state and derivate the next move. Everything else, from inputting the state to executing the move must be done by the operator.

As mentioned, due usage of switches this already violates the boundaries set by the question.

  • 1
    @Spektre Sounds like the perfect property for a 1940s Frankenstein film
    – Raffzahn
    Commented Mar 14, 2021 at 20:24
  • 1
    @supercat No many. I say 9 three way switches. One for each field (position), with three possible settings (empty, X, O). The real issue will be the number of contacts per switch, as they 'encode' the logic equations. Their number depends on how much the equations can be simplified (reduced), but will be rather high without intermediate (memory and multiplexer) elements. MENACE does prove that a static solution is available, which is the most important part here. As more I think about as more it seems doable. even with minimal hardware - but not without switches/contacts.
    – Raffzahn
    Commented Mar 14, 2021 at 21:57
  • 1
    @SimonF There was one by Tandy? So far I only knew about the Logix clone in the US. Googling brought up a video. Cool. I need to add that. thanks.
    – Raffzahn
    Commented Mar 16, 2021 at 14:35
  • 1
    @Raffzahn Saw the video. Strange, that's different to what I had... maybe they had cost-reduced it by the time I bought mine. For example, the one in that video had springs to connect the wires. I'm sure mine only had holes you poked the wires into, a little bit like a breadboard... but cheaper and nastier!
    – Simon F
    Commented Mar 17, 2021 at 14:04
  • 1
    @Raffzahn Found one! radioshackcatalogs.com/flipbook/1976_radioshack_catalog.html Go to page 94 bottom right corner!
    – Simon F
    Commented Mar 17, 2021 at 14:18

Not the answer you're looking for? Browse other questions tagged .