The Golden Age

Whilst I'm busy with buzzing out connections, I thought it might be interesting to drop in some information about other dedicated tic-tac-toe machines.

I'd argue that the era of "interesting" tic-toe-toe playing machines is quite well defined.  It starts in 1949 with the independent work of Donald Davies, Robert Haufe and William Keister that leads to machines used for public demonstrations in the 1950s and 60s.  And it ends in 1978 when Parker Brothers released Merlin, designed by Bob and Wendy Doyle, and Wendl Thomis.  Perhaps it ends a bit earlier with the the publication of David Ahl's book, "BASIC Computer Games" in 1973 which includes a couple of tic-tac-toe opponents.  Regardless, at the point that most people can afford a handheld electronic noughts and crosses opponent like Merlin, I think we can definitely say it's no longer the subject of awe and mystery.

Donald Davies

We have to start with Donald Davies, who put together a tic-tac-toe machine in early 1949.  This was built in his spare time while working at the National Physical Laboratory (NPL) in London.  Like the Computox, the Davies machine was built from telephone exchange parts.

Donald Davies' machine 

Following its demonstration at the Royal Society in May 1949, his feat was documented in newspapers of the time and on Pathe news.

In addition to creating a bit of a stir, the Davies machine seems to have had a varied life.  The last mention of it I've found is a note written by Davies, quoted on page 91 of his Royal Society memoir.  

"A by-product of my electromechanical work was to finalise my quest for a machine to play noughts and crosses. In the evenings I built a machine using Post Office relays. NPL put it forward as an exhibit at the Royal Society Soirée in 1949. Its clicking noise and nice display attracted people, so it received more attention than it deserved. I felt rather bad for the other exhibitors with real scientific achievements to show. Next day I was on the Daily Express front page and much more publicity followed. I appeared on TV with Richard Dimbleby at NPL and then at Alexandra Palace, I was interviewed on a children’s TV program with it. This came full circle from the radio children's hour which started off my quest. It was even used as a ‘prop’ in a popular magic show at Lime Grove Studios.

This machine did not use the whole strategy, so I built a second model which was partly electronic, with a wire threaded through magnetic cores to represent each rule. This machine was a regular feature of the NPL children’s party. Forty years later I heard that it was being revived (again) for this party."

Unfortunately, I don't believe it's known whether either of Davies' machines still exists, or where they are if they do (I have tried contacting NPL, without response so far).

Robert Haufe 

At the same time as Davies was working in London, Robert Haufe was building a similar machine at CalTech.  This was demonstrated in August 1949 at the Pacific General Meeting of the American Institute of Electrical Engineers, and written up in the paper "Design of a Tit-Tat-Toe Machine", published in volume 68, issue 10 of Electrical Engineering (sadly this doesn't seem to be available on the internet).  The machine's existence was subsequently reported in a short article within the January 1950 edition of Popular Science.

 

 

Robert Haufe's machine

Haufe filed for a patent of a commercial "coin-op" version his machine in April 1950, which was granted in 1957.  Like most patents of the time, it describes the working of his machine in detail and includes a full circuit schematic.

Haufe's machine as portrayed in his patent

Again, unfortunately it's not known whether Haufe's original machine, or an example of his coin-op derivative survives today.

William Keister

The final member of the 1949 relay club is William Keister.  He had mapped out the boolean logic for tic-tac-toe earlier in 1937.  While working for Bell Labs, Keister constructed a machine embodying this logic for public demonstrations.

William Keister with his machine in 1958

It's not totally clear when Keister's first machine was put together, but Bell also filed for a patent covering the machine in November 1950, which suggests it was a reasonable time before this.  The patent was finally granted in early 1959 (what is it with the US patent department, surely one of the Haufe or Keister machines should have been prior art?).  Like both Davies and Haufe, Keister built his machine out of telephone exchange relays and the patent provides a full description of its operation along with a schematic.

Keister's machine as portrayed in his patent

According to Keister, several of his machines were constructed by Bell Labs, who used them for promotional work, one of them appears to have been displayed over a decade later at the 1964-65 World's Fair in New York.  Bill Keister can be seen reminiscing about his work in a family video on YouTube.

Perhaps in part because multiple machines were constructed, at least one example of Keister's machine is known to exist today.   Both a display unit and relay chassis are to be found in the collection of the Computer History Museum in Mountain View.

The publicity around this early work seems to have kicked off a minor craze for building dedicated tic-tac-toe machines, and although there are others that could be discussed, for example "Bertie the Brain" or "Relay Moe", I'm going to pause here and pose an obvious question...

Is one of these a direct ancestor of the Computox? 

At first glance, the Drof Computox, with its 47 relays and two uniselectors, looks similar to all three of these machines.  Is one of them a distant relation?, well...

• Whilst the Davies machine also has two uniselectors, and is constructed around six columns of relays, it appears to be more complex with a total of 81 relays.  Additionally, the naming of the relays visible in the Pathe video differs from that of the Computox.
• The Haufe machine seems to have a single uniselector and only 38 relays, but the patent does use the same A-H & J position naming scheme as the Computox.
• The Keister machine looks to have 48 relays, very similar to the Computox, but it has no uniselectors at all.  Again, the patent uses the same A-H & J position naming scheme.

So, although any single or combination of these machines could have inspired the creator of the Computox, it doesn't appear to be a clone of any of these earlier creations.  Nonetheless, given the similarity of the Keister and Haufe machines, the open nature of patent literature and the scale of the task of brute force reverse engineering the Computox, it's probably sensible to go through the cryptic descriptions of the Haufe and Keister patents, just in case they provide a monumental shortcut.