the totalisator

It has been an unfortunate turn in the software industry, one of many as of late, that gambling is once again one of its primary engines. With the rise of almost nationwide online sports betting, not to mention prediction markets, making odds on real-world events and extracting the money of suckers is no longer limited to island nations. It is a great American pursuit, or at least, that's what modern television sports coverage leads you to believe.

There has always been an uncomfortable relationship between software and the manipulation of marks. Techniques developed by casinos became a fundamental part of consumer software, while the software industry wholeheartedly embraced "gaming" as a market (the older meaning of the term here, meaning gambling). We can readily point to a couple of reasons: first, gambling is profitable, and technology is first and foremost a means of accumulation. Second, gambling is mathematical, or at least arithmetical, in nature. Most forms of gambling involve some sort of complex calculation with real-world stakes.

Gambling predates history, or it might be better to say that gambling has been around for as long as recorded history has been able to observe it. Most early gambling seems to have been based on card or dice games, but humans have been betting on animal fights for more than a thousand years. As sensibilities and resources changed, animal fighting has mostly given way to animal competition. The most famous of these wagering opportunities is horse racing, a form of gambling with such a long and pervasive history that it has often achieved a unique regulatory status as one of the only legal sports betting venues in the US. Well, at least, before Murphy v. National Collegiate Athletic Association.

The earliest recorded horse races were held in England in 1539, and bets were placed. By 1666, horse racing had reached such prominence that King Charles II—himself a jockey—commissioned and then won the "Newmarket Town Plate." That event's eccentric history gave way to the King's Plate, a broader 17th-century racing series whose royal remit made up the first formal rules for the sport. Queen Anne founded the racetrack at Ascot in 1711; while it took decades for permanent facilities to be built at the track, only stands for the royal family came before a betting office. As British empire expanded around the world, horse racing spread with it. Likewise, horse racing spread throughout Europe. By the 19th century, horse racing could be found almost anywhere.

For most of the history of horse racing, betting was based on the setting of odds. A bookmaker would apply historical knowledge, experience, and no small amount of "guesstimation" to set odds of various events—say of a specific horse winning. A bookie might set odds of 7/1 on a given horse, read as "seven to one against." A bettor placing $1 on this horse will make $7 if it wins the race, which of course implies that the bookmaker thinks the horse's chance of claiming first is around 14% or less ¹.

This method of betting is known as "fixed odds," and the fact that it relies on the bookmaker to set those odds is a significant limitation. First, there is the constant possibility of error. A bookmaker who accidentally sets odds too favorably could ruin themselves, which creates a natural pressure towards odds that are less favorable to bettors. At the same time, competition between bookies drove odds the other direction. This tension between bookmakers and bettors, and between the bookmakers themselves, was a constant source of conflict. Besides, despite the upper-class connotations of British horse racing, gambling has always touched on the unsavory. Bookmakers were not always known to be honest. A host of possibilities, from collusion to ignorance, could leave bettors with no good options. Every bet offered might be a bad one—even more so than the spread collected by bookmakers would suggest.

Of course, despite their prominence, horses did not originally inspire the revolution in gambling that would soon transform racetracks—roosters did.

Josep Oller was born in Catalonia in 1839, but his parents emigrated to France when he was a child. His parents seem to have found some regret that he grew up without Spanish, and perhaps the opportunities in Spain were better. In any case, as a young adult, he moved to Bilbao to study. Accounts vary on what happened next: depending on who you trust, he either merely observed cockfighting, or he was enthusiastically involved in promoting it. I tend to suspect the latter, as Oller was an enthusiastic promoter of many things and not too concerned with vice. Decades later, after his return to France, he co-founded Moulin Rouge. That wasn't even his first entertainment venue, just the most famous. But that's a later story, in another industry. In Bilbao, in the 1860s, Oller watched the cockfights. And then, he watched the bettors and bookmakers haggle, argue, and fight.

The vagaries of fixed odds, the questionable motives of the bookmakers, and the general atmosphere of debauchery made it all rather ugly. Oller thought there must be a better way: the calculation of payouts according to fixed rules, with the impartiality and precision of mathematics. When he returned to France and began to promote his new betting system to racetracks, he named it Parimutuel.

Here's how it works: parimutuel betting is not against the bookmaker or house; it is against the other bettors on the same event. Take an event with a list of possible outcomes, like a horse race and the set of horses that might take first place. Each bettor wagers a sum of money on a specific horse, and a parimutuel teller records the details of each of these wagers.

After the race, when the outcome is known, the parimutuel teller sums the total money wagered on the question, which is called the "pool." Some amount, usually a percentage, is subtracted from the pool to cover taxes and a house take (profit margin). The remaining majority of the pool is then distributed to all of the people who bet on the winning horse, proportionally to their wagers. For clarity, let's work an example: Secretariat, Shecky Greene, and Warbucks are running a race. Being a clever person well educated on Kentucky Derby outcomes, you put $100 on Secretariat. Your friend, similarly informed but cash poor, bets $50 on the same outcome. At the completion of the race, the betting office adds up all of the wagers to $900 (I guess it wasn't a popular race day). There is a state racing commission tax to keep in mind, and of course the owner of the race track wants a share, so they remove 20% from the pool leaving $720. That pool needs to be distributed among the winning bettors. Let's say that you and your friend were the only two Secretariat fans present. Having bet $100 and $50 for a total of $150, a parimutuel teller works out that you receive 2/3 of the pool and your friend 1/3, proportional to the original wagers. You walk away with $475 and your friend with $237, give or take some change. A good day at the races.

There are a few things to observe about this system. First, there is no estimation of odds involved (properly called handicapping). The payout on bets is calculated based on the bets placed, regardless of what anyone expected the outcome to be. But, like the prediction markets with which Silicon Valley is so infatuated today, the general expectation is that bettors will place bets proportional to the likelihood of the horses winning... this means that if a popular horse, widely expected to win, does indeed take the podium (a big podium that horses fit on), the wagers on that horse will have made up a large portion of the total pool and the payouts will be proportionally lower. In the most extreme cases, it is possible to place a bet, win, and still lose money: if everyone bet on the same outcome, they all just get their money back, but minus the house take. Exactly how these situations are handled varies by jurisdiction, but you might be relieved to know that in some rule systems there are scenarios where the house can similarly lose money on edge-case outcomes.

An implication of this fact is that the returns on a wager, assuming that it wins, are not exactly known until after the race has begun, since the tellers continue to take wagers (that change the size and distribution of the pool) up to that point. It's also, just, a little bit complicated? Picture yourself at the racetrack, decadent and depraved, and no doubt several beers deep. You place a bet on your favorite horse—if it wins, what do you get? Parimutuel payouts are relatively difficult for gamblers to understand, and that could limit sales and satisfaction. That's an especially big problem when parimutuel coexists with simpler fixed odds options.

Oller came up with a neat solution: advertise parimutuel betting as if it were fixed odds. Oller worked mostly out of carts, portable parimutuel offices, with big display boards over the teller window. As bets were placed, Oller would calculate the expected payouts on each option and convert them to odds. The equivalent odds were displayed on the board—just like the fixed-odds bookie had. Of course, these odds are only estimates rather than exact, and they need to be updated as the race approaches so that everyone knows what they're getting into.

The exact design of these boards changed over time, Oller himself was actively experimenting for the first decade. While later versions would to display the equivalent odds and perhaps a bet count, Oller's first scheme made the math easier (for the clerk) by displaying the bet count and totals on each outcome instead (conversion to payouts was left as an exercise for the customer). Since the main point of the calculation was to determine the totals, Oller called the process a totalisator. The board, where the totals could be read, was often shortened to the tote board ².

We should also discuss what people are actually betting on. So far, we have stuck to the simplest case: betting on which horse will take first place. That's called a bet to win, or a straight bet. It is also common to bet on a horse placing either first or second ("place"), or first, second, or third ("show"). There are a lot of other scenarios you can bet on as well, like the composition or order of the placing/showing horses or the place that a specific horse finishes, but these more complex bets also get to be less common and may not be offered as parimutuel (one property of parimutuel is that, as a practical matter, you need to be able to attract a certain minimum size of betting pool for every option you offer). The point is that there can be a lot in the air for any given race, with multiple separate betting pools—and on top of that, a busy track might be taking bets for multiple races.

Within a few decades after Oller's invention, parimutuel was taking over as the norm at horse tracks. Along with the betting system came the boards: huge tote boards were a main feature of "modern" race tracks by the 1880s. Oller's carts gave way to "tote houses," dedicated offices with teller windows at the front, a business office (for totalisators) at the back, and the tote board on top. As clerks updated the totals, they sent the new numbers "upstairs" to a clerk who worked behind the tote board, changing out number cards.

It's hard to tell where the first mechanical calculators for parimutuel emerged. There were probably multiple parallel inventions, since the idea is obvious, and it appears that many different ideas were pursued. Oller himself probably experimented with means of automation, since the first devices probably came from France. By 1880, at least one German company advertised totalisator machines. Not very much is known about these first devices, but they mostly fell into two broad categories.

First, there were techniques and devices for larger operations to coordinate multiple tellers and speed up payout determination. Most of these were not exactly calculating machines as we think of them today, but more like derivatives of other 19th-century office technology like cash ball systems. For example, in one system that seems to have found use in several countries, parimutuel tellers sold "tickets" of a fixed face value. For each ticket sold, they took a ball like a steel bearing or marble and set it on a rail for the corresponding horse. The ball rolled down the rail to the end of the tote house, and fell into a bucket. At the end of the race, the winning horse's bucket was weighed to determine the number of tickets sold and, thus, the payout on each winning ticket.

On the other end of the spectrum were small mechanical machines intended to speed the work of a single parimutuel teller. Once again, the history of these devices is not well documented, but during the late 19th century and at least in Europe there seem to have been several parimutuel machines marketed. These were probably more like adding machines, dispensing tickets and keeping the sum of tickets sold. Their major limitation was that, so far as I can tell, none of this generation of devices tried to suit situations where multiple tellers were involved.

This highlights a dichotomy between two schools of parimutuel technology: devices sold to single-person operations, running out of carts like Oller's, and devices sold to large parimutuel offices at race tracks. The former did the math but left it to the user to combine sales between tellers. The latter combined the sales between tellers, but left most of the math as a manual exercise. Neither family of devices updated the tote board—at best, they made it a faster manual process (e.g. by regularly weighing the buckets). It should also be said that none of these solutions were reportedly that good. Most were short-lived, and bettors don't seem to have had much trust in them. Marble-bucket-type solutions were known in particular for having a certain margin of error... a margin of error that, to a slightly less scrupulous bookie, became a margin of profit.

Of most interest to us, considering later events, is the early history of parimutuel automation in New Zealand. Parimutuel, as a system of gambling, landed in New Zealand around 1880. The first parimutuel devices and machines soon followed, imported from European manufacturers. An 1884 issue of Southland, New Zealand's Western Star gives some of the flavor of this new pastime, describing the happenings at an amateur steeplechase meet:

The British love for backing one's fancy was specially provided for in a manner the fairest possible to the public, in the shape of a pari mutuel. This machine was worked by two gentlemen whose suave manner was inimitable, to the satisfaction of all parties, who volunteered their services, and have handed over legitimate commission for presentation to the Southland Hospital.

This is a fascinating passage that reveals a few things. First, in New Zealand, parimutuel betting as a concept had been popularized mainly by people who were importing and selling machines—as a result, New Zealanders thought of parimutuel betting as closely coupled to mechanized calculation ³. The perceived impartiality of the machines (and, as it turns out, the suave manner of its operators) did much to reinforce the idea of parimutuel as a fundamentally more trustworthy form of gambling.

Second, the mention of a "legitimate commission" to the hospital reflects the legal situation in New Zealand at the time. The history of gambling and its legality is a fascinating topic that I am trying very hard not to digress too far into, but many of the fits and starts in development of gambling technology result from widespread recognition of gambling as a vice and subsequent efforts to regulate, restrict, or outright ban it. This might feel like a modern issue, but in the 19th century it was already at a steady boil, and some of the parimutuel industry's emphasis on "fairness" was clearly an effort to maintain legal status. Much the same interplay (between regulation or banning of gambling and claims that mechanization mitigates the harm) is visible in our more traditional gaming industry today (e.g. slot and pachinko), although the online sports books are evidently free of the need to defend themselves.

Despite legal complexities, parimutuel was a hit in New Zealand. In 1908, the Trentham Racecourse had a large tote house with eleven mechanical calculators, presumably operated by eleven tellers with additional clerks to calculate sums and update the manual tote board that made up much of the building's facade. Mechanical calculators were indeed available but expediency, and the chaos of a busy race, meant that much of the odds calculation was done in the minds of the clerks updating the board. There was a margin of error involved, and by newspaper accounts a lot of stress and conflict. Still, this was the dominant model for the first decade of the 20th century: New Zealand racetracks replaced an ad-hoc system of bookmakers and parimutuel carts with large, central tote houses where a single, track-wide pool was maintained. The staff of a busy tote house could be 30 or 40 people, all running around, copying numbers between slates, and shouting updates to tote board's loft. Add another challenge of racetrack betting: the tote houses, even as large as they were, struggled to keep up. One can imagine the impatience of a gambler who was just won it big, running up against a harried teller whose mind has already moved on to the next event.

This problem was all the worse because of a problem clearly illustrated by our modern prediction markets: the matter of closing bets. At a roulette wheel, the croupier ritually declares bets closed before the ball can touch the wheel. Otherwise, the last to place their bets have more information on the outcome. Similarly, in a horse race, any bets placed after the race has begun are better than those placed before. The takeaway is that, under most rules, the tote house had to close bets and complete final odds calculations before the race could begin. A busy tote house meant that races started late—and then later, and later. This was, apparently, a very serious problem in the nineteen-oughts.

Relief would come not quite from New Zealand, but certainly nearby. George Julius was born in England in 1873, moved to New Zealand in 1889, and soon enrolled at Canterbury College to study railway engineering. After his studies, Julius took a job with the Western Australia Government Railways, the start of a career that found him in Sydney in 1907. By this time, he had two sons, and a clear mechanical inclination. While he set up a practice as quite possibly Australia's first consulting engineer, he spent evenings with his sons on a bit of a side hustle: an automatic vote tabulator.

This part of the history has become obscure, and the order of events is not completely clear to me, but it went something like this: while Julius had been working in Western Australia, a friend had complained about the slow process of elections and suggested that some kind of machine could do most of the work. The idea stuck with Julius, and in Sydney, he built a prototype and perhaps even applied for patent on a "foolproof mechanical voting machine." Australia has a long history of interesting and innovative election practices, so Julius's invention might have started a new chapter in the history of election administration had it not been a total failure. The prototype must have worked, but Australian authorities were unimpressed and declined to purchase the system.

It might have moldered among so many other not-quite-revolutions had it not been for yet another friend, this one in Sydney, who was familiar with parimutuel betting and the current state of the art in machines. This friend told Julius that his election machine met similar ends (adding up numbers and displaying the results) and explained the operational details of the large tote houses typical of New Zealand racetracks (but, as far as I can tell, not yet popular in Australia—at least not in Sydney, as Julius later wrote that this was the first he had heard of parimutuel). What followed was another series of late nights, but one that ended with more success: prototype in hand, Julius incorporated as Totalling Mechanisms Ltd and made his first sale.

In 1913, the first Parallel Automatic Totalisator was installed at the Ellerslie Racecourse in Auckland, New Zealand. This machine was called "parallel" because it was a single, large mechanism that all of the parimutuel tellers operated simultaneously. It was "automatic" in that it performed all of the calculations end to end. Not only did it record the tickets sold by each teller, it recorded the totals (in currency) for each horse and a grand total. All of these numbers were presented to bettors by a brand new kind of tote board: a mechanical one, made of chain-driven number wheels showing through windows.

Julius's machine was installed in the existing large tote house at Ellerslie, and fitted with 30 teller stations. Different from today's approach, Ellerslie's parimutuel operated on fixed-value tickets sold in different denominations at different windows. A person wishing to place a 20 shilling bet would queue for one of the 20 shilling tellers, but once at the window could buy any number of tickets on any number of horses. As the teller dispensed the tickets, the tote board above updated in real time.

While quite trivial today, it was a tremendous accomplishment for 1913. Conceptually, it was closely based on the established type of mechanical counter that had been sold, for example, by Hengstler. The difference was one of scale: these wheel counters were huge, as they were directly read by the public. The tote board actually was the machine, in other words, and the number wheels behind it were not just displays but the actual mechanism by which the totals were calculated and stored.

At the back of the tote house, an array of concrete weights hung from chains over sprocket wheels. Gravity pulled the weights, which turned the sprockets, which rotated drive shafts for each horse. Or, at least, tried. The drive shafts were normally blocked by pawls. Each of the parimutuel tellers sat at a station with a set of levers, one for each horse. When a bet was placed, the teller took a ticket for the corresponding horse out of their drawer. They inserted it into their machine and pulled the corresponding lever, which both caused a pattern of holes to be punched into the ticket (validating it as properly issued) and pulled on a long metal wire that ran up through the ceiling into the tote board. That wire pulled on the levers of the clockwork-like escapements that locked the driveshafts, so that each pull of the wire caused the counter for the selected horse to advance by one.

Each of the individual counters was connected by chain to a "shaft adder," basically a series of sprockets coupled to the shaft by one-way clutches. When each individual counter advanced by one, it also pulled the shaft adder along with it, which drove the grand total display at the top of the board. As the machine ran, the weights descended, with those representing the most popular horses falling the fastest. A catwalk rail, up in the tote board, accommodated a worker who carried a crank handle from driveshaft to driveshaft. They would crank the weights back up as needed, and reset all of them to the top between races.

The first Julius machine was big and complicated, with the attendant inertia and friction. My description of its workings is more than just a bit simplified, but most of the complications have less to do with parimutuel than the practical realities of machinery. There were several extra layers of mechanisms like escapements and shaft adders, required so that all 30 of the teller stations could advance the machine with a reasonable degree of effort (although the size of the teller's levers suggests that they still required some heft) and without the machine jamming, miscounting, or otherwise misbehaving when two tellers happened to pull levers at the same time. Given his background in railroad engineering, I suspect that Julius drew from techniques used in railroad interlocking machines, which at that time also made great use of levers that pulled cables to drive chains.

Despite later patent applications, the history of this machine is not as well documented as you might hope. An excellent description from the University of Auckland notes that the actual design of the counters is unknown, particularly how they performed carries. This isn't all that big of a mystery as similar counters (then often called totalisers) existed at the time and Julius must have used one of several known techniques, but still, it is always disappointing to have mysteries in the details of such a notable invention.

This first parallel automatic totalisator operated until 1918, although it was modified and reworked several times in that span. At least early on, the tote house continued to keep manual totals for payout calculations, mistrusting the machine where it really mattered. There are indications that it was out of service during some races, and it may not have been all that reliable in general. Still, it worked. After its first proper operation during a major race, the New Zealand Herald published a report that conveyed its success and its complexity:

A small knowledge of the extraordinary demands that the work of a totalisator imposes upon its parts shows clearly how [the] mechanism might have failed in a dozen ways. But in the meantime alterations have been made, the machine has become more familiar to its operators, and it worked on Saturday and yesterday with fine regularity, and to everybody’s satisfaction....

[An] obstacle of importance was raised by the fact that, in view of the facility with which many investments could be put upon a single horse at a time, the machinery was apt to be run at times so fast that the starting and stopping of the rotating wheels, which, in spite of their light construction, have considerable inertia, set up almost destructive strains and shocks. This trouble has been overcome with an ingenuity altogether admirable. The counter releases a spring-driven gear, which can go as fast as it will irrespective of the motion of the wheel on which the figures are shown to the public, and the wheel simply runs leisurely and smoothly ahead until it overtakes the gear. Even then the wheel, free to make several rapid revolutions, would be difficult to stop at the right place if its speed were not controlled by a governor.

The Ellerslie totalisator might not be called a complete success. Julius charged the racing club £4,000, but spent over £11,000 on construction and modifications. The result was so complex, and required so much maintenance, that the staff of the tote house actually expanded after its installation. Still, the immediate display of totals was a huge draw for bettors, and it succeeded as well in speeding up the closing process. I am always hesitant to apply superlatives, but the Ellerslie totalisator has been called the largest calculating machine to date. I believe it to also represent a major step in the history of displays: a very early, if not first, implementation of a billboard-sized data display that automatically updated in real time. In any case, the machine brought in money. On the totalisator's first full race day, the grand total counter went through £41,514. An estimated 83,000 bets, 10,000 on some single races.

On the back of the Ellerslie machine's success, Julius restyled Totalling Machanisms Ltd as Automatic Totalisators Ltd, or ATL. The First World War caused a great deal of disruption to, well, just about everything, including the racetracks. It halted progress in totalisators for several years. The lull didn't last long, and by 1922 ATL had built improved totalisators for a half dozen New Zealand racetracks. These used a simplified design, benefiting from newer technologies such as electricity.

While the totalisator would remain mechanical for decades to come, the Ellerslie machine is the only one to have been completely mechanical. It seems that Julius began to contemplate electrical designs before he even finished the Ellserslie installation, and all later machines (including an "upgrade" installed at Ellerslie after 1918) used an electric drive system instead of the concrete weights and their crank-wielding attendant. More significantly, though, they also electrified the teller machines: instead of pulling cranks, the tellers pressed buttons, which energized circuits that operated the escapements in the tote board via solenoids. This made it possible, for the first time, to locate tellers for a single parimutuel pool at different locations throughout the racetrack.

The greatest of ATL's electromechanical totalisators was installed at Longchamp, Paris, in 1928. This machine, occupying a building that one could easily mistake for a grand hotel were it not for the numbers behind each window, supported at least 270 tellers (some sources say more). Sales numbers don't seem to have made it to the modern age, but with so many tellers working it must have hit six-figure ticket volumes for popular races. It remained in use until 1973.

Later ATL totalisators were reduced in size and price (truck-mounted portable versions for small racing clubs, for example), and improved in features. We learned early on that some manual tote boards had displayed equivalent odds, while the early Julius machines showed only totals. That limitation wasn't permanent, and ATL machines switched over to estimated payout odds in the 1930s. These took the form of "barometer" displays, or vertical bar graphs, that rose as the odds approached 1/1 (a horse favored to win) and fell down to 100/1 (decidedly long odds). Many of the early ATL systems were upgraded to this later design, and some of the "win/lose barometer" machines saw service, like Longchamp, into the 1970s when computers presented a cheaper alternative.

Totalisator technology became somewhat stagnant after the 1930s, perhaps a result of saturation. Many large markets like the UK had legal restrictions on race betting that prohibited large totalisators (although an oddity of UK law meant that they were allowed for dog races, where the Julius totalisator is best known in that country). In countries with laxer gambling rules, like New Zealand and France, there simply weren't that many large racetracks and by the end of the '30s most of them already had ATL machines that would operate for decades more.

The more innovative designs were seen in portable systems, which were a big market for the many racing clubs that held seasonal racing at different venues. They could not afford to build a large tote house, and didn't have the attendees to keep one busy anyway, but a semi-trailer machine could be towed around with the race days, to be set up and run a small crew in the tradition of carnival rides. These portable systems demanded miniaturization, which was difficult with chain-driven gear shafts but much easier with the relays, and relay logic, that had emerged from the telephone industry. Similarly, the transition to more electrical totalisators made them easier to assemble and disassemble, since it was a matter of hooking up wires rather than aligning sheaves and tensioning cables

The burgeoning telephone industry also meant that relay systems and all manner of electric devices were better understood and easier to manufacture than ever before, inviting new manufacturers into the market. British firm Bell Punch, founded to manufacture machines for validating (punching) railroad tickets, saw ATL's products and responded with their own line of totalisators that were almost entirely electrical. These were particularly popular at smaller tracks, since they were much more compact and easier to install.

In the United States, electrical engineer and race horse breeder Harry L. Straus founded the American Totalisator Company, or AmTote, in 1928. It took several years for AmTote to get off the ground, but their 1933 system at Arlington Park in Illinois established them as the dominant US manufacturer. Their machines were entirely based on telephone relays, and appear to have been the genesis of incandescent numerical displays. Straus's design used a grid of light bulbs, 4 wide by 6 high, with a cabinet of relays that would activate the correct bulbs to show the digits 0 through 9. AmTote numerical displays were an innovation in their own right and found widespread use outside of totalisators (the prices on the original set of The Price is Right are a notable example), but the totalisator business was good, and AmTote installed hundreds of machines around the world.

AmTote also brought other innovations. Despite the historical practice of bets on "win" or "place," ATL machines had only handled a single pool of bets for the race. AmTote's more compact machines could afford the extra logic, almost a duplicate machine, to run a second pool. That meant that parimutuel tellers at AmTote ticket machines could select either a "win" or "place" ticket for each horse, and odds and payouts were calculated appropriately. Instead of preprinted tickets, the AmTote machines operated a mechanical stamp to mark the ticket with the bet information including a "code word" that changed for each race to discourage modification or forging of tickets. A central control panel, in the tote house, "locked" the ticket machines to close bets at the start of each race. Later generations of AmTote machines even used modular (connectorized) electrical wiring and connected all ticket machines to a shared bus, which made portable systems much easier to set up and encouraged owners of fixed systems to add more teller stations at locations convenient to bettors.

The totalisers used in AmTote machines are similar to, and presumably derived from, the Strowger rotary or step-by-step switches used in telephone exchanges. They were fitted with solenoids that would advance them not just in single steps, but in units of 5 or 10 steps as well, for atomic issuance of tickets at different denominations. Much else was taken from the telephone industry: the "brains" of an AmTote machine were installed on relay racks, and extensive diagnostic features were fitted, including controls to produce void "test tickets" at each ticket station and a self-test routine that should advance the counters to known values. AmTote even used a distinctly telco power arrangement: motor-generators converted mains power to DC, which floated the lead-acid batteries that actually powered the machine.

ATL was not to be left out of these innovations, and in particular designed an impressive relay-logic system that fit into a truck called the "Totemobile." A number of these systems were built, and ATL operated them as a service for smaller racetracks. For a commission, ATL would send its own staff to your club's track, and wire the Totemobile to ticket machines installed wherever convenient. They took bets, paid out winnings, and at the end of the day they drove away to the next job.

George Julius lived until 1946, by which time he was Sir George Julius and better known for his influential tenure as chairman of Australia's government R&D organization CSIRO. At CSIRO, he turned towards electronics, as did the company he had founded. After the Second World War, ATL started on a major rework of their product line that introduced relay logic and led to the company's expansion into relay and analog computers. The post-war global economy led to new totalisator installations around the world, and with that surge of new interest came new features. "Quinella" bets, that two horses will place first and second but in either order, were newly popular in the 1950s and ATL designed the first machines that could efficiently manage them and display the odds. Also during the 1950s, ATL designed a new generation of ticket machines that issued tickets as small segments of punched paper tape.

In 1966, ATL designed a totalisator for the New York Racing Association with 550 ticket issuing machines, two primary incandescent light bulb odds displays, and twenty smaller odds displays located around the stands. Installed at the Aqueduct racetrack in NYC, the machine handled over $700 million in a season. Here, for the first time, there were no chains, no cables, no rotary selectors. Instead, the tote house was home to two Honeywell 200 computers. A partnership of ATL, Honeywell, and a software firm called Data Trends had reinvented the totalisator as an early online computer application, the computers polling the ticket machines and updating the displays. Over the following years, ATL would adopt the PDP-8 as the core of their totalisators. From 1970 on, totalisators were almost universally computers.

North London's Harringray Stadium, a greyhound track, operated an ATL mechanical totalisator until its closure in 1987. There was, reportedly, a mechanical totalisator operating in Caracas as late as 2005. In the UK, where many smaller greyhound tracks had totalisators installed, there are several examples that are disused but still standing. At many older racetracks in New Zealand, the UK, and Europe, bets are still placed at teller windows under the enormous tote board that once housed an ATL machine.

George Julius changed the "totalisator" from a room full of people to a machine. His company, ATL, would later change it from a machine to an application. The era of electronic totalisators is its own story, with its own ups and downs and anecdotes, one that I plan to cover in better depth one day. In particular, electronic totalisators were a major impetus of improvements in numerical displays. Display systems designed for totalisators went on to be used in all kinds of applications, from scoreboards to transit destination signage. Starting in the 1980s, off-track betting led to nationwide networked totalisator systems and all of the fascinating asides that that entails. AmTote is still around, a major developer of gambling technology. ATL has scattered to the winds, but makes up part of the modern Light & Wonder (formerly Scientific Games) and Sportech.

Still, I don't find these later machines quite as romantic as Julius's cruder invention. There's something about the fusion of the totalisator as a system, a building, a process, and a machine. The tellers at their windows, pulling on wires to influence the workings of a great machine hidden away in the ceiling.

The lever pulls a long wire cord; the cord releases the counter. The gearing revolves a little, and the counter wheels turn. Sometimes the wheels move at long intervals; and then, as a horse becomes popular, its set of wheels will commence to spin frantically, so that one might imagine the count terribly apt to fail. But the ‘tens’ tot up the units, the ‘hundreds’ go up in their turn; and grand total wheels turn phlegmatic somersaults, and perform mathematical prodigies. (New Zealand Herald)

It is funny, then, that one of the faults of the earliest electronic totalisators was their struggle with arithmetic. But that's a story for later.