On Feb. 21, Aqueduct’s racetrack was shrouded in a dense fog. That didn’t stop a 3-year-old maiden, Ground Control, from floating over the muddy track in a most peculiar way and apparently setting a six-furlong world record when the time for the fifth race was put on the board at 1:06.07. “Ladies and gentlemen, I don’t believe those times are correct,” said an incredulous John Imbriale, the track announcer, just after the horses crossed the finish line. Imbriale was right. Minutes later, the time was corrected, adjusted up almost seven seconds to a rather pedestrian 1:12.89. The next day at Gulfstream Park in south Florida, Normandy Invasion, a highly regarded 4-year-old making his first start since a fourth-place finish in the 2013 Kentucky Derby, won an allowance race by nearly eight lengths, posting a final time of 1:34.78. Soon after, though, the time was adjusted, but unlike the case of Ground Control, the adjustment worked in Normandy Invasion’s favor. The result: a track record of 1:33.13. The two timing mishaps, coming just 24 hours apart at two of the most prestigious tracks in the country, have cast a spotlight on how races are timed in the modern era of racetracks. Though mistiming incidents appear to be rare among the roughly 45,000 races run each year at U.S. tracks, when the incidents do occur, many influential people take notice, and they usually are not shy about making their opinions known. While racetrack veterans often caution clock-watchers that “time only matters in jail,” the final times and internal fractions of races are used as the raw data in the construction of widely used speed figures, pace ratings, and track variants, critical tools for many handicappers (and the livelihood for many past-performance data producers). So, if any part of a race time is incorrect, the error cascades through a long list of calculations designed to give horseplayers accurate metrics to compare performances among horses and tracks, making the recording of both final times and internal fractions incredibly important to the sport’s lifeblood – bettors. Currently, only three companies provide official time-keeping data for U.S. racetracks. Two have been doing the job for decades, relying on beams and sensors to determine when horses have passed the places on a track corresponding to the internal fractions for a race at that distance. The other, the relative newcomer Trakus, relies on an array of local antennas and transmitters forming a network analogous to a GPS system to determine when a horse reaches a point on the track to register a fractional time. There is no consensus about whether one system is more accurate or reliable than the other, though individual operators of the systems claim that their systems are better at certain aspects of timing. In the case of the two mistiming incidents in late February, the Aqueduct race was timed by a beam-based system run by American Teletimer, the largest of the timing companies, while the Gulfstream race was timed by Trakus, the official timer for two U.S. tracks and a handful of tracks overseas. The beam-based systems used by American Teletimer and the other U.S. operator, Teleview, are relatively straightforward. As the horses make their way around the track, the first horse to reach each point of call breaks a beam running straight across the track to a sensor on the other side, triggering the timing system. The sensors are approximately 4 1/2 feet high, with the beam expanding to approximately a vertical foot at the crest of the track, so that the beam “is aiming for the nose,” said Joel Rosenzweig, the president and chief executive of American Teletimer. A time is recorded at each point of call, including the finish, where the final time is verified by the time stamp on the picture generated by the photo-finish camera. There are some nuances. For example, under the system, the sensors that will be triggered during the running of the race are active only for approximately 20 seconds. That limits the amount of time that the sensors can be triggered by objects other than the horses in the race, such as outriders, ambulances, birds, a player’s program hanging over the rail, or even a wind-tossed umbrella (it has happened). The most significant nuance – and the biggest source of controversy surrounding the timing of races – is when the timing system actually starts. In North American racing, for reasons that are not entirely clear, almost every race is run at a distance greater than the official distance listed in the program. The extra distance between where the starting gate is placed and the official start of the race is called the “run-up.” (In two recent columns, Andrew Beyer, the speed-figure publisher and handicapping expert, has begun arguing for the elimination of run-ups from North American racing.) Run-ups vary widely from distance to distance, from course to course, and from track to track, and there is no rule as to what a run-up should be at any specific distance for all tracks. As a result, run-ups at U.S. tracks are at a hodgepodge of distances. For example, the run-up for a one-mile race at Santa Anita is 172 feet, according to charts; the run-up for a one-mile race at Gulfstream is only five feet; and the run-up for a race going one mile and 40 yards at Tampa Bay Downs is 24 feet, which means horses in a one-mile race at Santa Anita run farther than horses in a race at a mile and 40 yards at Tampa Bay Downs. Different run-up distances are one of the reasons why some tracks seem to consistently post faster or slower times than other tracks, especially at shorter distances. If horses have a long run-up, they are more likely to be at full speed when the timing of the race starts than if there is a short run-up. All timing systems disregard the run-up, whatever the distance, since it is not part of the official distance of the race. As to why run-up distances differ, it mostly has to do with specific track configurations that limit where starting gates can be placed. To determine when the timer should start, the operator of a beam-based system works backward from the finish line to identify the sensor that corresponds to the listed distance of the race. That sensor is designated the starting beam, and when that beam is broken, the timing of the race starts. If there is not a sensor for the official start, then the operator manually starts the timer when the first horse reaches the official starting point of the race. Ron Couturier, the chief operating officer of American Teletimer who has worked for the company for 40 years, said he believes the run-up is a holdover from the days when horse races were run on dusty American streets, well over a century ago. In those days, the horses were usually given a running start, eliminating any starting-line vagaries or misbehavior that might obscure the horse with the most raw speed. When the starting gate was introduced in 1939 – and was nearly universally adopted by Thoroughbred tracks within a year – tracks were still accustomed to granting a run-up, so the practice stuck. “It all began in the cowboy days, when the distance was measured by the place where the guy dropped the hat,” Couturier said. Nevertheless, Quarter Horse races are timed from the break, with no run-up. There also are timing complications regarding turf courses, where rails are often set up in different lanes to avoid damage caused by the overuse of a portion of the course. By simple geometry, moving a rail from the inside toward the outer rail lengthens the course because horses will now have to cover more ground while running around the turn and while getting from the start of the turn to the end of it. As a result, beam-based systems have to install multiple sets of sensors in order to have a beam at each possible point of call based on the lanes that can be used. According to Rosenzweig, Del Mar needed 120 sensors to cover all of its course configurations, including the main track, and this was prior to the track’s turf course being ripped out for a widening project. According to officials, every U.S. track has a set of sensors for every turf course configuration, with one exception, Gulfstream, where the turf course has as many as 10 different lanes. Although Trakus is now the official timer at Gulfstream, the previous timer, Teleview, adopted a unique method to estimate internal fractions for the configurations that lacked a set of sensors. If a course configuration did not line up with an existing set of sensors, Teleview would time the race using the set of sensors closest to the course configuration, but the official time would not be posted until a computer calculated the estimated difference between when an actual beam was broken and when a beam at the proper location would have been broken had it been in the right place. The recalculation happened in real time, according to Nate Roberts, the director of timing and photo finish for Teleview, with no discernible delay on the posting of fractions during the race. The computer performed the calculation automatically, Roberts said, with no human tinkering. “It was all built into the system,” Roberts said. When Trakus took over as the official timer at Gulfstream in December 2012, it ushered in a new era of timing races, at least in the United States. Originally marketed primarily as a charting and real-time visual-enhancement tool, Trakus uses an array of antennas and transmitters placed in horses’ saddlecloths to determine a horse’s position on the racetrack at all times during a race. The system is in place at many prominent racetracks, including Keeneland, Santa Anita, Del Mar, Saratoga, Belmont, and Woodbine, and it is familiar to many racing fans through the “chicklets” the system generates on the bottom of a live race feed. To time races, Trakus generates a “geometry” of a racing surface that serves as a template for the specific surface’s different distances and configurations. To determine when to begin timing the race, minus the run-up, the system works backward from the finish line to the official distance, and it fills in the locations for the internal fractions from there. It then places imaginary lines on the template to determine when the timer should start and when internal fractions should be registered as the horses reach the lines. For highly variable turf courses, Trakus generates an individual “geometry” for every course configuration. The system requires the operator to enter in the correct configuration prior to the start of the race in order to generate accurate times. For Gulfstream, there are 10 different configurations, officials of Trakus said. Michael Ciacciarelli, the chief operating officer of Trakus, said the most common error leading to the mistiming of a race is when the wrong configuration of a turf course is entered into the system by the operator or miscommunicated to the operator. However, the correct time can be generated after the race by using the live-race data to rerun the race over the correct geometry template, added Pat Cummings, the business manager of Trakus. “We can go in and say, ‘We knew this was wrong,’ and we can rerun the race as if it was in Lane 2 instead of Lane 3, and get the right time,” Cummings said. Trakus also has run into problems when a horse’s position is not continuously tracked during a race because of signal loss, especially if that horse is leading at one of the internal fractions. In those cases, Trakus uses the data generated by other horses to extrapolate the time or times of the lost horse. “Because we’re tracking all the horses in the race, rather than just the leader, we have total situational awareness, so to speak,” Ciacciarelli said. In the case of the Normandy Invasion race, Trakus officials said the race timer started too soon due to the short run-up distance (five feet) and because of a “tricky spot” on the Gulfstream property for antenna coverage. Because the antenna coverage is spotty, the locations of the horses when they are at or near the top of the chute can “jump around,” Ciacciarelli said, in the same way as the location of a GPS device drifts from place to place in an area with spotty satellite coverage, such as in woods. So, when the system was armed just after all the horses were loaded into the gate, one of those “jumps” of a horse’s location triggered the timer five feet away at the imaginary line designating the start of the race, approximately one second before the horses actually reached the line, Ciacciarelli said. It was not the first problem Trakus has had with the one-mile chute at Gulfstream. “That distance acts up on us,” Ciacciarelli said. The company has not yet figured out how to improve coverage in that location due to property limitations. Operators of beam-based systems say the overwhelming majority of timing problems occur when a sensor is blocked or is triggered by something other than a horse. For blocked sensors, there are fail-safes, such as alarms alerting the operator that sensors are not registering. In those cases, it’s most often an outrider, photographer, or ambulance blocking the sensor, officials said. “Most of the track personnel are very good at that, but if they do [block a sensor], you get the stewards on the radio, and you say, ‘Hey, that guy can’t stand at that pole or location,’ and it gets fixed pretty quickly,” Rosenzweig said. If the sensor is triggered prior to the horses reaching an activated beam, the operator can reset the system if he notices that the sensor was triggered too early. If not, the operator relies on a distinctly analog backup: a stopwatch. For every race, the timing-system operator hand-times the race from the official start through the internal fractions to the finish, and if the system does not record a time for any reason, the hand-timed figures are used. Hand-timing off the replay of the race is used to verify the times. In the case of the Ground Control race at Aqueduct, where the official time was adjusted upward by almost seven seconds from the first posted time, the timing system was activated earlier than normal because the operator could not see the field through the thick fog, Rosenzweig said. The start timer was then triggered too early by an object the operator also could not see, leading to a cascade of erroneous data for the internal fractions and final time. The error was corrected relatively quickly, Rosenzweig said, calling it a “very rare error caused by severe weather conditions.” Trakus also was timing the Aqueduct race, and its fractional times were almost identical to the revised times, although its final time for the race was a half-second quicker than the official time. Trakus officials said that at tracks where the company is generating race data but is not the official timer, track officials will often ask the company to provide its times as a “reference” if the official timing system had a gap or produced a suspicious time. The final backup is Equibase, the official record-keeper for North American racetracks. Equibase receives its data from field operators, and when the data are entered into the system, a computer program scours the official times, including internal fractions, for “any figure that is outside of an acceptable range,” said Tom Roentz, Equibase’s track and field manager. If a time is flagged, Equibase asks the track to verify the figure, with the time omitted from the official chart until the time is changed or verified. Roentz said the system rarely flags times as being unacceptable, but if it does, it means that the time is wrong, since the system is set up primarily to flag data that are “theoretically impossible,” such as a sub-20-second quarter-mile. So, when Equibase asks a track to re-time a race, tracks almost always comply. “If it’s something that our system says is a wrong time, then I have a hard time believing that a track would stand by something that is so obviously ludicrous,” Roentz said. Correction: An earlier version of this article misstated Normandy Invasion's adjusted running time for a mile when he set a track record at Gulfstream Park. It is 1:33.13, not 1:33.71.