Twenty-Eight Metres of Security: Mexico City Metro, the 50% Drop, and the 2026 World Cup Test at Azteca

On 7 September 2026, at 16:40, I stood in the eastern corridor of Tasqueña station — the southern terminus of Mexico City Metro Line 2, where tens of thousands of passengers change trains every day to continue toward Xochimilco or turn back toward the centre. Two Metro police officers stood exactly twelve metres apart, their eyes fixed on opposite directions. The distance between them is what I will spend most of this article measuring. Three weeks later, on 28 September 2026, the Sistema de Transporte Colectivo (STC) released a figure that made several sports newsrooms in Brazil reopen their files: robberies with violence (robo con violencia) across the Mexico City Metro system were down 50% year on year. Between 1 January and 7 September 2026, only 12 investigation files (carpetas de investigación) were opened, against 24 in the same period of 2026. Four consecutive months — March, April, May, June — recorded zero files. I read that release three times. The first time as a sports reporter. The second time as a data analyst. The third time as a 44-year-old woman who has sat through enough tactical meetings to know when a number is being told as a success story and when it actually is one. Mexico City, Guadalajara and Monterrey are Mexico's three host cities for the 2026 World Cup — the first 48-team edition, running from 11 June to 19 July 2026, co-hosted by the United States, Canada and Mexico. Estadio Azteca was assigned the opening match and several knockout fixtures. After its renovation, Azteca is the largest stadium across the three host nations. What does that mean for the Mexico City Metro? The STC operates 12 lines, roughly 226 km of track, and carries over 4 million passengers a day under normal conditions — the highest figure in Latin America and among the highest in the world. On matchdays, Line 2 (Tasqueña–Cuatro Caminos) and Line 3 (Indios Verdes–Universidad) carry direct pressure because both have stations near spectator gathering zones. The Estadio Azteca station on Line 2 sits a few hundred metres on foot from the stadium gates. This is where a football analyst has to read differently. The stadium is not where the match begins. The match begins when the spectator leaves home, and for most Mexico City spectators, it begins at the Metro turnstile. Around mid-2026, the Mexico City government launched Operativo Quetzalcóatl — a security strategy focused on the Metro, coordinating Metro Police (Policía del Metro) with supporting units. According to information the STC provided to the press, roughly 5,800 Metro police officers were deployed in patrol and control operations under this strategy. Five thousand eight hundred people. I wrote that number in my notebook, next to 12 files and 24 files. Three numbers, three different layers of meaning. Start with the unit of measurement, because this is where many sports articles go wrong. Robo con violencia in the Mexican justice system does not mean every act of property taking. It is a specific offence requiring an element of force or direct threat of force against the victim, and it must be recorded as a carpeta de investigación — an official investigation file opened by the prosecutor's office. A bag snatched without a threat does not fall into this category. A phone lifted from a pocket does not fall into this category. An argument that ends with someone being pushed down to have their wallet taken does. With that understood, return to the numbers. Twelve against twenty-four across roughly 250 days, from 1 January to 7 September. Twelve files spread over more than eight months means around 1.5 files per month in 2026; the equivalent figure for 2026 was around three files per month. This is where I must state clearly what op-eds tend to skip: at a frequency below five events per month, every statistical conclusion sits in the noise band. If one month has 1 file and the next has 0, a 100% reduction appears — but its practical meaning is close to zero. If one month has 3 files and the next has 1, a 66% drop is arithmetically correct but says nothing about trend. Statisticians call this fluctuation around a low base. At a high base — say, 3,000 monthly files falling to 1,500 — a 50% drop carries structural meaning. At a low base, the same percentage may be no more than the difference between two coin tosses. That is why I did not make 50% my argument. I made 12. Four consecutive zero-file months — March, April, May, June — are the more interesting point. The probability of four straight months landing at zero, assuming independent months with a mean of 1.5 events, is not high. It suggests one of two possibilities: either Operativo Quetzalcóatl genuinely produced a period effect, one that needs independent verification; or the recording of files changed, for example by shifting some conduct into a lighter offence category to relieve the prosecutor's caseload. I do not have enough data to conclude which is true. But I have enough to say that a sports article should not turn 50% into a headline without independent corroboration. Next comes the 5,800 variable. This is the number I tracked longest, because it is the only variable in the release that can be quantified as density. If 5,800 officers are spread across 12 lines, 226 km and more than 190 stations, average density would be roughly 25 to 26 officers per station, before accounting for major interchange stations that are far busier. The problem is that such a figure is never distributed evenly. In every studied urban security system, density concentrates at central stations, terminal stations, interchanges and stations near tourist zones — precisely because those are where crowds gather. Estadio Azteca station, when there is no match, sits in the low-priority group. On a matchday, it becomes the highest-priority group. This is where I need a comparison that people who only watch football on television are not used to: security density in the Metro operates much like a team's pressing structure. Pressure is not standing close to the ball. Pressure is standing in the right place to cut off the next pass. A good pressing team does not put all 11 players around the ball; it places players in defensive layers, cutting the third and fourth passing options, forcing the opponent backwards. PPDA — passes per defensive action — is the standard measure, counting how many opponent passes are completed before each defensive action. Metro police work the same way. Two officers at Tasqueña station will not reduce robbery if nobody stands at Pino Suárez station, where a thief can hop a train and switch lines. Three officers at a terminus help nothing if thieves choose crowded peak hours at mid-line stations. Standing in the right place matters more than standing in numbers. The Operativo Quetzalcóatl data released by the STC do not disclose the deployment structure of the 5,800 officers. They disclose the total and the result — fewer files. This is the same situation as a club announcing it has improved its defence while offering only goals conceded, no PPDA, no expected goals against, no description of the back line's positioning. The result does not lie. The result simply does not tell enough. I have followed various national leagues for more than twenty years, and one of the lessons I hold tightest comes from the 2026 season in Brazil. Matchday 23 of the Brazilian national league, Corinthians 1-0 Santos. Corinthians midfielder Maycon, number 8, dropped twelve metres deeper than his average across the previous five matches. I wrote a short blog post about how he stretched Santos's midfield and opened the space for Jadson to score in the 67th minute. A male commentator laughed: women only see handsome players. But the Santos assistant coach, Mr Cuca, sent a message confirming the analysis was correct and invited me to a tactical meeting. The lesson: people do not argue about results, they argue about mechanisms. To defend a judgement, you must be able to measure the mechanism. Back to the Mexico City Metro. What is the mechanism? If Operativo Quetzalcóatl genuinely worked, the mechanism would have at least four layers. Layer one: deploying personnel at stations with high passenger density and a history of incidents. Layer two: improving cameras and inter-station communications. Layer three: coordinating rapid response between station police and prosecutors to shorten file-opening times. Layer four: social impact — constant presence reducing the opportunity for offenders to act. The release only describes layer one. The other three layers do not appear in the published data. Without those three layers, the 50% effect could come from layer one plus another uncontrolled variable: lower ridership, changing reporting habits, or a change in offence classification. In urban security data analysis, these are the three most common confounding variables, and none of them appears in the release. There is another detail I want to bring in because it bears directly on the 2026 World Cup. Large sporting events create two kinds of pressure on an urban security system: crowd-flow pressure and crime-mix pressure. Crowd flow is easy to grasp — tens of thousands of people on the same Metro line in the same hour. Crime mix is subtler: travelling spectators carry cash, phones and cameras, are often unfamiliar with routes, often move in groups, and often pay less attention to their surroundings. That is an ideal victim profile for violent property crime. I was in the Moscow press room during the 2026 World Cup as one of four female analysts. Before France 4-3 Argentina, I predicted that France's pressing line would exploit the gap between Argentina's defenders and midfielders. Griezmann's 13th-minute opener followed exactly that pattern. My article was republished by L'Équipe. A male colleague said: she was just lucky. I collected data from 12 group-stage matches, proving France's pressing model was consistent. He said nothing more. I tell this story because it relates to how I read the STC release. Data from one match cannot prove a structure. Data from a season can. Data from an eight-month window of a Metro system that has operated for decades, in a World Cup year, says nothing yet about long-term trends. Here is the point I found when I read the release a third time: the data period ends on 7 September 2026. But the 2026 World Cup ran from 11 June to 19 July 2026. In other words, the reporting window covers the entire World Cup, except for a few days in early June before the kick-off. That is more important than the 50% figure. It means the 12 files in the reporting period reflect September, the World Cup month of July at Azteca, and the June build-up to the opening match. If, during a window containing a World Cup, the file count still sits between zero and two per month, that is a genuinely positive signal — positive in the structural sense, not the numerical sense. But the STC release does not separate World Cup data from ordinary annual data. That is its biggest weakness. In sports-event analysis, the event period must be separated from the baseline period. A team performing well in a stretch when its main striker has returned cannot be evaluated on the same scale as one without him. A Metro system during a World Cup cannot be compared directly with a period without a World Cup. If I had month-by-month data — and I do not, because the STC publishes only the total — I would split it three ways: January to May, with no major sporting event; June to July, with the World Cup; August to 7 September, post-World Cup. Those three periods have entirely different passenger profiles. Merging them into a single figure of 12 destroys information. Here I return to the line I still use when analysing matches with small samples: luck repeated twelve times is called a model; repeated once, it is not called anything at all. With the Mexico City Metro, we have one instance. One instance in which the World Cup variable has not been separated out. There is a counter-intuitive reading I think should be offered, even if it displeases newsrooms that want a positive security story. If Operativo Quetzalcóatl were genuinely effective and file counts fell from 24 to 12, that does not necessarily mean the Mexico City Metro became safer for 2026 World Cup spectators. Crime is not a static block; it shifts. When police density rises at some stations, offenders may move to other stations, other hours, other lines, or shift into lighter offence categories that are not counted at the same threshold. Crime displacement is widely documented in urban criminology. In football terms, it is like a strong defensive side forcing the opponent from wide attacks into central attacks — the number of chances may fall while the conversion probability per chance may rise. I do not have enough data to assert this happened in Mexico City. I have enough to say it has not been excluded from the analysis. The only way to exclude it is to compare the distribution of files by station, by hour and by line — data the STC does not publish. Anyone reading the 50% figure as a complete conclusion is working with incomplete data. One further point deserves careful handling: investigation-file statistics depend on victims' reporting behaviour. When a system adds police and adds a communications campaign, reporting rates usually rise, not fall. If reporting rates rise while file counts fall, the structural signal is stronger. If reporting rates fall for some reason — victims not trusting the process, or lengthy resolution times — then a fall in files does not reflect a fall in crime. The release does not say which is true. This is where I want to recall something learned from more than two decades in the trade: gaps do not lie. The gap here is the gap between what was published and what needs to be published — between 12 files and the distribution of files, between 5,800 officers and their positions across each time window. Looking into that gap, I see a different story that has not yet been told. It is also worth adding that Mexico City's municipal politics in 2026-2026 had clear reasons to want the Metro security story told positively. The World Cup was the largest image-promotion event the city has had in decades. A release announcing a 50% fall in Metro crime, timed just after the World Cup, serves two purposes at once: reassuring prospective visitors and shaping favourable opinion ahead of upcoming local elections. That is an observation about the media context, not an accusation about the data. But an analyst must distinguish the two. Back to the purely sporting angle. From a World Cup logistics standpoint, the Mexico City Metro data has three direct consequences. First, if Metro security density genuinely rose during the World Cup window, clubs and organisers could consider Metro-based travel for some spectator groups instead of relying solely on dedicated buses — reducing surface traffic pressure. Second, any security incident in the Metro on a matchday could be amplified by international media, producing a psychological effect larger than its actual scale. Third, if the August and early September data show files rising again, that would signal the World Cup effect was temporary and the system needs to adjust its security model for recurring sporting events. The Mexico City Metro match in 2026 is not over. The World Cup has just passed, leaving a short but valuable sample — the first time in years that the Mexico City Metro carried a recurring, structured flow of international spectators daily for more than a month. If the STC publishes data broken down by station and by time window for the World Cup period, we will have a sample large enough to start talking about structure rather than about a number. If it does not, we will keep reading 50% as a reference point that does not hold — and keep waiting for next season to learn whether Operativo Quetzalcóatl is a model, or merely one lucky March multiplied into four. We do not need to believe the number. We only need the patience to measure.