Martin Fayulu has reason to thank the voting machines of the Congo that he once feared

The people of the Democratic Republic of the Congo may have just experienced one of the biggest electoral frauds in recent history.

Last week the country's electoral commission announced that the opposition leader Felix Tshisekedi is the winner of a historic election marking the end of 18 years of President Joseph Kabila. But an analysis of the Financial Times of two separate collections of voting data shows that it has not won.

At 3 am in Kinshasa on January 10, the commission told the Congolese with wide-eyed eyes that the prognostics of the main candidates were as follows: Tshisekedi had 38.6%, opponent opponent opponent Martin Fayulu 34.8% and the party in command took Emmanuel Shadary 23.8%.

According to the electoral data seen by the FT, which represent 86% of the total votes cast throughout the country, the result was very different. Mr. Fayulu got 59.4% of the votes, the data show, while Tshisekedi should have finished a second distant with 19% and Mr. Shadary who got 18%.

The FT compared the figures with a separate set of voting results manually collected by the country's respected Catholic Church, which ran the largest election observation mission. The two sets are almost exactly related and would have been almost impossible to fake, experts say.

The group of Catholic bishops, known as CENCO, was explicit by the vote. Five days after the election, he said that the voters expressed "a clear choice at the polls" and called on the electoral commission to report the result accurately. In private he told the diplomats that Mr. Fayulu was in an unassailable guide and after Mr. Tshisekedi was declared victorious, he told the UN Security Council that the commission's result did not match his own findings.

These percentages, which represent 43% of the turnout and hand-picked from 28,733 election points, were seen by the FT and show that Mr. Fayulu obtained 62.8% of this vote.

In contrast, leaked data figures are electronic records from 62,716 election machines across the country and are said to have been obtained from the electoral commission's central database before the results were announced.

The electoral commission introduced untested electronic voting machines in December for the first time, claiming that the devices would help reduce costs. The machines printed ballot papers that were counted by officials on election night, but also stored electronic counts.

During the election campaign, Fayulu warned that the machines would be used to make the vote. In the end, the devices may have given the opposition leader the transparency he requested.

An FT analysis of the electronic narratives in the leaked document shows an almost perfect match with the Church's partial results – with a correlation coefficient ranging from 0.976 to 0.999 for each of the three major candidates (1 represents a perfect match).

The electoral commission denied that its results were fraudulent. Barnabé Kikaya Bin Karubi, chief diplomatic adviser to Kabila, said that it will be up to the constitutional court to decide on the validity of the elections and refuse to comment on any potential fraud. Gilbert Kankonde Nkashama, a spokesman for Tshisekedi, said it was impossible for Mr. Fayulu to have won the election and questioned the independence of the Catholic Church.

Fayulu is trying to overthrow the result in the constitutional court, although the court's impartiality has been questioned.

The Congo, a former Belgian colony of 80 million people, held only four elections since independence in 1960 and has never had a transfer of power through the polls. Kabila should have resigned in 2016 but the elections were postponed until the street protests and regional pressure forced him to hold the vote.

Kabila's ruling coalition had tried to keep control of the presidency through Emmanuel Shadary, his chosen successor. Fayulu supporters said that when the voters failed to get out to Mr. Shadary in sufficient numbers – he finished third – the electoral commission was tasked with installing Mr. Tshisekedi instead.

According to the results of the FT, Fayulu received more than 9.3 million votes, 3 million more than the announced electoral commission, and won in 19 of the 26 provinces of the Congo, including the capital Kinshasa and the heavily populated eastern provinces of the North Kivu and South Kivu.

On the contrary, Tshisekedi has obtained 3 million votes, the data indicate, 4.1 million less than the electoral commission, while Mr. Shadary obtained 2.9 million votes, or 1.5 million less than the published count.

The data show 15.7 million votes out of the 18.3 million expressed on election day, but the missing votes could not have led to a different winner. Malfunctions in voting machines meant that not all voting feedback was transmitted to the central database, said the person with knowledge of the database.

The file takes the form of a comma-separated value spreadsheet, a format used by many software packages to store tabular data and runs on more than 49,000 lines. Each polling station is identified by a unique six-digit number.

The FT analysis also found no significant evidence that the data diverged from Benford's law, a statistical test commonly used in fields such as forensic accounting to identify fraudulent data.