It was a classic illustration of the power of numbers.

A year ago, seeking to stifle growing worry about crime, Mayor Rendell trumpeted statistics that, he said, showed the critics were all wrong.

Rendell marched into a Convention Center hearing in September 1997 called by a bipartisan group of state legislators who had been turning up the heat for an overhaul of the Police Department.

While his aides papered the room with charts and statistics, Rendell announced a surprise: Crime in Philadelphia had dropped a remarkable 17 percent during the first six months of 1997.

"Since they're using statistics to make some less-than-favorable comparisons to Philadelphia, we wanted to use them to defend the Philadelphia police," Rendell said afterward.

What Rendell didn't know was that the favorable numbers partly reflected artful massaging of the statistics by police rather than a real decline in crime. The mayor now says that while he still believes crime was down, "it's hard to say that any statistic is accurate at any time."

The episode reveals a little-known aspect of the Philadelphia Police Department's culture of sanitizing statistics. Inquirer stories on downgrading crime have described how police -- contrary to department policy and FBI rules -- swept major offenses off the books or reclassified them as minor incidents.

The 17 percent drop Rendell reported stemmed partly from something different -- an effort by Police Commissioner Richard Neal, then at the end of his tenure, to follow FBI rules to the letter, so that police reported as few major offenses as the bureau's "green book," a volume of crime-coding guidelines, permitted.

As part of that effort, police in 1997 were using new counting methods that reduced the crime tally in two important categories: auto theft and property crime.

The new methods were sanctioned by the FBI. The problem was that Philadelphia police used the new counting method for the first half of 1997, then made comparisons to the 1996 figures without adjusting the previous year's figures, which had been calculated a different way.

These were the numbers Rendell brought with him to the Convention Center that day.

Police did not tell the mayor or the public about the new counting techniques. So neither Rendell nor the 250 people at the Convention Center forum knew that the 17 percent reduction he reported was based on a flawed comparison.

Police changed their tallying methods in two key ways, police commanders say:

First, in totaling car thefts, the department excluded hundreds of cases in which autos had been stolen by people who had signed them out from rental agencies. Those incidents were classified as credit-card fraud instead.

The shift was significant, because auto theft is one of the seven "Part I" major crimes that go into a city's crime rate and form the basis of the FBI's Uniform Crime Report, the agency's nationwide survey of crime in America. Credit-card fraud is not a Part I offense.

The other change was to bundle separate instances of car break-ins and other property crimes into single episodes if there was reason to believe the incidents were committed by the same person. So, if four cars were broken into on the same block, police would enter that on the books as one offense rather than four.

Together, the statistical changes reduced the city's count by as many as 2,500 major crimes for the first half of 1997, compared with the same period a year earlier. That accounted for about a third of the 17 percent drop touted by Rendell.

Rendell, in an interview this month, said he had not known police were using the new counting methods.

But he added that robberies, burglaries and other crimes not affected by the new methods had dropped in the first half of 1997, suggesting that the new techniques "didn't skew the entire package."

Still, the mayor said he was no longer prepared to say that crime had dropped 17 percent.

In a brief interview, Neal said Philadelphia had been playing by the rules in adopting the new counting methods, which are still in use.

"That's nothing more than what the [ Uniform Crime Report ] permits you to do," he said.

Neal, now a security consultant for Drexel University, declined to discuss the unfolding controversy over police coding.

"I am no longer in the department," said Neal, who was commissioner from 1992 until March 1998. "I have nothing to do with what we're submitting" to the FBI.

Neal's emphasis on taking maximum advantage of FBI rules began in late 1996, after the FBI's annual city-by-city crime report showed Philadelphia lagging other cities in cutting crime.

That year, while major crime for big cities was down 6 percent, in Philadelphia it was up 8 percent, the FBI said.

Neal responded with a drive to retrain the officers who classify crime -- about 800 officers in a force of 7,000.

For months, detectives, captains and numbers-keepers on captains' staffs trooped into coding classes at the Department's Advanced Training unit at the Progress Plaza shopping center on Broad Street. FBI experts helped train the police instructors.

The lesson dealt in part with FBI rules allowing the bundling of certain property crimes if they took place in roughly the same location and at the same time, under the theory that they were by the same criminal.

The classic example is a thief who moves down a block, breaks into car after car, looting each of them.

Such calculations are not always popular with crime victims. In early 1997, Philadelphia police coded as a single crime a dozen instances in which car windows were shattered with BB shots in a four-square-block area in the city's Wissinoming section.

One of the vandalized cars belonged to Richard B. Costello, president of the Fraternal Order of Police. He said teenagers in the neighborhood had been competing to shoot out windows.

"Crime goes down if you can take 12 offenses and make them one," he said afterward. "These figures have been made to dance more than the Rockettes."

An Inquirer review of crime data shows the impact of the shift in calculations with regard to car break-ins. Police reported about 2,400 fewer break-ins in the first half of 1997 than in the same period the year before -- a drop of 10 percent.

Not every police department consolidates crimes this way, even if the FBI allows it.

San Diego police are uncomfortable assuming that one criminal was responsible for a rash of crimes, even if they occurred in the same area, said Pat Drummy, commander of the department's crime-analysis section.

"How do you measure it?" Drummy said. "Do you go by one block or two blocks? Everything that takes place in five minutes or 10 minutes? You just get into trouble."

In his city, Drummy said, "if nine cars are broken into, that's nine victims" -- and nine crimes.