By Jason Clayworth

America's popular and lucrative lottery drawings, in which computers randomly select numbers that turn a lucky few into instant millionaires, may not be as random as they seem.

In dozens of the games across the United States, identical winning numbers have been generated within weeks or months of each other — sometimes in consecutive drawings, a Des Moines Register investigation shows.

Lottery officials, even some who have previously acknowledged concern with the national lottery system, contend the repeated numbers are nothing more than chance.

"While such repeats are rare and uncommon, there is no reason to suspect these numbers were not drawn reasonably," said Patricia Mayers, a spokeswoman for the Wisconsin Lottery, which had several duplicate draws over more than a decade.

But critics of random-number drawings contend that the identical draws, coupled with records identifying problems in several states with "random number" computer generators, demonstrate that the fairness and integrity of the nation's $80.5 billion annual lottery system are compromised.

The problems are aggravated by the lack of national oversight, those critics contend.

"Lottery directors and state legislators are running these rigged games with blinders on," said Les Bernal, director of Stop Predatory Gambling, based in Washington, D.C. "They're shielding their eyes from the dishonesty of these games."

Math experts who have researched lottery drawings told the Register it's difficult to truly know whether the identical draws indicate a serious problem without an in-depth inspection of the lottery software, but they say the draws deserve a closer look.

"It suggests that there are outliers, and it suggests it's worthy of investigation," said math professor David Austin, of Grand Valley State University in Michigan. Austin has researched and written about the difficulties associated with creating number generators.

Already, several states, including Iowa, Arizona, Connecticut and South Carolina, have identified problems with random number-generated lotteries, and some have ended games because of the problems.

Professors review drawings, and find outliers

Using data of winning drawings in 37 states collected by the website Lottery Post (www.lotterypost.com), the Register identified more than 100 drawings over the past 25 years where the same game generated identical winning numbers within 365 days of each other.

(The Register reviewed only those lottery games that select at least five numbers, because those games generally have a smaller probability of having the same numbers drawn twice.)

In six instances, the same winning numbers were generated in consecutive drawings in the same game in Arizona, Rhode Island and Colorado, the Register found.

Austin, the Grand Valley State math professor, reviewed some of the same-number draws the Register identified as possible anomalies in Wisconsin and West Virginia. 

Austin — who ran more than 10,000 simulations as part of his review — determined that although the duplicate drawings in West Virginia seemed consistent with statistical odds, several examples in Wisconsin appeared to be outliers "worthy of investigation."

"Convention would say that the Wisconsin Lottery results are statistically significant and even close to being highly significant," Austin said.

Robert Molzon, a retired math professor at the University of Kentucky, concurred.

It's impossible to directly test whether lottery computer-generated numbers are truly random without gaining access to the systems, Molzon said.

However, it is simple to compute probabilities that a given number of "doubles" (identical numbers) will appear over a fixed time period as a way to assess whether a problem may exist, he said.

Like Austin, Molzon — in reviewing sets of winning strings of lottery numbers called multiple times in Wisconsin and West Virginia over a multiyear period — concluded that outliers exist in some of the Wisconsin drawings identified in the Register's investigation.

"My conclusion is that it is highly unlikely that the chosen numbers in either of these lotteries are truly random," Molzon said.

A scammer spots the problem

The Register launched an eight-month investigation into lottery draws with identical numbers after the conviction and sentencing last year of Eddie Tipton, a former Multi-State Lottery Association security worker in Urbandale who admitted masterminding the largest lottery fraud in U.S. history.

Tipton — in a confession following his 2017 lottery fraud conviction — told investigators that in 2006 he had warned his employer that computer software he had written to randomly draw numbers had an alarming flaw: The draws weren't truly random.

He was tipped off to the problem after two Wisconsin SuperCash! games drew the same six numbers twice within 10 days in 2006.

The odds of that happening: 1 in 326,262 — akin to flipping a coin and having it come up heads 18 times in a row — Wisconsin Lottery mathematician JonMichael Rasmus told the Associated Press in 2006.

Multi-State Lottery is an umbrella gaming organization that is owned and operated by 36 member lotteries, including the Iowa Lottery. Many of its members use its random number-generating software.

Neither the Multi-State Lottery nor the Wisconsin Lottery notified the public of the reported flaw, and they say they have no reports showing investigations were conducted into the matter.

The Multi-State Lottery chose not to order a rebuild of the random-number-generating software that ultimately was used in at least 17 states, according to transcripts of Tipton's confession to prosecutors that the Register obtained last year.

"It just kept growing and growing," Tipton said, according to court transcripts. "It became spaghetti codes, unfortunately."

Iowa ends several lotteries over problems

Steve Bogle, Iowa Lottery's vice president in 2015, recommended the state suspend four popular games, including Hot Lotto, because of serious flaws with the software that randomly generated the winning numbers, according to records a Polk County District Court judge released following the Register's request.

As a result, the Iowa Lottery no longer uses Multi-State Lottery computer software to draw game numbers.

The Multi-State Lottery Association is the subject of multiple lawsuits from players who contend they were bilked out of winnings by Tipton's scam.

The repeated numbers and decision to discontinue the use of Multi-State Lottery random number-generated software indicates the nation's lottery system has little oversight, said Gary Dickey, a Des Moines attorney for one of the players.

"You have to ask, 'Who does MUSL answer to?'" Dickey said. "It's not regulated by a federal agency or any single state. It just sort of exists in a patchwork of state laws."

Lottery officials say they were unaware of warning

The random-number glitch in the software Tipton said he warned Multi-State Lottery officials about in 2006 is separate from the code he inserted into the system that allowed him to predict numbers and scam multiple state lotteries. 

Multi-State Lottery and Wisconsin officials indicated in response to Register questions that they did not know that Tipton allegedly warned their staff about a random-number problem.

Multi-State Lottery Association Director Bret Toyne said — other than in Arizona — his organization has not identified or been notified "of an instance when a similar hardware failure occurred, including the 2006 Wisconsin game."

And the Wisconsin Lottery said it has no documentation or institutional knowledge of Tipton's warnings.

"Nor does his allegation make much sense," said Mayers, the Wisconsin Lottery spokeswoman. "There is no reason why he would have been monitoring the numbers drawn in Wisconsin in the manner he suggests."

In Wisconsin, the Register identified 10 sets of numbers that were called twice within five years of each other in the state's SuperCash! game, including the two 2006 draws that Tipton said he had warned his bosses about.

Wisconsin began using the Multi-State Lottery's random-number generators in 2004 but stopped in 2016 because of concerns involving Tipton. (Wisconsin now uses a different random-number system that costs the agency $125,000 a year, versus the $15,000 it had previously paid for the Multi-State Lottery's services.)

Five of the 10 duplicate sets of numbers the Register identified were drawn before or after the Multi-State Lottery random-number system was used in Wisconsin.

"While such repeats are rare and uncommon, there is no reason to suspect these numbers were not drawn reasonably," Mayers said.

State lotteries identify internal problems

At least three state lotteries have acknowledged random-draw malfunctions, assuring the public that steps were taken that fixed the problems.

• Arizona officials in 2017 identified more than a dozen draws linked to four separate lottery games where they suspected computer software had malfunctioned and called duplicate strings of numbers, including in consecutive draws.
• Connecticut suspended two employees in March 2018 following revelations that a machine used to randomly select winners for a high-profile game had been erroneously programmed.
• South Carolina officials in May declined to pay more than $35 million in wins after it was discovered that a game played on Christmas day had printed thousands of winning tickets because of a computer error.

As with any technology, including ball machines, the opportunity exists for hardware to fail, said David Nunn, a spokesman for the Arizona Lottery.

The agency used an independent third-party firm and a statistics expert to evaluate draw history after the duplicate draws.

The Multi-State Lottery Association — Tipton's former employer and the agency that provided Arizona with the software — also conducted an investigation. An executive summary of that investigation showed the software had been installed in July 2015, which was about six months after Tipton's arrest.

A Multi-State Lottery employee flew to Arizona and replaced an internal counter and concluded the machines were repaired.

Arizona Lottery generated $816,231 in revenue on the four games during the affected time frames and issued $106,017 in refunds to players — in addition to paying winning tickets.

"Arizona Lottery has the responsibility to minimize that risk, protect the integrity of our operations and be transparent to the public when a failure occurs," Nunn said. "It is our obligation to be able to earn the public's trust, and it is our objective to operate that way in all aspects of the products that we provide to the public."

Dan Zitting an executive of ACL, a software company based in Vancouver, Canada, said state lotteries should follow Arizona's lead when duplicate numbers are drawn and independently investigate whether problems exist.

Zitting, whose company has worked with more than 900 state and federal government agencies, said he believes computer-generated drawings generally are fair and efficient.

But he also believes national standards are necessary to set how computer-generated drawings are used and tested throughout dozens of state lotteries.

"A whole lot is premised on the integrity of the lotteries," Zitting said. "Even if this isn't crisis mode yet, I would really be on top of this before it hits that state."

How this story was reported

For its analysis, the Register compared lottery drawings with five numbers or more. It identified hundreds of instances between 1992 and 2017 where identical numbers were generated in the same game within five years of each other, using winning data from 37 states collected by the website Lottery Post.

A more narrowly defined review of identical numbers drawn within 365 days of each other identified more than 100 duplicate drawings, confirmed by other online sites that collect winning numbers.

To help assess whether those duplicate drawings could be the result of chance, the Register asked two math professors to review some of its findings.

The statistical probability of lottery drawings producing the same winning numbers rises as the number of drawings increases, meaning that a short window of time with fewer drawings is less likely to produce duplicate drawings than a longer window of time with more drawings. The Register's review reflects this principle.