Lottery Management Consulting, LLC
Building Insight from Information
applied data science
A new look for lottery
And I mean a new ‘looking for’ as well as a new ‘look’. Rather than sticking to game mechanics that have simple, obvious math defining the outcome probabilities, let’s ‘look for’ play value first. Get the probabilities by numerical simulation (teaching a computer to play). The grid above could be worth significant money, in the game I describe in NASPL Insights February 2019!
In December, as I have done for several years now, I review what happened in Powerball and Mega Millions in the past year, and use math models to make some predictions about future behavior. The games became more similar in October 2017, when Mega Millions raised its price and changed its matrix. My outlook for 2018 is based on players responding much as they did when Powerball made similar changes in 2012. A key point from simulations is how broad is the range of “likely” outcomes- give or take half a billion in profit is about right!NASPL Insights December 2017
How is grand-scale research on lottery advertising like grand-scale research in agriculture? No one does it, who must thrive or fail according to the outcome. With my friend Jade I discuss tests of smaller scale, and we agree that there is usually something to be learned from a well-constructed test of advertising, even if the results are not what everyone wanted. After all, don’t we want to know what not to do? NASPL Insights August 2017
After January 2016, we saw sales for jackpots in the range $200 to $400 million significantly lower than before. My math models support estimating the effect of these changes on the development of big jackpots in the future. I estimate the likelihood of a jackpot reaching $1 billion in 2017 at about 10%, down from about 17% at the start of 2016. NASPL Insights December 2016
I combined data from two sources to get a view of how much the “average citizen” lost in legal gambling, across 41 U.S. jurisdictions. The per capita annual loss to lottery is much more variable than the loss to all legal gambling combined. Lottery thrives in states where other legal gambling is limited (and even in neighbors of those states). Total annual per capita losswhich seldom exceeds $500 per capita annually. The potential for improved lottery yields is likely constrained by this combined limit. All Legal Gambling 2016
I used math models to explore how matrix changes in Powerball will likely increase sales (as we all want) and increase year-to-year variability (as nobody wants), and estimated the likelihood of a Powerball jackpot exceeding $1 billion in 2016 at about 17%. A few weeks after this publication appeared, that event did happen. As I remarked to some at the time, same odds as Russian Roulette (just saying). NASPL Insights December 2015
Pio is an economist educated in the US Ivy League, who serves his native island of Zoa in many capacities, including Director of the Lottery. I was introduced to him at a meeting of economists in Seattle, and he opened my eyes to a factor that drives our year-on-year increases in Instant game sales. Economists agree: if money is getting cheaper, it takes more of it to express the same value, year after year. NASPL Insights June 2015
NASPL Insights December 2014, I assessed the behavior of the Powerball and Mega Millions games in the eleven months since California joined the Powerball game. I found that while sales at jackpots below $100 million had remained fairly steady, the sales response to jackpots approaching $300 million was less than half what it had been during the “base period” between Feb 2012 (when Powerball raised its price to $2) and April 2013 (when California joined). I went on to project the probability distribution for jackpots greater than $300 million in the coming year. The difference between the games was stark: Powerball was more likely than not to have three or more such events, while Mega Millions was more likely than not to have no more than one.
In my continuing discussion with Veronique, Jade, and their colleagues at the ad agency, Jade emerged as the champion of the view that the end-effect of actual sales is likely the most sensitive measure for demonstrating the combined effects of all kinds of lottery advertising and promotions. Advertising efforts may aim to influence sentiment, and they may be effective, but the player is more likely to express this by spending an extra dollar, rather than by changing the way she responds to a survey question. Further, although our efforts may be focused on a particular game at a particular time, we hope the effect of these efforts is more diffuse and longer-lasting. This suggests a measurement strategy aimed at detecting improvements across the portfolio, which I am glad to implement as the Dashboard According to Jade.
Math modeling can seem arcane, but I believe it is most effective when it incorporates intuitive or common-sense features. In describing the way people play the big jackpot games, for instance, we intuitively recognize that there are some people who play nearly all the time, and others who rush in when the jackpot gets to a certain size. I reassured Veronique’s agency that a well-crafted model can incorporate these features, in NASPL Insights August 2014.
