This evening the CBS News show 60 Minutes reported on an impressive example of design detection in the on-line poker world.
Online gambling has grown in a few short years to a 16 billion dollar a year industry, and a big part of that growth has come from internet poker. Recently several professional gamblers at one of the larger internet poker sites, Ultimatebet.com, noticed that some of their opponents were playing extremely poorly, yet winning consistently. They suspected cheating.
One of the professionals obtained tracking data on one of the suspected cheaters, and after running the numbers determined that the suspect’s winning hand percentage was 13 standard deviations away from the mean percentage. This is equivalent to winning a 1,000,000 to 1 lottery six times in a row. The professionals took their findings to the licensing authority. Denyse, you’ll get a kick out of this. Most internet poker sites are licensed by a sovereign Indian nation near Montréal, Canada, the Mohawk Kahnawake tribe. The tribe hired a professional gaming expert to investigate, and sure enough there was cheating. One of Ultimatebet’s employees had gotten an administrative password, which gave him the ability to play poker at the site while looking at the other players hands! In all, the employee stole more than $20,000,000. Read the whole story here.
How does this relate to ID? The investigation was pure scientific design detection. Here is how the investigator employeed the scientific method to reach his conclusion.
Step 1: Decide on a question one would like to explore. In this case, the investigator suspected cheating, but it was just a gut feeling. The poker players among us know that in any given hand the worst player in the world can beat the best player in the world by pure dumb luck. I have personally seen a player win a hand in which the probability that he was going to win was only 1%. But blind luck like this succeeds only in the short run. In the long run, the better player will always come out ahead. Here, the investigator saw data that seemingly contradicted that maxim. A player (let’s call him Joe) who was playing very poorly, constantly taking foolish risks, was nevertheless winning not only in the short run, but also in the long run.
Step 2: Form a hypothesis. This was easy enough. The investigator hypothesized that the Joe was cheating.
Step 3: Test the hypothesis. The investigator gathered data about Joe’s history and performed a statistical analysis to test his hypothesis. He determined that Joe was winning at a rate that was 15 standard deviations above the mean. In the story the investigator is quoted saying, “Now, this sort of stuff just doesn’t happen in the real world.” In other words, the investigator cannot rule out random chance in an absolute sense, but as a practical matter, he is certain that Joe is cheating.
Step 4: Form a conclusion. The data indicate that Joe is cheating.
Acting on his scientific findings, the investigator reported Joe to the licensing authority, which performed its own investigation and found that Joe had in fact been cheating by using the administrative password to look at the other players’ hands while he was playing.
How is design detection in this instance different from the design detection employed by ID proponents? As far as I can tell, not at all.