US Would Be 28th In 'Hacking Olympics', China Would Take The Gold (infoworld.com)

After analyzing 1.4 million scores on HackerRank's tests for coding accuracy and speed, Chinese programmers "outscored all other countries in mathematics, functional programming, and data structures challenges". Long-time Slashdot reader DirkDaring quotes a report from InfoWorld: While the United States and India may have lots of programmers, China and Russia have the most talented developers according to a study by HackerRank... "If we held a hacking Olympics today, our data suggests that China would win the gold, Russia would take home a silver, and Poland would nab the bronze. Though they certainly deserve credit for making a showing, the United States and India have some work ahead of them before they make it into the top 25."
While the majority of scores came from America and India, the two countries ranked 28th and 31st, respectively. "Poland was tops in Java testing, France led in C++, Hong Kong in Python, Japan in artificial intelligence, and Switzerland in databases," reports InfoWorld. Ukrainian programmers had the top scores in security, while Finland showed the highest scores for Ruby.

Comments

Um, baloney

[110010001000]

Has anyone actually ever used Chinese software?

Re:Um, baloney

[Anonymous Coward]

Hacking != software development

Re:Um, baloney

[Hognoxious]

The first one I worked with was pretty good. Nice bloke too.

The dozen after that ... stupid, incompetent, dishonest too.

Hillary promises military retribution

[Citizen of Earth]

Not to worry, Hillary has promised military retribution for cyber attacks. I guess she'll either start wars with Russia and China on day one of her presidency or draw red lines and then run away from them like Obama has.

Self-selection sampling bias

[Solandri]

Once upon a time, a city was considering expanding its subway system. To determine if this was a worthy use of public money, they decided to find out how many hours a week the average person rode the subway. The agency tasked with collecting these statistics thought about the problem. Asking random people on the street seemed like it would waste a lot of time since most of those people might not even ride the subway. Then they realized if they just asked people riding the subway how many hours a week they rode, it would neatly filter out the non-riders and dramatically simplify their job. So that's what they did.

When the city got the statistics, it said there needed to be 10x as many trains as they currently had. That obviously couldn't be right since the trains were only occasionally full. So what went wrong?

• First, the statistical gathering method filtered out non-riders. This skewed the average (both mean and median) up, since they didn't have a bunch of "0 hours" in their statistics.
• Second, asking people riding the subway gives you a time normalized sample, not a population normalized sample. Say only two people use the subway, one of whom rides it 1 hour a week, while the other rides it 10 hours a week. If you randomly hop onto the subway at any give time, you are 10x more likely to encouter the 10 hr/wk rider. Your statistics end up saying more about who is riding the subway at any given time, rather than how much each person usually rides the subway.

Likely, the only thing these HackerRank statistics are measuring is that there are just a lot more job opportunities for mediocre programmers in the U.S. and India. While there are fewer such opportunities in China, Russia, and Poland, so the few people who pursue programming careers there tend to be the cream of the crop. To normalize it, you'd have to survey to find out how many total programmers there are in each country, compare to their total populations, then assuming a normal distribution of "skill" for the entire population of the country, map each countries results to that distribution. Then for the countries where the number of people taking the test are overrepresented relative to the total population, truncate their distribution to match that of underrepresented countries. e.g. If, say, only 0.01% of Poland's population tried the HackerRank tests, while 0.1% of the U.S. population did, then you'd have to compare Poland's results with the top 10% of the U.S. results (0.01% of the U.S. population matching the 0.01% of Poland's population) to get an apples-to-apples comparison. But that's a lot of assuming and normalizing for me to be comfortable with using the data to draw conclusions.