Learning from Brazil's Defeat

Reference to Context

For those who followed the worldcup, Brazil going down to Germany by 6 goals must have been a shock. However, what many don't know is that a Goldman Sachs team of number cruncher's had predicted twice that Brazil would win the world cup. Making this prediction worse is the fact that they did not get just the outcome wrong, but many of the teams they earlier predicted did not even make it past the 2nd round. Surely, a thrashing was predictable if the system had any credibility.

Link to their earlier prediction: http://www.goldmansachs.com/our-thinking/outlook/world-cup-sections/world-cup-book-2014-statistical-model.html
Link to the revised prediction: http://www.goldmansachs.com/our-thinking/outlook/world-cup-sections/world-cup-prediction-model-update-6-26-2014.html

Inference

I have always believed, using statistics to predict where you may fail, is more easier than predicting where you may succeed.
One should always listen to numbers & reason but where do you draw the line?
To me what comes clearer is, failure is easier to predict using mathematical models but success is not a number thing. The human will and intuition is a more powerful tool.

People may say, well isn't it two sides of the same coin? It's not ! That's because there are many faces to the coin, and the outcome of one is not a simple inverse of the other.

Risks/Failures can be assessed by probability better, I feel. This is because, failure requires the first breaking point. Means if there are multiple related events, failure is a result of ANY one event failing. Think of a for { } loop with a break on the first fail condition.
This is easier to approximate and calculate and hence trust (for me).... success on the other hand is NOT impacted by failure and neither do related or unrelated events of success guarantee further success. Success is harder to prove or trust!

Even in web-site analytic's, I have observed that you can predict why people are NOT buying something better, than why they WOULD be buying something. Luckily in E-Commerce consumers are more interested in their reasons for failure to improve their conversions, than focus on whats working for them.

Summary

Arguably, you can be sure of what WON'T work than what WILL work. Though one should backup all facts by numbers, as far as predictions go I'd be careful not to guarantee or trust predictions just on the basis of numbers.

Other References:

• Germany's 12th Man: http://blogs.wsj.com/cio/2014/07/10/germanys-12th-man-at-the-world-cup-big-data/