During graduate school I’ve somehow been fortunate enough to avoid undergraduates. I ended up talking with one yesterday. He was a finance/cs guy and wanted to impress me with his big project on the side. He talked at length about how he could predict stock movements using algorithms on historical data, day trade on it, etc. etc. Bored quickly, I asked: “so what exactly are you data mining to predict on?” He answered “Historical analyst performance” – he’d find statistically good analysts and follow them. He’d then use some sort of monte-carlo to generate predictions of movements.
Growing less impressed by the second, I gave him this thought exercise as a warning: “Be careful with drawing assumptions off of that. Let’s take 1024 people and put them in a tournament structure with 2-person brackets. Now let’s let them win a game and advance if they can call if a fair coin comes up heads or tails. The ‘winner’ would have won 10 coin tosses in a row. Now would we say that he has coin-toss calling ‘talent’?”
“Well, no. Somebody has to win this game 10 times in a row, but did he win the most games because of ‘talent’?”
“Yes. And I can use math to predict who that guy will be.”
That was about as wrong of an answer as one could give. Was I that way as an undergrad? I remember spending most of it drunk and chasing girls.