Controlling for over-fitting... and Renaissance Technologies

Discussion in 'Strategy Development' started by mizhael, Jan 19, 2011.

  1. Heard that Renaissance Technologies never hire people having Wall Street experiences... and all they hire are quants.

    So they are not fundamental traders.

    Their trading strategies are not based on fundamentals... instead, they are based on statistics...

    Therefore, they must have a lot data-mining and fitting there.

    Maybe their success is about how to control for over-fitting in their statistical data-mining process? They must have put up a framework, to weed out bad strategies...

    Any thoughts?
  2. nLepwa


    A large part of what they do is exploiting dependencies in portfolios.

    It hasn't much to do with curve fitting.

  3. poyayan


    They start from the commodity market. That should tell you something.
  4. Please don't take my "curve fitting" narrowly...

    How do you exploit dependencies?

    Using statistical data-mining, right?

    Then you have the danger of over-fitting...
  5. What's special about commodity market?
  6. nLepwa


    Not necessarily.

    Statistical arbitrage goes beyond data-mining. :)

  7. Exactly because people with wall street experience are brainwashed enough to produce mean performance. They go for the extrtemes, hopefully the one to the right.