Other Forms of Statistical Arbitrage

Discussion in 'Strategy Development' started by jones247, Jun 9, 2009.

  1. Other than Pairs Trading and Dispersion Trading, I am hoping to begin a "think tank" discussion on the feasibility of various high probability trade set-ups...

    It's reported that Jim Simons, one of the best traders/fund mangers over the last 15+ years (based on his alleged performance - $3 billion in personal earnings last year - #1 ranking), has over 100 PhDs (primarily mathematicians) working for his Renaissance fund. He stated that he automates almost exclusively, taking advantage of rapid burst and repeatable price patterns. Basically, his staff has two primary functions: (1) R&D to discover high statistical probable events in price movements (i.e. Statistical Arbitrage); (2) Programming of high probability set-ups.

    Some of the high statistical probabilities I've noticed are: (1) when the price reverses and moves strongly in one direction without any material retracements (i.e. trends 3x ATR without 1/2 ATR retracement), then a retracement or reversal of at least 1x ATR is more than 50% probable; (2) If price bounces off of support or resistance 3 consecutive times, then it's more than 50% likely to break through on the 4th thru 6th approaches; (3) if the market moves opposite to the opening, 2 days in a row, then it's likely to move in the direction of the opening on the 3rd or 4th consecutive day (i.e. the market opens at $30; by 10:00 AM Est it moves to $32; the market ends the day at $28... if this dynamic occurs 2 consecutive days, then on the 3rd or 4th day its probable to move in direction noted in the intial 30 minutes of the trading day); (4) if the price reverses and moves 10% of the ATR, then it's more than 50% likely to continue another 10% of the ATR

    Perhaps we can get a good discussion going on HIGH PROBABILITY set-ups, whether it's frequent as 30 times per day or as rare as 2 times per month...

  2. You're describing a statistical approach to trading based on the probability of price movement based on prior price movement. It's viable but not really arbitrage.

    But still I'm game, since I'm currently looking at this.

    I'm a verbal thinker, so I like to start my testing out with an assertion such as:

    "If price moves x pips within m minutes, it moves y pips more p percent of the time without returning to the starting price."

    That's the one I'm looking at right now. Next I want to explore how to describe some kind of counter-trend movement. Something along the lines of:

    "If price moves more than x pips within m minutes, and then retraces y pips, price will move at least z pips more p percent of the time without returning to y."

    This can all be analyzed with tick data pretty easily.
  3. This is not what most people would term statistical arbitrage, which usually involves running some kind of balanced book.

    This is pattern trading. Nothing wrong with it, it's just a very different beast. Quite a few CTAs practice it.
  4. i think this whole mean reversion thing is a good idea. things tend to revert to their mean . . .there are some laws of financial gravity.

    BUT. and i think this is a big but

    there can be pricing analomies that persist. and they probably persist for some fundamental underlying reason. like a stock can stay detached from its peer group, it can literally go away and just stay there, on and on . .. at least i think so

    the market can remain inefficient longer than you can remain solvent :D

    what i like about pairs trading though is this 'if you buy somethin you gotta sell something' concept which is built into this whole concept