Building a GAP Trading Strategy

[jmiranda]

Originally Posted by atto I have a decent amount of experience here, so will share some if anyone wants to look into this. I have an automatic strategy that trades gaps on ES that has a fairly impressive track record. This isn't my primary method of trading, but was one of my first statistical looks at markets, and has helped smooth out the P/L curve. It trades using statistical biases I've found, and only actually attempts roughly 20-25% of the time (but it's actually highly accurate, and has a good r:r). In turn, many of the "gap and go" days are filtered out. For those who are statistically inclined, here are some things to look into regarding statistical biases for gaps. Remember, if the gap filled or not is only one side of the coin. Also look at the bias in context of the gap size and risk. Is the general market trending or congesting? How big is the gap? (I have filters for gaps that are too big, and too small) How does gap size compare to ATR? (same) Have the previous x gaps filled? (look at both the direction of the gap, and against) How does gap direction compare to the trend direction? (look at a few timeframes) Where is the gap positioned compared to the previous day's value area? (ie, is there liquidity close by?) Where is the gap located compared to yesterday's High and Low? (inside/outside) What day of the week is it? (yes, there's a bias) How volatile has the market been? Is the gap near S/R? (this is one of the hardest to automate, so sometimes you end up cutting corners) This list is by no means complete, and I'd encourage someone who wants to delve into this look into these suggestions along with any other ideas they may have. I simply got a decent amount of historical data (think a couple years), and did some testing for each of my potential biases in Mathematica (Excel also works well for this) with a portion of the data. Once I had a good idea of a potential bias, I tested it on out of sample data. When combined, these biases give you a much better idea of the statistical edge you may have by trading any given gap. I programmed the strategy to enter anywhere from 9:00 to 9:45 on a technical trigger (for brownsfan's case, he looked at hammers; you may prefer some type of indicator or other method). Exits are enabled within a certain % of the gap fill, and exit on a similar technical trigger (so they may go past the gap). Hope this gives budding statisticians some ideas. In addition to biases, stop strategy is very important (and could be a bias in itself.. if price goes x% against, what's the chance it'll go back and fill?). Something mine does not do is the "bullets" approach, and there could be some validity in that. Hope I didn't barge in on your candlestick discussion, browns

Dear Browns

You said : "but was one of my first statistical looks at markets"

I am very interested in this strategy, but organizing the data and KNOWING where and what to look for is very unclear, Can you point me in the right direction to learn how to get useful data ?

I really will appreciated it !

Regards

Jaime

[trade-samarai]

Yesterday I watched a webinar with Frank Lesh. He mentioned the "Gap and go" setup. "Gap and go" is the opposite of "gap fill" and an "impulse" (momentum) move. (He didn't explain it at all.)