Hello Ced,

In my mind, you are suggesting a set of rules, which would need to be refined a bit before testing them via a backtest. When you talk about an 8% or more move, do you mean (close - open)/open >= 0.08 (where I am using daily prices)? Or something more complicated? Also, when you say "I would check back and the stock would be down maybe 2-3%" the day after, when do you check back, and are you measuring the drop relative to the prior day's close?

Grant

Yes, everything in daily price moves. In fact the equation you provided ((close - open)/open >= 0.08) would be the basis for my idea. If that rule is respected(+/-8%) the day before, I would take the other side of the move the next day right at open. A problem is that when large caps move so much they sometimes drop alot in after market and pre-market trading as a reaction to such large moves.
More detailed example of what I would want my system to do:
Monday: TWTR closes at 4 p.m. with a 10% gain(daily move).
Tuesday: I short TWTR at 9:30 a.m. market open assuming TWTR will scale back from the crazy rally it just had on monday. From what I have seen(this is not backed by data-thats what I am trying to do...Simply a very random observation) the stock would probably go down 2-3%, depending on the size of the prior day's move.

I want to see if data backs my observation and if this could be perhaps turned into a trading system. Again, a problem is that when these stocks move alot during 1 session, they seem to reverse during after amrket and pre market.

Thanks alot for the response, it is really appreciated.

@Ced H. If you're looking for statistical evidence of this behavior then you don't need Quantopian. Pick 5 or 10 stocks, fetch their data from yahoo and drop into Excel. All the data to determine the move and the return metric is right there. Or some work in R would allow you to collect and plot the stats for a set of securities. Statistical analysis can be done in Quantopian, (see the Candle patterns - research strategy post), but it would be easier to do in Excel or R or maybe the research project Quantopian is trying to put together (soon to be offered?)

[of course Grant or some other may just knock this out with a simple collection strat...]

Makrte Tech.. Thanks for the response. True, data from yahoo Fin and exported into excel could fit, but I just don<t really know where to start. What kind of formula would I need?

Given this file:

http://real-chart.finance.yahoo.com/table.csv?s=TWTR&d=0&e=20&f=2015&g=d&a=10&b=7&c=2013&ignore=.csv

~ 300 daily return periods for TWTR

I created this excel:

https://dl.dropboxusercontent.com/u/217910322/Finance/PctReturnsAnalytics.xls

From which the following statistics were realized:

Pct trigger: 8%, Pct return (next day): -2% -- 0 count
Pct trigger: 6%, Pct return (next day): -2% -- 0 count
Pct trigger: 4%, Pct return (next day): -2% -- 8 count (Edit: total trigger count: 27. So 8 / 27 = 29% of the time a 4% daily move resulted in a -2% retracement).
Pct trigger: 2%, Pct return (next day): -2% -- 15 count

Pct trigger: 8%, Pct return (next day): -1% -- 1 count
Pct trigger: 6%, Pct return (next day): -1% -- 4 count
Pct trigger: 4%, Pct return (next day): -1% -- 13 count (Edit: 13 / 27 = 48%)
Pct trigger: 2%, Pct return (next day): -1% -- 22 count

You'll have to check the formulas and cell locations but it looks about right.

Have fun

WOW-Thanks alot for that. Truly appreciated.