First off, my background is not economics or related to stock it all; it is Applied Physics and Computer Science; so any naivety comes in regard to economics from pure ignorance. I'm currently using yahoo stock quote API for datafeeds for acquiring approximately minute by minute data for all stocks on the Nasdaq and NYSE. I then inject this into a SQL database were I have algorithms that essentially analyzes the entire NASDAQ and NYSE data from the previous day and spits out a stock which will make 0.5% the next day. Running this algorithm against historical data from 8-20-2009 through 8-19-2010, I was right 135 out of 187 times (or 72.19% of the time...) [note that I do not take into account Mondays or days after holidays] The HUGE caveat is that while 72% of the time I am guaranteed to make money if I put a limit order at 0.5%, I also have to put a lower limit order not to lose alot of money as while its highly likely to gain 0.5%; it will also throughout the day be very highly likely to drop a few % ....
Lol yah, Currently im trying to analyze the likelihood of it hitting the .5% before going below .5% (my cutoff point). Unfortunately, if it sounds too good to be true, it probably is...but i've been writing and tweaking the algorithms for 3 years... In the next few months im going to test it IRL and then probably cry
Im trying to get access to all tick by tick or at least minute by minute data for the NYSE, Nasdaq and AMEX to run my simulation engine against a more complete dataset.... Do any of you guys take and analyze your own data programatically?
What is the success of getting .5% before losing .5%? If you are over 50% you may have something? Divide your data in 3 sets, training data, validation data, test data. Do not use the same data sets! Then try it on new data. Mind the commissions!
That is the thing, I need more complete data to see how often .5% is gained before it lost...because while I know it is likely it will hit .5%; they typically do hit below 5% throughout the day and then somewhat recovering Yes, I actually have infinte data sets in that the algorithm for testing is a moving algorithm.... To really test it completely, I need tick by tick or minute by minute data...they only thing is tickData charges $25,000 for what I want and IQfeed is only like $400 but I have to code against their API and god knows how long it will take to pull all the data from their servers. With commissions included at $8.00 per trade, the break even startup point is about $12,000...anything below and the commissions eat your gain and you go to 0, anything above and smiley faces everywhere...
Not sure where to get cheap tick data. I have seen http://www.tickdata.com/ suggested here. Can someone suggest a source? selrod0324 can you point me to materials about a "moving algorithms"?