As I said, I don't need tick data, 5 seconds would be enough, but 1 minute is not enough, and I doubt there exists anything between 1 minute and tick resolutuion. The strategy is based on some rare condition and requires quick action, but has very high winning rate. I tested it on 3 year's worth of tick data downloaded from opentick. Sorry, can't provide any more details.
I'd think most competent software could build second bars from tick data. Also, I can't remember if you said how long you were going to hold these positions... but with a 2k account, watch out for the pattern daytrader rule.
Personally, I would look for a higher time frame filter if possible that you wouldnt have to look at each tick for 2000 symbols. I'm not sure your really going to be able to get around what davidmadison posted about. I guess the other way to look at it is if you did go with something like what he posted, you would have a seriously baddass high freq setup for future use. No way your going to be able to do this though on the cheap.
Thank's for your input, let's focus on API now. I found a broker with an API - Just2Trade. But their site looks like they are going out of business soon. Don't know why, just smells like a dead corpse.
One thing to watch out for, is that some such API's don't actually encrypt anything. Consider this in conjunction with whatever ISP type you're using -- e.g., it probably won't be a good idea to use such an API on a shared T1 as your neighbors might be able to snoop your password and/or order flow. I remember just2trade as having such an unencrypted interface, although it's been a while so I may be wrong and/or things may have changed. On the other hand, their prices seem pretty good. Some other brokerages that may be suitable for a small, automated account include CoolTrade, sogoinvest (a "small acount" version of Genesis), ChoiceTrade, MB Trading, ThinkOrSwim, Interactive Brokers, and SpeedTrader.
Tradestation has the Radarscreen app that can update 1000 symbols at a time with history and 2000 symbols with no history. Probably you can get it to do all 2000 symbols with history with some dll's or something. You can interface the broker API from Radarscreen with dll's too..
That's 5 signals per symbol, correct ? If not, it's my firm belief you do not have a feasible strategy that will be profitable over the long haul. It's unheard-of to trade second bars with that kind of low frequency. And besides, you need 200 MONTHS of backtesting for 1000 trades to make your backtesting results even valid. Finally, a $2k trial investment ? If you need a custom, multithreaded C++ trading application, consider that you will be spending about $100k soup-to-nuts. Good luck. You're going to need it !!
That's BS about 1000 signals; even 9 out of 10 is statistically significant enough, out of sample of course.
I'd be rather cautious about the 9 out of 10 thing... Significant, maybe, but a lot of chances of missing your black swan with a sample of 10. The black swan, along with slippage, the spread, and commission are some real important things.