Enough being loco... (Though, it's fun every once in a while...) I warned and asked kid lolat... about trashing the thread. He kinda said it's fine for me to kill the thread. Seriously... I've been very very nice until he starts making dumb assumptions. lolatency... seriously good luck trading. Anyways... I'm back into hiding but before I did so I posted my blotter in the P&L Thread yesterday but someone deleted it... hrmmmm... Ciao, ET.
lolatency, can we go back to your journal? btw... I don't... think... asiaprop uses "..." alot... like TSGannGalt... except sometimes in the beginning and at the end of the post... so most likely they are different guys unless... dealing with a real crazy here. would you recommend other reading besides tsay's book? Thanks.
I read the first page which made sense, then skipped to the last page and have no clue what you guys are even talking about anymore. Anyway to the OP, I went through everything you did, tried out all the vendor stuff like ninjatrade, tradestation etc but they are just too limiting. Also dont think starting off with high frequency trading where everything depends on milsec accuracy is a good idea especially with your one-man infrastructure support team unless you already have extensive knowledge coding such system and truely understand the business logic (edge) from that bank you worked at. Instead try for strategies that trade on 10+ min bars or more but using tick data as the source, on highly liquid products like the es mini futures. That way your slippage / issues will be kept to a minimal, and your backtesting result will mimic actual very closely. The best way to do that I found out is to use IB's api (java for me) then code your own framework. If you architecture it correctly, it should be very modular so later on you can just replace IB api with better stuff as they come without touching the rest of your app. Basically use IB's quote + order execution, and code the framework/logic yourself. Once ready you should let it run for a month or two as if it's in production but not actually submitting the final step - order execution, just log it to a mysql db instead. Trust me you need to do this to avoid many bugs, some of them you can never think of no matter how much exception handling code you write. Cost is minimal, i use eclipse for development, and pay $150 a month for a dedicated server hosting and run my java processes / sql db etc.. all from there. For security, put pieces of the business logic into the mysql database as encrypted garbage data, so even if someone is able to decompile the classes they still cant connect the dot without additional effort.