March 5th, 2008, 10:40 am
statTrader, i have used the systems with personal money but scalability is a problem, in this case combining a high volume strategy with under-capitalization results in a nice equity curve but low in absolute CG... this is not surprising.. and re-coding the MM isn't an option - I have tried this. Another barrier is overheads - for instance a dedicated, powerful server with very high % uptime is a must(I am currently renting a Linux VPS running a mod'd version of one of my systems to communicate with POSIX (Unix API)... another cost barrier is data feeds for testing, for instance the most realistic hypothetical forward / back tested results came from using the Bloomberg API but i only had access to this for a short period of time when i was on a free trial (which i managed to get extended over an over again until the Bloomberg rep got tired of me).and one other thing, most funds have the advantage of economies of scale to make deals with their brokers to lower transaction costs. Institutional transaction costs were taken into account before i began prototyping, and as said in my initial post, the results are adjusted for realistic institutional transaction costs. This system was always just designed to imitate institutional level systems to show what im capable of, nothing more. Therefore running the system with transaction costs offered to say, an individual account, seriously offsets profit.
Last edited by
alpha1 on March 4th, 2008, 11:00 pm, edited 1 time in total.