Completely agreed, but noise filtering and data verification is the part of system stability. A would add to this list physical disconnects, servers downtime, slippage, spread variations, broker tricks and so on. IMO accumulation of real data is the best approach. Not mention then it differs from broker to broker.
strange, i seem to understand everyone on this thread. 1. searching for an experienced backtester in order to get something more advanced than TS or whatever other off the shelf testing product. 2. criticising that search as naive is reasonable as well. i made that experience several times within my team. someone who is just coding usually does not get you too far if you yourself do not really, really know IN DETAIL what you want. when you are looking for something automisable, you have to go back and forth, and back and forth. better strategies, better walk forward testing, better execution speed, and this goes on and on. so, a pure coder is lost, cause there is so much interaction required between the one who "knows" and the coder, that it can hardly be done if the two functions are completely seperate. i have now a hybrid of this setup. there is one guy who is doing statistical analysis of the time series, gets some idea out of that and starts simple backtesting on one market in tradestation (with all its limitations). once that looks fine we move it to the second guy, who is programming it in C++ and integrates into our C++ platform which optimises, tests walk forward, combines different parameter sets while taking care of correlations between them and some other things and finally sets it on semi-auto, which means in our case that the engine at our execution broker gives a signal and they fill it by phone (slippage reduction is my main motivation for doing it like this right now - i wnat my orders to be "worked"). now, the important point is, that both ghuys are developers. a pure coder does not get you too far. you most probably get either set up by the constant stream of nasty questions or you end up with an overfit your coder produced without being aware of it. 3. i understand why someone critisises nasty and aggressive conduct. but, it it nononsense, that is his style. do not take it personal. he is a pro in the end (at least i give him the benefit of doubt), so rather take his essence seriously. actually i think the thread is misplaced in auto, that is all. i am the good one. i understand everyone ... please don't bash, i am just joking!
thank you, Man, very isightful. I also am looking at building a futures team. how suceesful have you been in developing profitable systems?
You won't be successful (with whatever team you are planning on running) unless you know about system trading very well. Too many ways to fake it. In terms of back-tester and execution platform, it won't be hard to find.
we have several systems running, trade a sharpe of about 2 and i hope we can trade 3 from january 2007 on. next in our pipeline is a forex system (actually fx futures). hope to set it ontrade by october. building a team is a real challenge, especially if you do not have the money to just hire some established pros. though i would say i have quite a strong team now, it literally took me years to grow it. and i am not sure if i am not just lucky at the moment. the problem is that the skill set for system building is very difficult to detect. i, for various reasons, was forced to hire more or less unexperienced people in terms of trading. i interviewed IT guys as well as stat-math-physics. you could tell after a month if someone would ever make it as a systemguy, but i still find it difficult to tell in advance. system building is a very tough job. the most important aspects IMO are endurance and a certain kind of natural curiosity. and this undescribable instinct to do the right twist at the right time. i had someone who had studied math and could program. yet he never developed any feeling for "whereIsIt" ... in short: building a team from scratch is an endeavour of its own. ah yeah and i forgot to mention that you rather have your contracts straight from day one. you would not believe what reasons people find to think you are an %$!&* once they have made backtest with a sharpe above 3 ...