in my opinion you need constand research as near as possible to what you know in terms of technology be it hardware, software, connections or algorithm and as far away as possible in terms of correlation.
Dictionary definition of robust: Etymology: Latin robustus oaken, strong, from robor-, robur oak, strength 1 a : having or exhibiting strength or vigorous health b : having or showing vigor, strength, or firmness <a robust debate> <a robust faith> c : strongly formed or constructed : STURDY <a robust plastic> d : capable of performing without failure under a wide range of conditions <robust software> 2 : ROUGH, RUDE <stories...laden with robust, down-home imagery -- Playboy> 3 : requiring strength or vigor <robust work> 4 : FULL-BODIED <robust coffee>; also : HEARTY <a robust dinner> 5 : of, relating to, resembling, or being a relatively large, heavyset australopithecine (especially Australopithecus robustus and A. boisei) characterized especially by heavy molars and small incisors adapted to a vegetarian diet -- Forget about my opinions, I started the thread very broad. In terms of "robustness" I still don't have the answer to it. I think it's a "Holy Grail" question for system traders, so I started the thread. As I mentioned above, I view market and structural risk as the same. Credit (Portfolio management/sizing) and Liquidity Risk is included too. I quantify risk to keep the performance and the decision process as consistant as possible. When I am under-performing, it allows me to pinpoint the problem, more than under a discretionary decision process. So getting back into topic, how would I define robustness? Very subjective but my answer so far is: A measure of the capability of performing without failure under a wide range of conditions - What measure? What conditions? What kind of performance? I'm too stupid to even know if I'm close to the answer or not. ARRRGGGHHHHH Also, I wrote this in the ATS forum because "most" discretionary oriented traders don't deal with it, or look at this topic as relevant to their trading.
"Ever ask the question: is it robust?" My answer: definitely! However I have my own ideas about what "robust" means, and my own opinions about what is crucial vs. what is trivial. They diverge from yours ... and that's okay ... it takes a difference of opinion to make a market after all. Is there one single "true path" which is best for all automatic traders of all ages in all financial circumstances in all countries around the world? I am uncertain.
I guess I'm being a bit too broad and subjective... How about I start off with historical backtesting and/or data itself. - Back in the internet bubble days, NatWest (now RBOS) made $100+ million loss, due to the options traders using the wrong data. Because they were using the wrong data, their options pricing model was giving the wrong figures, and ended up losing lotsa money. - On a personal experience level, I use Forex as part of my portfolio of products. The problem with Forex is they don't have a centralized exchange, this causes each individual liquidity provider with their own set of data. - Another experience is, around 4-5 years ago, a mainstream real-time data provider was filtering the data they stream to their customers due to bandwidth issues. Along with the data provider, depending on the hardware setup you have, there was a risk where your computer was not catching all of the data. It's pretty obvious that inaccurate data is a big problem for us, traders. During live trading, you'll be entering trades where you shouldn't be. When you're researching or running tests on historical data, you're going to be getting inaccurate results. Let's say you have a trading system that is working brilliantly, as it should. There can been times, where you would get outlyer trades just because the brokers are having a hardware problem. Which can then trigger your contingencies to cut your trading system. Along with the above, you still have the risk of the market always changing and systems always fading. The historical data you use may not be representing the live market condition. How do you all deal with the operating risks of market data? How you do you deal with the market risks of trading? How do you differentiate between the two?
Good trading systems should only loose their edge slowly. The speed with which a system degrades can be measured. Simply measure it then tune your scripts accordingly. Runningbear