Hi Runningbear No one can learn to program in 2 months, it takes years and years of hard training and constantly analyzing and improve your skills. You can compare it to trading However you can program something that maybe works, maybe it dont in a couple of months and maybe you dont need more than that. For instance... there is no need to learn how to build a Ferrari Car if all you need is a Volvo. To compare a programmer with 2 month of experience with a programmer with 10 years are to compare a 2month baby with a grown up teenager but still the baby maybe can crawl, maybe it can't.
I just wanted to comment on your comment. You are 100% correct. Nothing is more frustrating than spending 3-5 months on one of those retail packages and then finding out that it doesn't quite do what you want, or there's some very specific use case that isn't covered. Those packages are for people who are doing very, very basic technical analysis or very basic order entry systems, maybe even using market orders. This forum is too quick to suggest these packages for general use. They are almost all fail-whales in the long run.
Patience. Being a programmer is like being a janitor, except your "shit" is memory management and correct algorithm implementation. Instead of endless toilets, you have endless debugging sessions. Moreover, being a trader is like working at a recycling facility. You're sifting through other people's shit on a conveyor belt looking for a can you can recycle or something. Combine the two, and you are talking about a whole lot of shit. The typical human wants an easy life, not a whole lot of shit sifting. Just tell yourself you have to sift through the shit to do anything worthwhile, and the knowledge will fill your head.
My opinion on the retail products: the better ones like OQ or Neoticker all do a fairly good job of simulating low/mid frequency daytrading strategies where latency doesn't play a big role. They all work basicially in the same way, you write code that reacts on a new tick or new bar, can look at historical/rt quotes and act accordingly. The weak link often times is the marketdata and broker interface (and maintainance thereof), gracefully recovering from a network or broker outage and supported order-types. If you want something like that, there is no point in reinventing the wheel, especially if you are a one-man show you have to focus on trading. Its however possible to use an existing product and only redo the parts that suck. I.e. i use Neoticker only for analysis and decision making, and a custom OMS that handles all the rest (execution tactics, connectivity, a persistance layer, position sizing/risk etc.). Its much more manageable than starting completely from scratch.
I gather you don't use any part of NeoTicker order interface and send orders directly to the broker. Do you still run your trading strategies as NeoTicker indicators when traidng live?
Yes, decisions are made inside neo-indicators, who send out messages (using WCF) to my custom OMS application. Nothing is reported back to Neoticker tho. I found the neoticker order-interface is too simplistic, since it has to be the lowest common denominator of all supported brokers ordertypes.
I understand your trading systems are in C#. Have you done any performace benchmarks with NeoTicker compared to Delphi or C++? Or is speed of backtest totally immaterial for you? I disagree that NeoTicker order interface "has to be the lowest common denominator of all supported brokers ordertypes". It could provide information regarding what specific functionality is supported by each broker. Say, it would have methods like SupportsOrderModification, SupportsNativeOCOOrders etc. Further it could simulate more complex order internally similar to how order interfaces can simulate stop orders so that the stop is never placed with a broker. Only a market or limit order is placed by NeoTicker at the point the stop price is reached. Order emmulation is not ideal but would be a large step forward. My guess is just historically anyone who needed more complex order handling did what you have done... especially given NeoTicker is quite friendly to this approach. I assume TickQuest probably gave up on improving order interface given trading systems see orders rejected by a broker as "open".
Yes, i agree, they could be alot smarter about their order-interface. But there are some things i would want to do myself anyways and not rely on any TQ interface, such as basic sanity checks before sending an order. I didn't test .net vs. c++/delphi performance, i understand its significantly faster due to .net's COM interop being slow. C# gets alot better if you cache as many things as possible locally (i.e.not using Heap objects) and minimize calls to NT objects. It wasn't a big issue for me, since i dont test on tick-data much.