Well, I have the zen-fire api.... working with it a little (i am in a crunch project a the moment). I like it from the data, but it is not complete. It lacks historical data (ninja has its own storage mechanisms). I do not really want a historical provider, but being able to get the last 5 minutes when your connection is flaky, or one needs a fast software reset... not good to need a data provider just for that. What I also miss is metadata (get a list of symbols for an exchange, get trading hours etc.) And finally I miss CORRECTIONS. Zen-Fire is not handling those at all. So, you get a tick, it gets cancelled, you never know it. I will use mirus for my own stuff, zen fire for data, and if things go as planned and hoped, will switch over to nxcore when the time has come (mostly when things work well enough to make sense. If Ninja 7 is really 5 steps forward I may go for them Working on my own stuff only on the side - I think I want something more flexible long term. If things work not out that well (which I assume at the moment), I will move part time into developing my own framework, all while trading and developing algorythms on ninja (6/7) while my own slowly comes up Thanks I have time and resources for that
Obviously this depends on your timescales/exposure etc. but if I ever lose my datafeed I just flatten all my positions immediately. No way will I risk having open positions when I don't know what's going on. Then once I get reconnected I just wait for the strategy to have collected enough data before it starts generating signals again. Not ideal I know, but how often do you lose connectivity? Once in a blue moon for me. P.S. I'd much rather have bad ticks than tick bundling. Bad ticks might screw with you once in a while, but tick bundling is like having a persistent slightly nasty smell in the room.
Well, I am not talking of long time. I mostly talk of "ok, I need to restart the connector", or "I need to reset my router". As a result, I take a 30 second downtime in my feed. Plus imagine - connection goes down for an hour. VERY unlikely for me (actually my trading computer runs in a high end data center), but imagine that happens. I call mirus for taking out my positions. But i loose an hour of tick data? Damn, not good. Backtests are screwed. Longer term indicators may mean I am out of the market for a trading day. With other providers the system could pull in historical tick data (nxcore does that atomatically), then simulate forward and enter a position when appropirate (i.e. the simulator has a position, and the entry can be done at that price or better). I just dream here I will popssibly integrate nxcore rather soonish.
You could take a look at Neoticker. Its got its caveats, like every other software, but i came to the conclusion its the best for me, out of the non-institutional products. You can use any COM-compliant language with it, including C# of course (that's what i use too). What i am doing is that i use a custom built data-server to feed data to Neo and a custom built execution server to process orders. I use Neo only for order-generation and nothing else, the execution server takes care of the rest. That way i don't have to write my own testing/trading framework but still have direct control over a large part of the whole process. In my experience, where most of the retail software fails is the broker-connection. Staying connected, recovering gracefully from an outage, syncing open positions/orders at the broker vs. trading system position after an outage, sanity checks before sending an order, and so on. Or supporting the broker/feed of your choice in the first place. That's the part i want to do myself, because that's where most of the reliability issues can be solved, and most of the catastrophic scenarios prevented. And Neo is one of the few retail-packages that allows me to do that.
Well, to add more to my with list: give me position, loss, order limits configurable on an ACCOUNT level. Assumption: the trade management platform is better tested than my strategy So, if I sa "not more than 3 contracts, daily loss X USD" in it, it can shut down a wild running strategy trying to find out how many contracts the account can handle before critical damage is done
It looks like there's only 4 or 5 users of tickzoom. A post just appeared on their forum that they're closing access to the public in 24 hours. They say they want to keep it secret and restrict access to the public. That's a very typical attitude of automated traders it seems.
I can only speak of the charts since that is all I use from NT. There are some major drawbacks. One of the most frustrating things about their charts is that trendlines do not line up or scale uniformly across different time spectrum. For example, say I draw a trendline on the 5-minute chart and the identical trendline is drawn on the 30-minute chart. On the former, I might see the high of the candlestick ABOVE this trendline; on the latter, the same candlestick might be BELOW the trendline. There are many other annoying glitches, which were all promptly raised in the NT forum. But the answer is always the same: wait until the next release.
Well, Chaostheory, it is simple. I would like to be able to set risk parameters outside a strategy. if I have a stupid error in my strategy order handling and the things tarts opening contrats like mad (like a stupid loop), then the outside risk managemnt would kill the strategy before it wears out the margin on the account.
Ahh. Got it. That makes a lot of sense. You probably need something open source to do that or else just roll your own entirely. It seems there's a guy adding that exact concept into one of the open source trading platforms out there in C#