That was Jan 2010 the blog was written. Wonder there has been any update to his quest of stock chart. Meanwhile, anyone has had experience with infragistics in the line of commercial chart?
I remember that "open source" project TICKZOOM, afterward he decided to commercialize it. However, it did sparkled quite some good discussion. Apparently the developer of TICKZOOM didn't learn anything from QuoteTracker's success.
I checked the reviews of quotetraker here (http://www.elitetrader.com/so/?action=view&SR_ProductID=33). It looks like quotetraker focuses more on visually technical analysis and to have convenience to manually trade the security. I did not see anything very related to fast backtesting and algorithm trading.(Tell me if I were wrong) From the thread of creator of TICKZOOM, i noticed that he focused on how to quickly load data and make the backtesting algorithm going through the tickdata faster. It is quite different angle of view on the trading. How doyou think?
Hi shewenhao By A mistake I was logged in with my old user... I've experienced some weird behavior here at ET for a couple of years ago My primary user should be: januson , you can find me at traderslaboratory and MultiCharts forum as well... and probably other places Anyway.... I'm developing my own Unfortunately I don't have very much time, so progress is not what I defines as fast The least of my worries are the execution speed, . Take a look at my speed tests here, where I compare TickZoom against my own: http://www.elitetrader.com/vb/showthread.php?s=&threadid=229155 The test are very simplistic and was only supposed to give myself a hint of the event framework that I developed. In between trying to develop a high performance framework and drawing some candlesticks I suddenly realized that the charting is low priority.... I mean... what do I gain by looking at some charts when I'm actually trying to develop a piece of trading software Certainly it would be nice with fancy charting where I could see candlesticks/ OHLC, indicators, volume etc., but as I said, low priority, and when I've the time it should absolutely be decoupled from the trading engine. Every trading application I've seen/ evaluated seems to have a strong coupling against the charting and thereby dependent on the rendering speed. most of the vendors are probably doing it threaded, but that is still not the best solution imho I'm very inspired by fullautotrading, I think this is a nice way of illustrating how the trading engine/ algo performs: http://www.elitetrader.com/vb/showthread.php?s=&threadid=202078&perpage=6&pagenumber=5 , scroll to bottom to see an examle.
Hi shewenhao I wouldn't use db4o when I'm able to develo my own storage fitted for my needs. db4o is very intesting when dealing with a lot of different objects with huge graphs/ entities, but for something as simle as a Tick or Quote I would rethink it Take a look here : http://www.elitetrader.com/vb/showthread.php?s=&threadid=147374&perpage=6&pagenumber=9 , the second last post: (snippet) If we calculate with your above 5 tick properties, each property roughly can fit into 32bit, that would be calculated to 160 bits => 20 bytes +/- some Using 485218 ticks it would require roughly 9 mb, which easily could be written to a low performing disk in about 1 second. Pulling data out and create 485218 objects is about the same speed + 1 second for processing I would guess. Please note this is very low performing hardware, I'm sure a high-performing disk can do it 10 times faster
Dear januson, thank you for your reply. It is good to focus on trading but not on charting or other stuff. Our goal is to find strategy and excute it, isn't? I just have a simple question for you: How do you define the event framework? What kind of key components you think is necessary for this kind of framework? Something like what openquant did? Like you always make decision based on the event about pricebar changing, whether your order submitted, whether order get filled from your brokers? Like this? My idea is also to build some object including all aspect of strategies, like entry, check order, profit, exit. But it is certainly far from any kind of "Framework". Again, please share some ideas from your mind about parts of framework
Yes event driven, sounds almost like a sales person trying to sell something old in a new box Well... events aren't so mysterious as many people think. My little framework raises an event whenever the a tick is received, as a tick could require an action of some sort, i.e. price change, depending on the compression, Open, high, low, close, and/or volume and/ or simply the time , furthermore events are raised when orders are executed etc. One thing to bear in mind.. when you develop your tick-factory, the component that receives and handle the ticks, remember to work in the time-factor as soon as possible. The system must not stall when no ticks is received... I.e. the system receives ticks, each tick maybe goes into a compression / candle/ OHLC, the system should be able to Close the bar without waiting for a tick to close the previous bar, the tick that defines the time, time for next bar... understand ? The ticks must not define the progress of the bars/ compression
Yes. Very true. Event-driven sounds like rocket engine for most people. Very true. I agree with your way. When related data arrives, event should be fired and corresponding function should work on to update some object or data strcture. Here are something I do not understand. I know we should set up priority thing in thread pool to make sure that information from tick update should not be delayed for the sick of slippage. But what do you mean " the system should be able to Close the bar without waiting for a tick to close the previous bar" I simply do not understand the english here. By the way, did you ever use the tickzoom? How do you think about the current condition of this software?