Hi I have a little question... When talking about aggregating 11.000.000tick to hour bars, how many bars does that produce? 5years * 220 * 8 hour = 8800 bars?, Which means that one bar is consisting of 125 ticks, which again can translate to 2ticks pr. minut ? Which stock produce such small transactions? Wouldn't it make more sense to use 3-4 ticks pr. second? Just thinking
Excellent question. TickZOOM performance has opened up all kinds of new possibilities that we're exploring in answer to this question. First, here's the direct answer of the data used in that benchmark test. The 11 million ticks of data comes from a relatively reliable download of the USD/JPY currency pair for a 5 year period from Gain financial's website. That tick data has been filtered by Gain to only include every change in the bid or ask price. When I personally capture tick data from a live USD/JPY real time feed, I record a tick for every change to four levels of the DOM. That can be millions of ticks for only a week or less. Now, you can expect similar from any active stock if you want T&S (Time and Sales) data, plus volume and 4 levels of DOM in every tick. Why capture so much detail? It allows for true "market simulation" for replay or historical testing where you can "see" how the market really behaves and reacts to certain situations. Still, with TickZOOM we have been experimenting and developing ways to optimize even further. For example it now supports bar data, which means only 4 ticks per bar for the open,high,low,and close and so the data engine will fill in missing ticks for limit and stop orders between ticks. There's several more ways of optimizing or compressing tick data in TickZOOM. That's probably enough info for this post. Wayne
One more thing to answer about the bars. 11 million / 5 = 2.2 million per year 2.2 million / 220 business days = ~10,000/day We use this benchmark on TickZOOM, to verify that any new changes haven't adversely impacted performance. And as a comparison to other platforms (we haven't found one that comes anywhere close yet.) When running it, we use hour bars, but the data engine also generates daily bars, session bars, weekly bars, monthly bars, and yearly bars. Those can't be disabled because parts of the platform rely on those to statistics reporting. Since USD/JPY is a 24/5 market, that means about 416 ticks per hour. However, there's far fewer ticks per hour in the evening EST and much more in the morning EST during London and New York business hours. In fact, I would have to check but I may have filtered the file to only include ticks during the New York and London market hours since that's only when I personally trade. Sincerely, Wayne
Hi, Concerning Tickzoom, you told that it can backtest faster than any backtester on the market ... Just to get the idea, could you tell me how long it takes to backtest a simple strategy for a year for instance on a standard PC ? Thanks a lot, Manu
Sorry, what is a standard PC? And what kind of data? Are you referring to tick or bar data? It might be far easier to run the software on your PC yourself to find out. It comes out of the box with 12 days of tick data of 11 different symbols. And if that's important to you, I can send the "benchmark" data we use to test performance in each version. It has 11 million ticks and takes less than 60 seconds on my PC here which is a recent Pentium (not Xeon which is faster) and quad core but this was using only a single CPU. I don't know how that compares to yours. Sincerely, Wayne
The project needed funding to focus on full time development. Until May I was working a "day job" full time and tickzoom was a weekend and night job. The progress of adding features greatly accelerated after a hedge fund offered $15,000 to get a copy and to allow me to work on it full time. So since June it has fixed all the bugs from beta and the hedge fund as a near full time developer testing and converting strategies to verify the results. Plus it has added multi-symbol support and support for bar data. Most people were kind of surprised to find it didn't support bar data originally. Are you? I personally have no use for bar data. Tick data revolutionized my trading and I won't personally use bar data any more. But we added support for it and that has been tested so that it works with bar only data but can still be switched over and run live on tick data. Right now it has several customers that are hedge funds and several prospects that are also. After asking the customers and prospects they do not wish it to be open source and free. As you may know hedge funds operate "more secret than the CIA" so they're happy to keep it a pay only service. But it's still open source to paying customers. We're now calling it commercial open source. Which is a more accurate term. But open to other terms to be more clear. Sincerely, Wayne
You hit the nail on the head. Money had to come from somewhere. In some industries corporations fund projects and encourage them to remain open source and free to the public. Unfortunately, nobody in the financial industry is so generous. I personally don't mind giving away the code for free. The bad part is the time sink in supporting and answering questions, fixing, or enhancing for non-paying users. That's why we created a members only site and forum for that kind of communication. Sincerely, Wayne
Perhaps some posters assume trading systems should be given away for free because of one or more of the following lies: 1. Elite Trader is free, isn't everything discussed herein also free? 2. The burden of proof is never on the trader to prove a system. 3. America is going socialist, so the taxpayers should pay for it. 4. Professional traders are too cheap to keep their options open. 5. ? I know that I would not give away my systems for anything less than my retirement. I might rent the signals out, though, for the right price, and not some $499/mo either.