If you are talking about hardware, that's a different discussion. I am talking about data structure. When IB is sending 3~4 aggregated quotes per second, that rate is constant, regardless of market volatility. Compared to a tick-based data feed, when market becomes more active, the data stream can get overloaded because there are more ticks to be sent. (there is no need to discuss this phenomenon, it has been dissected to death here on ET and else where.) Your computer can be a source of delay too, because it has to process ticks individually. Theoretically it can be done. But practically you are opening a new can of worm.
I'm completely with you on this. I had a post on the Ninja forum asking for RELIABLE NTP servers. Every server I've tried has been off by a few seconds....and I "sync" every hour. What do I have to do to get Win/XP to tell me the correct time ? It's totally amazing that in this day and age, computers can't tell the right time down to the second. WAIT ! I just checked and today, at this moment, I'm spot on ! FINALLY !!!!
The API derives time from the server time (same time that is displayed on the lower right corner of TWS), not the PC clock.
or the fact that like me, they cannot find any reliable NTP servers as well. Today, I'm off a whopping 30 seconds. Some days I'm "spot on". My current server is: 2.us.pool.ntp.org Someone on the NinjaTrader forums suggested the real problem lies with Microsoft's W32TIME api...which is the one that controls the updates from the NTP server. I'm starting to believe this now. I'm supposed to get updates every HOUR, but I don't think this is happening. IN FACT, I AM RIGHT ON THIS...HERE IS THE MESSAGE FROM MY EVENTS LOG: The time service is now synchronizing the system time with the time source 2.us.pool.ntp.org (ntp.m|0x1|192.168.1.6:123->131.128.183.247:123). This occurred at 9:30 this morning.....in fact it (the update)happened twice....SO THAT'S A BUG. Also, the update at 10:30 NEVER HAPPENED. Another bug. Conclusion: Windows/XP cannot tell time. Wow. Scary if traders are relying upon it.
I am using time.nist.gov. You can change the polling interval to be more frequent too: http://support.activehost.com/index.php?_m=knowledgebase&_a=viewarticle&kbarticleid=300
BEEN THERE, DONE THAT: 1) Used that server before. 2) Fooled around with polling before. IT STILL DOES NOT WORK RELIABLY. The W32TIME api is buggy....and the lack of technical documentation on it is the big "tell". No one at MSFT has looked at it in YEARS.
It is exceedingly difficult to synchronize your computer to any standard via the internet. Using a program such as D4 (http://thinkman.com/dimension4/index.htm) you can get a time sync from any of the world standards available via the internet. However, the length of time between the transmission of the sync info and its receipt by your computer is a variable based on internet routing and internet loading. This latency is not a constant. Further, IB aggregates and samples the quotes you receive. You are not getting an exchange provided feed, that is one of the reasons IB does not have to charge exchange fees for its data feed, you are not getting a bit-by-bit retransmission of exchange data. If you run TWS you've probably seen the time in the lower right corner of the TWS. As an experiment set your computer clock to an hour or two away from the correct time and then start TWS. Then shutdown TWS, reset the computer time to an internet-synced time standard and restart TWS. If you use D4 (or some other similar utility) the time difference between the timestamp you apply to the IB datastream and the actual time IB receives data from the exchanges and ECNs should be within 200 to 400 milliseconds depending on: The latency of the internet sync signal The latency of the quote stream from the exchange to IB IB's 100-300 millisecond processing delay The latency between your machine and the IB server. IB could time stamp its data but why should they? As a trader, even a fully automated trader, it should make no difference to you. The only reason I could think of for needing to know the actual time was so that I could compare the quote (or tick) time against my machine time. If I detected a significant deviation between the normal latency and the current latency I could stop trading if my system would be affected by such a change. I don't know for sure, but I believe IB's API does have a function that allows a call to the IB clock. You could use such a call to monitor latency variations, if it mattered. Jack
Told you, IB is using watches made in China. When I buy a watch, I look for two small words on the watch: Japan Movement.