Question: Were you only looking at the ES ( or other futures contracts ) and not stocks? Just curious.
Thanks. I have been using the DTN "IQ-Feed" for equities on a Watch-List and charts displayed via Quotetracker and it has been working pretty well . . . much better than Ameritrade's "streamer" data-feed, that's for sure!
This has been a great thread...thanks to all that have contributed. This data feed area is certainly one where there has been few reviews and much vendor hype....a recipe for disaster for technical or systematic traders !
Yeah, it can be a REAL mind-field out there with far too much hype, and not enough meat that actually sizzles. One other thing that I noticed about DTN and their "IQ-Feed" is that their customer support staff is always fast, and well informed. This is certainly not always the case, and I can think of another data-feed vendor by the name of "BS-Signal" that has a business plan in which customers are treated as virtual beta-testers and downright "guinea pigs" . . . with one version after another version released with "bugs" and more "bugs". I used them several years ago and every time they had issues with "freeze-ups" and "latency" they would say that they were in the middle of an "upgrade" to their server farm. Funny how they never once sent out an e-mail to warn people of their impending "upgrades" or that such a transition would always occur during oh-so-critical market hours. Just ridiculous. To this day, people still have tons of issues with them. Buyer beware!
It seems to me that the ideal is to combine IB's feed with another feed. Using Sierra Chart as my platform I partly do this. IB has two feeds, the second one of which most people seem unaware of and many platforms ignore. The first is their live feed which sends a packet every 100ms (or 200ms, I'm not sure which) - its great because it never lags, and you see the last price in that 100ms but not so good because you might not see or record other prices in the 100ms and you don't get a count of every transaction so your ticks charts are "out." The second is their 5-second or true feed which sends a packet every 5 second which includes the OHLCVT for the last 5 seconds - its great because its accurate but not so good because its not in sub 5 sec intervals, its still not a true tick count and it comes at the end of the 5 seconds. So with SC I generate my time bars to include both sets of information. As the live data comes in my bid/ask and last price are updated. Also new bars are started on the live data (and old ones closed off). Then, as the 5 second bars arrive I correct for any inaccuracies in the bars caused by the possible loss of data in the 100ms intervals. So I get bars with true highs, lows, opens and closes and, never get lag. It seems to me that dtn or other 100% complete data could be combined with ib's up to date data in the same way. That way you get 100% good charts and timeliness. I've not bothered simply because IB data is perfectly tradable the way I use it for the strategies I've developed but it wouldn't seem all that hard for someone else to do it if they cared.
Believe it or not, you can incorporate multiple data-feed vendors in Quotetracker. For example, you can use one feed for all Watch-List quotes, charts, and time and sales while using another vendor for Level-2 quotes. DTN only charges the Quotetracker user $45 per month + the exchange fees. ( there is no $50 "start-up" fee for Quotetracker users ). Level One fees are only $4 per exchange for non-professionals. Nasdaq Level-2 is an extra $20 per month + a $9 exchange fee. Currently, the only drawback of using DTN's "IQ-Feed" is that they ONLY have Nasdaq Level-2 quotes, and no Level-2 quotes for NYSE. If enough people show interest, they will eventually change this.
Getting two feeds is interesting but not useful. Most packages can do that or you can run two instances (at least I'd hope they can). But it doesn't combine the two to produce one chart that is both timely and accurate. Can QT combine the two feeds to create one better feed? That's what you can do in Sierra Chart and means you can get ib's responsiveness + accuracy. One of the reasons I left qt years ago (to ts2000i originally) was that you couldn't do things - there wasn't a rich programing interface or dll interface. Edit: What I'm really arguing is that the trader wants to look at one chart and see information that is both timely and accurate.
How did you determine this ? In other words, how can you say it's accurate, yet the tick count is not "true" ? Plus, how can the 5 seconds "lag" have an effect on your trading system assuming your average time-in-trade is at least 1 minute ?
The tick count is not true. But the volume is correct as are the OHLC of the bar. So if you must have the same tick count as someone else then the data is incorrect. But you can use tick charts if you use them because they spread market action at busy times (they can just never spread it by more than 100ms intervals). The 5 sec data is used to correct for any missing hs or ls in the live data. Live data gives me the current update. If you are using 1 minute bars (say) to decide then most people use a break of the bar to determine the trade signal point ... and that price will possibly be inaccurate for up to 5 seconds iff the high or low was in that period. In the past I compared HSI and ES (two of the most active futures contracts) and found that 2-5 bars per day were in error with live data. There are 60 seconds in a minute so (assuming) the highs or lows are made evenly in the 60 seconds the chances are 5/60 that the high or low will be made in the last 5 seconds. Thus with truelive data you have an expectancy that 2/(5/60) - 5/(5/60) bars will be inaccurate for 5 second each day until corrected. That is, this will happen to between 0.8 and 2.1 bars per week. If you traded with 5 minute bars then it would be 0.6 to 2 bars per month that were inaccurate for a whole 5 seconds. Double that if you think the distribution is weighted to the last 5 seconds ... its still a very good result. Personally to have 1 to 2 bars per week incorrect for 5 seconds is ok for me. And if you trade something slower like spi or stw (as i do now) it might be 1/10th of that. But, the question missed my point: My point is that you could combine IB's live data with ANY data feed and get the benefit that it was as fast and free of lag as IB'd live data and was corrected for any error by the other feed ... with whatever lag was in that feed only impacting the speed of any required correction (which is only going to affect 2-5 bars a day).