I keep seeing these advertisements for unfiltered data.....
Isn't the whole point of filtering the data to ensure that a rogue/erroneous data point won't ruin your day? If you have a stop in place and a bad piece of data comes in, that exceeds your stop, most platforms treat stop orders as "market if touched or exceeded" orders.
What's the whole issue with having "unfiltered" data?
"A dumb man never learns. A smart man learns from his own failure and success. But a wise man learns from the failure and success of others."
The following user says Thank You to RM99 for this post:
Site Administrator Swing Trader Data Scientist & DevOps
Manta, Ecuador
Experience: Advanced
Platform: My own custom solution
Trading: Emini Futures
Posts: 49,412 since Jun 2009
Thanks: 32,068 given,
96,713
received
Here is my opinion.
Filtering isn't about getting rid of bad ticks. It's about latency and completeness (accuracy).
Interactive Brokers is widely known for filtering data. It's snapshot data, meaning that they aggregate info, and send you a 'snapshot' of what it looks like every 250ms or so. This means you are not getting all the ticks, not getting all the L2 depth changes, not even getting all the volume. If you trade daily bars or even minute data, then you won't care about this. But if you are using tick based data: ie tick, range, renko, volume or other exotic bar types, then you should not be using IB for data.
On top of that, IB has big restrictions on their historical data. For this reason, my conclusion is that IB filters data simply due to lack of resources on their end, and probably because overwhelming majority of their customers are using minute based or daily based data.
IB is a combination execution + feed provider.
Next we have the Rithmic's and Zen Fire's of the world. They send every tick. They are well known in the industry as unfiltered data. They also have a reputation for being ultra low latency (fast). This is very appealing to a large amount of more 'advanced' traders that want to use tick based charting.
However, they don't have their own historical servers (some platforms provide it for their users, it varies). They also do filter some L2 depth changes. In order to be as fast as possible, and get you that tick as fast as possible, the trade off is to not send some of the L2 changes. Even most of the 'advanced' traders today are not measuring L2 changes so pretty much no one is complaining about this or cares about it.
Rithmic/Zen Fire is an execution/risk system more than it is a data feed. The focus on the feed is execution, which is also why the focus on the feed is latency. They do this extremely well. I've heard rumors for many years that Zen Fire is branching out and adding other exchanges or breadth tools soon. Maybe they will? But for now, it's really designed as execution only, although a lot of people use it for data as well and are completely happy with it, provided they also use a platform (like NinjaTrader) that gives them historical data.
Next we have the IQFeed's and Kinetick's (oem) of the world. They send every tick. They also have an extensive historical database and ticker plants. They all major exchanges. They have something like 500 breadth tickers.
IQFeed's (and Kinetick) business is data. Not execution. Since their focus is on data, they send every L2 update. They send every tick. They send everything, and they also send it in the right sequence. It is the best quality feed currently available to a retail trader, in my opinion. They do give up some latency when compared to Rithmic/Zen Fire, as there is a price to be paid for completeness.
IQFeed also provides historical bid/ask backfill. If you don't know what that is, you don't need it. Only some platforms support it, such as Sierra Chart, Investor/RT, MarketDelta. NinjaTrader 8 is supposed to support it. If you need this then IQFeed is not only an excellent choice, but pretty much the only choice.
As for TradingTechnologies TT FIX, it is an execution feed. It's focus is on execution, although many do use it for data as well.
As for which feed is right for you, only you can answer that.
Thanks Mike, so as long as I'm not scalping or involved in the weeds, then filtered data (like Tradestation) should be sufficient for my purposes?
My trading strategies aren't made or broken based on the accuracy of the next few hundred miliseconds or my order in the book at the next ask/bid.
My charts and indys are populated historically and use end of bar calc's, so I should be good right?
I'm not shopping, I was just curious and the whole data quality/filtered/unfiltered thing is new to me.
I would like to take advantage of data timestamped down to the second rather than the minute however, which would allow me to explore some multi-timeframe ventures on advanced charts.
So all that I've read about rogue/errant data points is just hype?
From what little research i did, I saw horror stories about people using unfiltered data and bad ticks coming in and screwing with their DOMs/Matrixes, etc.
"A dumb man never learns. A smart man learns from his own failure and success. But a wise man learns from the failure and success of others."
Site Administrator Swing Trader Data Scientist & DevOps
Manta, Ecuador
Experience: Advanced
Platform: My own custom solution
Trading: Emini Futures
Posts: 49,412 since Jun 2009
Thanks: 32,068 given,
96,713
received
If you are using TradeStation, then you can choose any data provider you want, so long as it is TradeStation. You can also choose any broker you want, so long as it is TradeStation. TradeStation is in all-in-one solution, and as such isn't really a leader in any field, in my opinion.
I feel the pain of all that, what I'm asking is that if I'm not scalping or dependent upon hyper quick calculations and execution if it's going to make an appreciable difference for me?
I guess for everyone's benefits, which "types" of traders benefit most from having unfiltered data and which types is it really a non-issue?
"A dumb man never learns. A smart man learns from his own failure and success. But a wise man learns from the failure and success of others."
The following user says Thank You to RM99 for this post:
Site Administrator Swing Trader Data Scientist & DevOps
Manta, Ecuador
Experience: Advanced
Platform: My own custom solution
Trading: Emini Futures
Posts: 49,412 since Jun 2009
Thanks: 32,068 given,
96,713
received
Having never used TradeStation, I can't say. But my understanding there is an option in the menu somewhere to disable filtering and send you all the ticks.
Asking about scalping is the wrong question. I mean, I guess it is possible to scalp on a minute chart. The better question is tick data, or not tick data. I cannot comment on TradeStation's data feed.
As for which traders benefit, I believe I covered everything in my original reply. ?
TradeStation is ok for time based charts [Longer term]
It doesn't build volume charts correctly [but close enough for govt. work]
{edit: didn't a few years back when we tested and compared ((7.x)). I no longer use volume or tick charts, so can not speak for current build} - but TS is TS...
If one requires higher quality data, then your recommendation for IQF is correct.
Regarding option to disable filtering and send you all the ticks, I never could find a specific filter, but believe in File/Preferences / Tradestation Nework preference / Streaming Data Optimization / "none" is as close as it gets.
SS's attached:
The following 9 users say Thank You to NoBias for this post:
Anything you could add about CQG? I guess the only thing I'd add that's worth taking into account with the data feeds is that with CQG you can store OCO orders on the server, which is not offered by Zenfire (nor Rithmic?).
I would imagine CQG would be comparable to Rithmic / close to IQ Feed ? Or would I be wrong ?
Site Administrator Swing Trader Data Scientist & DevOps
Manta, Ecuador
Experience: Advanced
Platform: My own custom solution
Trading: Emini Futures
Posts: 49,412 since Jun 2009
Thanks: 32,068 given,
96,713
received
It's going to look different, because TDA is a filtered feed as far as I am aware, while Zen Fire is unfiltered tick data.
As for "manipulation" -- this is futures, not forex. In the futures world, it's a centralized and regulated market. Brokers filter their feed in order to cut bandwidth and resource costs of delivering unfiltered data, not in order to some how game the system and make profits from "manipulation".
Site Administrator Swing Trader Data Scientist & DevOps
Manta, Ecuador
Experience: Advanced
Platform: My own custom solution
Trading: Emini Futures
Posts: 49,412 since Jun 2009
Thanks: 32,068 given,
96,713
received
You need to read post #3. "Filtered" is really the right word, not "inaccurate". But luckily for everyone, you can choose your broker and your data feed, and pick the one that you like best.
For example, I do have an IB brokerage account, but I do not use their data.
well, this is interesting. i have been scratching my head, indeed tearing my hair out trying to understand tradestation time and sales "filtered from chart" and "out of sequence" conditions. i don't know why it is impossible to get a definite explanation from the trade desk. they say it "might be" part of a "spread". that doesn't explain why the prices are obviously old however - prints at the *previous day's* price levels nowhere to be found on the matrix.
perhaps it is due to this "filtering".
Du sublime au ridicule, il n'ya qu'un pas. ~Napoleon Bonaparte
Site Administrator Swing Trader Data Scientist & DevOps
Manta, Ecuador
Experience: Advanced
Platform: My own custom solution
Trading: Emini Futures
Posts: 49,412 since Jun 2009
Thanks: 32,068 given,
96,713
received
Kinetick is just a NinjaTrader branded version of IQFeed. Yes, it offers historical bid/ask. However, NT7 doesn't support historical bid/ask, so if you want historical bid/ask backfill in NT7 then no data feed is going to help you. Instead, you need GomRecorder.
If you are not using NT, then you need IQFeed (Kinetick is NT only). With Sierra Chart as an example, there is historical bid/ask backfill with IQFeed.
Elite Members receive a discount to IQFeed. Click 'Elite Members' top of any page for all the discounts.
I trade with Range bars and I would think that even if IB data is filtered....it would still spit out the new range tick data so that it should be fairly accurate. Even if there are 10 missing ticks....one tick should still print the range value within that bar to make it accurate?
I was hoping I could use range bars with Ninjatrader and an IB feed.
Site Administrator Swing Trader Data Scientist & DevOps
Manta, Ecuador
Experience: Advanced
Platform: My own custom solution
Trading: Emini Futures
Posts: 49,412 since Jun 2009
Thanks: 32,068 given,
96,713
received
It's snapshot data. You are missing ticks, and also only receiving them in a "burst" fashion (every 250ms if I recall, or maybe it's every 100ms). You aren't missing just a few ticks, you are missing a bunch of them.
If you use exotic bars, like range bars or non-time based bars of any kind, you really should use a better data feed.
I guess I will have to look at the charts side by side on the Ninjatrader/ AMP/CQG feed vs the Ninjatrader/IB feed on the range bars to see how bad they look. I just assumed that IB would at least print a high or low in a range which would print.... giving a similar looking bar...since its based on a range of a low and a high of a 15 tick range in may case.
It would just be convenient and keeping my machine lean and trim if I could package IB and Ninja using Bracket Trader to handle the details of my trade. Otherwise I will have to run AMP/CQG to fill the charts on one computer and IB on another computer to do the trades with Bracket Trader.
I'm a client with IB and using ninja. What is the cheapest way I can get unfiltered data for ES, CL, and forex? I know kinetick is an option but they are fairly expensive for me right now considering im a new trader and am not working with much capital.
Denver, CO, June 30, 2014 NinjaTrader Group, LLC has announced the launch of NinjaTrader Brokerage following their acquisition of Mirus Futures. The new brokerage will leverage the best of both award winning teams to address unfulfilled traders expectations …
I don't know what "use with IB" means. But it sounds like you think you can use a data feed other than IB with TWS, and that is definitely never going to happen.
If you mean "use with IB" as in IB for execution, and data feed for data, then of course that is fine so long as you don't use TWS. You can do it with Sierra Chart for example (which I use) or with NinjaTrader for example (which many others use). Both of which are off-topic for this thread, which is supposed to be about unfiltered data.
Hi is someone able to provide any more recent info on this:
I am looking at the options offered by AMP Global. I am looking for an unfiltered feed and the options are CQG , TTnet and Rithmic.
Latency of the feed would be less of a concern on my side. More of a concern would be all the ticks should be non aggregrated. Accurate Bid / Ask trade classification
Can anyone advise if these feeds are all unfiltered? Any views on the Bid Ask trade classification
I use the CQG from Amp option, as provided to it's Ninjatrader users (now only available to new NT7 users that have purchased their own NT7 license.). The feed is slightly different from the full (and much more expensive) CQG datafeed: it is made up of realtime incoming data and is provided by CQG directly - it is completely unfiltered, meaning you get all trade ticks and all the bid/ask info in that is available. This allows me to have completely accurate footprint charts showing exact buyers and sellers at every price level (which I record via GOM, for the entire opening duration of the market through each and every week). I have compared my footprint charts to the other datafeeds you have mentioned, and they are almost always exactly the same.
The difference with the Amp CQG feed for NT7 users is that whenever NT7 makes a request for historical data, it comes from NT's own housed servers, from data they have recorded for that instrument. They used to have issues where their servers would go down every so often on the weekend or an overnight and then when you request historical data, it is not available to you: this problem appears to have been rectified by them and it has been quite some time where this has now not been an issue.
The following 3 users say Thank You to Beljevina for this post:
Thank you Beljevina for providing some more insight.
I am wanting to use it with IRT who routes CQG data through the marketdelta trader platform. Interestingly I got a reply from IRT which said I should go with Rithmic.
With IRT there is the option to backfill through I believe DTN. The only concern with Rithmic is the additional per side pricing fee which adds up over time.
When you have tested you CQG feed, have you eyeballed it ( which I tend to do) or have you dug into data on different feeds with excel or something similar. What sort of discrepancies have you noticed or picked up with regards any of this?
I am still having a problem with NT7 & NT8 requests to the Ninja servers. They have been missing some data and I have been getting an error on the front month contract since the last week of December. @NinjaTrader siad there is a problem , but they do not have an ETA on a fix. It has been a week already so I don't think they are in a hurry.
I don't know enough about MarketDelta and how it does it's cumulative delta/footprint (if you are using that feature), and hence how it builds, collects, stores that data, either real-time or historically; I mention this as a contrast to GOMi's fabulous tools for NT footprint and delta, which I completely depend on, and find essentially flawless in their operation. Hopefully you have a technical understanding of why they would recommend Rithmic; I mention that because help-desks usually have a bias and often steer their customers one way, for non-technical reasons.
My CQG/Amp usage over the past 4+ years I'd say is somewhat beyond generic eyeballing: I've compared the footprint chart's bid x ask prints in entire bars, trying to focus in on what I knew to be fast moving periods in the market, and comparing colleague's charts either on SierraChart and/or MarketDelta, and matching up those bid x ask levels. I've done this for ES, CL, ZN, 6E and others. Yes, there are rare differences, but, on say a $CL print of 720 x 674 at $34.70, if it's 5 to 10 to 20 contracts more or less, I don't see a reason to be concerned; usually I can find the discrepancy in the tick above or below, which, is indicative of an incredibly fast moving market as well as uncontrollable software where the GOMi indicator 'slotted' the up/down ticks perhaps in the opening or closing seconds into a price level above or below, or, it is in the next bar if that was the opening or closing tick. In other words, because delta is not part of the datafeed, there can be minor software impacts that will make a 5 minute footprint chart appear different on NT vs SC vs MD. My sense is that it is better than 99% similar. And, if one is different versus the other (by 1% at an incredibly fast point in the market), the question is ... so what? Will it impact one's profitability? I don't think so. My advice is that there is a far greater spectrum of impact we as a trader introduce to the mix, making potentially bad decisions completely unrelated to incredibly small variances in a footprint or delta print: note I am assuming your chart's individual bar's high, low and other non-footprint data is completely the same. Also, things like OCO-server side, and latency frankly I feel are more important. I hear you on the pricing differences, but brokers can adjust those for their clients to be more competitive (just ask them) and, with ES at $12.50 per tick, I will put in limited effort/concern towards a 20¢ full-turn price variance.
I'm not aware of any chart gaps, but, I could be wrong and not seeing it, or, because I also run my NT7 during every waking minute of the futures markets and my charts will get built from historical data when I do restart NT7, perhaps I am not impacted. If you want to PM with details, I'd be happy to see if I'm experiencing the same issue, or, if I could help or somehow shed light on the topic.
The following 2 users say Thank You to Beljevina for this post:
You've obviously spent some time with this by the looks of things , and bring through some great understanding.
You point as well on the tick size and price per trade is a valid one as well. i sometimes tend to forget its just the cost of doing business.
Thanks for the lucid explanations, Good luck with your trading
This is a helpful post. I was looking at charts/data in tradestation and then going into Interactive Brokers to place some trades and I realized that the high and low of day columns did not match. At first I thought it was due to IB considering the premarket and tradestation using the actual trading day.
I then looked at 5 min charts in both platforms and those don't match either. I find that disconcerting since many traders use 5 min bars and there is no way to know which one is accurate.
I downloaded about a year's worth of daily prices of BIIB from both brokers' software and compared them. I am surprised to see that even on daily bars there is a decent amount of difference, particularly in the opening prices. I have attached the sheet with differences. Tradestation is the first set, IB is in the second set starting in column G and the IB Value less the respective Tradestation Value is the third set starting in column M.
As an aside, I find that Tradestation and ThinkorSwim data matches for whatever that is worth.
The following user says Thank You to mks212 for this post:
I am not exactly sure how to determine the precise price for an equivalent CQG set of instrument and data feeds; their page at Products | CQG, Inc. doesn't seem to specifically offer pricing solely for the datafeed, but may bundle it with their other products. In general terms and without doing specific research for you, the cost for datafeeds can be as little as $60/month (Kinetick?, but may not include markets like the Russell, Brent, FDAX, etc) and from past memory, Rithmic, which is likely upwards of $90). Bear in mind too, that a datafeed is one thing, but, entering and exiting trades is determine by whatever broker you choose, and thus their infrastructure to successfully route your trade to your instrument's exchange.
My point simply in my posts(s) was that there are many bundled solutions from brokers or other sources, that will provide for an incredibly high quality datafeed at minimal cost. And, my advice is to pick a reputable unfiltered data source, and not spend silly amounts of time exploring that avenue, when, it is the quality and success of each of our abilities to successfully trade in to and out of positions, that are the real cornerstone of making it or breaking it as a trader.
The following user says Thank You to Beljevina for this post:
I am using Ninja with CQG and I trade with a 6500 volume chart, during the day I will reload the historial data maybe 5 times to make sure I am getting all the correct info. I copy my chart and look the next day at the trades and always the chart candles change. So if I get Kinetick can do I need 2 ninja lic to run one with CQG to trade and one to see Kinetick feed to see the correct data? If I get kinetick when I download the historical data with the chart look the same tomorrow as it will live?
Thanks
Happy Rick
Your tick chart is (slightly) changing every time for different reasons; there is nothing wrong with the data. Rather, when dealing with tick and volume bars, each bar is made up of a fixed number of ticks or trades but they are different trades because the chart is started based on what is defined under "Data Series" to load for example 7 days worth of data; if you are loading lets say a 500 tick chart, (I believe) Ninja calculates backwards from now, going back exactly 7 days, and starts plotting that first bar with whatever the price was exactly 7 days ago. But, each time you reload or bring up even the same instrument on the same value tick chart, because time has advanced, you are prone to get a different starting tick (thus a different traded price) for each and every bar on your tick bar chart, and hence the high/low/open/close will be at different points, and bars will have different wicks on both the top and bottom of the bar.
Unless you are having disconnects from your datafeed, I don't see a need to reload historical data, ie 5 times per day. The fact that you are seeing different bars, just represents different trades, from varying bar start times, being stuffed into each bar, and the actual trades going into each bar, are prone to change. Hope that makes sense.
This post is a bit old and was just wondering what is the fastest data provider for NTB. I don't see an option on their site for Zen/Rithmic right now so what would be the next best choice?
Your excellent data about data feeds for tick charts was exactly what I was looking for. Since this was originally written in 2009 do you know if this data is still accurate? Any updates you know of would be appreciated.
Thank you!
The following user says Thank You to Larryeh1 for this post: