Which Futures data feed is the quickest?
Today’s post is for futures traders. Specifically, I would like to talk about data feeds.
Have you ever wondered what happens when you click the mouse to enter an order? Where does the order go? How long does it take to get there?
Have you ever wondered if Broker A is faster than Broker B?
When you open a futures trading account, and the broker gives you a choice of routing providers, which should you choose? Is CQG faster than Rithmic? Is Trading Technologies faster than everyone? And what about TradeStation?
Or Transact, IQ Feed, Interactive Brokers, CTS?
Whom in the hell is the fastest? Does it matter?
The natural evolution of the Futures trader
As Futures traders, we all go through the same process. And we all ask the same questions.
Typically, we all begin with the assumption that we need the fastest futures broker. And then so we stumble through the internet, asking Google ‘long string’ questions like:
- Who is the fastest Futures broker?
- Which trading platform is fastest for Futures traders?
- Is Tradestation faster than NinjaTrader?
- Is Rithmic faster than CQG?
- etc, etc, etc.
Google will then serve up a mishmash of differing opinions and articles. You will find articles and opinions on some of the most popular trading blogs like Futures.io or EliteTrader.com. But these are just the opinions of random people. Whom is correct? Whom is full of BS? Can we trust any of these random voices?
Additionally, Google will serve advertisements with tempting headlines like, “Ultra-fast futures trading only .50 cents per trade!” or “Strike like a Ninja with Ninja Trader” or “Trade with award winning TradeStation!”, etc.
It’s all really confusing. But one thing we know for sure, faster is definitely better than slower. Another thing we know for sure…everyone claims to be the fastest and most award winning.
Do I need a faster internet connection?
Another question that Futures traders eventually ask, “Do I need a faster internet connection?”
Does it make sense to upgrade my internet service from Ultra Fast at $49 per month to Super Ultra Premium at $99 per month?
Eventually, this leads to “Should I use a virtual private server?” Or, “Should I install my trading software on a virtual private server located at, or near the exchange servers?”
Once again, really confusing questions with no easy to find answers. Yes, you will find all sort of internet chatter on message boards and online trading forums, but is any of this chatter even real?
Do I need a faster computer?
This is another fundamental question that everyone wonders about. Do you need the latest Intel processor with 7-cores, or does a 5-core processor work just fine?
Is an AMD processor faster than an Intel processor? How much RAM or memory do I need?
Do I need to have the latest and greatest video graphics processor?
Should I have 6 monitors stacked into a fancy looking ‘war chamber’? Does this massive display of visual information help me to better understand what is happening in the markets?
Once again, more confusing questions. Our default position is YES! We need the fastest, the latest and greatest, the most hyperactive jet fueled technology to give us slightest edge.
An industry with salespeople at every level
The Futures trading industry is vast and murky.
There is a salesman at every level. The trading platform salesman promises his platform using C# is faster than the fastest cheetah. That you must be using C#, or .Net, or Java, or Easy Language, or Python, etc.
The broker promises that all orders are “exchange level” and “our data servers are across the street, next to the taco shop, within spitting distance to the Chicago Mercantile Exchange!”
The computer salesman swears that a $2,000 personal computer with two graphics cards and a multi monitor display is going to give us the decisive edge!
The indicator salesman promises an algorithm pulled straight from the mathematics department at MIT or NASA. The faster the trading indicator processing time, the more profitable the trading. So they promise.
The internet provider swears that your orders will arrive sooner to the exchange by simply upgrading the internet speed.
It’s all so confusing. And everyone has their hand out.
All of us go through the same myriad of questions, at least on some level.
Today, I would like to talk about how we can clear the clutter, and focus on the most important question. Who has the fastest data pipe?
What in the hell is a data pipe?
I want to break this down into the most simple and basic explanation. The data pipe is the connection between the computer that you are using to enter and exit trades and the Chicago Mercantile Exchange.
In other words, when you are sitting on your computer, and you hit the ENTER button…what in the hell is happening to your order? And who is actually delivering your order to the exchange the quickest?
Example: Suppose that you have an account with Interactive Brokers and you are trading the Emini SP500 futures contract. You place a trade on Interactive Brokers…how long does it take before that order is processed and appears at the Chicago Mercantile Exchange as a live order?
Or, suppose that you have an account with NinjaTrader and are also trading the Emini SP500 futures contract. You place a trade with Ninja Trader that is using the CQG data pipe, how long does it take before the order appears at the CME as a live order?
Should you even care? Is the question even important? If you are day trading futures contracts, this is absolutely critical.
Some might argue that it is really not that important. That a few milliseconds here, and a few milliseconds there is not going to make a big difference on any individual trade. However, I would argue the exact opposite. The cumulative effect of hundreds of trades, over a long period of time could be the decisive factor of whether you are a net winner or yet another loser. Speed matters.
And another question? Who is policing these data pipes? While everyone claims to be the fastest, with the lowest latency and lowest cost, where are the independent studies that either prove or disprove?
With no solid answers to these questions…none of these companies will be held to account. Why should data pipe ‘A’ invest in additional infrastructure or bandwidth without the proper motivation?
Beyond knowing who is the fastest, I want to know who is the slowest! And I want the world to know that they are the slowest. To embarrass them. And put them on full display, in a way that only TradingSchools.Org can.
Test Them All
This is where I need help from the audience. I need to know who has the largest market share. I need for TradingSchools.Org readers to complete the following survey. Which contains only one question…which data provider do you currently use? The survey will give me a priority list on whom I should be focusing immediate attention.
If you are not currently trading, then select the last used.
The Testing Procedure
If you are looking for an answer today…its not happening.
Instead, this test is going to take several months. But, I will be rolling out the results in monthly installments.
The reason this is going to take so long? I need a live trading account set up for each data feed. Simulators will provide no useful information.
Once the live accounts are set up, then I will be setting up automated order entry ‘bots’ that will trigger both limit and stop orders based upon predetermined price levels.
For instance, 5 trading machines with identical hardware and software, using the same internet connection will attempt to enter hundreds of trades over a sample period of time.
Each of these orders will be time stamped by the exchange. We will simply report who was able to register a working order at the CME, and the average time placement for each order.
At the end of the test, we will be writing an authoritative guide on which data feed is consistently the quickest.
Someone should have done this, a long time ago. Thanks for reading. Would love to read your opinion below. Am very curious who the audience believes is currently the fastest data feed.
What are you talking about – 7 cores and 5 cores CPU? Do you think that’s what Intel i7 and i5 stand for? This really made me question your entire article. Because you’re completely wrong, my friend.
How goes the testing?
https://www.tradingschools.org/reviews/battle-of-the-data-feeds-and-how-to-waste-1302-on-trading-commissions/
Good work,
I’m looking for a good broker to scalp through.
Hopefully reduce the slippage I experience with my broker currently.
Thanks for the Tradovate suggestion.
How’s the test going ?
Hi M. Moore,
It’s been 18 month since you’ve began this test, nice and welcomed work.
Even if you don’t charge audience for the result it sounds weird no result came out after such a long time….
Trading Technology’s will win hands down, as it is used by the prop houses and investment banks. It is the professionals choice. Their ladders have the fastest data-feeds and execution. They are not really chart providers, but do give charts. It is execution based software, and the support is FAST.
I use TT and I will never go anywhere else, as there is now-where else to go.
I trade ladders, no charts. Pure orders in and out looking at the flows. This is their edge and they have no equal at the time of writing. If you want to trade the book with no charts, then this is it.
Here is a video of my personal trading
https://youtu.be/CS3KvuqlOG4
I conducted a very unscientific study. TT and Rithmic had a tiny edge over CQG. It wasn’t even worth writing about.
But I can tell who is the slowest! Interactive Brokers.
Any results? Maybe a small update would be nice to read about. Thanks for testing!
Latency discussion. If any one is using a data feed on remote computer/server not co-located in Chicago for CME tradematch- Aurora, ICE-Chicago-Cermak. They should leave this conversation. Latency to trading server, to cme gateway and then to tradematch also know as tick to trade. FYI, if you are measuring in 200 milli-seconds. You should really think if you are a low latency trader. The fast guys are measuring in micro-seconds.
Data pipes….are like a fire hose containing data. Low data, minimal flow easily handled. when at capacity, only so much can go through the fire hose at full capacity. But if book updates, jam the pipe or hose, data packets are lost because the pipe and or your server cant keep up. This is analogous to everyone getting into a sporting event or movie theater at the same time. Everything is backed up.
Lets say you have a fast data pipe for the latency you believe you require. Your system needs to process the market data, then transfer it to a cpu, where you run your strategy. If you have an intensive strategy multiple symbols and logic that takes time or increases your latency. If you are running the strategy on an excel plug in….you do not have a latency trade you are too slow and should make sure your strategy can account for this lag.
Lastly, the true fast guys will utilize a direct ilink, decode raw market data, traverse the pci express, build books and maintain status while calculating strategy, get a decision, traverse the pci express again and release the order to the exchange. ultra low latency players do this under 5 micro-seconds using hardware/software. All hardware on an fpga is sub 1- 3 micro-seconds.
Curious about the results, but found out that feeds, technical equipment, level 2, and overcomplex chart are not needed for trading success.
I have interactivebrokers and tradestation and IQfeed.
Love Iqfeed.
I’ve used CQG and Rithmic and I trade tons of news events. Rithmic beats CQG hands down!
Great piece.
I have been with 3, Rithimic , CQG, IQ .. Rithimic was best overall, but when volumes are high and volatility is nuts, they all bog down almostly identically.
As far as hosting closest to the exchange … unless you get dedicated gear with multiple NICs, it wont make a difference.
Many of these hosting services provide VM’s (virtual machines) which are segmented portions of large gear. So although you may be right next to the exchange, if your on a VM with 500-1000 other bozos, your throughput wont be what you’d expect being close to the exchange. In fact you can do better further away if the network & hardware is right.
The cost of a dedicated host for what could give you an edge, would price most people out unless your a huge swinging DICK trader …
Every one has made some great points, and should lead to a great follow up .
Emmett, good article, and many valid points. I would like to add a few things that your readers may find interesting as well.
1) A trader needs to distinguish between executable and non-executable data. For example. There is no point of comparing DTN IQ and Rithmic because DTN does not send orders to the exchange. This leads me to the second point.
2) When comparing between data providers, you need to consider the data you are receiving. In other words, latency and execution are important but the quality of the data you are receiving is equally important. The quality of your data allows you to build and backtest the method you are basing your decision on.
3) Some platforms allow you to change data, and some do not. In essence, there are data neutral platform, and some are not. So the functionalities of the platforms have to be taken into considerations along with the data.
If someone has a question, I would be more than happy to answer.
Regards,
Matt Z
http://www.optimusfutures.com
There is a substantial risk of loss in futures trading. Past performance is not indicative of future results.
I’m with Matt on this one. The majority of data providers focus their efforts on being good brokers, having good trade execution…etc. But the quality of their data is an after thought.
Take IB and IQFeed for example. IB’s data is an aggregate Snapshot feed. Instead of sending every single trade to the tick with true volume, they send bundled data every ~250 milliseconds. IQFeed is the only (that I’m aware of) vendor that doesn’t try and cut costs on their telco fees and sends true data.
If you are doing robot day trading then I would think you want to rent a server right next to the exchange to get fastest execution. Also in addition to the reasons you gave speed is important as it gets you into the cue faster. So on a touch you might get a fill while the person 2 ms behind you does not.
Most my executions these days is via IB so it will be interesting to see how it compares to the rest. Are you testing stocks and futures or futures only?
If you are trading a “robot” I would use direct API, and yes, I would use a server. However, There are two locations in Chicago for hosting: Cermak and Aurora.
You need to decide where to host your method.
Thanks,
Matt Z
http://www.optimusfutures.com
There is a substantial risk of loss in futures trading. Past performance is not indicative of future results.
First, we really need to answer the question regarding Futures. And it should be pretty simple because we are only talking about one exchange.
But stocks, that’s going to be another animal. Multiple venues and liquidity providers. Yikes. I wish everyone would use IEX Exchange. Its the equivalent of the CME, but for stocks.
I will tell you there are huge differences in latency issues when the market is extremely volatile. I’m sure there have been some of you that entered a buy order with an oco attached only to be executed on the entry then to not know if you have been taken out by your oco or what happened to the trade because the data feed got so screwy because of the speed of data flowing through the system. It feels very bad to know you went long only to find out 30 minutes later that the price is 8 points against you and there was nothing you could’ve done about it. Speed is important, no doubt, but those latency issues can blow up your week during those volatile markets.
IMHO, Trading Technologies and CQG are the only feeds a pro should be trading with in order to stay sane. We shall see the actual numbers. This should be interesting. I can’t wait to look at the data.
Man, I have no idea how this is going to test out. Should be really interesting. For years I have wondered who is the fastest. Going to have a real horse race.
If I had to bet, I would bet that Trading Technologies is going win this. With Rithmic and CQG a close second.
The biggest fun should be the laggards. And if I were to bet on the slowest, it would be Interactive Brokers and Trade Station. I suspect that Interactive Brokers is routing everything through Connecticut. And Trade Station running everything through Miami.
We shall see. Setting up several retail trading accounts now.
Emmett,
With Trade Station I would have a stop in the system before a big econ release, it would trade through my stop down 10 points while my position still said 1 lot. I couldn’t get anyone to confirm if I still had the position so I would cover the position that my stop was supposed to accomplish. Then the volatility would slow after which my original stop execution would show up thus flattening me out. Unfortunately, I would be short because I flattened out my position that was supposed to be done by my stop. When I would call Trade Station’s customer service they would say how it was all my fault with the biggest attitude. They had the worst customer service along with biggest latency problems I have ever witnessed in the business. My friends from the CME and NYMEX use to tell me that was common with Trade station and why pros from the pit never used them.
This was many years ago, I don’t have any current experience with them. Emmett’s article brought up some really bad memories of terrible executions, bad feeds, and customer service with Trade Station.
I do not trade into uncontrolled situations. I execute on different platforms, and on every one of them, I exit unconditionally before news events.
Experience has taught me that the times that I get the lucky breaks are far fewer than the situations in which I lose money when trading through news.
Obviously, when day trading flatten up before a major report, but the volatility that comes around every so often can eat you alive when the platform fails to do its job. I’m not in anyway advocating playing a report for a measured move. That’s gambling insanity. I am referring to those volatile time periods where the market whips around enough to do some serious damage if on the wrong side coupled with the platform failing you.
Don’t feel alone, I think we all have experienced similar issues with platforms and data feeds over the years. I remember having the position I just bought immediately sold due to the platform’s buggy ATM.
This adds to my point when I tell folks Sim Trading and Live Trading is not the same.
Rob
I have been down similar situations due to sub par feeds. Well said, there are quite a few SIM trading superstars that never made it past AAA ball.
For futures it’s more Rithmic first then TT/CQG/IB (with an Interactive Brokers account in Chicago). While the IB data is not good at all the execution latency is surprisingly better than expected.
Interactive Brokers is not routing everything in Connecticut if your IB account is hosted in Chicago. It’s easy to verify this and if the account is not int the right location you can ask their support to relocate it.
TradeStation is not routing everything to Florida, only the risk management aspects (are this account allowed to trade this instrument, have enough margin, etc), the market gateways are in NJ/NY.
An easy way to test the execution latency (without real execution) is to write a small strategy which will send limit order far from the current price then cancel it while it’s in the order book, and see how long does it takes. Avoid this during news ;). In repeating this process during different hours and days of the week will give a good view of what kind of latency to expect.
The platform used will matters too (C API > C++ API > C# > Java > EasyLanguage) but the differences are not that significant compared to the total execution time. I should have kept examples programs for the CQG API, Rithmic API, IB API and NinjaTrader somewhere.
Last comment then I’ll stop spamming:
– there is no Intel “processor with 7-cores, or does a 5-core processor”, there are Intel i3, Intel i5 and Intel i7 but they have between 2 and 10 cores but never 3, 5 or 7 cores ;). In iX the X is not the number of cores.
– if it’s not too late for this I can maybe help you. I could provide identical machines and maybe help you to code similar testing strategies on different platforms.
– I agree that it’s not simple to test and it’s important to note that there are two separate things: the data feed on one side and the execution channel on the other one. Execution is not that bad on IB while their data feed is not good. IQFeed data are very good but their latency is as good as CQG for example.
– on a different but related subject it would also be interesting to measure trading software stability. I know we’ll have some surprises there…