Best Futures Data Feeds Compared (2026)

The best futures data feed depends on the job: the prop firm's own execution rail (commonly Rithmic or Tradovate) for funded trading, a full-depth feed such as Rithmic, Sierra Chart's Denali or dxFeed for order flow, Databento for backtesting, and Barchart's free delayed quotes for cheap charting. A futures data feed is the stream of trades and order-book changes a vendor relays from the exchange (for US futures, CME Group's Globex platform) into charting or trading software; an execution rail is the broker-side system that also routes orders and enforces risk limits.

Key takeaway

For a funded trader, the prop firm's execution rail decides the futures data feed, so the choice is usually made already. Between feeds, three things change what a trader sees: whether trades print as exchange summaries or as individual fills, whether updates are conflated in fast markets, and how far the data travels from CME's Aurora data center.

Futures data feed comparison sheet (2026)

The sheet uses only each vendor's own pages as read on 27 September 2026, and prices move with CME's fee schedule. Tick history = (27 September 2026 minus the stated start date) / 365.25 days; where a vendor gives only a month or a year, the count starts on its first day, so the figure is a maximum.

Provider (type)	Depth	Own claim on completeness	Tick history (computed)	Pricing model	Common connections	Best for (our judgement)
Rithmic (execution + data rail)	Full depth plus MBO	Says its feeds are un-throttled, with no filtering during volatility	From December 2011: up to 14.82 years	Through the FCM or prop firm	Over a dozen platforms by its own count, incl. NinjaTrader, Sierra Chart, Quantower, Bookmap, ATAS; listed by Apex	Prop-firm execution, order flow
CQG (execution + data rail)	Not stated	No unfiltered claim	Tick and best bid/ask history; no start date	Live via broker; Data Factory history "Price per symbol, per calendar month"	CQG platforms and broker-supplied software	Long daily history through a broker
Tradovate (platform and brokerage brand of NinjaTrader Clearing)	DOM via API; top of book with a funded account	No unfiltered claim	Tick charts via API; span not stated	Top of book listed as a plan feature for approved, funded accounts	Tradovate apps and API; listed by Apex	Prop-firm execution
dxFeed (data only)	Full Order Depth stated for ICE Futures; CME Market Depth sold to retail (levels not stated)	"Tick-level"; no unfiltered claim	Service from 1998: up to 28.74 years (all assets, not CME tick)	Historical by request; retail through platform bundles	ATAS, Bookmap, Quantower, NinjaTrader, MotiveWave and others	Order flow, research
Databento (data only, API)	MBO (L3), MBP-10, MBP-1	"Tick-by-tick with full order book depth"	CME from 2010-06-06: 16.31 years	Historical per GB ($125 free credit); live plans from $199/month	Developer API	Backtesting and research
Barchart (data only, delayed web + API)	Not stated	No such claim; web quotes labelled delayed	Tick, minute or daily via API; span not stated	Free delayed web; paid memberships; API by quote	Barchart.com and OnDemand API	Cheapest delayed charting
Sierra Chart Denali (data only)	Up to 1,400 levels per side; MBO on Package 12	"Tick by tick unfiltered data feed."	CME tick from 2011: up to 15.74 years; 1-minute from June 2008: up to 18.32 years	Integrated package $36 to $56/month plus non-pro CME fees $2.00 to $40.50/month (funded account required)	Sierra Chart only	Order flow on Sierra Chart
DTN IQFeed (data only)	Not stated on the pages checked	"TRUE, tick-by-tick datafeed"; "completely unfiltered"	180 days of tick; 1-minute E-mini from Sept 2005: up to 21.07 years	Core Service $108.15/month; "Exchange Fees Extra for Real-Time Data"	Developer API for third-party software	Tick charts that need only recent tick history

Each use turns on one column: execution on who carries the orders, order flow on depth (market by order or 1,400 levels per side, against "not stated"), research on history span and price per gigabyte, and cheap charting on whether a 10 or 15 minute delay is acceptable.

No feed wins every row. The right column shows what decides each one, and for execution on a funded account that decision was made when you chose the firm.

Which data feed do prop firms use?

Prop firms supply the feed of the execution rail their accounts run on, so a funded trader's data choice is made when picking the firm and platform. Apex Trader Funding's homepage (September 2026) lists Rithmic, Tradovate and WealthCharts among its platforms and notes that certain exchanges are available only on the Tradovate platform. Topstep's homepage names its own TopstepX platform rather than a data vendor. Rithmic's page for funding evaluators and prop firms offers server-side risk enforcement for every account in a firm's book and states that Rithmic processes approximately 20% of daily US futures volume, a figure it does not source. Firms on in-house platforms differ in ways the Rithmic vs ProjectX data feeds comparison sets out.

A trade copier does not change the feed. When one master account is copied to several firm accounts, each copy still executes on its own firm's rail and data, so a copier (Thor, this blog's own product, included) is not the fix for chart data a trader dislikes; a separate data subscription on the charting platform is.

What is the difference between Rithmic, CQG and dxFeed?

Rithmic and CQG are execution rails that route orders to the exchange and send market data back, while dxFeed sells data only and routes no orders. Tradovate, the trading platform and brokerage brand of NinjaTrader Clearing, LLC (a futures commission merchant, or FCM, the firm that holds customer futures accounts), serves data from its own servers and belongs with the rails; Databento, Denali, IQFeed and Barchart are data-only like dxFeed, so each must be paired with a separate trading connection.

Only the rails can carry your orders, which is why a funded account comes with one attached. A data-only feed can drive your charts, but the trade still leaves through a rail.

The rails sell reach and speed. Rithmic's technology page lists Aurora among its Americas connection points and puts tick-to-trade latency (time from receiving a price to sending an order) at typically under 250 microseconds for its R|Diamond API, a service that needs Linux, a dedicated co-located server and a compatible FCM. CQG's about page says it "provides Direct Market Access to more than forty-five exchanges through its worldwide network of co-located CQG Hosted Exchange Gateways". The Rithmic data feed explainer covers that rail end to end.

The data-only vendors sell their source. dxFeed describes its live service as tick-level data taken directly from global exchanges, and Databento's CME Globex dataset page advertises data "Direct from colocation facilities" and "Tick-by-tick with full order book depth", which Databento says it captures at Aurora DC3 with an FPGA-based network card and hardware timestamping.

Why do tick charts differ between data feeds?

Tick charts differ because vendors print the same exchange fills in different ways (aggregation) and some merge rapid updates into periodic snapshots (conflation). CME's feed reports an aggressive order's fills as summary trades, one per price level, and a vendor either keeps each summary or prints one trade per resting order matched: Sierra Chart's Denali documentation lists "Combining of Sub-trades into Exchange Reported Original Summary Trade: Supported for the CME futures symbols", so both forms exist.

In an illustrative case, a 10-lot market buy lifts resting sell orders of 3, 3 and 4 contracts at one price. An aggregated feed prints 1 trade of 10 and an unaggregated feed prints 3, so volume is 10 on both, but a 3-tick chart closes a bar on the unaggregated feed and leaves the aggregated bar one third built. If the buy instead takes 6 contracts at price P (orders of 2 and 4) and 4 at one tick higher, a per-price summary shows 2 prints (6 and 4) and a sub-fill feed shows 3 (2, 4 and 4). Footprint trade-count columns change with the representation; volume columns do not.

Feeds differ in how many prints one aggressive order becomes, not in how many contracts traded. That single difference changes tick charts and trade-count indicators while leaving volume untouched.

Conflation drops intermediate states instead of relabelling them. Rithmic says most data feeds throttle or filter during high-volatility periods, and DTN's IQFeed contrasts itself with products that "provide a snapshot of real-time data", warning that their users "could be missing more than 50% of the actual trades". Both are vendor claims about competitors with no independent test cited; the conflation and throttling explainer covers the mechanics.

Does a faster data feed matter for retail traders?

For a retail trader, distance from CME's matching engine in Aurora, Illinois sets a latency floor that no choice of vendor removes. Vendors agree on the site: Sierra Chart states that "The connectivity for CME Group data is directly within the CME Aurora data center", Databento says it captures CME data at Aurora DC3, and Rithmic lists Aurora as a connection point.

Light in optical fiber travels at about 200,000 km per second, so minimum one-way delay (ms) = great-circle distance (km) / 200,000 x 1,000, doubled for a round trip. Distances are great-circle (haversine, Earth radius 6,371 km) from an illustrative Aurora point (41.7985 N, 88.2462 W) to each city centre and are rounded to the nearest km; delays use unrounded values. Real fiber routes are longer and add switching time, so every figure is a lower bound.

From	Great-circle km	Minimum one way (ms)	Minimum round trip (ms)
Chicago Loop	52	0.259	0.518
New York City	1,195	5.976	11.951
London	6,397	31.983	63.967
Frankfurt	7,007	35.035	70.071
Tokyo	10,118	50.589	101.179
Sydney	14,824	74.120	148.241

Even from Chicago, a retail round trip is dozens of times slower than the races professional firms compete in. For a retail trader, the feed's location matters far less than how its data is packaged.

Microsecond speed decides contests between co-located firms, not between home screens. Using stock exchange message data, Aquilina, Budish and O'Neill (2022) found that latency-arbitrage races happen "about one per minute per symbol for FTSE 100 stocks", make up about 20% of trading volume and most often last 5 to 10 millionths of a second, and that "Race participation is concentrated, with the top six firms accounting for over 80% of all race wins and losses." The Chicago Loop round trip of 518 microseconds is 51.8 to 103.6 times such a race and 2.07 times the 250 microseconds in Rithmic's tick-to-trade claim.

A virtual private server (VPS, a rented machine in a data center) near Chicago removes most of that floor for automated strategies. For a discretionary trader clicking by hand, aggregation and conflation change the screen more than a few milliseconds do.

What is the best data feed for order flow?

Order-flow trading is best served by a full-depth feed that shows individual orders: among these eight, Rithmic and Sierra Chart's Denali carry market by order for charting, dxFeed sells CME depth through order-flow platforms, and Databento sells market by order as an API. Market by price (Level 2) shows total resting size at each price; market by order (MBO, often called Level 3) shows every order event behind that size, including queue position.

Rithmic says its market-by-order data shows individual orders in the book rather than only aggregated volume at a price level. Denali, the deepest stated book, carries market by order data from the CME, CBOT, NYMEX and COMEX, and Sierra Chart's market by order documentation states that "Access to this information is limited to Service Package 12", which Sierra Chart's package pricing lists at $56 a month; adding $13.50 a month of non-professional CME depth fees makes $69.50 a month ($834 a year), or $96.50 ($1,158 a year) with all four CME Group exchanges at $40.50. dxFeed sells a CME Market Depth subscription and reaches order-flow traders mainly through platforms its site names, such as Bookmap, ATAS and Quantower. Databento's MBO is a developer API rather than a charting feed. The Level 1 vs Level 2 guide covers what depth adds to a chart, and dxFeed vs Rithmic compares those two feeds head to head.

Where can you get historical futures tick data?

Databento is the most direct route to bulk historical CME tick data among these eight: its CME dataset starts on 6 June 2010 (16.31 years of history) and is sold pay as you go at a price per gigabyte, with $125 of free credits at sign-up.

Sierra Chart's Denali has CME tick data from 2011 (up to 15.74 years) and 1-minute data from June 2008 (up to 18.32 years). Rithmic states that its tick-level historical data goes back to December 2011 (up to 14.82 years). DTN's IQFeed developer documentation lists "180 calendar days of tick" plus 1-minute E-mini bars from September 2005 (up to 21.07 years), so Databento's 5,957 days of tick history is 33.09 times IQFeed's tick window. dxFeed's historical service dates from 1998 (up to 28.74 years), but that start covers the whole service rather than CME tick data, and pricing is by request. CQG offers "Decades of daily data and many years of intraday data, including tick and best bid/ask" with no start date.

For backtesting, what matters is tick history in the resolution your strategy trades on. Long one-minute history does not help test an order-flow idea that needs every print.

For why event order inside a bar matters in testing, see tick data vs bar data for backtesting; recording your own tick data builds a free archive from a feed already paid for.

What is the cheapest futures data feed?

The cheapest futures data is free and delayed: Barchart's futures pages label their quotes as delayed by 10 or 15 minutes, with times in US Central Time. Real-time CME data costs least when it rides on a funded account. Tradovate's pricing page lists "Access top of book market data with an approved and funded account" as a plan feature with no separate data price beside it. Sierra Chart's Denali charges $2.00 a month in non-professional CME exchange fees for top of book (the best bid and best offer) on top of an Integrated package from $36, so $38 a month ($456 a year), or $42 ($504 a year) for all four CME Group exchanges, and those non-professional rates require "a verified live funded futures trading account".

IQFeed's Core Service costs $108.15 a month before exchange fees, and Databento's live Standard plan, at $199 a month ($2,388 a year), is a research budget rather than a charting cost.

Many traders need nothing beyond what their firm or broker already bundles, and a second feed pays off only when a specific tool needs depth or history the bundle lacks. The free real-time futures data guide covers the no-cost routes, and Barchart vs Rithmic compares a delayed web feed with a live rail.

Go deeper

Frequently asked questions

Is Rithmic data better than CQG data?

No public, independent test shows either one is better for a retail trader. Each rail's data comes with its execution, so the practical choice follows the broker or prop firm.

Where does Tradovate get its market data?

Tradovate does not name an upstream source on its public pages. Its API documentation shows quotes, DOM and charts served from its own md.tradovateapi.com servers, and the often repeated claim that the data comes from CQG could not be traced to any Tradovate source.

Which feed decides my fills in a prop-firm evaluation?

The firm's execution platform decides them, not a separate charting feed. Evaluation accounts are often simulated, with fills computed by the firm's platform from its own data, so a second charting feed may not match a fill exactly.

Can I chart on one data feed and trade on another?

Yes, many platforms accept a separate data connection and trading connection. Expect brief disagreements in tick counts and last price between the chart and the order ticket, because they read different streams.

Do I need Level 2 data to day trade futures?

No, top-of-book (Level 1) data is enough for strategies built on price charts, indicators and footprint charts of traded volume. Depth earns its cost for methods that read resting orders, such as DOM trading or liquidity heatmaps.

Is delayed futures data good enough to trade?

Not for live decisions, because every price on screen is already stale when an order goes in. Delayed data suits study, end-of-day review and learning a platform before paying for a live feed.

What is the difference between a CME and a Full CME Group data package?

A CME package covers only the CME exchange (for example E-mini S&P 500 and E-mini Nasdaq-100 futures), while Full CME Group adds CBOT, COMEX and NYMEX. Traders of Treasury, Dow, gold or crude oil futures need the wider package.

What counts as a non-professional data subscriber?

Broadly, an individual using data for personal trading rather than for a business or as a registered financial professional, with the exact test set by the exchange's subscriber agreement. The label matters because Sierra Chart's Denali charges professionals, and users without a funded trading account, $53 to $145 a month per exchange product, against $2.00 to $13.50 for non-professionals.

Does an 'unfiltered' feed show every trade the same way as other unfiltered feeds?

No. 'Unfiltered' is vendor wording with no independent standard behind it, and two feeds that both avoid conflation can still disagree on trade count if one prints exchange summary trades and the other prints sub-fills.