Measuring ICE futures liquidity beyond the lit book
When 39% of cotton futures trading occurs via ICE implied order book, building an exhaustive view of executable liquidity becomes critical.
By Etienne Mercuriali, Director of Research, BMLL
As traders roll a futures position from one delivery month to the next on ICE, much of the liquidity in the new contract was never entered there as a direct order. ICE's matching engine derives firm, executable prices from related outright and spread orders and publishes them as implied liquidity. Any feasibility or cost analysis built on the outright book alone understates what a trader could actually execute.
For example, across the May-to-July 2026 Cotton No. 2 roll, a 25-lot marketable buy in the July contract was executable within ten ticks 66% of the time on the outright book, and 99% of the time once implied liquidity was considered. For a 50-lot order, feasibility rose from 15% to 83%.
Does your historical analysis of ICE futures markets consider implied orders liquidity?
ICE operates implied matching across its futures markets, deriving executable prices from orders resting in related contracts. An implied-in price is a spread price derived from outright prices in two delivery months. An implied-out price is an outright price derived from a calendar spread and an outright price in another delivery month. During a roll the second mechanism matters most: activity in the calendar-spread market generates executable liquidity in the incoming contract, even though no trader entered a native order at that price.
For pre-trade analysis, access to these two data sets affects estimates of accessible depth, sweep cost and order size. For post-trade TCA, it affects the benchmark used to judge an execution. An outright-only touch may not represent the best executable price maintained by ICE’s matching engine.
Two historical feeds capture the full picture
BMLL captures the market through two structurally different datasets. IFUS is the full-tick, message-by-message feed used to reconstruct the outright order book. Its trade records identify executions resulting from implied matching, and the book can contain implied quantities disseminated within ICE’s standard outright feed.
@IUS is BMLL’s ICE’s Full Implied feed. It is a conflated price-level view of executable implied liquidity rather than an order-by-order record of the spread and outright orders from which that liquidity was derived.
To construct the combined executable book without double-counting the implied quantity already disseminated in IFUS, we subtract the IFUS quantity identified as implied, retain the remaining native outright quantity and add the full Implied quantity once to reconstruct the order book.
A view of how much of the roll trades as implied liquidity
Our study period runs during the May to July roll from 8 to 24 April 2026 and captures the effective roll out of the May Cotton No. 2 contract. Over those 13 trading dates, May open interest fell from 108,703 to 569 lots, a reduction of 99.5%.
Implied matching represented 67.7% of notional traded in the May contract over our roll period. Across the wider February-to-May sample of the March, May and July contracts, implied executions represented 39.0% of traded notional.
We selected three specific dates for analysis, the 10, 13 and 14 April recorded the largest daily reductions in May open interest. Together, those three dates accounted for an open interest reduction of 57,546 lots, or 53% of the contract’s open-interest reduction over the study window.
Download chart
Figure 1. May 2026 Cotton No. 2 open interest; notional-weighted share of IFUS executions marked FromImplied; and continuous-session notional divided between native and implied executions. The shaded area marks the 8–24 April roll window, during which May open interest fell by 99.5%.
Source: BMLL Data Lab. Sample: 3 February to 22 May 2026. ICE Cotton future for March, May and July 2026.
The implied order conflated order book completes the liquidity picture
IFUS and the implied feed should not be expected to produce identical books. IFUS is reconstructed from full-tick order messages, while our implied feed is a conflated price-level representation of the full implied liquidity.
The feeds nevertheless overlap substantially at common prices. On 13 April, across 709 matched snapshots and 6,217 shared offer levels, The implied feed quantity had a correlation of approximately 0.90 with both native and total IFUS quantity. Quantity identified as implied within IFUS was present at only 3.7% of those shared levels.
The incremental value of @IUS consequently comes from completing a view of what is executable in the market.
Below the surface: filling a 25-lot clip during the roll is more feasible than at first view.
Cotton trades typically in relatively small lot sizes. Over the roll period, the typical May / July cotton trade was 1 lot, and 99% were 10 lots or smaller.
Up to about 15 lots, the on-screen book on its own has enough visible liquidity for 94% of trades to be tradable within 10 ticks.
The analysis of execution feasibility becomes materially different when considering orders above 20 lots. While the lit book would indicate that only in 80% of samples could a sweep order be filled within 10-ticks, the actual feasibility was up at 99.9% when the merged book was considered.
At 25 lots, pure lit feasibility seems to be 61.5%. The standard IFUS view, including the limited implied quantity disseminated within it, was feasible in 66.4% of observations. On the implied feed alone feasibility was 83.0%, while the merged native-plus-Full-Implied book showed sweep feasibility was actually achieved in 99.3% of cases.
At 50-lot size, native outright feasibility was 13.8%, the standard IFUS view 14.9%, the implied order book alone saw feasibility at 46.7% while the merged book showed a sweep order would be feasible at 82.6%.
An analysis built on outright liquidity alone understates the capacity available to a larger marketable order, and overstates how often an execution algorithm appears unable to complete a clip within a cost limit.
Download chart
Figure 2. July 2026 Cotton No. 2 marketable-buy feasibility by clip size across four book views: IFUS as disseminated, native IFUS outright quantity, @IUS conflated full implied order book and the merged native-plus-Full-Implied book. Feasibility is the share of one-minute snapshots in which the clip can be completed at a volume-weighted price no more than ten ticks above the contemporaneous IFUS ask.
Source: BMLL Data Lab. Sample: 10, 13 and 14 April 2026, the three dates with the largest reductions in May open interest. Analysis window: 02:30 to 14:20 ET. Feeds: IFUS and @IUS conflated implied feed. Instrument: ICE Cotton No. 2 future for July 2026.
How much depth do implied orders add, and at what cost?
The full implied feed adds quantity both at the touch and deeper in the book. At the best offer, native outright quantity averaged 2.75 lots and the implied feed averaged 5.73 lots, producing 8.48 lots in the merged book.
Within five ticks of the IFUS offer, native outright depth averaged 16.99 lots and the implied depth 18.88 lots. The merged book therefore carried 35.87 lots. Within ten ticks, average depth reached 31.28 lots in the native book, 34.60 lots on the implied feed and 65.88 lots in the merged book.
The practical effect is clearest when the native book is insufficient. At 25 lots, native outright liquidity failed the ten-tick test in 819 of 2,129 sampled minutes. Adding full Implied liquidity made 804 of those minutes executable, recovering 98.2% of the opportunities that appeared unavailable in the native book.
In those recovered observations, the median merged-book sweep cost was 3.28 ticks, equivalent to $16.40 per lot. The 90th-percentile cost was 7.36 ticks. At 50 lots, the merged book recovered 1,465 of 1,836 native-infeasible minutes, or 79.8%, at a median cost of 4.88 ticks, equivalent to $24.40 per lot.
More than 40% of the time implied orders are better placed than outright limit orders
Across the 13 roll dates we study the implied feed showed implied offers were inside the standard IFUS / outright ask in 3,844 of 9,215 matched one-minute observations, or 41.7% of the sample. When the implied feed showed the better offer, the median difference was two ticks.
Conclusion
An analysis built on outright liquidity alone understates the capacity available to a larger marketable order, and overstates how often an execution algorithm appears unable to complete a clip within a cost limit. Contact BMLL to arrange a demonstration and trial of both datasets, and extend your analysis of executable liquidity across ICE futures markets.
Download chart
Figure 3.Share of matched one-minute observations in which the Implied offer was inside the outright book (IFUS) ask, with the median improvement in ticks when it was inside. Across the 13-date roll window, implied feed showed the better offer in 41.7% of 9,215 observations, with a median improvement of two ticks. Source: BMLL Data Lab. Sample: 8 to 24 April 2026. Analysis window: 02:30 to 14:20 ET. Feeds: IFUS and implied feed. Instrument: ICE Cotton No. 2 future for July 2026.
Universe and method
The analysis covers ICE Cotton No. 2 futures for delivery in March, May and July 2026, using the IFUS outright feed and the @IUS Full Implied feed.
The roll window runs from 8 to 24 April 2026 and contains 13 trading dates. Over this period, open interest in the May contract fell from 108,703 to 569 lots. Detailed book analysis uses 10, 13 and 14 April, the three dates with the largest daily reductions in May open interest.
Books are sampled once per minute between 02:30 and 14:20 ET. At each snapshot, a hypothetical marketable buy in the July contract is swept through four book views:
- IFUS as disseminated;
- native IFUS outright quantity after removing quantity identified as implied;
- @IUS Full Implied quantity; and
- the merged book, consisting of native IFUS quantity plus @IUS quantity.
Clip sizes range from one to 50 lots. A sweep is feasible if it can be completed at a volume-weighted price no more than ten ticks above the contemporaneous IFUS ask. One tick is 0.01 cent per pound, equivalent to $5 per contract.
Who is the full implied ICE dataset for?
For research, execution and market-structure teams, the two datasets make it possible to reconstruct liquidity generated through the relationship between outright and spread markets. They allow users to measure how much trading results from implied matching, where Full Implied liquidity sits relative to the native outright book, how much it changes executable capacity and how it affects the reference price used for pre-trade and post-trade TCA.
Ready to see the market differently?
Discover how BMLL helps trading desks optimise execution and market insight
