Build an open interest profile and state its limits

Allocate OI increases to price buckets and keep decreases apart

Start with a fixed time range and price bucket size. One transparent teaching method assigns each positive OI change to that interval's volume-weighted price bucket. Another distributes it across traded prices using volume weights. These methods can produce different profiles because OI is sampled over time, while trades occur at many prices between observations.

A positive sampled change is already net of openings and closures inside that interval. For example, 200 newly opened pairs and 80 closed pairs produce the same +120 observation as 120 openings with no closures. Here “additions” means the sum of positive sampled changes, not gross new contracts. The sampling interval therefore affects the profile even when the underlying trades are identical.

A net decrease is harder to locate. The feed does not identify which earlier cohort closed. Removing it from the oldest, newest or nearest-price bucket imposes a model rather than revealing a fact. A conservative display can keep positive sampled changes by price and show unallocated reductions separately. Label it as such; do not title it “remaining entries.”

OI profile vs volume profile vs trader cost basis

A volume profile measures traded activity by price under its chosen data method. An OI-change profile allocates changes in outstanding exposure. A true inventory cost-basis ledger would require participant-level position history that public aggregate data does not provide. Similar-looking histograms do not make these quantities interchangeable.

Flat OI does not imply that the same cohort remains. Transfers can replace a position holder without changing the total. Openings and closures can also offset within a sample. Therefore even a careful OI profile is a map of an estimation procedure, useful for generating hypotheses but unable to certify who is trapped or where stops are placed.

Example: +120 and +180 additions, then an unallocated −90

During a fixed range, one interval adds 120 contracts around the 100 price bucket and another adds 180 around 102. A later observation shows a decline of 90 contracts, with no information identifying their original entries. Positive sampled changes sum to 300 contracts; net growth is 210.

The defensible profile shows 120 and 180 additions, plus a separate −90 reduction. Subtracting all 90 from the 100 bucket would leave 30 there; subtracting them from the 102 bucket would leave 90 there. Both are possible model choices, but neither is established by this dataset. Retain the ambiguity when planning a retest.

Mistake: moving the range start until a peak fits the reversal

Changing the range start until a peak lines up with the eventual reversal selects the answer after the fact. Save the range, bucket width, venue set and allocation method at decision time. Then inspect how robust the level remains under nearby reasonable settings. A hypothesis that disappears after a small boundary change deserves less confidence.