JPMorgan spoofing: reconstruct the order lifecycle before counting patterns

365 ALTERAM analysis

A row is not necessarily an order

A market-data export can contain a convincing-looking sequence of entries, changes and cancellations. Before interpreting that sequence, a reviewer must establish what each row represents. Counting messages as separate orders, combining records under an incomplete identifier or treating a request as a completed event can distort the behaviour being assessed.

The practical question is whether the underlying data support the claimed lifecycle of each order and its relationship to other activity. The JPMorgan precious-metals proceedings provide a public setting for that question. Their outcomes demonstrate proven misconduct; the review method below does not suggest that those outcomes resulted from a data error.

Public record

Corporate admissions and individual verdicts are separate

On 29 September 2020, the US Department of Justice announced a deferred prosecution agreement with JPMorgan Chase & Co. concerning wire-fraud charges and more than $920 million in monetary relief. The admitted conduct included precious-metals orders placed with an intention to cancel before execution, creating misleading signals of supply and demand. The resolution also covered a separate US Treasuries scheme. [1]

On 10 August 2022, DOJ announced that a jury convicted Gregg Smith and Michael Nowak of offences including fraud, attempted price manipulation and spoofing. Those individual verdicts should be identified on their own evidential and procedural footing, rather than inferred from the earlier corporate agreement. [2]

On 20 August 2025, the Seventh Circuit affirmed the convictions challenged by Smith, Nowak and Christopher Jordan. In addressing Smith and Nowak’s intent arguments, it focused on intent when an order was placed and upheld the sufficiency of the evidence. It also explained that the trading platform permitted cancellation before execution. Cancellation alone and the proven deceptive conduct were therefore distinct propositions. [3]

365 ALTERAM analysis

Establish the relationships before running the query

For an authorised review, obtain the available source files, format documentation and extraction history. Identify the meaning of each message type and the boundaries of each identifier. Determine whether an identifier is unique across the dataset or only within a session, account, venue or trading day. A join should reproduce the documented relationship, not merely connect matching numbers.

Construct an event chain for each relevant order. Distinguish submission, acknowledgement, amendment, rejection, execution and cancellation where the preserved records support those states. Keep original fields alongside interpreted values. If a replacement receives a new identifier, record the evidence linking it to the earlier order rather than silently treating it as either unrelated activity or the same unchanged instruction.

Check counts before and after each transformation. Duplicate exports, retransmitted messages and one-to-many joins can inflate apparent activity. Retain a reconciliation showing how many source rows were accepted, excluded or linked, and why. An unexplained mismatch is a question to resolve, not permission to discard inconvenient records.

365 ALTERAM analysis

Reconstruct time and the available market view

Identify what each timestamp measures: a local action, a gateway event, an exchange acknowledgement or a later export. Document precision, time-zone conversion and any supported clock adjustment. When two records cannot reliably be ordered, preserve that uncertainty. Sorting a spreadsheet does not establish the real sequence when the clocks or event meanings differ.

Compare the reconstructed activity with the market information available at the relevant time. State the coverage and limits of the retained feed, including any gaps or aggregation. Do not equate a later analytical reconstruction with the exact display a trader saw unless the evidence supports that equivalence. A specialist in the relevant market and systems should assess those distinctions.

365 ALTERAM analysis

Test the inference, including adverse explanations

A pattern analysis should explain how episodes were selected and what comparison set was used. Examine whether the same criteria capture activity with documented legitimate explanations, and whether materially different conditions were combined. Preserve contrary evidence, including repeated sequences or communications that strengthen the allegation. Independence requires testing the strength of adverse evidence as well as identifying weaknesses.

Attribution needs a separate evidential link. An account or system identifier should be assessed against authorised users, access records, operational arrangements and relevant communications. The data reconstruction establishes what the records show; it does not independently read a person’s intentions.

365 ALTERAM can examine data structures, reproduce analytical transformations and coordinate appropriately qualified trading specialists. A useful report supplies the event map, reconciliation, assumptions and alternative interpretations so another reviewer can check the route to each conclusion. Counsel retains legal strategy, and the competent court determines responsibility. The outcome may confirm, qualify or leave unresolved the proposed interpretation; no favourable result is promised.