Knight Capital's order router sends millions of unintended orders in 45 minutes, 1 August 2012

Changes to what this case says, with the reason. Every earlier version is kept.

• 2026-10-21: Inference: signed by. The interpretation's wording changed after the editor signed it (second, independent check of 21 october 2026); the signature is cleared so the editor signs the corrected wording.
  Show the change

  Before: Saurabh Bhatnagar

  After:
• 2026-10-21: Inference: value. Found in the editor's second, independent check of the interpretations, 21 october 2026: the earlier sentence was wrong: the order itself links the missing threshold to the continued orders, paragraph 24: 'because knight did not link the 33 account to pre-set, firm-wide capital thresholds'; paragraph 37 says this contributed to knight's failure to mitigate the effects.
  Show the change

  Before: The Securities and Exchange Commission's order shows failures from deployment to recovery: a deployment that missed one of eight servers with no second review, 97 reject emails before the open that were not designed as alerts, dormant code moved in 2005 and never retested, no firm-wide capital threshold linked to order entry, and no procedure to halt the router. The order names three controls that it says could have averted or identified the problem: a double-check of the deployment, treating the reject emails as alerts, and retesting the code in 2005. It does not say that a capital threshold would have bounded the loss; that is the Register's reading, because a threshold limits exposure whatever the cause. The record does not say how the orders were stopped

  After: The Securities and Exchange Commission's order shows failures from deployment to recovery: a deployment that missed one of eight servers with no second review, 97 reject emails before the open that were not designed as alerts, dormant code moved in 2005 and never retested, no firm-wide capital threshold linked to order entry, and no procedure to halt the router. The order names three controls that it says could have averted or identified the problem: a double-check of the deployment, treating the reject emails as alerts, and retesting the code in 2005. It also finds the router kept sending orders because no firm-wide capital threshold was linked to order entry, but does not say how much of the loss a threshold would have prevented; that it would have bounded the loss is the Register's reading, because a threshold limits exposure whatever the cause. The record does not say how the orders were stopped
• 2026-10-21: Inference: value. Found in the editor's second, independent check of the interpretations, 21 october 2026: the order does not describe failures in layers, and the case's own mandate fields show none; abbreviation spelled out.
  Show the change

  Before: The SEC order shows a failure at every layer: a deployment that missed one of eight servers with no second review, 97 reject emails before the open that were not designed as alerts, dormant code moved in 2005 and never retested, no firm-wide capital threshold linked to order entry, and no procedure to halt the router. The order names three controls that it says could have averted or identified the problem: a double-check of the deployment, treating the reject emails as alerts, and retesting the code in 2005. It does not say that a capital threshold would have bounded the loss; that is the Register's reading, because a threshold limits exposure whatever the cause. The record does not say how the orders were stopped

  After: The Securities and Exchange Commission's order shows failures from deployment to recovery: a deployment that missed one of eight servers with no second review, 97 reject emails before the open that were not designed as alerts, dormant code moved in 2005 and never retested, no firm-wide capital threshold linked to order entry, and no procedure to halt the router. The order names three controls that it says could have averted or identified the problem: a double-check of the deployment, treating the reject emails as alerts, and retesting the code in 2005. It does not say that a capital threshold would have bounded the loss; that is the Register's reading, because a threshold limits exposure whatever the cause. The record does not say how the orders were stopped
• 2026-10-20: Inference: value. Found in the editor's source re-check of 19-20 october 2026: the interpretation presented a capital threshold as 'the one control that would have bounded the loss' as if the record said so; the sec order (paragraph 41) names different controls that 'could have' averted or identified the problem (a deployment double-check, integrating the reject messages into monitoring, retesting the power peg code in 2005) and does not make the capital-threshold counterfactual. the new wording separates what the order says from the register's reading and says what the record does not show. the interpretation was not signed, so no signature is reset.
  Show the change

  Before: Every layer from inputs to recovery failed, yet the one control that would have bounded the loss was a capital threshold: a limit on how much the system could do before a check. The cheapest missing control can matter more than the longest chain of causes

  After: The SEC order shows a failure at every layer: a deployment that missed one of eight servers with no second review, 97 reject emails before the open that were not designed as alerts, dormant code moved in 2005 and never retested, no firm-wide capital threshold linked to order entry, and no procedure to halt the router. The order names three controls that it says could have averted or identified the problem: a double-check of the deployment, treating the reject emails as alerts, and retesting the code in 2005. It does not say that a capital threshold would have bounded the loss; that is the Register's reading, because a threshold limits exposure whatever the cause. The record does not say how the orders were stopped
• 2026-10-20: Controls: c-x1. Found in the editor's source re-check of 19-20 october 2026: the note named the control without the finding; the sec order paragraphs that state it are now cited in the note (bucket and confidence unchanged).
  Show the change

  Before: Firm-wide capital threshold

  After: Firm-wide capital threshold (SEC order, paragraphs 9(B) and 22: Knight 'did not have pre-set capital thresholds in the aggregate for the firm' and its position limits 'did not account for the firm's exposure from outstanding orders')
• 2026-10-20: Controls: c-x2. Found in the editor's source re-check of 19-20 october 2026: the note named the control without the finding; the sec order paragraphs that state it are now cited in the note (bucket and confidence unchanged).
  Show the change

  Before: Kill switch

  After: Kill switch (SEC order, paragraph 21: no procedures 'to halt SMARS's operations in response to its own aberrant activity'; paragraph 42: no clear guidance on when to disconnect a malfunctioning system from the market)
• 2026-10-20: Controls: c-d1. Found in the editor's source re-check of 19-20 october 2026: the note named the control without the finding; the sec order paragraphs that state it are now cited in the note (bucket and confidence unchanged).
  Show the change

  Before: Deployment verification across servers; Second-technician deployment review (absent); Regression testing of dormant code (absent)

  After: Deployment verification across servers; Second-technician deployment review (absent); Regression testing of dormant code (absent). SEC order, paragraph 9(A): no controls reasonably designed to prevent the entry of erroneous orders 'at a point immediately prior to the submission of orders to the market'; paragraph 21: no adequate controls in SMARS to prevent erroneous orders; paragraph 41: a double-check of the deployment 'could have identified that a server had been missed' and retesting the Power Peg code in 2005 could have identified the disabled cumulative-quantity function
• 2026-10-20: Controls: c-i4. Found in the editor's source re-check of 19-20 october 2026: the note named the control without the finding; the sec order paragraphs that state it are now cited in the note (bucket and confidence unchanged).
  Show the change

  Before: Alerting design

  After: Alerting design (SEC order, paragraph 19: the 97 'Power Peg disabled' emails were not designed as system alerts and were generally not reviewed; paragraph 41: integrating these reject messages into monitoring 'could have prevented the events of August 1')
• 2026-10-20: Outcome: as of. Found in the editor's source re-check of 19-20 october 2026: the outcome field now carries virtu financial's acquisition of kcg holdings, completed 20 july 2017 (virtu form 8-k exhibit 99.1).
  Show the change

  Before: 2013

  After: 2017
• 2026-10-20: Outcome: regime consequence. Found in the editor's source re-check of 19-20 october 2026: the summary omitted the cease-and-desist order; the sec order (section iv) orders knight to 'cease and desist from committing or causing any violations' and the recovery/accountability field already lists it.
  Show the change

  Before: $12M SEC penalty; censure; consultant review

  After: $12M SEC penalty; censure; cease-and-desist; consultant review
• 2026-10-20: Outcome: affected parties. Found in the editor's source re-check of 19-20 october 2026: 'counterparties' is not the record's word; the sec order (paragraph 18) says the erroneous executions 'influenced share prices' and 'these share price movements affected other market participants'.
  Show the change

  Before: ["Knight Capital", "counterparties"]

  After: ["Knight Capital", "other market participants"]
• 2026-10-20: Context: autonomy level text. Found in the editor's source re-check of 19-20 october 2026: 'human supervision after the fact' understated what the sec order shows: position monitoring 'relied entirely on human monitoring' (paragraph 25), senior personnel 'observed a large volume of positions accruing in the 33 account' and the technology team attempted to address the problem during the event (paragraph 27), but knight 'did not have procedures in place to halt smars's operations' (paragraph 21).
  Show the change

  Before: Machine acts alone at machine speed; human supervision after the fact

  After: Machine acts alone at machine speed; people watched positions by hand and tried to intervene during the 45 minutes, with no procedure to halt the router
• 2026-10-20: Fields: recovery: technical. Found in the editor's source re-check of 19-20 october 2026: 'human intervention after roughly 45 minutes' read as if people acted only at the end; the sec order (paragraph 27) says knight 'relied primarily on its technology team to attempt to identify and address the smars problem' during the event, that 'knight uninstalled the new rlp code from the seven servers where it had been deployed correctly' and that 'this action worsened the problem'. the order says only 'by the time that knight stopped sending the orders' (paragraph 1) and does not say how the orders were stopped.
  Show the change

  Before: Human intervention after roughly 45 minutes

  After: Knight's technology team tried to identify and fix the problem while the orders were going out; removing the new code from the seven correctly deployed servers made it worse by sending more parent orders to the Power Peg code (SEC order, paragraph 27). The order says Knight stopped sending the orders after about 45 minutes but does not say how
• 2026-10-20: Fields: outcome. Found in the editor's source re-check of 19-20 october 2026: the field stopped at the 2013 merger; virtu financial's press release filed as exhibit 99.1 to its form 8-k says it 'completed its acquisition of kcg holdings, inc. (nyse: kcg) on thursday, july 20' 2017. the loss figure is also in the sec order (paragraph 1: 'knight lost over $460 million from these unwanted positions').
  Show the change

  Before: {"value": "State: harm. Loss of more than $460 million; Knight merged with GETCO into KCG Holdings in July 2013. As of 2013.", "source_ids": ["SEC-PR", "KCG"]}

  After: {"value": "State: harm. Loss of more than $460 million; Knight Capital Group merged with GETCO into KCG Holdings on 1 July 2013; KCG was acquired by Virtu Financial on 20 July 2017. As of 2017.", "source_ids": ["SEC-ORD", "SEC-PR", "KCG", "VIRTU"]}
• 2026-10-20: Fields: action. Found in the editor's source re-check of 19-20 october 2026: the field took 'more than 4 million orders' from the sec press release; the sec order (paragraph 1), the primary record, says smars 'routed millions of orders' and 'obtained over 4 million executions in 154 stocks for more than 397 million shares', and gives the positions as a net long of about $3.5 billion in 80 stocks and a net short of about $3.15 billion in 74 stocks. both accounts are now stated; the order is a regulator's record and the press release is the same regulator's summary of it, so the field is not marked disputed.
  Show the change

  Before: {"value": "More than 4 million orders sent to fill 212; over 397 million shares traded; several billion dollars of unwanted positions", "source_ids": ["SEC-PR"]}

  After: {"value": "Millions of orders routed in about 45 minutes to fill 212 parent orders, producing over 4 million executions in 154 stocks for more than 397 million shares; a net long position of about $3.5 billion in 80 stocks and a net short position of about $3.15 billion in 74 stocks (SEC order, paragraph 1). The SEC press release summarises this as 'more than 4 million orders'.", "source_ids": ["SEC-ORD", "SEC-PR"]}
• 2026-10-20: Fields: mandate: objective. Found in the editor's source re-check of 19-20 october 2026: the sec order (paragraph 1) says '212 small retail orders that knight had received from its customers'; the press release says '212 customer orders'. wording and citation taken from the order as the primary record.
  Show the change

  Before: {"value": "Fill 212 small customer orders", "source_ids": ["SEC-PR"]}

  After: {"value": "Fill 212 small retail orders received from customers", "source_ids": ["SEC-ORD", "SEC-PR"]}
• 2026-10-20: Fields: mandate: principal. Found in the editor's source re-check of 19-20 october 2026: the sec press release does not call knight a broker-dealer; the sec order (paragraph 11) says knight 'is registered with the commission pursuant to section 15 of the exchange act', which is broker-dealer registration, and paragraph 9(e) refers to 'its market access'. re-cited to the order as the primary record.
  Show the change

  Before: {"source_ids": ["SEC-PR"]}

  After: {"source_ids": ["SEC-ORD", "SEC-PR"]}
• 2026-10-13: Fields and sources. Claims aligned to what each linked record says, found on review.
  Show the change

  Before: 'position limits not linked to firm-wide limits'; 'near insolvency and later acquired'; 'no kill switch'; human intervention cited to the press release

  After: wording from the SEC order (paragraphs 21 and 22); acquisition cited to KCG's filing; insolvency wording removed
• 2026-10-13: Controls: c-d1. A control has one state per case; entries in two states merged into the state the record supports.
  Show the change

  Before: failed: Deployment verification across servers; absent: Second-technician deployment review, Regression testing of dormant code

  After: failed: Deployment verification across servers; Second-technician deployment review (absent); Regression testing of dormant code (absent)