Scope and learning objectives

This article synthesizes source claims about backtest timing, modeled fills, practical tradability, software records, position reconciliation, and product translation. Examples tied to particular platforms, positions, or speaker workflows remain narrowly scoped.

  • Explain why sampled or end-of-day data may not reproduce an intraday trading path.
  • Distinguish analytical fill assumptions from evidence about live executability.
  • Evaluate whether a favorable backtest is practically tradable for the trader reviewing it.
  • Describe how actual executions can be used to reconcile campaign results and analytical records.
  • Recognize when software changes or product translation require position-specific review.

01

Sampled Data Is Not the Intraday Path

A backtest can display a precise observation without establishing that the same information or execution opportunity existed at that moment in live trading. End-of-day reconstruction, interval sampling, and delayed visibility can each separate the modeled path from the live path. [2][8][12]

  • End-of-day data may not reliably reconstruct how a strategy would have traded earlier, even when the test checks the position only once per day. [2]
  • When intraday profit and loss and delta move between sampled intervals, the reviewer should question whether displayed execution points were realistic and the strategy practically tradable. [8]
  • In the described software-specific case, exchange data labeled 3:30 in a backtester might not have appeared live until seconds later or around 3:31. [12]

02

Modeled Fills Need Separate Scrutiny

Analytical prices are modeling inputs, not proof of executable fills. This distinction is especially important for individual orders and near expiration, while any workaround used to represent an adjustment or delay must retain its illustrative, strategy-specific scope. [5][6][7][10][11]

  • Mid-price fills may average out reasonably over time in the speaker's analytical practice yet differ substantially from executable prices on a particular live order. [11]
  • Near expiration, rapidly changing mid-prices can make simulated execution assumptions unreliable. [6]
  • One speaker-specific check used very small live positions to test whether backtested prices were likely to receive actual executions. [7]
  • In one model, a call was added five minutes later at approximately 65 delta to approximate differences across many observations; the speaker acknowledged that the modeled fill might not be realistic for the specific interval. [5]
  • A separate software-specific procedure represented a written butterfly adjustment with a dummy order at an unusually favorable illustrative price. [10]

03

Judge Tradability, Then Reconcile Records

A favorable simulation is insufficient when practical constraints make the strategy untradeable for the person evaluating it. For trades that are executed, the cited workflows anchor campaign accounting in actual order executions and compare broker and analytics records to expose differences. [3][4][9][13]

  • A strong-looking backtest is not worth pursuing when practical constraints make the strategy untradeable for the trader. [4]
  • In the described Thinkorswim workflow, open-option profit and loss can differ from campaign profit and loss; campaign profit and loss is calculated from the actual execution prices of all campaign orders. [3]
  • For newer traders, the speaker recommends tracking live trades in both analysis software and broker records to understand discrepancies. [9]
  • One end-of-day process imports filled trades to verify profit and loss, strike inventory, and the risk graph, then checks inventory, profit and loss, and Greeks in a second analytics platform. [13]

04

Software and Product Translation Can Change the Representation

A position's analytical representation can depend on the software and product used. Historical software substitutions and translations to smaller products therefore require position-specific reconciliation rather than an assumption of identical modeling or simple proportional prices. [1][14]

  • The speaker reports changing the ROCK trade in 2020 because replacement software modeled its butterfly-and-call position less appropriately than the prior software. [1]
  • For the illustrated translation from SPX to XSP or SPY, the speaker adjusted option prices until simulated profit and loss matched because smaller-product option prices were not necessarily exactly one-tenth of SPX prices. [14]

Review

Key takeaways

  1. Treat timestamped and sampled backtest observations as items to verify, because they may not reproduce the information or trading path available live. [2][8][12]
  2. Separate aggregate modeling assumptions from the executability of an individual order, particularly when mid-prices change rapidly near expiration. [6][11]
  3. Practical tradability is a distinct review criterion even when simulated results look strong. [4]
  4. Use actual execution records to interpret campaign results and compare analytical representations with broker records. [3][9][13]
  5. Keep software substitutions, fill workarounds, and cross-product translations explicitly scoped to the positions and workflows from which they came. [1][5][10][14]

Self-check

Review questions

Why can a once-daily backtest still misrepresent an intraday trade?

End-of-day data may not reconstruct the earlier intraday path reliably, while profit and loss, delta, and data visibility can change between sampled or displayed times. [2][8][12]

What distinction should a reviewer make when a test assumes mid-price fills?

A modeled mid-price may behave reasonably across many observations yet remain substantially different from the executable price of an individual live order; the concern is sharper near expiration when mid-prices change rapidly. [6][11]

What does a strong simulated result fail to establish by itself?

It does not establish that the strategy is practically tradable for the trader evaluating it. [4]

How do the cited workflows investigate differences between analytical and trading records?

They use actual campaign execution prices, compare analysis-software records with broker records, and verify filled trades, inventory, profit and loss, risk graphs, and Greeks across analytical platforms. [3][9][13]

Why should a reviewer avoid assuming that a smaller-product position is a simple one-tenth price translation of SPX?

In the illustrated position, XSP or SPY option prices were not necessarily exactly one-tenth of SPX prices, so the speaker adjusted prices until simulated profit and loss matched. [14]

Traceability

Evidence index

Canonical source claims used in this guide. Open a session link to verify the underlying passage at its original timestamp.

[1]The speaker says changes were made to the ROCK trade in 2020 because the available replacement software modeled its butterfly-and-call position less appropriately than the prior software.
[2]The speaker warns that end-of-day data may not reliably reconstruct how a strategy would have traded earlier in the day, even when a backtest checks the position only once daily.
[3]In the described Thinkorswim workflow, open-option profit and loss can differ from campaign profit and loss; the speaker calculates campaign profit and loss using the actual execution prices of all campaign orders.
[4]A strong-looking backtest is not worth pursuing if practical constraints make the strategy untradeable for the trader.
[5]The speaker models adding the call five minutes later at approximately 65 delta to approximate, over many observations, the difference between likely live execution and backtested execution, while acknowledging that the modeled fill may not be realistic for the specific interval.
[6]Backtesting near expiration requires understanding which prices and executions are realistic in live trading because rapidly changing mid-prices can make simulated execution assumptions unreliable.
[7]The speaker uses very small live positions to test whether prices assumed in backtesting are likely to receive actual executions.
[8]A backtest should assess whether a strategy is practically tradable and whether apparent execution points are realistic, especially when intraday P&L and delta fluctuate between sampled intervals.
[9]The speaker recommends that newer traders track live trades both in the analysis software and at their broker to help them understand differences between the two records.
[10]To model a written adjustment in simulated trading, the speaker adds a dummy order at an unusually favorable price, such as entering a one-dollar order for a butterfly quoted at nine dollars.
[11]Mid-price fills used in analytical testing may average out reasonably over time but can differ substantially from executable prices on an individual live order.
[12]The speaker warns that exchange data displayed for 3:30 in a backtester may not become visible in live trading until seconds later or around 3:31, creating timing differences between backtests and live execution.
[13]The speaker's end-of-day process imports filled trades to verify profit and loss, strike inventory, and the risk graph, then checks inventory, profit and loss, and Greeks again in a second analytics platform.
[14]When translating the illustrated SPX position to XSP or SPY, the speaker adjusts the option prices until the simulated profit and loss matches because the smaller-product option prices are not necessarily exactly one-tenth of the SPX prices.