SPX 0DTE Slippage: Our Seven-Exit Live Experiment

The written reference below adds the calculation basis, parameter tables and measurement rules needed to interpret later results.

What is SPX 0DTE slippage—and what is it not?

Slippage only becomes a useful measurement when we state the reference price, the instrument and the time. Comparing a short option's fill with a vertical spread's stop is not a like-for-like calculation.

0DTE means zero days to expiration: the options expire that trading day. Slippage generally describes execution away from an expected or reference price. For this study, we separate three quantities:

A hypothetical buyback at $4.00 after a $3.51 stop has $0.49 of stop overshoot: $49 for one standard $100-multiplier contract, before fees. That difference can include market movement while the exit is being confirmed, submitted or filled. It is not automatically $49 caused by the broker.

The distinction matters most on fast sessions. A displayed mid-price is not a promise that someone will trade there, and a stale quote is not an executable benchmark.

Why did the May 18 loss lead to this experiment?

The May 18, 2026 event exposed a gap between the loss we intended to take and the prices at which some positions actually closed. In the video, I report losses exceeding $20,000 across the event; that aggregate is my reported experience, not an independently reconciled calculation of pure execution slippage.

One identified historical short call provides a concrete example:

Source: the historical trade example discussed in the episode, reconciled in our internal incident documentation. These are historical live economics, not returns from the new seven-arm pilot. The affected historical entry was not the new 14:03 template.

That example asks a research question; it does not answer it. A different order might have avoided one dislocated fill, remained open through a worsening move, or failed to execute altogether. Seeing a later price recovery does not prove that waiting was the better decision at the time.

What stays the same across the seven exit policies?

We use a shared entry specification so that exit behavior is the main subject of the experiment. Shared targets do not guarantee identical strikes or fills across brokers.

The experiment is EXIT-LAB-01, publicly called The Big SPX Slippage Study. It uses a fixed entry slot from our **MEIC—Multiple Entry Iron Condor—**research family. Each arm combines a call credit spread and a put credit spread; the pilot itself tests one scheduled entry, not a whole day's sequence of MEIC entries.

The documented setup does not establish the date every arm became active. Actual activation, account assignment and first fill must be captured from the platforms. Seven complete one-lot arms represent 14 vertical sides and 28 option legs: concentrated exposure to the same market, not diversification.

How is a 95% stop calculated?

A 95% loss stop means a closing debit equal to 195% of the applicable opening credit, before fees. Which credit counts depends on whether the stop monitors the short leg or the vertical.

Illustrative calculation—not a reported pilot fill: sell the short for $2.00 and buy the long for $0.20.

The long's economics must still be included in the short-leg case. This is why assigning a $3.90 stop to every arm would distort the comparison.

The SPX index level is a different quantity again. There is no fixed conversion from “the short option costs $3.90” to “SPX is at this level.” Time, volatility and option-chain prices matter.

What do the four OptionsApp variants test?

A and B begin with a vertical stop; C and D begin with a short-leg stop. That distinction separates trigger scope from the order used after triggering.

Single Bid is the configured trigger-method name. It should not be confused with a guaranteed fill at the displayed bid. There is no Double Bid arm in this final seven-policy lineup; earlier planning versions used different labels.

B is also not a vertical REL order. Its intact spread uses a market stop; REL applies after migration to the short. The migration's timing, trigger recalculation and remaining long must be recorded rather than assumed from the strategy name.

What do the REL offset and cap control?

REL means a relative, quote-linked order. The offset changes its pricing aggressiveness; the buy limit caps what it may pay. Neither parameter guarantees execution.

The documented B/C/D export uses a $1.00 offset and $11.70 absolute cap. The template calculation was:

min(3 × $2.00 target short premium × 1.95, 50-point width ÷ 2) = $11.70

These are experimental settings, not a universal recommendation. Importantly, $11.70 is fixed in the template; it does not automatically become three times the actual stop after every fill or migration. The live ratio must be measured.

A tighter cap limits the permitted buy price more strongly but can leave an exit unfilled. A wider cap permits more expensive execution but cannot promise liquidity. A cap is an order-price constraint, not a portfolio-loss ceiling. This export has no added automatic emergency market fallback for the REL arms.

Vendor reference: OptionsApp REL orders. Actual submitted order fields and the broker's order lifecycle are the authority for what happened in a specific trade.

How does the strike-distance condition change D?

It adds an underlying-price condition to the short-leg premium stop. Both must permit activation; the gate does not replace the REL cap.

The documented D baseline is factor 0.50, minimum distance 10 SPX points, and a five-minute median. OptionsApp derives a reference option price from the short's stop and factor, finds a reference strike in the option chain, smooths the resulting distances and applies the larger of that median distance and the minimum. For a put the activation level is above its short strike; for a call it is below.

This is not a fixed waiting period. It can block a premium spike while SPX remains far away, or delay an exit during a genuine adverse move. It is supported for individual short legs using REL, not an intact vertical stop. Threshold updates depend on OptionsApp's connection to TWS. OptionsApp's feature documentation explains the calculation and limitations.

What do the three TradeSteward speeds compare?

E, F and G keep vertical stop scope while changing confirmation and order-walking behavior. Those are two separate clocks, so the preset name alone is not enough to explain a fill.

The saved pilot baseline uses Bot Algo, Percent Loss 95, Verticals, with these presets:

Each subsequent attempt adds 10 percentage points of aggression, not $0.10. These are documented configuration values, not promises of total time to fill. Confirmation, submission, replacements and market fallback each have their own lifecycle.

The episode calls E “TradeSteward Default”; the preset recorded for E in this pilot is Lightning. G is Slow, not the Turtle comparator from earlier research. The Fast arm borrows an exit-policy comparison from our examination of Rob's setup; it is not a reproduction of his full portfolio.

The platform's stop-settings documentation explains the broader configuration. Account-level spread protection and current broker behavior also need to be logged.

What happens to the long option after a short-only exit?

The long does not disappear because the short closed. Its sale, remaining value or settlement is part of the experiment's economics.

C and D do not automatically liquidate the long under the documented export. A and B can also leave a residual long after migration. TradeSteward's three arms have “Sell Bidless Longs After Exit” enabled, but an enabled recovery feature is not evidence that a sale filled.

We therefore separate short-leg P&L, long-leg P&L, fees and total side P&L. A bidless long is not necessarily worthless at settlement. Conversely, its potential future value is not realized profit today.

Rare long payouts can change an exit policy's ranking. We will show their actual contribution and a sensitivity excluding the largest contributor, without deleting it from the real cash-flow record.

What operational failure must the study track besides bad fills?

A missing or rejected protective order is a different failure from an expensive fill. A successful import is not proof of working protection.

In a separate developer response, OptionsApp identified REL limit prices of $17.55 and $13.65 as invalid for the applicable IBKR price grid. The response said repeated rejection left affected trades without working stops. This incident must not be silently attributed to May 18.

Our documented $11.70 template cap is on that response's $0.10 grid. Future calculations and overrides still require contract- and order-specific tick validation after the calculation: two decimal places alone are not sufficient. Promised vendor fixes must be distinguished from an installed, verified release.

The monitoring chain is: configured → monitored → submitted → accepted → filled. Each step needs evidence. A conditionally held order or a platform-monitored stop also requires interpretation; absence from a simple open-order list does not by itself prove failure.

How will daily and weekly results be compared fairly?

We will report complete economics and protection reliability, not just the difference from a convenient reference. An unfilled stop that later expires profitably still counts as a nonfill.

Daily reporting needs one row per arm, even when an arm was disabled, failed entry or has unresolved exposure. Each row should include actual strikes and premiums, stop basis, stop events, submit/accept/fill times, net P&L, residual longs, fees, configuration version and evidence gaps.

Weekly reporting adds cumulative net P&L, end-of-day maximum drawdown, worst day, stop counts, delay statistics, rejects and long contribution. Intraday drawdown requires synchronized position marks; daily closing P&L cannot reconstruct it.

The most informative comparisons are A versus B, C versus D, and E/F/G. We will distinguish matching contracts from the full operational books and show exclusions. B's migrated exits deserve their own analysis.

Broker and platform change together across A–D versus E–G, so a cross-platform difference is not pure exit-policy causality. Seven correlated arms are also not seven independent market observations. A fixed 50-point-width pilot cannot establish the best settings for every width.

What can backtests add without pretending to reproduce the broker?

Backtests can expose sensitivity across normal days and stress sessions, but they cannot certify an unobserved live order lifecycle. A simulated fill remains a model output.

The historical extension should preserve reference settings, run the same strategies over the full available period and a defined trailing-year window, then examine stress dates separately. Historical date labels must be reconciled before comparison; older notes used inconsistent October/August references.

Our second-level historical data does not reveal every intrasecond quote change, queue position or broker decision. Previous exploratory replays also had material limitations involving vertical fill bounds, strike-distance approximations and retained-long contributions. Those rankings are not presented here as validated winners.

The next useful result is not “this cap would have saved that day.” It is whether a frozen configuration improves the complete distribution of outcomes, including misses and delays, on data not used to choose it.

Related reading

• SPX settlement: what happens to remaining options
• SPX versus SPY options
• How automated options trading works

Educational disclosure: This article documents an experiment, not a recommendation to copy its trades or settings. Options involve substantial risk. Historical live examples and hypothetical calculations do not predict future results. Backtests depend on data quality, transaction assumptions and selection choices; no order mechanism guarantees a particular fill or loss limit.