Notes

This is a research note from Convex, which includes delayed trade bars that are not executable quotes. The data starts at 2026-09-11T02:15:19.701710+00:00. The note includes stock bars data with a status of 200, which means the request was successful, and a cutoff time of 2026-09-11T01:55:19.701710+00:00. The data includes information on the following stocks: EPM, HAIN, HITI, and ZUMZ. For example, EPM's stock bar on 2026-09-09T04:00:00Z had a closing price of 3.69, a high of 3.75, a low of 3.63, and a volume of 296055. The note also includes option bars data with a status of 200 and a cutoff time of 2026-09-11T01:55:19.701710+00:00. The data includes information on the following options: HITI261016C00002500 and HITI261016P00002500. For instance, HITI261016C00002500's option bar on 2026-09-08T04:00:00Z had a closing price of 0.34, a high of 0.35, a low of 0.3, and a volume of 131. The note ends at 2026-09-11T02:15:20.760400+00:00.

A news scan found the headline is not available, but a progress report was found. Eight minutes have elapsed out of 43, and the original stage one ended within 8 minutes. The screening and testing process is progressing. There are 13176 listings in the census, 12147 daily bars, and 147 earnings discovery matches. Thirty-one option underlyings, which are the underlying assets of options, were checked, including 12 initial ones, SCHL, and 18 second ones, including an error with NXT.L. No option chain requests were made. PLAY has 12 primary AMC dates with current call and put coverage of 0 out of 12. Other promising low premium options, SFIX, VRA, and BB, require independent dates and testing. The TCOM primary PDF archive was recovered. CBRL's original September 9 date confirms a September 23 BMO date at 08:15 ET, not September 16, according to the calendar. SCHL's September 3 schedule lists a September 24 date at 16 ET, not September 17, according to the calendar. ABAT's preliminary FY Q4 August 20 data is already public, so it cannot be assumed to be a fresh September 17 event. All measured screen figures will be recorded before finishing. Initial census errors included 400 non-standard tickers, and 429 were fixed with 384 explicit exclusions and 2 second sequential cadence. No funds were allocated, and no orders were made. Existing collectors were verified at 192 out of 199 at 00:15, and this has been confirmed as the actual end.

The trading plan has a 43-minute time limit starting from 02:00UTC, with 5 minutes reserved at the end. The plan prioritizes handling stale quotes over arbitrage opportunities, excluding a monthly shortfall of $349.01. The decision tree is organized as follows: first, it considers universe coverage, listings source, and exclusions; then, it looks at option eligibility, catalog, and affordability. Next, it examines event discovery, original calendar, and news versus calls. The tree also evaluates event edge, historical dates, and comparable issuer events. Additionally, it assesses return horizon, directional coverage, and mispricing. The plan further considers business mechanisms, cash conversion, and valuation, as well as external forces, rates, and regulation. It also looks at overseas and crypto markets, listed proxies, and basis and event mismatch. The implementation section covers sizing, with a half-Kelly criterion and 3%, 5%, and 6% caps, as well as exits, with a focus on regular trading hours and prior session holds. The plan requires daily response resolution and minute synchronization for pricing. Lessons 724-726 are applicable, and there will be no retrospective claims for low selection or stale mark executable returns.

This is a research note from a trading fund manager. The note is about the stocks MLKN and ZUMZ. It says that the option outcomes for these stocks are delayed. The note started on 2026-09-11T00:15:27.870854+00:00. The data for the options and stocks of MLKN and ZUMZ is as follows: for the option ZUMZ261016P00015000, the price was 0.36 on 2026-09-09 and 0.9 on 2026-09-10. For the option MLKN261016P00017500, the price was 0.28 on 2026-09-09. For the option MLKN261016P00020000, the price was 0.71 on 2026-09-08. The stock prices for MLKN were 21.65 on 2026-09-09 and 20.8 on 2026-09-10. The stock prices for ZUMZ were 17.55 on 2026-09-09 and 16.71 on 2026-09-10. The note ended on 2026-09-11T00:15:28.239329+00:00.

This is a research note from Convex, which includes delayed trade bars that are not executable quotes. The data starts at 2026-09-11T00:15:20.531129+00:00. The note includes stock bars data with a status of 200, and a cutoff time of 2026-09-10T23:55:20.531129+00:00. The data includes information on the following stocks: EPM, HAIN, HITI, and ZUMZ. For example, EPM had a closing price of 3.69 and a high price of 3.75 on 2026-09-09T04:00:00Z. The note also includes option bars data with a status of 200, and a cutoff time of 2026-09-10T23:55:20.531129+00:00. The data includes information on the following options: HITI261016C00002500 and HITI261016P00002500. For instance, HITI261016C00002500 had a closing price of 0.34 and a high price of 0.35 on 2026-09-08T04:00:00Z. The note ends at 2026-09-11T00:15:20.961651+00:00.

The automated trading fund manager completed 12 minutes of a 43-minute ceiling and began the final records block at 00:10 UTC. The manager had no Convex holdings, orders, exits, or trade history, with $178.71 in cash and $850.99 in the account. The manager saved various statistical, technical, and fundamental data, as well as consequences, backtest, simulation, and strategy lessons. A new MLKN issuer confirmed an August 25 event, with a call at 08:30 ET on September 22, not September 23. The manager analyzed various stocks, including ZUMZ, which had a price correction, and MLKN, which was rejected. The manager also analyzed HITI and GRAL/SRRK, with preliminary rejects. The manager retained $220 out of $221 and resolved 231 issues. A new job was started with 72 runs, and the manager will check actual outputs at 02:00 UTC. The manager also fixed technical issues with local scripts and data files, including mlkn_test.py and audit.py. There are still gaps in the data, including true option backtest and executable bid ask, MLKN full historical date census, and ACB date access.

A news scan found the script output of a test of the ExactPython collect_new.py recipe. The test started at 00:10:41.835988 on September 11, UTC, and ended at 00:10:43.474780. The test used stock and option bars, both with 200 items and no pagination. The daily prices for MLKN and ZUMZ were 20.8 and 16.71, respectively. The prices for MLKN on September 9 and 8 were 17.5 and 20.71, respectively. The price for ZUMZ on September 10 was 15.90. This test extends a previous collector, number 192, which covered HITI, HAIN, EPM, and ZUMZ stocks, but not MLKN or ZUMZ options. The plan for the test was to run for 240 minutes, with 72 runs, over 12 days, and to output the start and end times and the last two bars, with a 20-minute cutoff. The test did not use executable quotes. The recipe used a Python script with the following code: python3 - <<'PY'.... The script imported several libraries, including requests, os, json, datetime, and concurrent.futures, and defined a function to fetch data from a website. The script then used this function to fetch stock and option bars for MLKN and ZUMZ, and printed the results, including the status code, cutoff time, and any errors. The script ended at the time it finished running, and printed the end time.

The trading plan starts at 00:00:24 UTC and reserves the final 5 minutes, with a 43-minute ceiling. Previously, it was noted that stale marks are more likely than free arbitrage, and a monthly shortfall of $349.01 was set aside. The plan involves a question tree with four main points. First, it considers the least predictable tree, which includes three subpoints: identifying the option contract by its size, strike price, and expiry date, synchronizing the trade minute with the stock minute, and setting bounds for the maximum payout and costs. Second, it looks for broader candidates based on the issuer date, affordability, and sampling of primary dates. Third, it examines attack theses for specific stocks, including HITI, HAIN, and ZUMZ, considering factors such as cash conversion, debt, and demand. Fourth, it assesses decision robustness by evaluating directional samples, sizing, and execution. The plan confirms an empty convex trade history with no holdings to reset exits. It will resolve daily event moves and synchronous minute checks, and only discover headlines. Lessons 705-707 have been applied. The plan seeks a new affordable dated candidate and will maintain a frozen HITI test.

This is a research note from Convex, which includes delayed trade bars that are not executable quotes. The data starts at 2026-09-10T22:00:20.653580+00:00. The stock bars dataset has a status of 200, a cutoff of 2026-09-10T21:40:20.653580+00:00, and no next page. The dataset includes bars for the following stocks: HAIN, HITI, ZUMZ, and EPM. For example, HAIN has two bars: one with a close price of 0.6418, high of 0.693, low of 0.6418, and volume of 568077 on 2026-09-09T04:00:00Z, and another with a close price of 0.6395, high of 0.676, low of 0.6273, and volume of 489968 on 2026-09-10T04:00:00Z. The option bars dataset also has a status of 200, a cutoff of 2026-09-10T21:40:20.653580+00:00, and no next page. The dataset includes bars for the following options: HITI261016C00002500 and HITI261016P00002500. For example, HITI261016C00002500 has two bars: one with a close price of 0.34, high of 0.35, low of 0.3, and volume of 131 on 2026-09-08T04:00:00Z, and another with a close price of 0.31, high of 0.34, low of 0.25, and volume of 121 on 2026-09-09T04:00:00Z. The data ends at 2026-09-10T22:00:21.201598+00:00.

At 21:04 UTC, the system confirmed that earlier estimates of 2 to 5 minutes for commentary were rough and overestimated. The actual start time was 21:00:25, and the work was completed in about 5 minutes, which is well under the 39-minute allowance. There are no current holdings, orders, or exits, and the cash balance is $178.71. The system saved 2318 statistical, 2319 fundamental, and 2320 strategy records, along with 22968 to 22976 observations, and recorded lessons learned. The system closed 222 and created a new midnight monitor, retaining 220 and 221. Prior records from 2301 to 2307 were read, including technical and vector backtest results, as well as NumPy simulations. No new strategy was tested or traded, so no redundant proxy backtest was needed. The apparent mispricing of HAIN was disproved within the same minute, with a put max of 2.7027x failing to meet the net 3R threshold, and no prices were available for K.5. A cheap, fresh candidate is needed before the 8-event study, and the full OCC notice and executable bid-ask gap remain. The HITI coverage is 1/9 each at C9.765625% and P8.59375% threshold, with a median of 5.494505%. The empirical put price ceiling is frozen at 40%. The price of 0.8065934 at S2.56 is not a fair value or trade instruction. The ZUMZ results were actual, with a price of $13.70 at 20:40, which is a decrease of 18.0132% from the previous price of $16.71. The intrinsic value of $15.36 and $1.30 is not a trade profit or loss. The system will verify the actual option exit prices on September 11 prospectively, rather than inferring from after-hours stock prices. The EPM is stale, with C2.5 at $95 and C5 needing 30.99%. The job was completed at 19:22:20 UTC, and the next run is scheduled for 22 UTC, with no duplicate job needed for existing HITI or stock watch. The original ZUMZ release and exact period of narrative remain unverified, with a CFO of 35.7m, while the H1 table shows a clear CFO of -27.276m. All source URLs and tested requests were recorded, and the local refresh.py and raw JSON files are in the current workspace, but may not persist. The cash reason failed eligibility and payoff, and there is no demand for certainty, so no FUND proposal was made. A news scan found the headline was not available, but the system confirmed that the job was complete.

A 39-minute time frame was used, including the last 5 minutes of records, starting at 21:00:25 UTC. Before this, it was thought that stale prices were more likely than free option arbitrage, so a plan was made to handle a possible shutdown shortfall. The most uncertain question was about HAIN, specifically: 1. Is the HAIN discrepancy real? This included looking at 1a, the identity of the multiplier and deliverables, 1b, the synchronization of option trade minutes and stock minutes, and 1c, the bounds of intrinsic value and costs. 2. Have existing signals changed? This included checking 2a, HITI for new closes and directional event coverage, 2b, EPM for the nearest strike and cap, and 2c, ZUMZ for the actual primary release, first news, and price response. 3. Is there an edge from comparable events? This included looking at 3a, eight or more issuer dates and session mapping, 3b, the distribution of signed moves and holdout, and 3c, valuation including time value, spread, and fees. 4. Are we operationally eligible? This included checking 4a, if trading is only done during regular trading hours and the next open, 4b, pre-event exit and hold exceptions, and 4c, using half Kelly, a 3 or 5 percent full loss, a 6 percent absolute cap, and limiting to three positions. The process started with 850.99 in equity and 178.71 in fund cash, with no trades made through query_trades, and only Convex was due 222. The goal was to review saved decisions, not to replace failed tests based on headline direction. The main limit was the closed session and unvalidated affordable payoffs. The resolution was based on daily and synchronized minute bars, and headline sentiment was not used as a trade signal. All prior statistical, technical, fundamental, simulation, consequences, and backtest analyses from January 2 to 7 were read, and lessons from January 6 to 8 were applied. A news scan found the headline, but it was not used to make trades.

This is a research note from Convex, which includes delayed trade bars that are not executable quotes. The data starts at 2026-09-10T20:00:22.157116+00:00. The stock bars dataset has a status of 200, a cutoff of 2026-09-10T19:40:22.157116+00:00, and there is no next page. The stocks included are HAIN, HITI, ZUMZ, and EPM. For example, HAIN had a closing price of 0.6418 on 2026-09-09 and 0.6375 on 2026-09-10. The option bars dataset also has a status of 200 and a cutoff of 2026-09-10T19:40:22.157116+00:00, with no next page. The options included are HITI261016C00002500 and HITI261016P00002500. For instance, HITI261016C00002500 had a closing price of 0.34 on 2026-09-08 and 0.31 on 2026-09-09. The data ends at 2026-09-10T20:00:23.991894+00:00.

The Convex trading fund manager ran for approximately 13 minutes on September 10 and did not make any trades, exits, or have any holdings. The account balance was $178.71 in cash and $850.99 in total. The manager performed 7 types of analyses, including statistical, technical, and fundamental, and reviewed lessons 696 to 698. A news scan found the headline that HITI's September 14 event was confirmed by the issuer on August 25, which was already pre-announced on August 4. The FrozenN9 thresholds were set at 2.5 with a price of $0.31 and a percentage change of +9.98043% for P2.5 with a price of $0.18 and a percentage change of -9.19765%, with coverage of 1/9 each versus 40%. The median absolute close was 5.494505%. A new put option with a price of $0.16 was still set at 1/9. The intrinsic-only mean payoff was $0.1667, which is not the true option value, and was $0.4719 times the premium. The Vectorbt training best T5 stock runup holdout was -2.6317% for N3, but no options backtest was available. The latest delayed S2.55 price at 19:31:51 did not change the rejection decision. The preliminary rejections for HAIN, EPM, and ZUMZ had no same-event study, so their missing tests were not reported as 0% wins. HAIN had an asynchronous price of $1.33 versus an intrinsic value of $0.3584, which deserves a fresh price and deliverable check, then 8 events, but not an arbitrage entry. EPM had a near call price of $51.06, which is above the $95 cap, and a cheap call needs a 31.5% increase. ZUMZ missed the default pre-event window tonight after the market closed. The monitors are set for September 11 pre-close, September 15 result review, and next run on September 21 at 21 UTC. The collector test ran both datasets in less than 1 second and is scheduled for 120 minutes with 60 runs, and the actual end time will be inspected. The full scripts are available in raw JSON format at /tmp/codex-run-B0b7gQ, and 9 verified backtest files will be used, while the obsolete N10 artifact has been superseded. No funds were proposed, and no independently validated outside-mandate operation was found. Cash was chosen because there was no measured qualifying payoff, and the next work will require fresh contract closes and issuer dates, not shutdown-driven risk.

A Python requests collector was tested on September 10 from 19:51:51 to 19:51:52 UTC. The test involved sending two parallel requests to collect stock and option data, specifically for stocks HITI, HAIN, EPM, and ZUMZ, and options HITI261016C00002500 and HITI261016P00002500. The requests were made to the /data/stock-bars and /data/option-bars endpoints with a SIP adjustment, a limit of 200, and a timeout of 20 seconds. The test also specified a start and end time, with the end time being the current UTC time minus 20 minutes, and the start time being 5 days before the end time. The script printed the last two daily bars, cutoff, HTTP status, and actual start and end times, but did not include tokens. The script is stored in a scheduled job that is set to run every 120 minutes for 60 runs, intended to span the September 14 event and September 15 response. The latest test, which ran through 19:31:51, showed prices of 2.55 for HITI, 0.6417 for HAIN, 3.855 for EPM, and 16.425 for ZUMZ, with call and put options for HITI at 9.31 and 10.16, respectively. Note that this is delayed trade history, not current bid or ask prices. Initial requests to the SIP endpoint returned a 403 error, and direct requests to the HITI endpoint on WordPress also returned a 403 error. The web tool reads original issuer pages, so scheduling a broken primary scraper is not recommended. The fact that the collector accepts requests does not guarantee success, so actual receipts should be inspected. Artifacts from the test, including raw bars, dates, analysis, and vectorbt script, are stored in /tmp/codex-run-B0b7gQ, and can be used to rebuild the workspace if it becomes unavailable.

The trading fund started on 2026-09-10 with $850.99 in equity and cash, and $178.71 in convex, with no prior history or holdings. The goal is to create a long-option payoff strategy with a proven advantage before entering a trade. The main constraint is the quality of the data and the whole-contract premium, with a maximum budget of $51.0594, which is 6% of the total. However, the fund will use a tighter budget of $25.5297, which is 3%, and a high-conviction ceiling of $42.5495, which is 5%. The fund can hold a maximum of three positions. The fund notes that cheap call options for low-price stocks can be cheap because the strike prices are far away or the earnings are already factored into the price. The fund will evaluate potential trades using a question tree, starting with the least predictable factors. The questions are: 1. Can executable contracts be found, considering factors such as price freshness, affordability, and identity, including deliverables, expiry dates, and corporate actions? 2. Is the event distribution comparable, looking at sample sizes, directional gaps, and robustness? 3. Is the catalyst incremental, examining fundamentals, anticipation, and consequences? 4. Can the exit strategy work, considering pre-event sessions, hold-through strategies, and sizing? The fund will use daily adjusted stock bars and individually priced contracts to make decisions, and will only use headlines for discovery purposes. Since there is no trade history, there is no claimed execution win rate, and no prior fund lessons apply.