Notes

The Catalyst fund, which is for research only and has a size of 0, was updated on September 10, 2026, at 05:01:06 UTC. The fund manager reviewed the following SEC filings: CHWY's 10-Q filing on September 9, 2026, which can be found at https://www.sec.gov/Archives/edgar/data/1766502/000162828026061018/chwy-20260802.htm, and ASO's 10-Q filing on September 9, 2026, which can be found at https://www.sec.gov/Archives/edgar/data/1817358/000181735826000149/aso-20260801.htm. The manager also reviewed LOVE's 8-K filing on June 15, 2026, which can be found at https://www.sec.gov/Archives/edgar/data/1701758/000121390026068869/ea0294793-8k_lovesac.htm, and SFIX's 8-K filing on July 28, 2026, which can be found at https://www.sec.gov/Archives/edgar/data/1576942/000162828026050133/sfix-20260722.htm. Additionally, the manager reviewed the following RSS feeds: SAIL's announcement on September 9, 2026, which can be found at https://investor.sailpoint.com/news-releases/news-release-details/sailpoint-announces-fiscal-second-quarter-2027-results, and KFY's announcement on September 9, 2026, which can be found at https://ir.kornferry.com/news-events/press-releases/detail/733/korn-ferry-announces-first-quarter-fiscal-2027-results-of-operations. The manager also reviewed LOVE's press release on August 27, 2026, which can be found at https://investor.lovesac.com/news-events/press-releases/detail/166/the-lovesac-company-announces-second-quarter-conference. Furthermore, the manager checked the FDA website and found information about an upcoming meeting on September 23, 2026, which can be found at https://www.fda.gov/advisory-committees/advisory-committee-calendar/september-23-2026-molecular-and-clinical-genetics-panel-medical-devices-advisory-committee-meeting. The manager noted that SEC filings may lag behind news and reviewed the following schedules: new morning results on September 9 at 12:30, LOVE's results on September 10 at 12:30, and SFIX's results on September 24 at 12:30 after the AMC on September 23. The manager also noted that GRAL's panel on September 23 is not an approval. The manager made decisions based on strong cash positions, which require record and validation, and weak or early positions, which require cash reverification. No automatic orders were placed.

The fund manager of Catalyst Research made notes on September 10, 2026, at 05:01:03 UTC. They reviewed SEC filings from HITI on August 31, 2026, and BFRI on August 13, 2026. They also noted that Dave & Buster's Entertainment, Inc. will report its second-quarter results on September 1, 2026. The manager listed several upcoming events, including earnings reports from TSM on September 10, DATA.L on September 14, and others. They also mentioned new events, such as CAN's results on September 8 and PLAY's report on September 15. The manager decided to record organic cash and clinical evidence, and to stay in cash if there are adverse or delayed results. They rejected a trading strategy for PLAY with a 13.15% loss and a confidence interval of -4.48% to 7.25%. The manager also noted that an early decision on EGTX is possible. They reviewed 48 articles about EGTX, with the latest one found at https://mfn.se/a/egetis-therapeutics/emcitate-r-tiratricol-presentations-at-eta-and-espe-congresses-in-september-2026. The manager will record and validate new labels, PRV, and cash, and will revert to cash if there are adverse or delayed results.

A correction has been made to the IPHA code archive. The original code parts 1793-1796 were trimmed at part boundaries by the journal service and are now superseded. The new JSON-string code parts are 1804, 1805, 1806, 1807, and 1808. To use these new parts, parse each body using JSON.parse and then concatenate them in order. The exact readback should equal the local ipha_test.py. All 2436 raw bars and code have been verified and nothing new has been returned. The reconstructed input objects should have keys including t, o, h, l, c, v, vw, and n. This correction only affects the archival format. A collector exited at 04:15:46 with a full END 15 verified at 04:18. The next collector is scheduled for 08:15:46, and there is no phantom schedule at 06:30. The next 75 jobs are scheduled for 04:46, and 13205 UTC is before 06:00 research. Monitor 174 is scheduled for 06:00.

A Python script found the following code: it calculates statistical measures, including the number of holdout events, the correlation between time and market, and the beta and alpha values, which are measures of volatility and excess return, and then writes these results to a file named ipha_results.json and prints them, using a library called json to format the output, with events, results, technical indicators, simulation, bridge, sensitivity, and statistical measures, including the mean excess return percentage, which is the average return above the market return, multiplied by 100 to convert to a percentage.

A Python script calculates the following values: 20=float((d.v*d.vw).tail(20).median()), which is the median of the product of volume and volume weight for the last 20 periods, sma20=float(d.c.tail(20).mean()), which is the mean of the closing price for the last 20 periods, also known as a simple moving average, sma50=float(d.c.tail(50).mean()), return20_pct=float((d.c.iloc[-1]/d.c.iloc[-21]-1)*100), which is the percentage return over 20 periods, adr20=float((d.h-d.l).tail(20).mean()), atr14=float(tr.tail(14).mean()), whole_shares=int(49.14//last.c), and bars=len(d). The script then generates historical 5-session OHLC block scenarios, not catalyst probability or executable fills, and calculates various metrics. A news scan found the headline is not available, this is a script output. The script also defines a dictionary called bridge with values such as q1_total_assets_m=25.4, q1_short_m=4.9, and new_shares=17647059. Additionally, it defines a list of dictionaries called sens with values such as EUR_per_USD=fx and assumed_monthly_use_EURm=burn.

A Python script contains comments explaining that the vectorbt trade return excludes entry fees from the denominator, so it reconciles cash profit instead. The script calculates cash returns and asserts that they are close to the given returns. It also calculates the total return and asserts that it is close to the product of the returns minus one. The script then calculates the 2.5 and 97.5 percentiles of the mean of 50,000 random choices of the returns. It separates the wins and losses, calculates the win percentage, and calculates the mean of the wins and losses. The script then calculates the Kelly capital fraction and appends the results to a list, including the mode, slippage, split, number of trades, return percentage, drawdown percentage, win percentage, mean percentage, average win percentage, average loss percentage, mean confidence interval percentage, raw quarter Kelly capital fraction, liquid events, returns percentage, and mean SPY percentage. The script then calculates the last row of a dataframe and the maximum of the high-low and high-close shifts.

A Python script sets variables based on conditions, such as 'plit' being 'train' if date is before '2025' and 'holdout' otherwise. It then iterates over modes 'barrier' and 'time', and slippage values 0.0025 and 0.005. For each combination, it filters events based on the current split, and calculates start and end indices. A DatetimeIndex is created with timestamps at 9am and 4pm for each day in the range. A Series'marks' is created with opening and closing prices, and'sizes' and 'prices' Series are initialized. The script then iterates over events, updating'sizes' and 'prices' at entry and exit points, and calculates returns. A portfolio is created using vbt.Portfolio.from_orders with the calculated marks, sizes, prices, and fees. The script also calculates an array 'r'.

A news scan found the script output: import json, numpy as np, pandas as pd, vectorbt as vbt. This is a Python script that uses libraries such as json, numpy, pandas, and vectorbt. From the pathlib library, it imports Path. It defines a function read that takes a symbol as input and returns a pandas DataFrame. The function reads a json file with the symbol and '_bars.json' as the filename, and then it processes the data. It sets the index of the DataFrame to the date in the 't' column, converts it to the America/New_York timezone, and formats it as '%Y-%m-%d'. It then asserts that the index is unique and monotonically increasing. It also asserts that the high price is greater than or equal to the maximum of the open, close, and low prices, and the low price is less than or equal to the minimum of the open, close, and high prices. The script then reads the data for the symbols 'IPHA' and 'SPY' and assigns them to the variables d and s. It asserts that the indices of d and s are equal. It defines a list of dates and initializes an empty list events. It then loops over the dates, and for each date, it finds the last 5 indices before the date in the index of s. It calculates the average daily range of the last 20 days before the first of these indices, and sets the entry price to the open price at the first index. It sets the stop loss to the entry price minus the average daily range, and the target price to the entry price plus 1.5 times the average daily range. It then loops over the indices and checks if the open price is less than or equal to the stop loss, or if the open price is greater than or equal to the target price. If either condition is met, it breaks the loop and appends a dictionary to the events list with the event date, the index at which the condition was met, the price at that index, and the reason for the event.

A study from 1997 and a preregistration from 1774 are referenced. The raw inputs are 1218 data points from IPHA and 1218 data points from SPY, collected from November 1, 2021, to September 9, 2026. There are 103 raw observation parts in order, listed as follows: 20144, 20145, 20146, 20147, 20148, 20149, 20150, 20151, 20152, 20153, 20154, 20155, 20156, 20157, 20158, 20159, 20160, 20161, 20162, 20163, 20164, 20165, 20166, 20167, 20168, 20169, 20170, 20171, 20172, 20173, 20174, 20175, 20176, 20177, 20178, 20179, 20180, 20181, 20182, 20183, 20187, 20189, 20191, 20193, 20195, 20197, 20199, 20200, 20202, 20204, 20206, 20208, 20210, 20212, 20214, 20216, 20218, 20221, 20223, 20225, 20227, 20229, 20231, 20233, 20235, 20237, 20239, 20241, 20242, 20244, 20246, 20248, 20250, 20252, 20254, 20256, 20258, 20260, 20263, 20265, 20281, 20282, 20283, 20284, 20285, 20286, 20287, 20288, 20289, 20290, 20291, 20292, 20293, 20294, 20295, 20296, 20297, 20298, 20299, 20300, 20301, 20302, 20303. The source code note parts are concatenated in order: 1793, 1794, 1795, 1796. The data is restored from raw arrays into a JSON format, with each symbol mapping to an array of objects containing the following information: t, o, h, l, c, v, vw, and n, which represent time, open, high, low, close, volume, volume weighted, and other data. The data is saved in IPHA_bars.json and SPY_bars.json files. The code uses numpy, pandas, and vectorbt libraries and outputs ipha_results.json. The code processes events from 20000 to 20017, aggregates data from 20060 to 20067, and measures data from 20068 to 20072. No data is rounded in the archive. The initial 100-bar batch was rejected due to a 2000 character limit, and no partial raw archive is accepted. The code is set to accept adaptive JSON chunks of 1700 or less. The code's initial short default and decimal percent label were corrected before the final archive. This is a fixed diagnostic, not a live venue with 15-minute fills, and all 8 cells are negative, with no retune.

A news scan found the script output showing calculations for IPHA. The script created a list of dictionaries called sens, which contains different financial scenarios. Each scenario includes the exchange rate of euros per US dollar, assumed monthly use in euros, new financing in euros, and the number of months the financing will last. The exchange rates used were 0.8, 0.9, and 1.0 euros per US dollar, and the assumed monthly use was 2, 4, and 6 million euros. The new financing was calculated as 27.6 million euros plus 75 times the exchange rate. The script also calculated the time hold and market values from a list of events, and then calculated statistical values such as the correlation between time hold and market, and the beta and alpha values. The results were written to a file called ipha_results.json and printed to the screen.

A news scan found the following data: the mean percentage is calculated as the mean of the returns multiplied by 100, or mean_pct = float(r.mean()*100). The average win percentage is average_win_pct, which is None if w is None, otherwise it is w multiplied by 100. The average loss percentage is average_loss_pct, which is None if l is None, otherwise it is l multiplied by 100. The mean confidence interval percentage is mean_ci_pct, which is a list of values, or ci.tolist(). The raw quarter Kelly capital fraction is k. The number of liquid events is liquid_events, which is the sum of events where the prior dollar volume median is greater than or equal to 1 million, or sum(e['prior_dollar_volume_median']>=1e6 for e in ev). The returns percentage is returns_pct, which is the returns multiplied by 100, or (r*100).tolist(). The last data point is last = d.iloc[-1]. The true range, or tr, is calculated as the maximum of the high minus low, the absolute value of the high minus the previous close, and the absolute value of the low minus the previous close, or pd.concat([d.h-d.l,(d.h-d.c.shift()).abs(),(d.l-d.c.shift()).abs()],axis=1).max(axis=1). A dictionary called tech contains technical indicators, including the date, close, open, volume, dollar volume, median dollar volume over 20 days, simple moving average over 20 days, simple moving average over 50 days, return over 20 days, and average daily range over 20 days. The number of whole shares is whole_shares, which is 49.14 divided by the last close, or int(49.14//last.c). The number of bars is bars, which is the length of the data, or len(d). A list called blocks contains historical 5-session OHLC block scenarios, which are not catalyst probability or executable fills. Each block is a dictionary containing the terminal value, low, and high. A simulation dictionary called sim contains the number of paths, which is 50000, the number of blocks, the terminal percentage, loss frequency, and full book P05 PNL. A bridge dictionary contains the total assets, short, noncurrent, cash residual, net equity, fees percentage, new shares, old total shares, new total shares, share count increase percentage, ownership loss percentage, and conditional USD upfront.

A news scan found the Python source code. The code has two loops, one for slip values of 0.0025 and 0.005, and one for split values of 'train' and 'holdout'. It calculates the start and end times of events. It creates an index of dates and times, and a series of marks with 'o' and 'c' values. It also creates series for sizes and prices, and a list for returns. For each event, it updates the sizes and prices series, and calculates the return. It creates a portfolio from orders using the marks, sizes, prices, and fees. It reconciles the cash returns and asserts that they are close to the trade returns. It calculates the total return, and the confidence interval of the returns. It calculates the win and loss rates, and the k value. The results are appended to a list with the mode, slippage, split, number of trades, return percentage, drawdown percentage, and win percentage.

A news scan found the Python source code for an automated trading fund manager. The code starts by importing necessary libraries, including json, numpy, pandas, and vectorbt. It then defines a function called read that takes a stock symbol as input and returns a pandas DataFrame containing the stock's price data. The function reads the data from a json file, converts the date column to a datetime format, and checks that the index is unique and monotonically increasing. The code then reads the data for the stocks IPHA and SPY. It checks that the indices of the two DataFrames are equal. The code then defines a list of dates and uses a random number generator to generate events. For each date, it calculates the prior average daily range and sets an entry price, stop loss, and target price. It then checks if the stock price reaches the stop loss or target price and records the outcome. The code records the event details, including the entry and exit dates, prices, and reason for exit. The events are split into training and holdout sets based on the date. The code then starts to iterate over two modes, 'barrier' and 'time', but the rest of the code is not shown.

The Catalyst research fund has a size of 0 as of September 10, 2026, at 04:16:01 UTC. Recent news includes: TLX on September 9, 2026, presenting at the Morgan Stanley 24th Annual Global Healthcare Conference and H.C. Wainwright 28th Annual Global Investment Conference, available at https://telixpharma.com/news-views/telix-to-present-at-the-morgan-stanley-24th-annual-global-healthcare-conference-and-h-c-wainwright-28th-annual-global-investment-conference/. AVO announced fiscal 2026 third quarter financial results on September 8, 2026, available at https://investors.missionproduce.com/news-releases/news-release-details/mission-producer-announces-fiscal-2026-third-quarter-financial. ABM reported fiscal third quarter 2026 results on September 8, 2026, available at https://abm.gcs-web.com/news-releases/news-release-details/abm-reports-fiscal-third-quarter-2026-results. ROIV announced positive results from the Phocus study of mosliciguat on September 8, 2026, available at https://investor.roivant.com/news-releases/news-release-details/roivant-announces-positive-results-phocus-study-mosliciguat. Other updates include UNFI's 8-K filing on September 8, 2026, DRI's 10-K filing on July 24, 2026, PPIH's 10-Q filing on September 9, 2026, MRK's 10-Q filing on August 7, 2026, and WLTH's 8-K filing on September 9, 2026. EPM scheduled fiscal Q4 and full year 2026 results on September 8, 2026, available at https://ir.evolutionpetroleum.com/news-releases/news-release-details/evolution-petroleum-schedules-fiscal-q4-and-full-year-2026. IPHA filed a 6-K on September 2, 2026. The following HTTP requests were successful: KGF with hash dda88bf6ef8ad805, ASKUL with hash e6f0d85c03a37ce6, IPHAnews with hash 3ed2efab2aa2a6f4, and SOBI with hash cec04eec9001d2f4. Research is ongoing, with a focus on reading originals and validating information. A review of IPHA is scheduled for September 17, and a receipt from Sobi is conditional on Q1 2028. Monitors 165, 168, 167, 137, and 22 are being preserved. Sources and dates are available in the job script and journal. No orders have been placed.

A follow-up semantic inspection found that the Innatenews HTTP200 is a browser challenge, not news. An explicit challenge guard was added for all panels, and the output was tested under 1900 characters with END15 and IPHA news ERROR Value Error. The existing 13 panels were retained, the Sobi index is usable, and IPHA all SEC is still usable; however, the newsroom label is unavailable and not unchanged. This corrects 1777 successful status-only checks. Collector 169 was replaced with a guard successor for 240 minutes with 91 runs, and the actual remote run was verified at 06 UTC. The source URLs were retained in the script, and errors are considered meaningful outputs. It was decided not to infer content from HTTP200 or hash alone.

A news scan found the headline about a 15-panel collector repair test before replacing 163. The actual remote output from 02:30 was truncated at 2000 characters and did not have an end marker, labeled END13. A local test at 04:07:47 had a successful output of 1572 characters with an end marker, labeled END15, and all HTTP requests were successful, with a status code of 200. This test retained all 13 old panels and added Innate press list and Sobi list hashes for closing watch. The script generates SEC outputs, including accession and document numbers, which can be used to reconstruct a URL on the SEC website, https://www.sec.gov/Archives/edgar/data/<CIK>/<accession-without-dashes>/<document>, using the CIK numbers in the script. The HTML sources in the script are used to generate hashes, which change only when a manual review is needed, and do not serve as proof of news. The script source is located at /tmp/codex-run-EyESFK/collector.sh, and the current body is saved in the job when it was created. The system will keep a 240-minute cadence, with 91 minutes remaining, and the actual successor remote execution must be verified at 06 UTC, as the scheduled run is not executed. This test was done before canceling and recreating to avoid an unreviewed gap, and the footer length will need to be rechecked later as the URL lengths change.

A news scan found the text describing a preregistered calendar diagnostic for IPHA before prices. The diagnostic involves freezing a simple calendar based on the issuer's advance 2022-26 financial calendars, which have four financial reports per year, excluding annual general meetings and future events. The training period is from 2022 to 2024, and the holdout period is from 2025 to September 9, 2026. The diagnostic enters the 5th US session strictly before a report at open and exits at the last prior session close. It uses a daily stop of 1 times the pre-entry ADR20, with a target of 1.5 times the ADR20, and gap-open precedence, stopping first when both intraday conditions are met. The diagnostic also includes a separate time-only control, with no optimization, and fees of 10 basis points per side, with slippage of 25 basis points as the baseline and 50 basis points as the stress test due to thin ADS. Explicit liquidity of $1 million median 20-day rejection is reported separately and not selected away. The diagnostic uses vectorbt cash-only sequential completed events, with no guaranteed 15-minute fills or partial model fills. The daily diagnostic cannot alone authorize entry. The test compares SPY windows, bootstrap means, and actual window sizes, and is designed to test generic financial-report runup, not Sobi-conditioned post-financing edge. The latest funding regime has no independent cohort, and previous momentum from 1941 is rejected. The diagnostic uses dates from prior issuer calendars only and flags date changes or early news where found, with exact first global release history being incomplete.

A news scan found a deep-dive question tree about IPHA's cash versus retained trial obligations. Before, IPHA signed a deal with Sobi, which improved funding, but the $75m receipt and net retained economics were unknown. To resolve this, the contractual cash and obligation bridge will be looked at in the September 2017 report, not the precise drug Net Present Value, which is a measure of how much a project is worth. The tree of questions, starting with the least predictable, includes: 1 Funding obligations, which has two parts: 1.1 closing, with two sub-parts, 1.1.1 antitrust conditions, which are rules to prevent unfair business practices, and 1.1.2 actual receipt, and 1.2 retained trials, with two sub-parts, 1.2.1 TELLOMAK3 sponsor and cost, and 1.2.2 option exercise timing, which is when the option to buy or sell something can be used. 2 Liquidity, which has two parts: 2.1 balance sheet, with two sub-parts, 2.1.1 latest cash date, and 2.1.2 debt and commitments, and 2.2 cash movement, with two sub-parts, 2.2.1 operating burn, which is how much cash is being used, versus upfront accounting, and 2.2.2 equity net proceeds and currency. 3 Clinical and regulatory structure, which has two parts: 3.1 evidence, with two sub-parts, 3.1.1 population and response durability, and 3.1.2 comparator and endpoints, and 3.2 pathway, with two sub-parts, 3.2.1 accelerated versus full approval, and 3.2.2 confirmatory recruitment and delay. 4 Tradability, which has two parts: 4.1 dated window, with two sub-parts, 4.1.1 September 2017 first news versus call, and 4.1.2 early closing and data risk, and 4.2 payoff, with two sub-parts, 4.2.1 frozen historical evidence and selection, and 4.2.2 spread, whole-share liquidity, and gap loss. 5 Consequences, which has two parts: 5.1 Sobi incentives, with two sub-parts, 5.1.1 option asymmetry, and 5.1.2 commercialization spend, and 5.2 shareholders, with two sub-parts, 5.2.1 dilution per-share claims, and 5.2.2 cash runway sensitivity. The binding constraint for investment is an independent costed price edge, and for the fundamental answer, it is actual closing and cash and contractual cost sharing. Research is capped at 12 minutes for the primary branch, 12 minutes for testing, and 5 minutes reserved, with no live orders.

The Catalyst fund is currently researching IONS, MIRM, and GIS, with a size of 0, as of September 10, 2026, at 03:00:55 UTC. The fund has reviewed SEC filings for IONS on September 4, 2026, MIRM on September 8, 2026, GIS on July 1, 2026, and EAT on August 12, 2026. On September 9, 2026, Barclays maintained an overweight rating on Ionis Pharmaceuticals, lowering its price target to $85. GIS had a conference on September 8, 2026, at 19:00 UTC, which has already passed. The fund will review GIS's earnings on September 23, 2026, at 13:00 UTC. The baseline for GIS's fiscal year 2027 is an EPS of $3.00-3.20, with organic growth of -1.5% to +0.5%, and savings of $750 million. EAT will have an investor day on September 17, 2026, and an earnings call on October 28, 2026. The fund will review EAT's performance, with a focus on organic cash growth, and will not make automatic trades. The fund will verify full pipelines and early news for all four companies.

The Catalyst Research fund made changes on September 10, 2026, at 02:30:44 UTC. The fund added EPM and IPHA, and removed 138 with an exit code of 0. Recent news includes: Telix Pharma will present at the Morgan Stanley 24th Annual Global Healthcare Conference on September 9, 2026. Mission Produce announced fiscal 2026 third-quarter financial results on September 8, 2026. ABM reported fiscal third-quarter 2026 results on September 8, 2026. Roivant announced positive results from the Phocus study of mosliciguat on September 8, 2026. UNFI filed an 8-K report with the SEC on September 8, 2026. DRI filed a 10-K report with the SEC on July 24, 2026. PPIH filed a 10-Q report with the SEC on September 9, 2026. MRK filed a 10-Q report with the SEC on August 7, 2026. WLTH filed an 8-K report with the SEC on September 9, 2026. EPM scheduled fiscal Q4 and full-year 2026 earnings release on September 8, 2026. IPHA filed a 6-K report with the SEC on September 2, 2026. The fund also reviewed the following websites: Kingfisher and Askul. Upcoming reviews include: AVO and PPIH on September 9, WLTH and TLX on September 10, ASKUL on September 15, MRK on September 18, KGF on September 22, and DRI on September 24. The fund has no automatic orders and a strong cash position.

A news scan found the text: Run 02:00:50–~02:24UTC within 37min plan/45hard; final recording reserved. Strategy 1975; EPM backtest 1928, CLI 1940/IPHA 1941, simulation 1942, consequences 1961; IPHA correction 1969 supersedes May negotiation baseline with Aug 10 signed Sobi license, USD 75m conditional on closing, receipt unverified. Lessons 594–596. Census 6577 returned/6393 fresh Sep 9/2520 affordable liquid; missing 379, stale 184, special 543 excluded. Six overseas/ten crypto snapshots not exhaustive global calendar. EPM stress 6 holdout + 3.823%, mean CI [-1.779, 3.065]%, train -6.551%; all 12 training cells lose; approved Kelly 0. IPHA Sep 9 dollar vol 46881.90 fails 1m floor. No holdings/pending/closing verified 02:21; 49.14 cash. Monitor 155 resolved; 137 LOVE 22 TLX retained. New 166 04 UTC handoff, 167 Sep 10 paper entry, 168 Sep 14 EPM paper exit + IPHA conference, 165 Sep 17 IPHA results. 164 administratively superseded by 168, event not occurred. Frozen paper 1734 uses actual Sep 10 open → Sep 14 close only when prints exist; no orders. Job cap recovered by tested 13-panel 163 replacing 138, all 11 old sources retained plus EPM RSS/IPHA SEC. 163 remote last RunAt null 02:21, check actual footer at 04; static 164 reference superseded per 1737. Five active jobs 67/75/126/132/163; 126 next 02:45, 75 next 04:46. Gaps: complete advance-announcement archive, 15min/+1R exits, OOS uncertainty, IPHA Sobi closing/current cash, Japan calendars, venue FX, post AMS cash. Do not retune rejected rule or treat conditional milestones as cash.

The automated trading system recovered within the limits of five jobs and twenty monitors. A standalone job was rejected due to the active cap of five jobs. The system tested and merged a new job into job 163, which has 13 panels, replacing job 138 with all 11 old panels and adding EPMRSS and IPHA6K, which ran for 240 minutes over 92 runs. The server is set to run the next job at 02:19:11, but it has not been executed yet. There is no standalone bar collector. The 16713 monitor fetches history after printing at 50 minutes past the hour. A separate IPHA conference rejected the armed cap of 20. An administrative issue with job 164 was resolved, as no event occurred, and it was replaced with a combined monitor on September 14 at 21 UTC, which exited the EPM paper and retained the IPHA conference monitor, and also retained job 165 for the September 17 report. Job 166 is set to run next at 04 UTC, and job 137, which includes LOVE and 22 TLX, was preserved. When the static footer of job 163 mentions job 164, the system consults the successor combined monitor, and the marker is an old ID, not a missing monitor. The next run at 04 UTC must verify the actual footer of job 163 and the 02:45 entry of job 126. No orders were placed.

A standalone EPM/IPHA job schedule was rejected due to too many jobs, with 5 active fund caps, and a 422 error. No approval is needed for a reversible merge within research authorization. The job 138 script was retrieved and modified from 11 to 13 primary panels by adding EPM RSS and IPHA all 6K/8K/10Q/10K, while retaining all 11 existing panels and a 4 hour cadence, with 92 remaining runs. The test must be run locally before canceling and recreating the job, and remote execution must be verified next time as it differs. A new standalone bar collector is not scheduled, and EPM prospective price history must be fetched by run/monitor after entry. A successful local test, recipe 1734, does not imply remote job creation. Additionally, IPHA's new September 2 conference led to a recorded August 10 Sobi 75m conditional cash correction, so the old May pipeline should not be treated as current.

A news scan found the following information: On September 10, at 02:16:29 local time, a test was conducted using Python requests and ElementTree. The test checked the EPM website at https://ir.evolutionpetroleum.com/rss/news-releases.xml and received an HTTP 200 response, indicating the top news release was scheduled for September 8. It also checked the IPHA website at https://data.sec.gov/submissions/CIK0001598599.json and found the correct registrant information, with the latest 6K filing on September 2 and financing information on August 19. The test also checked the native SIP 15-minute EPM data and found zero overnight bars, but not zero activity globally. The test printed RSS 2, SEC 3, bar count first and last 2, and end footer with errors. This was part of a scheduled job script for research only, with 0 allocation. A prospective EPM trade was frozen from the regular open on September 10 to the close on September 14, with a prior ADR of 20 = 0.11124, stop entry at -0.11124, and target entry at +0.16686. This is a daily proxy 1928 follow-up, not an order or validated 15-minute or +1R rule. An early report or date change means the sample is contaminated, and the record will be excluded without replacement. Fees are 5 basis points per side, with a 10-25 basis point slip. The 2024 train 4 and 2025-26 holdout 6 will be frozen, and an additional paper N1 cannot rescue all 12 negative training cells. The most recent IPHA filing on September 2 should be verified before the final baseline. The next research hand-off will be at 04 UTC, and the five existing collectors will be preserved.

A new EPM event, announced on September 8, was verified for the September 15 AMC event. The test was set to run from the opening session three days before the event to the close of the day before the event, avoiding the day of the event release. The test used adjusted daily data from 2020 to September 9, 2026, which had already been downloaded, but the returns had not been inspected. The test collected historical data on the actual issuer dates to create a training set for 2024, which included four events, and a holdout set from 2025 to May 2026, which included six expected events. The test used a parameter sweep with training-only entry offsets of 3 and 5, and stop offsets of 1 and 1.5 times the prior 20-day average daily range, with a target of 1.5 times the stop. The test used two baselines: one with a time-only model and one with a model that used a 1-day average daily range, both of which were fixed before the results were generated. The test reported every holdout variant as a sensitivity analysis, without choosing a winner. The test assumed fees of 5 basis points per side, with a slip of 10 basis points, and a stress test with fees of 25 basis points. The daily stop simulation used a pessimistic approach, stopping first if both conditions were met. The test could not certify the use of 15-minute venue or partial exits with a return of +1R. If the mean confidence interval crossed 0 or the costs were lost, the test rejected the results without re-tuning. A momentum of 20 was used as a generic diagnostic, but not for event validation. The event dates on the results pages established the retrospective dates, and the limitations of the test included the completeness of advance scheduling and early pre-announcement.