Notes

The issuer's second quarter results for July 2016 show EBITDA of 901 million, which is an increase of 306 million from the previous quarter. Operating cash flow was 608 million and free cash flow was 422 million. Aluminum production was 636 thousand tons, a 5 percent increase from the previous quarter, while alumina production was 2.2 million tons, a 6 percent decrease. For the third quarter, it is expected that tariff cost improvements of 10 million will be offset by unfavorable alumina costs of 10 million in the aluminum segment, but there will be a net improvement of 10 million in the upstream alumina segment. A direct fetch of 403 was attempted, but a full read of the issuer's information was successful through the web. https://news.alcoa.com/press-releases/press-release-details/2026/Alcoa-Corporation-Reports-Second-Quarter-2026-Results/default.aspx

The company's annual report for FY2025, re-read on September 11, shows the following power supplier and concentration exposures: Grundartangi provides 70% of the power, which is fully indexed to the London Metal Exchange, and 30% is fixed with a London Metal Exchange component. Sebree gets its power from the MISO market via Kenergy until May 2028. Mt. Holly's cost of service is in effect through 2031. In terms of sales, Glencore accounts for 54.0% of 2025 sales, and the company's year-end ownership is 36.4%. Sales are based on the London Metal Exchange price plus regional and product premiums. The company also has limited suppliers for coke, pitch, and cathodes, and has take-or-pay obligations that can persist even after production is curtailed. These are existing risk exposures, not new failures. https://www.sec.gov/Archives/edgar/data/949157/000162828026013788/cenx-20251231.htm

The issuer of CENX reported its Q2 results on August 6. The company had shipments of 130,632 tonnes, which is higher than the 122,865 tonnes in Q1. Revenue was $752.1 million, compared to $649.2 million, and adjusted EBITDA was $326.9 million, compared to $231.4 million. The company guided $325-345 million for Q3. Improvements were driven by restarts and favorable realized metal and power prices, but were offset by raw materials costs. A refund of $94.3 million from July 45X is expected in a later period. As of June, the company had $343.4 million in cash and $480 million in debt, with 98,993,295 shares outstanding. The consolidated figures include alumina, and the revenue divided by aluminum tonnes does not represent the pure realized metal price, which is a price that does not include other costs or factors, also known as the price at which metal is sold. https://centuryaluminum.com/insights/century-aluminum-company-reports-second-quarter-2026-results/

A news scan found the headline about EGA R91104 Recycling, which has a gross capacity that is a blend of metal. The operator page was observed on September 11, and it showed an undated production capacity of 185,000 tons per year, a melting furnace capacity of 90,000 tons per year, and a scrap sorting capacity of 150,000 tons per year. This facility blends recycled and primary aluminum, which means that the full 185,000 tons per year cannot be counted as new primary substitution. https://www.ega.ae/en/about-us/operations/al-taweelah-recycling-plant

The issuer reported that as of April 3, 2026, Al Taweelah is expected to produce 1.6 million tonnes of cast metal and 2.4 million tonnes of refinery output, with an initial recovery rate of up to 12 months. The on-water, UAE, and overseas inventory buffers are not quantified. The annual cast tonnes may not be equal to the cell hot-metal capacity, which refers to the maximum amount of hot metal that can be produced by an electrolytic cell, a device used to extract metal from ore. https://media.ega.ae/update-on-egas-al-taweelah-site/

The issuer's report for the first half of 2026 shows the following results: cast metal production was 1006 kilotons versus 1420 kilotons, and sales were 939 kilotons versus 1373 kilotons, and alumina production was 602 kilotons versus 1142 kilotons. Earnings before interest, tax, depreciation, and amortization, which is a measure of a company's profitability, were 4506 million AED versus 4065 million AED, and operating cash flow was 1484 versus 3441. As of Monday, August 10, there were 227 cells. The company is using alternative ports to reduce stocks in the UAE, and forecasts that pre-incident shipping levels will return if the Hormuz Strait reopens. The company's inventory of strategic inputs is affecting its cash position, and management has an incentive to emphasize the company's resilience. https://media.ega.ae/ega-delivers-resilient-h1-2026-performance-maintaining-operational-and-supply-chain-continuity-amid-regional-disruption/

The issuer reported that as of August 26, 2026, 315 out of 1262 cells were restarted at the Al Taweelah facility. It will take about one week for these cells to stabilize, and some hot metal will be used for subsequent starts. The company forecasts that it will reach full production by the first quarter of 2027, but this is not guaranteed. The refinery is currently operating at around 50% capacity, and the smelter's recovery does not depend on the refinery being fully restarted. https://media.ega.ae/egas-al-taweelah-smelter-restoration-reaches-25-completion-milestone/

A news scan found the note that on 2026-09-11 at 04 UTC, the provider closest to data construction was examined. This provider constructs bars, which are groups of trading data, and includes the middle point of quotes, which is the average of the best buy and sell prices. It also allows bars with no trading volume to have prices, and late trades can update these bars. The current documentation for streaming data lists the following sources: Alpaca in the US, Kraken in the US, and Kraken in Europe. However, an older version of the historical data page, version 1.1, only mentions Alpaca. The discrepancy should be resolved by considering the version and location of the data, rather than looking at contradictory market evidence. Additionally, the location of the worker feed must be verified separately, and it should not be assumed that the bars provide information about the executable depth, which is the amount of trades that can be executed at a given price, or the total volume of trades on an exchange. https://docs.alpaca.markets/us/docs/real-time-crypto-pricing-data ; https://docs.alpaca.markets/us/v1.1/docs/historical-crypto-data-1

A data scan found the following information: there are three collector receipts. The first receipt has an id of 192, it was last run at 2026-09-11T02:15:16.289Z, which is a specific date and time in Coordinated Universal Time, or UTC for short, and it ended at 2026-09-11T02:15:20.760400+00:00. The next run is scheduled for 2026-09-11T04:15:16.289Z. The second receipt has an id of 199, it was last run at 2026-09-11T00:15:16.134Z and ended at 2026-09-11T00:15:28.239329+00:00, with the next run scheduled for 2026-09-11T04:15:16.134Z. The third receipt has an id of 206, it has no last run time or exit code, which is represented as null, meaning not yet verified or not successful, and its next run is scheduled for 2026-09-11T02:24:16.203Z. A local test of recipe 206 ended at 02:22:29.535548 on September 11, with a result of 200, which is a success code, and there was no pagination, or dividing of data into pages, needed. Convex list_jobs; tested https://runway.adityaperswal.workers.dev/data/stock-bars and /data/option-bars

A news scan found the following market data retrieved on September 11 at 02:03 UTC. The prices and percentage changes are as follows: 9961.HK was at 305.8 HKD with a decrease of 6.081081% at 01:48:24, 7203.T was at 3005 JPY with a decrease of 2.939276% at 01:48:43, SAP.DE was at 177.82 EUR with a decrease of 4.181485% on September 10 at 15:39:50, NXT.L was at 14915 GBp with a decrease of 1.907267% on September 10 at 16:12:40, TLX.AX was at 16.08 AUD with a decrease of 2.367942% at 01:43:37, BTC-USD was at 76896.74 with a decrease of 2.804887% at 02:03:40, ETH-USD was at 2450.94 with a decrease of 1.60653% at 02:03:43, and SOL-USD was at 99.29 with a decrease of 4.405888% at 02:03:44. These prices are based on secondary vendor quotes and unsynchronized sessions, and do not include any option edge or spot order information. The data comes from sources including https://finance.yahoo.com/quote/9961.HK/ and corresponding tickers. Yahoo Finance via sources.js world quote

A news scan found the headline 'TCOM261016P00035000 Catalog correction with September 10 delayed trade history'. The trade history for TCOM261016P00035000 on September 9, 2026, had a closing price of 0.4, a high of 0.45, a low of 0.35, 40 trades, an opening price of 0.38, a total volume of 331, and a volume-weighted average price of 0.410574. On September 10, 2026, the trade history had a closing price of 0.48, a high of 0.5, a low of 0.4, 21 trades, an opening price of 0.4, a total volume of 92, and a volume-weighted average price of 0.454348. This data was retrieved on September 11. The session date may not reflect the exact trade minute, and there is no executable ask claim. The full premium is 100 times the close price. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline 'MANU261016C00025000 Catalog correction with September 10 delayed trade history'. The trade history for September 11 retrieval is as follows: on September 8, 2026, at 04:00:00Z, the trade had a close price of 0.35, a high price of 0.35, a low price of 0.33, 2 trades, an open price of 0.33, a volume of 2, and a volume weighted price of 0.34. This is a delayed trade, the session date may not be the exact trade minute, and there is no executable ask claim. The full premium is 100 times the close price. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline 'CBRL261016P00037500 Catalog correction with September 10 delayed trade history'. The trade history for this stock option, which is a contract to buy a stock at a set price, is as follows: on September 9, 2026, the stock option's closing price was 0.27, its highest price was 0.49, its lowest price was 0.27, there were 2 trades, its opening price was 0.49, and its average price was 0.38. On September 10, 2026, the closing price was 0.5, the highest price was 0.5, the lowest price was 0.4, there were 5 trades, the opening price was 0.45, and the average price was 0.455556. This data was retrieved on September 11. The session date may not reflect the exact minute of each trade, and there is no claim on the ask price, which is the price at which someone is willing to sell. The full premium, which is the total amount paid for the stock option, is 100 times the closing price. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline 'ABAT261016P00003000 Catalog correction with September 10 delayed trade history'. The trade history for September 9 and 10 is as follows: on September 9, the closing price was 0.5, the highest price was 0.5, the lowest price was 0.47, there were 3 trades, the opening price was 0.49, the total volume was 4, and the volume-weighted average price was 0.49. On September 10, the closing price was 0.57, the highest price was 0.57, the lowest price was 0.54, there were 10 trades, the opening price was 0.54, the total volume was 20, and the volume-weighted average price was 0.5475. This data was retrieved on September 11. The session date may not match the exact trade minute and there is no executable ask claim. The full premium is 100 times the close price. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline 'ABAT261016C00003000 Catalog correction with September 10 delayed trade history'. The trade history for ABAT is as follows: on September 9, 2026, the closing price was 0.23, the highest price was 0.25, the lowest price was 0.23, there were 4 trades, the opening price was 0.25, and the total volume was 12, with a volume-weighted average price of 0.233333. On September 10, 2026, the closing price was 0.15, the highest price was 0.2, the lowest price was 0.15, there were 12 trades, the opening price was 0.18, and the total volume was 34, with a volume-weighted average price of 0.165588. This data was retrieved on September 11. The session date may not reflect the exact minute of the trade, and there is no executable ask claim. The full premium is 100 times the close price. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline 'TCOM261016C00045000 Catalog correction with September 10 delayed trade history'. The trade history for September 9, 2026, shows a closing price of 0.4, a high of 0.48, a low of 0.4, with 2 trades and a volume of 11. The trade history for September 10, 2026, shows a closing price of 0.32, a high of 0.35, a low of 0.3, with 31 trades and a volume of 102. This data was retrieved on September 11, and the session date may not reflect the exact trade minute, with no executable ask claim, and a full premium of 100 times the close. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline 'PLAY261016P00007000 Catalog correction with September 10 delayed trade history'. The trade history for September 9, 2026, shows a closing price of 0.47, a high of 0.47, a low of 0.32, 4 trades, an opening price of 0.32, and a volume-weighted average price of 0.4275. On September 10, 2026, the closing price was 0.5, the high was 0.5, the low was 0.5, there was 1 trade, the opening price was 0.5, and the volume-weighted average price was 0.5. This data was retrieved on September 11. The session date may not match the exact trade minute and there is no executable ask claim. The full premium is 100 times the close price. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline 'PLAY261016C00010000 Catalog correction with September 10 delayed trade history'. The trade data is as follows: on September 9, 2026, at 4:00:00 AM, the trade details were: the closing price was 0.45, the highest price was 0.6, the lowest price was 0.44, there were 16 trades, the opening price was 0.6, and the total volume was 54, with a volume-weighted average price of 0.499444. On September 10, 2026, at 4:00:00 AM, the trade details were: the closing price was 0.4, the highest price was 0.45, the lowest price was 0.4, there were 7 trades, the opening price was 0.45, and the total volume was 12, with a volume-weighted average price of 0.435. This data was retrieved on September 11. The session date may not reflect the exact trade minute, and there is no executable ask claim. The full premium is 100 times the close price. https://runway.adityaperswal.workers.dev/data/option-bars

A news scan found the headline about a cash-constrained half-Kelly calibration for C91102. The base rule for a 20-day period with a 15-minute holdout for Bitcoin, Ethereum, and Solana returns was analyzed. A grid of 10001 fractions from 0 to 1 was used to maximize the mean of the log of 1 plus the fraction times the return. The cash-constrained optima were found to be 0, 1, and 0, with half-wealth fractions of 0, 0.5, and 0. For a fund of 235.86, the hypothetical full notionals were 0, 117.93, and 0, with quarter probes of 0, 29.4825, and 0, and a capital of 212.7475. However, the actual values were all 0 because there was no entry and there were validation gaps. A matrix was used to calculate a score of 0.258924 for Ethereum, which is a risk unit binary average score, not a wealth fraction, as noted in note 2143. One operator had a liquidated Bitcoin trade in the live journal that could not be used to calibrate the edge. Six reused Ethereum trades and a confidence interval cross of 0 made the log-growth estimate unstable, so a bound of f less than or equal to 1 was used to avoid margin. https://runway.adityaperswal.workers.dev/data/crypto-bars ; local empirical log-growth calculation

A news scan found the following data: Observed on September 11 at 02:21 UTC, a fixed target was computed with a 10 basis point slip before a 20 basis point slip execution. This corrects the same terms stress, unlike recomputed target variants 23955-23996. The strategy uses a 20-day high and 10-day low, with a 2 ATR risk and a net 3R baseline, and a +2R breakeven with a 2 ATR trail and a 20-day timeout. A long-only 25 basis point credit asset fee is applied, with training from 2024-25 and reused in 2026 holdout. The data includes the symbol SOL, a 1-hour signal time frame, 15-minute execution, and a holdout part, with a 20 basis point slip and a trail. The return percentage is -10.84105866362728, the drawdown percentage is 34.000140794467214, and the win rate is 28.57142857142857. The sample size is 7, the mean percentage is -1.1365409527872063, the average win percentage is 13.912969321861102, and the average loss percentage is 7.15634506264653. The half Kelly field is a risk-unit score, not a notional fraction, as noted in note 2143. This is a no-quote, off-run implementation proof. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: Observed on September 11 at 02:21 UTC, a fix target was computed with a 10 basis points slip before a 20 basis points slip execution. This is a type of risk management, where basis points refer to a unit of measurement for interest rates or percentages. The same terms stress, unlike recomputed target variants 23955-23996, was used. The strategy included a 20-day high and 10-day low, with a 2 ATR risk, which stands for Average True Range, a measure of price volatility, and a net 3R baseline, where R refers to a unit of return. There was also a +2R breakeven, 2 ATR trail, and a 20-day timeout. The asset fees were 25 basis points for long-only credits. The data was trained from 2024 to 2025 and reused for 2026 holdout. The results included a signal time frame of 1 hour, execution minutes of 15, and a trail of true. The return percentage was 22.649221308842883, the drawdown percentage was 33.0116872332518, and the win rate was 41.66666666666667. The sample size was 12, the mean percentage was 2.3013284030441468, the average win percentage was 14.305373014291831, and the average loss percentage was 6.272989176418485. The half Kelly field, which is a risk-units score, not a notional fraction, was 0.08043580550975488. The raw half Kelly was also 0.08043580550975488. The confidence interval was between -3.826905858777564 and 8.48012973575873. There were 47215 bars, a maximum gap of 150 minutes, and 3 armed trades. The buy and hold percentage was 586.352957997662. The target slip was 10 basis points. This is a data line. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: Observed on September 11 at 02:21 UTC, a fixed target was computed with a 10 basis point slip before a 20 basis point slip execution. This is a type of stress test, also known as a simulation to see how a system performs under difficult conditions. The same terms stress test was used, unlike some recomputed target variants, which are different versions of a target, from 23955 to 23996. The test used a 20-day high and 10-day low, with a 2 ATR, or Average True Range, which is a measure of price volatility, risk and a net 3R baseline, which is a measure of profit. It also used a +2R breakeven, which is the point at which the profit equals the loss, 2 ATR trail, which is a way of setting a stop-loss order, and a 20-day timeout, which is a period of time after which a trade is closed. The test was long only, meaning it only bought assets, with a 25 basis point credit asset fee, which is a fee paid for borrowing assets. The test was trained on data from 2024 to 2025 and reused on a holdout set from 2026. The results of the test were: the symbol was SOL, the signal time frame was 1 hour, the execution minutes were 60, the part of the test was the holdout, the slip in basis points was 20, the trail was true, the return percentage was -11.56%, the drawdown percentage was 33.71%, the win rate was 28.57%, the sample size was 7, the mean percentage was -1.26%, the average win percentage was 13.45%, the average loss percentage was 7.15%, the half Kelly field, which is a risk-unit score, was 0, the raw half Kelly was -0.04, the confidence interval low was -7.78, the confidence interval high was 6.70, the number of bars was 6073, the maximum gap minutes was 120, the number of armed trades was 2, and the buy and hold percentage was -20.19. The half Kelly field is a risk-unit score, not a notional fraction, as seen in note 2143. There is no quote and no off-run implementation proof. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the headline about a SOL/USD trading strategy. The strategy was observed on September 11 at 02:21 UTC. The target price was computed with a 10 basis points slip before a 20 basis points slip execution. This is a type of risk management called a stop-loss, which helps limit losses. The same terms stress, unlike recomputed target variants 23955-23996, were used. The strategy used a 20-day high and 10-day low, with a 2 ATR risk, which stands for Average True Range, a measure of price volatility, and a net 3R baseline, which refers to a risk-reward ratio. The strategy also included a +2R breakeven, 2 ATR trail, and a 20-day timeout. The ATR trail is a type of stop-loss that adjusts to the price movement. Long only positions were taken, with a 25 basis points credit asset fee. The strategy was trained from 2024 to 2025 and reused in 2026 for holdout, which means it was tested on unseen data. The results included a return percentage of -2.8397, a drawdown percentage of 33.7744, and a win rate of 33.3333. The half-kelly field is a risk-units score, not a notional fraction, which refers to a measure of risk. The results also included a sample size of 12, a mean percentage of 0.3512, an average win percentage of 14.7013, and an average loss percentage of 6.8238. The half-kelly and raw half-kelly values were 0.0119. The confidence interval was between -5.7103 and 6.6642. The strategy used 11810 bars of data, with a maximum gap of 180 minutes, and 2 armed trades. The buy and hold percentage was 586.353. The target slip was 10 basis points. This is a data line from a trading strategy report. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the headline about SOL/USD C91102 frozen-target stress 1440min holdout. Observed on September 11 at 02:21 UTC, the fix target was computed with a 10 basis points slip before a 20 basis points slip execution. This corrects the same terms stress, unlike recomputed target variants 23955-23996. The strategy uses a 20-day high and 10-day low, with a 2 ATR risk and net 3R baseline, and a +2R breakeven 2 ATR trail with a 20-day timeout. ATR stands for Average True Range, which is a measure of price volatility. The long-only strategy has a 25 basis points credit asset fee, trained from 2024 to 2025 and reused for 2026 holdout. The results show a return percentage of -9.31%, a drawdown percentage of 32.40%, and a win rate of 16.67%. The half Kelly field is a risk-units score, not a notional fraction, as explained in note 2143. The data also includes other metrics such as sample size, mean percentage, average win percentage, and average loss percentage. The data is as follows: {"symbol":"SOL","signal_tf":"1Day","execution_minutes":1440,"part":"holdout","slip_bp":20,"trail":true,"return_pct":-9.311181002529565,"drawdown_pct":32.40499103083092,"win_rate":16.666666666666664,"sample":6,"mean_pct":-0.8370456351239212,"avg_win_pct":26.811745604764138,"avg_loss_pct":6.366803883101532,"half_kelly":0,"raw_half_kelly":-0.015609681806304843,"ci_low":-9.061755447312118,"ci_high":10.682732771200797,"bars":253,"max_gap_minutes":1440,"armed_trades":1,"buy_hold_pct":-20.81869466583686,"target_slip_bp":10}. This is a raw data output. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the headline about a SOL/USD trade with a frozen-target stress test that lasted 1440 minutes. The test was observed on September 11 at 02:21 UTC. The target was computed with a 10 basis points slip before a 20 basis points slip execution. This is a type of risk management, where basis points refer to a unit of measurement for interest rates and slips refer to the difference between expected and actual prices. The same terms stress test was used, unlike other recomputed target variants 23955-23996. The test used a 20-day high and 10-day low, with a 2 ATR risk, which stands for Average True Range, a measure of price volatility, and a net 3R baseline, where R refers to a unit of risk. The test also included a +2R breakeven, a 2 ATR trail, and a 20-day timeout. The asset fees were 25 basis points for long-only credits. The test was trained from 2024 to 2025 and reused for holdout in 2026. The results included a return percentage of 26.93%, a drawdown percentage of 27.31%, and a win rate of 44.44%. The half-kelly field is a risk-units score, not a notional fraction, as explained in note 2143. This was an off-run implementation proof, meaning it was not a live trade, and no quote was provided. https://runway.adityaperswal.workers.dev/data/crypto-bars