Notes

A news scan found the headline 'CSTM R91104 costed L10 train slippage0.001'. The trading results for the symbol CSTM are as follows: the return percentage was -61.29714714645047, the drawdown percentage was 66.69543938093223, and the win percentage was 34.883720930232556. The mean percentage return was -1.861651470978222, the average win percentage was 7.875420407128093, and the average loss percentage was 7.077939977106605. There were 43 trades, with 1 target, 24 stops, and 18 time exits. The half Kelly criterion was 0. A data line found the following information: Frozen2178, vectorbt longonly, with two accounting points per day allowing for same day entry and exit. Original open and close marks were used, except for the exit price. Fees were 10 basis points per side, plus the listed slippage of 0.001. A daily touch and gap model was used, but not based on 15 minute venue data, and there were zero ambiguous bars. No parameter was selected from the holdout set. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the headline 'CSTM R91104 costed L20 holdout slippage0.002'. The symbol is CSTM, which is a stock ticker. The lookback period is 20, meaning the system looked at 20 days of data. The period is called the holdout period, which is a term used in finance to describe a set of data used to test a system. The slippage is 0.002, which refers to the difference between the expected price of a trade and the actual price. The return percentage is 27.889096029296635, the drawdown percentage is 19.94679403354033, and the number of trades is 15. The win percentage is 46.666666666666664, the mean percentage is 2.1162182531996745, the average win percentage is 11.767464607178923, and the average loss percentage is 6.328622306532168. The half Kelly criterion, a formula used to determine the optimal size of a bet, is 0.08991819069966198. There were 2 targets, 5 stops, and 8 time exits. The ambiguous count is 0, meaning there were no unclear trades. The block mean confidence interval percentage is between -2.692580486873163 and 7.417179296151888. The system used is called vectorbt longonly, which is a type of automated trading system. The system allows for two accounting points per day, which means it can enter and exit a trade on the same day. The original open and close marks are used, except for the exit price. The fees are 10 basis points per side, plus the listed slippage. The daily touch and gap model is not based on a 15-minute venue, and there are zero ambiguous bars. No parameters were selected from the holdout period. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the headline 'CSTM R91104 costed L20 holdout slippage0.001'. The symbol is CSTM, which is a stock symbol, a unique code to identify a stock. The lookback period is 20, meaning the system looked at 20 days of data. The period is holdout, which is a part of the data used to test the system. The slippage is 0.001, which is the difference between expected and actual trade prices. The return percentage is 31.783908071503248, and the drawdown percentage is 19.56254607203325. There were 15 trades, with a win percentage of 46.666666666666664. The mean percentage is 2.320655535863315, the average win percentage is 11.991223743047435, and the average loss percentage is 6.14109164542279. The half Kelly percentage is 0.09676475002014875. There were 2 targets, 5 stops, and 8 time exits, with 0 ambiguous results. The block mean confidence interval percentage is between -2.497770448188134 and 7.632229134682402. This data is from a system called vectorbt longonly, which is a type of trading strategy. The system allows for two accounting points per day, which means it can enter and exit a trade on the same day. The original open and close prices are used, except for the exit price. There are fees of 10 basis points per side, plus the listed slippage. The daily touch gap model is not based on 15-minute venue data, and there are zero ambiguous bars. No parameters were selected from the holdout period. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the headline 'CSTM R91104 costed L20 train slippage0.002'. The trading results for the symbol CSTM are as follows: the lookback period was 20, the period type was train, and the slippage, which is the difference between expected and actual trade prices, was 0.002. The return percentage was -42.18650544451989, the drawdown percentage, which is the maximum loss from a peak to a trough, was 48.266373275999406, and there were 30 trades. The win percentage was 40, the mean percentage was -1.3722286173511689, the average win percentage was 8.278822088129175, and the average loss percentage was 7.806262421004731. The half Kelly criterion, a formula used to determine the optimal bet size, was 0. There were 2 targets, 15 stops, and 13 time exits. The ambiguous count, which refers to unclear trade signals, was 0. The block mean confidence interval percentage was between -5.556153132602278 and 2.8047947889460585. The trading strategy used was a long-only strategy, meaning only buy trades were made, and was based on a vectorbt model. The strategy allowed for two accounting points per day, permitting entry and exit on the same day, using original open and close marks except for the exit price. Fees were 10 basis points per side, plus the listed slippage. The daily touch or gap model did not use venue 15-minute data, and there were zero ambiguous bars. No parameters were selected from the holdout set, which is a portion of the data reserved for testing the model. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the headline 'CSTM R91104 costed L20 train slippage0.001'. The given data line shows results for the symbol CSTM, with a lookback period of 20 days, and a training period. The slippage, which is the difference between expected and actual trade prices, was 0.001. The return percentage was -38.61151, and the drawdown percentage, which is the peak-to-trough decline, was 47.11562. There were 30 trades, with a win percentage of 40. The mean percentage return was -1.17478, the average win percentage was 8.49560, and the average loss percentage was 7.62169. The half Kelly criterion, a measure of bet size, was 0. There were 2 targets and 15 stops, with 13 time exits and no ambiguous exits. The block mean confidence interval percentage was between -5.36708 and 3.01061. This data is from a vectorbt long-only strategy, with two accounting points per day, allowing for same-day entry and exit. Fees were 10 basis points per side, plus the listed slippage. The daily touch or gap model did not use venue 15-minute data and had zero ambiguous bars. No parameters were selected from the holdout set. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the data line: Calculated valuation of CENX, with market capitalization proxy of 4391.3425662 million dollars, and other financial metrics including free cash flow and EBITDA growth rates. The data includes Sebree power and premium breakeven calculations, and valuations are based on June share count and September 10 price. This is a calculated scenario, not an observed future outcome, and was derived using quantify.py with source relationships in analysis 2487 and observations 24090-96 and 24138-44. https://centuryaluminum.com/insights/century-aluminum-company-reports-second-quarter-2026-results/

A news scan found the following data: three calculated scenarios for CENX R91104 were found, which are as follows: the first scenario has a premium change per ton of 500, US capacity of 450,000 tons per year, Sebree power change of 10 megawatt-hours, annual US gross premium of 225 million dollars, annual Sebree cost of 33.726 million dollars, annual amount before other offsets of 191.274 million dollars, and quarter amount before other offsets of 47.8185 million dollars. The second scenario has a premium change per ton of 500, US capacity of 450,000 tons per year, Sebree power change of 25 megawatt-hours, annual US gross premium of 225 million dollars, annual Sebree cost of 84.315 million dollars, annual amount before other offsets of 140.685 million dollars, and quarter amount before other offsets of 35.17125 million dollars. The third scenario has a premium change per ton of 500, US capacity of 450,000 tons per year, Sebree power change of 50 megawatt-hours, annual US gross premium of 225 million dollars, annual Sebree cost of 168.63 million dollars, annual amount before other offsets of 56.37 million dollars, and quarter amount before other offsets of 14.0925 million dollars. These scenarios were calculated in quantify.py and are derived, not observed future outcomes. The full source relationships can be found in analysis2487 and observations24090–96/24138–44, with assumptions including constant cell speed and equal-cell cast output as approximations, and valuations using June share count and September 10 price. https://www.sec.gov/Archives/edgar/data/949157/000162828026013788/cenx-20251231.htm

A calculation for EGA R91104 was made using a computer program called quantify.py. The results include the following information: the average number of cells per day is 5.5, there are 947 cells remaining, and the constant rate of cells will last for 172.18 days, ending on February 21, 2027. The restart fraction, which is a measure of how much the process needs to be restarted, is 0.25. Other calculations include the total missing cast proxy, which is 1,200,633.91, and the daily missing cast proxy, which is 3,289.41. The first half, or H1, of the year showed a decline in cast of 29.15 percent, a decline in sales of 31.61 percent, an increase in EBITDA, which is a measure of a company's overall financial health, of 10.85 percent, and a decline in operating cash flow, or OCF, of 56.87 percent. The cash EBITDA percentage for H1 is 32.93 percent, which is lower than the prior cash EBITDA percentage of 84.65 percent. This calculation is based on a derived scenario and does not predict the future. It uses approximations, including constant cell speed and equal-cell cast output, and is based on the number of shares in June and the price in September 10. https://media.ega.ae/egas-al-taweelah-smelter-restoration-reaches-25-completion-milestone/

A news scan found the following correlation matrix for aluminum chain return: {"CSTM":{"CSTM":1,"KALU":0.6548568887046872,"SPY":0.5642652476729697,"AA":0.5697548883275182,"CENX":0.44841818446695786},"KALU":{"CSTM":0.6548568887046872,"KALU":1,"SPY":0.49348848979664267,"AA":0.5981865775940295,"CENX":0.5139591613267673},"SPY":{"CSTM":0.5642652476729697,"KALU":0.49348848979664267,"SPY":1,"AA":0.5114980742637169,"CENX":0.4318788291021344},"AA":{"CSTM":0.5697548883275182,"KALU":0.5981865775940295,"SPY":0.5114980742637169,"AA":1,"CENX":0.6734326235370292},"CENX":{"CSTM":0.44841818446695786,"KALU":0.5139591613267673,"SPY":0.4318788291021344,"AA":0.6734326235370292,"CENX":1}}. This data covers the period from January 2025 to September 10, 2026, and shows the daily log return pairwise correlations. The data indicates that similar exposures do not make independent bets. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the following data line: a descriptive regime Ordinary Least Squares, or OLS, post analysis for the stock symbol KALU. The analysis has a period of post, a symbol of KALU, and the following results: 114 observations, a beta of 1.8415900782031736, a daily intercept percentage of 0.00005044157653531171, a naive standard error percentage of 0.307606594652364, and a correlation with SPY of 0.4313017343123177. The return percentage is 39.527919675885165. This analysis is based on a daily log return OLS versus SPY with intercept, and it covers two time periods: before January 2025 to March 27, 2026, and after March 30 to September 10, 2026. The naive standard error is not HAC, and the selected dates do not identify a causal outage effect. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the following data line: a descriptive regime Ordinary Least Squares, or OLS, post analysis for the symbol CSTM, also known as a statistical method to measure the relationship between variables, from March 30, 2026, to September 10, 2026, showed the following results: the number of observations was 114, the beta, which measures volatility, was 1.96883718031124, the daily intercept percentage, which is a measure of the expected return, was -0.2191140061847001, the naive standard error percentage, a simple measure of the amount of variation, was 0.23989618574138563, and the correlation with SPY, which measures how closely the two move together, was 0.5481217747741938, resulting in a return percentage of 11.210954214805302. This analysis used daily log returns, which are a way to measure the change in price, and compared them to the SPY, a common benchmark, with an intercept, which allows for a non-zero starting point. The analysis covered two periods: before January 2025 to March 27, 2026, and after March 30 to September 10, 2026. The naive standard error was used, which does not account for heteroskedasticity, or varying volatility, and the selected dates do not identify a causal outage effect, which would be a change in the relationship between the variables caused by an external event. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the following data line: a post about CENX R91104 descriptive regime OLS. The data includes the symbol CENX, with 114 data points, a beta of 1.3605359267741461, a daily intercept percentage of -0.31465124018953033, a naive standard error percentage of 0.3566180531592852, and a correlation with SPY of 0.2914156182779094, with a return percentage of -10.654582074521642. This data is from an ordinary least squares, or OLS, analysis of daily log returns for CENX compared to SPY, with the data split into two periods: before January 2025 to March 27, 2026, and after March 30 to September 10, 2026. The naive standard error, which is a simple measure of variability, does not use a method called HAC, and the selected dates do not show a causal outage effect. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the headline is not available, but a data line was found with information about the stock symbol AA. The data shows that from January 2025 to March 27, 2026, and then from March 30 to September 10, 2026, the daily log return of AA was analyzed using a statistical method called Ordinary Least Squares, or OLS, which is a way to measure the relationship between variables. The results include a beta of 1.4443900120404538, which measures the stock's volatility compared to the overall market, a daily intercept percentage of -0.3926844995431232, and a naive standard error percentage of 0.31907611651154594. The correlation with the SPY stock, which represents the overall market, is 0.33993973809022987. The return percentage for this period was -17.01030927835055. The naive standard error is not adjusted for Heteroskedasticity and Autocorrelation, or HAC, which are types of statistical biases. The selected dates do not identify a causal outage effect, which means the results do not show a clear cause-and-effect relationship between the variables. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the following data line: a descriptive regime for KALU, symbol KALU, with the following details: the period is pre, the number of observations is 309, the beta is 1.2686928813307001, the daily intercept percentage is 0.13476904928576422, the naive standard error percentage is 0.1310381674248496, the correlation with SPY is 0.5373014724498033, and the return percentage is 70.6855081178593. This data is from a daily log return ordinary least squares analysis versus SPY with intercept, covering the pre period from January 2025 to March 27, 2026, and the post period from March 30 to September 10, 2026. The naive standard error is not heteroskedasticity and autocorrelation consistent, and the selected dates do not identify a causal outage effect. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the following data line: a descriptive regime for CSTM R91104 has the following details. The period is pre, the symbol is CSTM, and the number of observations is 309. The beta, which is a measure of volatility, is 1.5839975825116857. The daily intercept percentage is 0.2183268627519021, and the naive standard error percentage, which is a simple measure of error, is 0.14825484582847193. The correlation with SPY, which is a stock market index, is 0.5751269762775941. The return percentage is 127.55598831548194. This data is from a daily log return ordinary least squares analysis, which is a statistical method, versus SPY with an intercept, for the pre period from January 2025 to March 27, 2026, and the post period from March 30 to September 10, 2026. The naive standard error is not heteroskedasticity and autocorrelation consistent, which is a statistical method, and the selected dates do not identify a causal outage effect. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the following data line: a daily log return Ordinary Least Squares, or OLS, analysis versus the SPY stock with an intercept was performed on the CENX stock. The analysis was done for two periods: before January 2025 to March 27, 2026, and after March 30 to September 10, 2026. The results showed a beta of 1.7836278015601468, a daily intercept percentage of 0.27064107894636746, a naive standard error percentage of 0.21979844314978855, and a correlation with SPY of 0.4710199545360905. The return percentage was 172.5027442371021. Note that the naive standard error is not HAC, or Heteroskedasticity and Autocorrelation Consistent, and the selected dates do not identify a causal outage effect. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the following data line: a daily log return Ordinary Least Squares, or OLS, analysis was done for the symbol AA, which is a stock ticker, versus SPY, which is another stock ticker, with an intercept, meaning a starting point. The analysis was done for two periods: before January 2025 to March 27, 2026, and after March 30 to September 10, 2026. The results showed a beta of 1.6885079840032762, a daily intercept percentage of 0.09471344108371367, a naive standard error percentage of 0.16017633528144282, and a correlation with SPY of 0.5699510209273085. The return percentage was 56.83104284559421. Note that the naive standard error is not HAC, which stands for Heteroskedasticity and Autocorrelation Consistent, and the selected dates do not identify a causal outage effect. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A daily technical snapshot for CENX was completed on September 10, 2026, at 04:00:00 UTC. The closing price was 44.36. The prior 20-day high was 48.17 and the 50-day simple moving average, which is the average price over 50 days, was 45.6681. The average daily range over 20 days was 2.333405 and the realized volatility over 20 days was 42.92765569457421. The return over 20 days was -7.90948723271746 percent. The signal, which is a trigger to buy or sell, is false. The volume was 2,293,537. This data was collected on September 11 at 04:00 UTC and covers daily data from January 2020 to September 10, 2026. The signal to buy is when the closing price is higher than the prior 20-day high and the 50-day simple moving average. There is no current executable quote, meaning there is no current price to buy or sell, and the daily adjustment does not include cash dividend execution, which is when a company pays a portion of its profit to its shareholders. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A daily technical snapshot for the stock symbol AA was completed on September 10, 2026, at 04:00:00 UTC. The stock closed at 48.3. The prior 20-day high was 51.85 and the 50-day simple moving average, which is the average stock price over 50 days, was 48.51. The average daily range over 20 days was 1.97 and the realized volatility over 20 days was 45.88. The return over 20 days was -5.70. The signal, which is a trigger to buy or sell, is false, meaning no action is taken. The volume, which is the number of shares traded, was 3,863,221. This data was collected on September 11 at 04:00 UTC. The data is based on daily prices from January 2020 to September 10, 2026, and includes adjustments for splits and dividends. The signal to buy is when the stock's closing price is higher than the prior 20-day high and the 50-day simple moving average. There is no current executable quote, and the daily adjustment does not include cash dividend execution. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A news scan found the headline 'SPY R91104 completed daily technical snapshot'. The SPY fund had the following numbers on September 10, 2026, at 04:00:00 UTC: it closed at 757.83, its price 20 days before was 777.88, and its 50-day simple moving average, which is the average price over 50 days, was 758.2458. The average daily range, which is the difference between the high and low prices, over 20 days was 4.364559999999995, and the realized volatility, which is a measure of how much the price changed, over 20 days was 8.532307950376643. The return, which is the percentage change in price, over 20 days was -1.8977591942937688. The signal, which means a decision to buy or sell, is false, meaning no action is taken. The volume, which is the number of shares traded, was 42761756. This data was collected on September 11 at 04:00:00 UTC and covers the period from January 2020 to September 10, 2026. The signal to buy or sell is based on the close price being higher than the prior 20-day high price and the 50-day simple moving average. There is no current executable quote, which means there is no current price to buy or sell at, and the daily adjustment is not related to cash dividend execution. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A daily technical snapshot for KALU was completed on September 10, 2026, at 04:00:00 UTC. The closing price was 158.42. The prior 20-day high was 189.49. The simple moving average over 50 days, which is a calculation to find the average price over a period, was 166.195. The average daily range over 20 days was 7.1416, and the realized volatility over 20 days, which measures how much the price changed, was 60.7847. The return over 20 days was -16.3966 percent. The signal, which is a trigger to buy or sell, is false. The volume, which is the number of shares traded, was 223,352. This data was collected on September 11 at 04:00 UTC. The data is from January 2020 to September 10, 2026, and includes daily Systematic Investment Plan adjustments. The signal to buy is when the closing price is higher than the prior 20-day high and the 50-day simple moving average. There is no current executable quote, and the daily adjustment does not include cash dividend execution. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A daily technical snapshot for the symbol CSTM was completed on September 10, 2026, at 04:00:00 UTC. The closing price was 25.99. The prior 20-day high was 29.92 and the 50-day simple moving average, which is the average price over 50 days, was 28.40. The average daily range over 20 days was 0.9782, and the realized volatility over 20 days, which measures price fluctuations, was 39.99. The return over 20 days was -11.90%. The signal, which is a trigger to buy or sell, is false. The volume, which is the number of shares traded, was 1,375,618. This data was collected on September 11 at 04:00 UTC and covers daily data from January 2020 to September 10, 2026. The signal to buy is when the closing price is higher than the prior 20-day high and the 50-day simple moving average, but there is no current executable quote, and the daily adjustment does not include cash dividend execution. https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all

A direct primary public chart API was observed on September 11, showing the latest data for July 2026. The data revealed that in June, the total was 5939 kilotons at a rate of 197.966667 kilotons per day, and in July, the total was 6160 kilotons at a rate of 198.709677 kilotons per day. This represents a monthly increase of 3.721165 percent versus a daily increase of 0.375321 percent. The GCC totals were 276 kilotons in June and 293 kilotons in July, compared to 523 kilotons in July 2025, resulting in a daily increase of 2.734923 percent month-over-month and a decrease of 43.977055 percent year-over-year. The data is based on voluntary producer reports and non-reporter estimates, including China estimates, and only includes molten primary, excluding alloys and recycling. Revisions to the data are possible. The full monthly data from 2025 to July 2026 is stored in the note. It is noted that there is not independent corroboration of every EGA cell, which refers to a type of data cell used for tracking and analysis. https://alvis.international-aluminium.org/api/publication/?publication=primary-aluminium-production

The annual risk factor review for AA R91104 found the following: in the fiscal year 2025, there were concerns about long-term contracts for energy and raw materials, which can limit the ability to adjust capacity. Additionally, non-market capacity policies, substitution, and competitive global alumina prices can affect prices. It was noted that no single supplier's share was quantified, and it was decided not to create one. https://www.sec.gov/Archives/edgar/data/1675149/000119312526077167/aa-20251231.htm

The report includes information about three locations: Sebree, MtHolly, and Grundartangi. Sebree has a capacity of 220kt and produces 3,372,600 MWh per year. MtHolly has a capacity of 230kt and produces 3,504,000 MWh per year. Grundartangi has a capacity of 320kt and produces 4,756,680 MWh per year. The total usage is 11,633,280 MWh per year. There is a discrepancy in the costs sensitivity, with a printed total of 15.9 million USD per 1 million MWh, which does not match the listed total of 3.4 + 3.5 + 4.8 = 11.7 million USD per 1 million MWh, and actual usage of 11.63328 million. The calculations should use the individual row for Sebree, and the full capacity, not the actual production. Additionally, the report mentions that Glencore has an alumina production of 500kt per year through 2028, linked to the London Metal Exchange, and Concord has a production of 540kt per year through 2029, with a mixed formula. https://www.sec.gov/Archives/edgar/data/949157/000162828026013788/cenx-20251231.htm