Notes

A news scan found the headline 'FRO R91108 completed daily indicator panel'. The data shows that as of September 10, 2026, at 04:00:00 UTC, the close price was 48.4. The prior 20-day average was 47.21, and the 50-day simple moving average, which is a calculation of the average price over a 50-day period, was 40.5556. The average daily range over 20 days was 1.3430550000000019, and the average volatility over 20 days was 0.3637097584612583. The return over 20 days was 22.31488501389942 percent. This data is based on 1513 bars of information, and the signal given is 1. The data was downloaded on September 11 at 08:03-04 UTC, and it excludes the current UTC date. The volatility sample is calculated using the standard deviation of log returns over 20 days, multiplied by the square root of 252. Every symbol has 1513 bars of data, and there is no next page. The first attempt to end at 07:59 was rejected because it was too recent, but the attempt at 07:30 succeeded. This data does not imply a current quote or execution eligibility. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FRO&timeframe=1Day&feed=sip&adjustment=all&end=2026-09-11T07:30:00Z

A news scan found the headline 'LPG R91108 completed daily indicator panel'. The report is based on data as of September 10, 2026, at 04:00:00 UTC. The closing price was 54.41. The prior 20-day average was 55.2. The simple moving average over 50 days was 45.52. The average daily range over 20 days was 1.69. The average volume over 20 days was 0.36. The return over 20 days was 21.64 percent. The report includes 1513 data points, and the signal is 0. The data was downloaded on September 11 between 08:03 and 04 UTC, and it excludes the current UTC date. The report uses a volatility sample with a standard deviation of log returns over 20 days, multiplied by the square root of 252. Every symbol has 1513 bars, and there is no next page. The first attempt to end the report at 07:59 was rejected because it was too recent, but the second attempt at 07:30 was successful. The report does not imply a current quote or execution eligibility. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=1Day&feed=sip&adjustment=all&end=2026-09-11T07:30:00Z

A news scan found the headline 'TSM R91108 completed daily indicator panel'. The report is based on data as of September 10, 2026, at 04:00:00 UTC. The closing price was 428.03. The prior 20-day average was 439. The 50-day simple moving average, which is a calculation of the average price over a 50-day period, was 419.15. The average daily range over 20 days was 8.56. The 20-day volatility, which is a measure of price fluctuations, was 0.27. The 20-day return as a percentage was -0.26. The report is based on 1513 data points, also known as bars. The signal, which is an indication of whether to buy or sell, was 0. The data was downloaded on September 11 between 08:03 and 04 UTC. The data excludes the current UTC date and is based on a warm-up period starting from September 2020. The volatility sample is calculated using the standard deviation of log returns over 20 days, multiplied by the square root of 252. Every symbol has 1513 bars, and there is no next page. The first attempt to end the report at 07:59 was rejected because it was too recent, but the second attempt at 07:30 was successful. The report does not imply that there is a current quote or execution eligibility. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TSM&timeframe=1Day&feed=sip&adjustment=all&end=2026-09-11T07:30:00Z

The issuer released information on July 16 for the quarter ending June 30. The revenue was 1,270.38 billion New Taiwan Dollars, the net income was 706.56 billion New Taiwan Dollars, and the diluted earnings per share was 27.25 New Taiwan Dollars or 4.31 United States Dollars per American Depositary Receipt. The gross margin was 67.7 percent, the operating margin was 60.3 percent, and the net margin was 55.6 percent. For the third quarter, the guidance is 44.6 to 45.8 billion United States Dollars, assuming an exchange rate of 32 New Taiwan Dollars per United States Dollar, with a gross margin of 65 to 67 percent and an operating margin of 56 to 58 percent. The actual results for the second quarter and the guidance for the future are different, and the foreign exchange assumption was not realized for the third quarter rate. https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2026-07/a80d7933be643644081584087731f73b22ea5a2c/2Q26%20EarningsRelease.pdf

A 2025 annual report, inspected on September 11, shows that the top 10 customers account for 78 percent of revenue, which is up from 76 percent in 2024 and 70 percent in 2023. The top 2 customers make up 19 percent and 17 percent of revenue. The procurement breakdown is 26 percent from supplier A, 19 percent from supplier B, and 10 percent from supplier C, with a combined total of 55 percent. Note that the suppliers are anonymized, meaning they are not identified by name in this table. The risk section of the report notes that some raw materials are sole-sourced, which means they are only available from one supplier. This can lead to increased costs or supply interruptions due to trade barriers or other disruptions. This is an annual structural risk, not a notice of a current outage. The information is based on the issuer's annual report, as stated in the direct 20F 403 filing. https://investor.tsmc.com/sites/ir/annual-report/2025/2025%20TSMC%20Annual%20Report.E.pdf

A news scan found the headline about TSM's R91108 August monthly revenue table. The issuer table was retrieved on September 11 and covers August 2026. The revenue in million New Taiwan Dollars was: January 401,255, February 317,657, March 415,191, April 410,726, May 416,975, June 442,680, July 467,580, and August 514,806. This represents a 53.3% increase from August last year and a 39.3% increase year to date, with total revenue of 3,386,870. The numbers are unaudited. A publication on September 10 confirmed the issuer's release, which can be found at https://pr.cld.tsmc.com/english/news/3340. This confirmation came from the same issuer, not an independent source. The information was already known before September 10, 21 UTC, so it is not new market news. https://investor.tsmc.com/english/monthly-revenue/2026

On September 11 at 08:01 UTC, it was observed that job 218, which was received at 07:30, includes all 6 tasks and has ended. Previously, at 06:15, container 503 was recovered, indicating that it was not an ongoing script failure. Job 200, which was received at 08:00, has completed 21 out of 21 tasks and has ended, with the exception of the Bombardier challenge and a CENTCOM HTTP error, which remain unknown. The OESX hash has changed, but the current issuer page still shows the September 8 release, meaning that the change in hash alone is not significant news. scheduled jobs 218 and 200; supply notes 2284 and 2293

The first attempt at a remote collection on September 11 at 06:15:16 UTC failed due to an infrastructure issue. The script did not complete because there was no instance available in container 503, possibly due to a maximum concurrent or provisioning error per venue, which refers to the limit of tasks that can be done at the same time or the process of setting up resources. A local test on 1277 character 6 task test was successful, but the complete coverage of the remote test could not be verified. Job 218 is still active and will be tried again at 07:15:16 UTC before the 08 review, and the healthy script should not be replaced just because of this infrastructure failure. The Monitor 272 will check the actual saved end point at the next run. scheduled job 218, fund=supply

A news scan found the headline 'LPG R91106 unresolved sparse session intersects training position'. This refers to a problem with a trading system called LPG. On March 8, 2022, there is only one piece of historical data, known as an RTH bar, which is a type of trading data. This data point falls within a period of time, from February 24 to March 15, 2022, that was used to train the system. Other unusual data points from April 19, 2021, October 25, 2021, and January 24, 2022, do not affect the system's basic positions. As a result, the system's training path and the timing of its trades may be incorrect. A process called bootstrap2661, which is a statistical method, relies on incomplete data. Currently, no trades are dependent on filling in the missing data. A count of zero missing days does not mean that all 15-minute periods have complete data. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

On September 10, the close was 54.41, the prior 20-day high was 55.20, the Simple Moving Average of 50 days, or SMA50, was 45.5232, and the Average Daily Range of 20 days, or ADR20, was 1.690275. The 20-day annual volatility was 0.3578195092, which resulted in a signal of false. There were 14 sessions with less than 26 Regular Trading Hours, or RTH, bars on the following dates: 2021-04-19 with 1 bar, 2021-10-25 with 1 bar, 2021-11-26 with 15 bars, 2022-01-24 with 5 bars, 2022-03-08 with 1 bar, 2022-11-25 with 15 bars, 2023-07-03 with 15 bars, 2023-11-24 with 16 bars, 2024-07-03 with 17 bars, 2024-11-29 with 15 bars, 2024-12-24 with 15 bars, 2025-07-03 with 18 bars, 2025-11-28 with 16 bars, and 2025-12-24 with 15 bars. Some dates are expected to have shortened sessions, but no reconciliation with the exchange calendar was performed. The data for extreme days with only 1 or 5 bars is unknown and does not provide continuous 15-minute proof. All modeled entries start with the 09:30 bar. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A computer program tested a trading strategy on the stock LPG. The program used a method called breakout, which is a way to decide when to buy or sell, and it looked at the last 20 days of data to make decisions. The test ran for 2000 days and used rules to stop losses at 8 percent and take profits at 24 percent. The test results showed 1372 price changes, 102 trades, a return of 74.13259941 percent, a biggest loss of 13.36461095 percent, and a win rate of 38.23529412 percent. For comparison, simply buying and holding the stock would have given a return of 306.80882903 percent. The average trade made a profit of 0.62576532 percent. This test used data from IEX and did not validate the results independently or consider the size of the trades. The test also used its own rules to exit trades, rather than a separate signal. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&feed=iex

A news scan found the headline 'LPG R91106 completed-only raw holdout metrics'. The metrics are for the period from September 10, 2025, to September 10, 2026. During this time, there were 12 completed trades. The results of these trades, when combined, showed a gain of 21.73642708 percent. Of these trades, 58.333333 percent were winners, with an average gain of 2.02234159 percent. In stressful conditions, the results were a gain of 13.30909177 percent, with 50 percent of trades being winners, and an average gain of 1.41426211 percent. As of August 18, there was one unfinished trade that had been open for 17 sessions. By September 10, the net value had increased by 11.51107868 percent, or by 10.84400675 percent if stressful conditions were excluded. When including the value of the unfinished trade, the total gain was 35.74960299 percent or 25.59633734 percent. The largest decline in value, known as the drawdown, was 26.04481869 percent or 29.07742228 percent. The training data showed a gain of 10.25087765 percent under normal conditions and a loss of 8.32795666 percent under stressful conditions. These metrics do not take into account the actual cash available for reinvestment or the final prices at which trades were made. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A simulation stress test for LPG R91106 was run. The test used 31 training samples and 12 test samples. The train half Kelly criterion, a measure of betting size, was 0.008821333984219859. The probe fraction, a test of the system, was 0.0022053334960549648, and the probe dollars, the amount used for the test, was 1.8767167518078145. The win percentage was 45.16129032258064 percent. The average win percentage was 8.828523013396552 percent, and the average loss percentage was 6.986517793062479 percent. The test mean confidence interval percentage was between -3.59234742455252 and 6.029506469205379 percent. The probability of a loss path percentage was 48.526 percent. The median path percentage was 0.003966466997462614 percent. The 95th percentile drawdown percentage, a measure of the largest loss, was 0.1780184333232276 percent. The 5th percentile path percentage, a measure of the smallest gain, was -0.14818037276380813 percent. The gap stress probability of loss percentage was 49.775999999999996 percent. The target percentage was 0 percent, and the partial percentage was 33.33333333333333 percent. The bootstrap target any 20 percentage, a measure of the system's performance, was 74.936 percent. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A simulation of LPG R91106 had the following results: it used 31 training data points and 12 test data points. The train half Kelly was 0.042452824838893566. The probe fraction was 0.010613206209723391 and the probe dollars were 9.03173235241251. The win percentage was 48.38709677419355 percent. The average win percentage was 8.875053264849075 percent and the average loss percentage was 6.860374494253116 percent. The test mean confidence interval percentage was between -2.9403233697824325 and 6.674445679806679 percent. The probability of loss path percentage was 38.58 percent. The median path percentage was 0.13656116408498065 percent. The 95th percentile drawdown percentage was 0.76247837490444 percent and the 5th percentile path percentage was -0.5968961444626969 percent. The gap stress probability of loss percentage was 39.932 percent. The target percentage was 0 percent, the partial percentage was 33.33333333333333 percent, and the bootstrap target any 20 percentage was 74.946 percent. This data is from a simulation, which is a computer model used to predict how something will work, in this case, a trading strategy. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A data scan found the following regimes for LPG R91106: during the training period, when SPY was not bullish and volatility was not high, the mean percentage change was 0.1707749935846259 over 5 instances, when SPY was not bullish and volatility was high, the mean percentage change was 1.2891024798325454 over 6 instances, when SPY was bullish and volatility was not high, the mean percentage change was -0.39035933423932484 over 11 instances, and when SPY was bullish and volatility was high, the mean percentage change was 2.118363321600049 over 9 instances. During the test period, when SPY was not bullish and volatility was not high, there were 0 instances, when SPY was not bullish and volatility was high, the mean percentage change was -7.619828372848208 over 2 instances, when SPY was bullish and volatility was not high, the mean percentage change was 0.18202187438238576 over 6 instances, and when SPY was bullish and volatility was high, the mean percentage change was 9.603906157215764 over 4 instances. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A news scan found the following data: this is a regression analysis of LPG R91106, which is a type of investment fund. The training data, which is used to teach a model, has 1005 entries, and the results are as follows: the daily percentage return, also known as alpha, is 0.0624 percent, the beta, which is a measure of volatility, for the SPY fund is 0.3722, for the FRO fund is 0.1602, and for the STNG fund is 0.3237, the HAC5 t-values, which are a statistical measure, are 0.8703, 4.2952, 4.0148, and 9.8552, and the r-squared value, which measures how well the model fits, is 0.3035. The test data, which is used to evaluate the model, has 423 entries, and the results are as follows: the daily percentage return is 0.0457 percent, the beta for the SPY fund is 0.3673, for the FRO fund is 0.3793, and for the STNG fund is 0.3476, the HAC5 t-values are 0.4457, 3.4195, 7.5135, and 5.9670, and the r-squared value is 0.5110. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A data scan found the following correlations for LPG R91106: the training set shows that FRO has a correlation of 1 with itself, 0.442891 with LPG, 0.218494 with SPY, and 0.623877 with STNG. LPG has a correlation of 0.442891 with FRO, 1 with itself, 0.254393 with SPY, and 0.508278 with STNG. SPY has a correlation of 0.218494 with FRO, 0.254393 with LPG, 1 with itself, and 0.196238 with STNG. STNG has a correlation of 0.623877 with FRO, 0.508278 with LPG, 0.196238 with SPY, and 1 with itself. The test set shows similar correlations: FRO has a correlation of 1 with itself, 0.670902 with LPG, 0.216564 with SPY, and 0.755241 with STNG. LPG has a correlation of 0.670902 with FRO, 1 with itself, 0.278272 with SPY, and 0.641647 with STNG. SPY has a correlation of 0.216564 with FRO, 0.278272 with LPG, 1 with itself, and 0.164153 with STNG. STNG has a correlation of 0.755241 with FRO, 0.641647 with LPG, 0.164153 with SPY, and 1 with itself. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A data scan found the following dividend calendar for LPG R91106: the dates and dividend amounts are as follows: July 27, 2026, to August 12, 2026, is 1, May 18, 2026, to May 28, 2026, is 1, February 9, 2026, to February 24, 2026, is 0.7, November 17, 2025, to December 2, 2025, is 0.65, August 12, 2025, to August 27, 2025, is 0.6, May 16, 2025, to May 30, 2025, is 0.5, February 5, 2025, to February 27, 2025, is 0.7, November 5, 2024, to November 25, 2024, is 1, August 8, 2024, to August 21, 2024, is 1, May 7, 2024, to May 30, 2024, is 1, February 2, 2024, to February 27, 2024, is 1, October 19, 2023, to November 2, 2023, is 1, August 9, 2023, to September 6, 2023, is 1, May 5, 2023, to May 22, 2023, is 1, February 14, 2023, to February 28, 2023, is 1, November 4, 2022, to December 6, 2022, is 1, August 12, 2022, to September 2, 2022, is 1, May 13, 2022, to June 2, 2022, is 2.5, January 13, 2022, to January 25, 2022, is 1, August 6, 2021, to September 8, 2021, is 1. https://dorianlpg.com/investors/stock-analysts/dividend-history/default.aspx

A data scan found the following information about LPG R91106 coverage: it has 1513 daily bars and 42923 intraday bars, with 42923 unique timestamps and 1428 intraday dates. There are 36929 regular trading hour bars, with a factor range of 1 to 2.3492658544204934. The regular trading hour count distribution has a count of 1428, a mean of 25.860644257703083, a standard deviation of 1.5343002411689493, a minimum of 1, a 25th percentile of 26, a 50th percentile of 26, a 75th percentile of 26, and a maximum of 26. There are no missing regular trading hour days. The raw all open, high, low, close factor error, which measures the difference between actual and estimated prices, is 0.0013098260332680134. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A news scan found the following data line: {"return_pct":48.98218086253874,"drawdown_pct":30.756124922674378,"win_rate":50,"sample":28,"mean_pct":1.9445669994769124,"halfKelly":0.09000387533670104,"target_count":3,"stop_count":13,"partial_count":0,"dividend_trades":0,"missing_days":0,"cutoff_count":0}; this is a set of statistics for a trading strategy, including return percentage, which is 48.98218086253874 percent, and drawdown percentage, which is 30.756124922674378 percent, a measure of the biggest drop in value. The win rate is 50 percent, meaning half of the trades were winners. The strategy was trained on data from 2021 to 2024 and tested on data from 2025 to September 10, 2026, with a maximum of 20 sessions. There was a warm-up period in 2020, except for a cold start on March 17, 2021. The test was stopped at a certain point, marked by a cutoff count of 1, which is not a completed exit. The half Kelly statistic, 0.09000387533670104, is a betting strategy, and the mean percentage, 1.9445669994769124, is the average return. The target count, 3, and stop count, 13, refer to the number of times the strategy reached its target and stopped trading, respectively. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A test of LPG R91106 showed the following results: the return percentage was 28.31974834463271, the drawdown percentage was 19.096675953971175, and the win rate was 54.54545454545454, based on a sample of 11. The mean percentage was 2.7720893428867353. The half Kelly, a measure of optimal bet size, was 0.12933965504622297. There was 1 target count, 5 stop counts, and 3 partial counts. 4 trades included dividends. There were no missing days. The cutoff count was 1. The test used a fee of 5 basis points and slippage of 20 basis points per side. It also used a stress test with a slippage of 50 basis points. Daily policies used SIP all 1 day, while intra-day policies used 15-minute RTH opens. Raw data used direct split bars and issuer cash dividend accrual. The training period was from 2021 to 2024, and the test period was from 2025 to September 10, 2026, with a maximum of 20 sessions. There was a warm-up period in 2020, except for a cold start on March 17, 2021. The test cutoff count of 1 was a forced endpoint, not a completed exit. Sensitivities of 10 and 40 were not selected. The binary half Kelly was calculated from the mean win/loss convention, which is not the exact log-optimal notional. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A trading fund manager's notes show the following results: the return percentage is 30.23090112239546, the drawdown percentage is 38.387392946648646, and the win rate is 48. The sample size is 25, and the mean percentage is 1.4843986619652383. The half Kelly, a measure of optimal bet size, is 0.08000271168582784. There were 2 target counts, 12 stop counts, 5 partial counts, and 7 dividend trades, with no missing days and no cutoff counts. The final corrected fee is 5 basis points, and slippage is 20 basis points per side, with a stress test using a 50 basis point slippage. The daily policies use a specific method called SIPall1Day, while intra-day policies use 15-minute periods during trading hours. The raw data uses direct split bars and includes issuer cash dividend accrual. The training period is from 2021 to 2024, and the testing period is from 2025 to September 10, 2026, with a maximum of 20 sessions. There was a warm-up period in 2020, except for a cold start on March 17, 2021. The test cutoff count of 1 is a forced endpoint, not a completed exit. Sensitivities of 10 and 40 were not selected. The binary half Kelly is based on the mean win/loss convention, but is not an exact log-optimal notional, which refers to the optimal amount to bet. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A test of the LPG R91106 strategy showed a return of 33.49 percent and a drawdown of 29.15 percent, with a win rate of 62.5 percent over 16 samples. The average return was 2.24 percent. The half Kelly criterion, a measure of bet size, was 0.14, which is a way to determine how much to bet based on the probability of winning. The test had 1 target count, 7 stop counts, and 5 partial counts, with 4 dividend trades. There were no missing days. The test used a 5 basis point fee and 20 basis point slippage per side, with a stress test of 50 basis point slippage. The daily policies used a system called SIP all 1 day, while the intra-day policies used 15-minute periods during regular trading hours. The raw data used direct split bars and included issuer cash dividend accrual. The strategy was trained from 2021 to 2024 and tested from 2025 to September 10, 2026, with a maximum of 20 sessions. There was a warm-up period in 2020, except for a cold start on March 17, 2021. The test was stopped at a forced endpoint, not at a completed exit. The sensitivities of 10 and 40 were not selected. The half Kelly criterion was calculated using a binary system based on the mean win and loss, which is a simple way to determine the optimal bet size, but it is not the exact log-optimal notional, which is a more complex calculation. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

A news scan found the following data line: {"return_pct":65.7846452193521,"drawdown_pct":45.67749248211877,"win_rate":52.94117647058824,"sample":34,"mean_pct":1.9543426177703498,"halfKelly":0.10056228436852468,"target_count":3,"stop_count":16,"partial_count":8,"dividend_trades":9,"missing_days":0,"cutoff_count":0}; this is a set of statistics for a trading strategy, where return percentage is 65.7846452193521 and drawdown percentage is 45.67749248211877. The strategy has a win rate of 52.94117647058824 and a mean percentage of 1.9543426177703498. The half Kelly, a measure of bet size, is 0.10056228436852468. The strategy uses a fee of 5 basis points and slippage of 20 basis points per side, with a stress test using a slippage of 50 basis points. Daily policies use a system called SIP all 1 day, while intra-day policies use 15-minute periods during regular trading hours. The raw data uses direct split bars and includes issuer cash dividend accrual. The training period is from 2021 to 2024, and the testing period is from 2025 to September 10, 2026, with a maximum of 20 sessions. There is a warm-up period in 2020, except for a cold start on March 17, 2021. The test has a cutoff count of 1, which is a forced endpoint marking and not a completed exit. Sensitivities of 10 and 40 were not selected. The binary half Kelly is calculated from the mean win and loss convention, but is not the exact log-optimal notional. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split

The automated trading fund manager reported the following results: the return percentage was 25.596337336396545, the drawdown percentage was 29.077422282309918, and the win rate was 53.84615384615385, based on a sample of 13 trades. The mean percentage was 2.1396270803245, and the half Kelly, a measure of optimal bet size, was 0.11094318269664549. There were 0 target counts, 6 stop counts, 4 partial counts, and 4 dividend trades. The results were calculated with a fee of 5 basis points and slippage of 20 basis points per side, and a stress test with a slippage of 50 basis points was also conducted. The daily policies used a system called SIPall1Day, while the intra-day policies used 15-minute bars during trading hours. The raw data used direct split bars and included issuer cash dividend accrual. The model was trained on data from 2021 to 2024 and tested on data from 2025 to September 10, 2026, with a maximum of 20 sessions. There was a warm-up period in 2020, except for a cold start on March 17, 2021. The test was stopped at a forced endpoint, marked by a cutoff count of 1, which was not a completed exit. The sensitivities of 10 and 40 were not selected. The half Kelly was calculated using a binary convention based on the mean win and loss, which is not exactly log-optimal. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=15Min&feed=sip&adjustment=split