Notes

A diagnostic test was run on the IONS stock using a trading engine. The test, which is part of a training set, showed a 39.175422535202586 percent return. For comparison, simply buying and holding the stock would have given a 102.48272331712309 percent return. The biggest loss during the test was 18.281471774313054 percent. The test made 24 trades, with 37.5 percent of them being winners, and the average trade made a 1.5739425450652018 percent profit. The test used a momentum strategy, which is a way of making trades based on the stock's price movement, and it did not account for any costs or real-world trading issues. The test was run from October 2024 to the end of 2025, and the results from January 2026 to September 8, 2026, were set aside to see how the strategy would work in the future. The test also included a 20-day warm-up period, where the strategy was not actually trading, to get a baseline for comparison. A separate test will be run to see how the strategy works in a more realistic scenario, with costs and other factors included. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A fund called catalyst was analyzed for the symbol RARE. The analysis used 191 data points and found a return of 1.8975920520518441 percent. In comparison, a buy and hold strategy would have resulted in a loss of 59.17695473251029 percent. The maximum drawdown, which is the largest loss from a peak to a trough, was 22.621906250000023 percent. The win rate, or the percentage of successful trades, was 33.33333333333333 percent. The average trade return was 0.3544089355959463 percent. This analysis was based on 15 trades. The same settings were used for the CLI engine and momentum, with a stop loss of 5, a target of 10, and 20 bars for momentum. A review of the source code in the file engine.js showed that the analysis did not account for fees or slippage, and that trades were filled at the stop price even if the price gapped below the stop. The analysis also showed that the closing bar signal was used for entry. However, this analysis is not suitable as trading evidence. The training period was from October 2024 to the end of 2025, and the holdout period was from January 2026 to September 8, 2026, with a 20-bar warm-up period that was flat. The buy and hold strategy included the warm-up period and was not aligned with the comparator. A separate test of costed events was conducted, and there is no plan to transport the live rule. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A diagnostic test was run on the RARE stock using the Catalyst fund's train split, which consists of 314 data points. The results showed a return percentage of -14.57%, compared to a buy and hold return of -59.30%. The maximum drawdown percentage was 23.51%, and the win rate was 23.81%. The average trade percentage was -0.60%. This test used 21 trades. The test was run using the same CLI engine and momentum settings, with a stop loss of 5 and a target of 10. The test assumed no fees or slippage, and fills were made at the stop price even if the price gapped below it. The test used the same signal for closing and entry, but the results are not suitable as trading evidence. The test was run from October 2024 to the end of 2025, with a holdout period from January 2026 to September 8, and a 20-bar warm-up period. The buy and hold results include the warm-up period and are not directly comparable. A separate test of costed events will be run, and there are no plans to transport the live rule. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T10_T2 2026 slip0.003'. This refers to a review of a trading rule, called T10_T2, for the year 2026, with a slippage of 0.003. Slippage means the difference between expected and actual trading prices. The results show a return of 3.264020427779446 percent, a drawdown of 3.1391809797858605 percent, and a win rate of 50 percent. The average win was 9.583820034262192 percent and the average loss was 3.0557791787033217 percent. The mean return was 3.2640204277794354 percent, with a 95 percent confidence interval between -3.0557791787033217 and 9.583820034262192 percent. There were 2 trades that stayed within the rules, and no trades had ambiguous exit days. The initial capital was the same for all trades, and the portfolio value was updated daily, including any idle cash. The data came from 17,320 to 47 event observations. A test using 5 training events showed that 3 TLX events dominated, but the 2 holdout events were not enough for independent validation. A bootstrap test, which is a statistical method, was done 50,000 times with a seed of 20260909, and it only looked at the observed returns, not the uncertainty or extreme values. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T10_T2 2026 slip0.001'. This refers to a review of a trading rule, called T10_T2, for the year 2026, with a very small slippage of 0.001. The results show a return of 3.6779 percent, a drawdown of 3.0448 percent, and a win rate of 50 percent. The average win was 10.023 percent and the average loss was 2.667 percent. The mean return was 3.6779 percent, with a 95 percent confidence interval between -2.667 and 10.023 percent. There were 2 trades that ended inside a certain range and 0 trades with ambiguous exit days. The initial capital was the same for all events, including idle cash, and the portfolio value was marked daily. The source of the data is from event observations from 17320 to 47. The training set had 5 events dominated by 3 TLX events, but the holdout set of 2 events is not enough for independent validation. A bootstrap test with 50,000 samples, started on September 9, 2026, resampled the observed returns, but did not include uncertainty or tail bounds. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T10_T2 2025 slip 0.003'. This refers to a review of a trading rule, called T10_T2, for the year 2025, with a slippage of 0.003. Slippage means the difference between expected and actual trade prices. The results show a return of 6.112171852019843 percent, a drawdown of 2.1126681861037278 percent, and a win rate of 80 percent. The average win was 8.486993783812723 percent and the average loss was 3.387115875151736 percent. The mean return was 6.112171852019832 percent, with a 95 percent confidence interval of 0.3972638733356759 to 12.829195755652885 percent. The data came from 5 training events and 2 holdout events, which may not be enough for independent validation. A bootstrap analysis, which is a statistical method, was done 50,000 times, starting from September 9, 2026, and only looked at observed returns, not uncertainty or extreme values. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T10_T2 2025 slip0.001'. This refers to a review of a trading rule, called T10_T2, for the year 2025, with a very small slippage of 0.001. The results show a return of 6.537472416458501 percent, a drawdown of 2.072133225198769 percent, and a win rate of 80 percent. The average win was 8.921812701239578 percent and the average loss was 2.999888722665911 percent. The mean return was 6.53747241645848 percent, with a 95 percent confidence interval of 0.7996589261197484 to 13.281418340512976 percent. The data included five training events, which dominated three TLX events, but only two holdout events, which is not enough for independent validation. A bootstrap analysis with 50,000 resamples, started with a seed on September 9, 2026, was used to estimate returns, but did not account for uncertainty or extreme values. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T20_T6 2026 slip0.003'. The given data is about a trading rule, T20_T6, for the year 2026, with a slippage of 0.003. The return percentage is -2.2473906718738057, and the drawdown percentage is 4.929601298264319. The win rate percentage is 0, the average win percentage is null, and the average loss percentage is 2.247390671873811. The mean percentage is -2.247390671873811, with a 95% confidence interval of -3.9863904512559833 to -0.5083908924916392. There are 2 trades, with 0 inside and 0 exit day ambiguous. The initial capital is equal, and the portfolio value is marked daily, including idle cash. The source is event observations from 17320 to 47. The training set is dominated by 3 TLX events, but the holdout set of 2 is not adequate for independent validation. A bootstrap test with 50,000 resamples, seeded on September 9, 2026, was run on observed returns only, without considering posterior uncertainty or tail bounds. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T20_T6 2026 slip0.001'. The audit results show a return percentage of -1.8555954687694776 and a drawdown percentage of 4.639660087072195. The win rate percentage is 0, and the average win percentage is null. The average loss percentage is 1.8555954687694831, and the mean percentage is -1.8555954687694831. The confidence interval is between -3.6015652080989558 and -0.10962572944001048. The audit also reports that the initial capital event includes idle cash, and the portfolio value is marked daily. The source of the event observations is from 17320 to 47. The training data dominated 3 TLX events, but the holdout data is not adequate for independent validation. A bootstrap analysis with 50,000 resamples and a seed of 20260909 was performed, but it only observed returns and did not account for posterior uncertainty or tail bounds. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T20_T6 2025 slip0.003'. The audit results for the rule T20_T6 in the year 2025 show a slippage of 0.003. The results are based on 5 events, with a return percentage of -8.873296220957162 and a drawdown percentage of 10.28521390053363. The win rate percentage is 20, with an average win percentage of 0.9600347847707358 and an average loss percentage of 11.331628972389149. The mean percentage is -8.873296220957172, with a 95% confidence interval of -14.18908635999282 to -2.887616822779162. There are 0 events that are inside and 0 events with ambiguous exit days. The results are based on equal initial capital, with daily marked portfolio values and idle cash included. The source is event observations from 17320 to 47. The training set is dominated by 3 TLX events, but the holdout set of 2 events is not adequate for independent validation. A bootstrap analysis with 50,000 resamples and a seed of 20260909 was used to estimate observed returns, but did not account for posterior uncertainty or tail bounds. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

A news scan found the headline 'FDA calendar audit aggregate T20_T6 2025 slip 0.001'. The given data is about a trading rule called T20_T6 for the year 2025, with a slippage of 0.001. The results show a return percentage of -8.508057833350623, a drawdown percentage of 9.998850991777541, and a win rate percentage of 20. The average win percentage is 1.3646854391706098 and the average loss percentage is 10.976243651480925. The mean percentage is -8.508057833350618, with a 95% confidence interval of -13.845153808532212 to -2.4983876640653557. There are 5 training events, with 3 TLX events dominating, but only 2 holdout events, which is not enough for independent validation. A bootstrap analysis with 50,000 resamples, seeded on September 9, 2026, was done, but it only looked at observed returns, not uncertainty or tail bounds. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all

The fund had 3 stock trades that were closed by the operator, which cannot be compared to the usual calendar-based exits. The trades involved the stocks SNOW, which lost 5.39236448%, DELL, which gained 7.93994131%, and GTLB, which lost 5.13049905%. The win percentage was 33.3333%, the average win was 7.93994131%, and the average loss was 5.26143176%, with a mean of -0.86097407%. This is called the Kelly criterion, which is a way to measure the performance of a trading strategy. Because there were too few trades and they were different from usual, it is not possible to determine the probability of winning based on the calendar. There are 0 open trades, 0 pending trades, and 0 closing trades that have been verified independently. Since there are no benchmark quotes available at the same time, it is not possible to estimate the effect of the type of order or the trading session on the results, or how often the stop or target prices were hit. A news scan found the data line, which is a raw output of trading results. query_trades fund=quant all and separate status=open,pending,closing; measured2026-09-09~04:08UTC

A news scan found the headline 'ASPI Explicit arbitrary-prior helium economics sensitivity simulation'. The simulation used 100,000 numpy draws with a seed of 2026090904. The results showed revenue ranges of 11.30 to 27.94, and operating ranges of -17.03 to 2.42. The share of positive operating scenarios was 11.45 percent, and the share of revenue above 27 was 7.46 percent. Correlations were 0.79 for utilization and operating, 0.30 for uncontracted helium price and operating, and -0.53 for cost and operating. The simulation used independent triangular distributions for utilization, uncontracted helium price, LNG price, and annual operating cost. The model did not account for joint endogeneity, gaps, capital expenditures, taxes, interest, or non-controlling interest, and could not justify the Kelly criterion. https://www.sec.gov/Archives/edgar/data/1921865/000119312526352603/aspi-20260630.htm

A hypothetical scenario for RARE was analyzed with the following assumptions: a hazard of 0.1, an other return of 0.015, and a shock loss of 0.45860061919504647. This scenario was run 50,000 times with 10 events. The results showed terminal percentiles of 8.654673967411684, 30.41916489761206, and 57.028976141735804. The loss frequency was 0.64868 and the 95th percentile drawdown was 0.8315707816376913. The expected return per event was -3.236006191950465 percent. This analysis used a NumPy seed of 20260909. The scenario assumed a 1.5 percent win and hazard rates of 0, 1, 5, and 10 percent, which were not based on historical data. A shock loss of 45.8600619 percent was applied to a closed RARE trade. The analysis assumed 49.14 percent of the portfolio was reinvested per hypothetical event, with no actual capital at risk. runway:query_trades?fund=catalyst

A hypothetical scenario for RARE has a gap-hazard sensitivity of 0.05, which is a measure of potential risk. This scenario assumes a hazard rate of 0.05, meaning a 5% chance of a hazardous event, and an other return of 0.015, or 1.5%. The shock loss is 45.8600619%, which is the potential loss from a closed RARE trade. The simulation ran 50,000 paths and 10 events. The results show terminal percentiles of 16.225534029725072, 57.028976141735804, and 57.028976141735804. The loss frequency is 0.40296, or 40.296%, and the downside deviation at 95% is 0.6980273612877876. The expected return per event is -0.8680030959752325%. The simulation used a NumPy seed of 20260909 and assumed a 1.5% win and hazard rates of 0, 1, 5, and 10%, which were explicitly assumed and not based on historical data. The full 49.14 was reinvested per hypothetical event, with no actual capital used. runway:query_trades?fund=catalyst

A hypothetical scenario for RARE has a gap-hazard sensitivity of 0.01, which is a measure of potential risk. This scenario assumes a return of 0.015, or 1.5 percent, and a hazard, or risk of loss, of 0.01, or 1 percent. A shock loss of 45.8600619 percent occurred from a closed RARE trade. The scenario includes 50,000 paths, or possible outcomes, and 10 events, or occurrences. The terminal percentiles, which show the range of possible outcomes, are 30.42 percent, 57.03 percent, and 57.03 percent. The loss frequency, or how often a loss occurs, is 0.09472, or 9.47 percent. The expected return per event is 1.0263993808 percent. This scenario assumes a 1.5 percent win and hazard of 0, 1, 5, and 10 percent, which are not based on historical data. The shock loss of 45.8600619 percent is from a closed RARE trade. The scenario reinvests 49.14 percent per hypothetical event, with no actual capital used. The NumPy seed is 20260909. runway:query_trades?fund=catalyst

A hypothetical scenario for RARE has been created with the following assumptions: the assumed hazard is 0, the assumed other return is 0.015, and the shock loss is 0.45860061919504647. The scenario includes 50,000 paths and 10 events. The terminal percentiles are all 57.028976141735804. The loss frequency is 0, and the dd95 is 0. The expected percentage per event is 1.5. This scenario uses a NumPy seed of 20260909. It assumes a 1.5 percent win and hazard of 0, 1, 5, or 10 percent, which are not based on historical data. A shock loss of 45.8600619 percent occurs from a closed RARE trade. The full amount of 49.14 is reinvested per hypothetical event, with no actual capital used. runway:query_trades?fund=catalyst

A news scan found the headline related to the symbol RARE on 2026-09-02. The news was released at 2026-09-02T20:01:36Z. At that time, there were 95 bars, which are segments of time used to measure trading activity, with the last complete bar ending at 2026-09-02T19:45:00Z. The price before the market closed, known as the preclose, was 26.55. The next bar, or segment of time, started at 2026-09-02T20:15:00Z with an opening price of 14.9915. This represents a change of -43.53483992467043 percent from the previous price. The next regular open time was 2026-09-03T13:30:00Z with a price of 13.97, which is a -47.38229755178908 percent change from the preclose price. The highest price after the news, or first post bar high, was 15.06 and the lowest price, or first post bar low, was 13.8. The data was measured on 2026-09-09. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=RARE&timeframe=15Min&feed=sip&adjustment=all

A news scan found the headline for the symbol RARE on 2026-08-19. The news was released at 2026-08-19T21:22:01Z. There were 94 bars, which are periods of time used to measure trading activity, with the last complete bar ending at 2026-08-19T21:00:00Z. The price at the end of this bar was 26.76. The next bar started at 2026-08-19T21:30:00Z with a price of 27. The price changed by 0.8968609865470878 percent in the first bar after the news. The next regular open time was 2026-08-20T13:30:00Z with a price of 28.45. This represents a 6.3153961136023895 percent change from the previous day's close. The highest price in the first bar after the news was 28.8513 and the lowest price was 27. The data was measured on 2026-09-09. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=RARE&timeframe=15Min&feed=sip&adjustment=all

A news scan found the headline for IONS on 2026-06-24. The data for IONS on 2026-06-24 includes a news event at 18:37:29. There were 65 bars, with the last complete bar at 18:15:00 and a price of 76.995. The next bar was at 18:45:00 with an open price of 77.15. The first post-bar change was 0.20131177349178575 percent. The next regular open time was 13:30:00 on 2026-06-25 with a price of 76.69. The overnight change from the pre-close price was -0.39612961880642894 percent. The first post-bar high was 77.26 and the low was 76.88. This data was measured on 2026-09-09. Note that price windows are aligned to sponsor HTML metadata, and do not imply causal attribution or guaranteed fills. A bar is a unit of time, in this case 15 minutes, used to measure trading activity. The last complete pre-bar excludes any bars that straddle the news event, and the next bar open is the first recorded trade in that 15-minute bar, not the bid or ask price or an executable limit. Some bars may be absent in extended sessions, and no fill is inferred. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=IONS&timeframe=15Min&feed=sip&adjustment=all

A news scan found the data line: SOL weekly supply regression development/holdout. This is a statistical analysis of the SOL/USD price. The analysis used a method called ordinary least squares, or OLS, to predict the next 7-day close return. It looked at Monday observations and used the prior 7-day growth of USDC/USDT and the prior 7-day close return of SOL as inputs. The results are split into two parts: development and holdout. The development part had 51 observations and found the following coefficients: intercept, USDC, USDT, and prior, with corresponding values. The holdout part had 35 observations and also found coefficients for intercept, USDC, USDT, and prior, with corresponding values. The analysis also calculated the R-squared value, which measures how well the model fits the data, and found it to be 0.0377 for the development part and 0.0576 for the holdout part. Additionally, it calculated the correlation between USDC and the predicted value, and the correlation between the prior USDC value and the predicted value. https://stablecoins.llama.fi/stablecoin/2 ; https://stablecoins.llama.fi/stablecoin/1 ; https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the data: ETH weekly supply regression development and holdout. The data includes two sets of information for the symbol ETH, one for the development window and one for the holdout window. For the development window, there are 51 observations, and the coefficients are -0.01958809028395293 for the intercept, -0.5381954082563941 for USDC, 4.463685032327928 for USDT, and 0.06803258120446076 for prior. The HC3 heteroskedastic standard errors are 0.0201003348258576, 0.6133313507662133, 3.299626505800723, and 0.1295973032845246. The USDC confidence interval is -1.772058934933387 to 0.6956681184205987, the r-squared value is 0.07912951222423315, the USDC y correlation is -0.10222991526435132, and the prior USDC correlation is 0.03378930328931067. For the holdout window, there are 35 observations, and the coefficients are -0.002268027773152274 for the intercept, 1.0177186280696657 for USDC, 0.904474307096202 for USDT, and -0.058493507659819165 for prior. The HC3 heteroskedastic standard errors are 0.01667829404574369, 1.4702453042588886, 2.3810107949666635, and 0.1491524746707091. The USDC confidence interval is -1.9808664394675157 to 4.016303695606847, the r-squared value is 0.03647913127339408, the USDC y correlation is 0.180560278995183, the prior USDC correlation is 0.386971018318211, the frozen predictive correlation is -0.027972230136303004, the frozen mean squared error is 0.009438473581974905, and the zero forecast mean squared error is 0.008318990493684546. https://stablecoins.llama.fi/stablecoin/2 ; https://stablecoins.llama.fi/stablecoin/1 ; https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the data line: OLS y=next7dayclose return; Monday observations,USDC/USDT prior7daygrowth lag1day,ownprior7dayclose return. Nonoverlapping outcomes withinwindows. Coefficients intercept,USDC,USDT,prior; HC3heteroskedastic SE,not serial-correlation corrected. The data includes two sets of results for the symbol BTC, one for the development window and one for the holdout window. For the development window, there are 51 observations, and the coefficients are -0.0069679255649645775 for USDC, -0.16097693134808475 for USDT, 1.2084609383025249 for prior, and -0.07263593198504811 for the intercept. The HC3 standard errors are 0.013273274196253933, 0.2798285470617676, 1.5754537541435267, and 0.14604837789676384. For the holdout window, there are 35 observations, and the coefficients are -0.0014636990479258798 for USDC, 0.7468017073738606 for USDT, 1.7736213859972918 for prior, and -0.06398934356186571 for the intercept. The HC3 standard errors are 0.011430114360679564, 0.7412089292120739, 1.524275094712161, and 0.12255717698144836. https://stablecoins.llama.fi/stablecoin/2 ; https://stablecoins.llama.fi/stablecoin/1 ; https://runway.adityaperswal.workers.dev/data/crypto-bars

The QTR9090400 benchmark results from September 9 are based on derived data. This data includes benchmarks such as buy and hold strategies with different costs, and mid-month strategies. A buy and hold strategy is a type of investment approach where assets are held for a long time, and the costs are measured in basis points, which are units of measurement for interest rates and percentages. The return percentage and drawdown percentage are also measured. Drawdown percentage refers to the decrease in value of an investment. The sample size and win rate are also considered. The win rate is the percentage of times an investment strategy is successful. The mean return is also calculated. The simulation results are based on 20,000 paths and a 12-month horizon. The profit quantiles show the range of possible profits, and the probability of loss is 0.63785. The cross-asset correlation shows how different assets, such as EEM, EFA, IWM, and SPY, which are stock market indexes, move in relation to each other. The post-T1 results show the mean return and win rate after a certain point. The exit counts show how many times the strategy exited a trade due to a stop, target, time, target gap, or stop gap. The worst trade percentage is -6.75242744962471. The simulation used 20,000 paths and 12 monthly windows, sampled with replacement across assets from 83 holdout months, preserving contemporaneous correlation but assuming stationarity and independent months. The hypothetical full deployment of 245.69 is not approved for a venue feasible allocation, and fixed costs are excluded. The probability of not hitting 349 USD is not true zero probability. https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=SPY,IWM,EFA,EEM&timeframe=1Day&feed=sip&adjustment=all ; finish_stats.py

A news scan found the event DTX401 for the stock symbol RARE. This event happened in the year 2026 and followed the rule called T10_T2, which is a set of instructions for buying and selling stocks. The slippage, or difference between expected and actual price, was 0.003 and the fee, or cost of the transaction, was 0.001. The stock was bought on 2026-08-10 and sold on 2026-08-20. The price at the time of buying was 25.91 and the price at the time of selling was 25.32. The net return, or profit, was -3.0557791787033217 percent, meaning the stock was sold for less than it was bought. The event dd pct, or maximum loss, was 6.458471958456958 percent. The news exposure, or how much the news affected the stock, was inside, meaning it was not publicly known. The news was released on 2026-08-19 and the target, or expected outcome, was for 2026-08-23. The spy return pct, or return of the overall market, was -1.2943308309603885 percent. This event was observed and recorded in notes 17167 through 17188 and corrected in note 1347. The results were also checked using a system called vectorbt, which independently confirmed the calculations. This was not an actual trade, but rather a simulated one, and the execution was not calibrated, meaning it was not adjusted for real-world conditions. https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all