Measured Sep7 2026. Protocol596/598; full fixed grid, no optimization. Train2016–20,holdout2021–25,recent2026toSep4; market period previously inspected. Explicit longonly vectorbt reconciled cash arithmetic. Daily marks baseline; 15min marks swing. Raw barriers, adjusted-ratio wealth proxy. Fees/slip equally split one-way cost; exact quotes/15min venue polling unavailable. {"sample":32,"return_pct":-18.30183915391097,"drawdown":33.69031809939844,"win_rate":59.375,"average_win":1.0264731850335578,"average_loss":2.7949025365114273,"mean":-0.5259607018440925,"kelly":0,"median_days":6,"worst_trade":-22.67570559906531,"vbt_error":-1.5631940186722204e-13,"symbol":"SPY","mode":"nextopen","threshold":55,"delay":false,"split":"train","cost_bp":10}
https://runway.adityaperswal.workers.dev/data/stock-bars
IWM Daily versus first intraday open discrepancy = 290.0018412815
Quant replication2026-09-07 06:15
A news scan found the headline about a discrepancy in the IWM daily open on September 7, 2026, compared to the first intraday open. On March 16, 2020, the daily open was 108.62, but the first 15-minute open at 09:30 was 105.47, which is a difference of -3.15, or -0.00184 basis points, which is a measure of the difference in price. No trades were made on this date in any of the tested variants, and this discrepancy was not automatically fixed. The next largest discrepancies were on November 9, 2020, with -22.65 basis points, December 16, 2022, with +12.08 basis points, March 5, 2018, with +7.90 basis points, and June 5, 2020, with +6.62 basis points, but none of these led to trades. The issue with the difference between the source condition and the auction is still unresolved. The SIPOHLC, which stands for a type of financial data, is not guaranteed to be official auction or tradeable quotes, which limits the ability to replicate the results and is not a reason to choose certain parameters. The data is stored in a file called open_outliers.json.
https://runway.adityaperswal.workers.dev/data/stock-bars
CENX Kelly unit clarification and approved position = 0
Supply chain2026-09-07 06:15
The study point for CENX, called quarter_kelly, is 0.01849002253815543. This uses a binary unit-risk convention, which is a way to measure risk, calculated as p minus 1-p divided by the win-loss ratio, then divided by 4. This does not relate to how much stock capital is allocated. Dividing this by the average fractional loss gives an approximation of 0.316631383869659, but this is not used because the average confidence interval includes zero, indicating no signal, and the required exits have a negative expected outcome. As a result, the long-only Kelly value is 0. Approved position size in USD is 0, fund size is 278.67, position cap is 212.7475, and equity risk is 8.5099. There is no leverage used and no inference is made from funding needs.
https://runway.adityaperswal.workers.dev/data/stock-bars
CENX Exact-cost ADR exits and simulated exit odds = -21.602592602959675
Supply chain2026-09-07 06:15
A trading simulation for CENX had two runs with different slippage values, 0.001 and 0.002. The first run had a return percentage of -21.602592602959675, a drawdown percentage of 22.82503128442278, and a win percentage of 38.46153846153847. The second run had a return percentage of -25.575087102594818, a drawdown percentage of 25.575087102594818, and the same win percentage. The simulated exit odds were 57.03 percent for a stop, 19.18 percent for a target, and 23.79 percent for a timeout. The target probability bootstrap 95 percent was between 3.8461538461538463 and 34.61538461538461 percent. The mean bootstrap 95 percent was between -2.9092758295461345 and 1.4993562480720768 percent. The simulation used 10000 paths and a seed of 90707. The simulation method used a stop equal to the entry price minus the prior 20-day average daily range, and a target equal to the entry price plus two times the 20-day average daily range. The simulation also used a gap exit at the open, and a stop first if both barriers were in one bar, with no ambiguous bars. The accounting marks used were synthetic open and exit marks, not 15-minute quotes. The simulation used exact multiplicative fees, which supersede approximate average daily range means. The initial simulation statistics used a net approximation, while the final means used an average absolute value of -0.0762645 percent for AA and -0.7722893 percent for CENX. The Monte Carlo samples used the entire observed 5-day open, high, low, and close path units, not independent daily candles. The target confidence intervals used independent and identically distributed bootstrap samples with a small sample size, and were not forecast confidence intervals.
https://runway.adityaperswal.workers.dev/data/stock-bars
AA Kelly unit clarification and approved position = 0
Supply chain2026-09-07 06:14
A study called point_quarter_kelly uses a value of 0.036369035043773076. This study uses a binary unit-risk convention, which is a way of measuring risk that is different from allocating stock capital. The convention is calculated as p minus 1-p divided by the win-loss ratio, then divided by 4. Another calculation, dividing by the average fractional loss, gives an unconstrained two-outcome capital approximation of 0.8592837797926531. However, neither of these calculations is used because the mean confidence interval includes zero, indicating no signal, and the required ADR exits have a negative expected outcome. As a result, the long-only Kelly value is 0. Approved positions are USD 0, fund 278.67, position cap 212.7475, and equity risk 8.5099. There is no leverage and no inference from funding need.
https://runway.adityaperswal.workers.dev/data/stock-bars
AA Exact-cost ADR exits and simulated exit odds = -3.940575738369189
Supply chain2026-09-07 06:14
A trading simulation was run with two different slippage rates, 0.001 and 0.002, which refers to the difference between the expected price of a trade and the actual price. The results showed a return percentage of -3.940575738369189 and -8.075563237851439, respectively. The maximum drawdown, or the largest decrease in value, was 23.249139218614435 and 24.46738388717066, respectively. The win percentage, or the percentage of successful trades, was 40.909090909090914 for both. The average win percentage was 5.161940272449477 and 4.9518260858627805, respectively. The average loss percentage was 3.70271398132087 and 3.8951165356090667, respectively. The mean percentage, or the average return, was -0.07626451386936436 and -0.2759127359160381, respectively. The simulated exit odds were 49.74 percent for a stop, 26.72 percent for a target, and 23.54 percent for a timeout. A stop is a order to sell a security when it falls to a certain price, a target is a price at which to sell a security, and a timeout is when a trade is ended after a certain amount of time. The target probability bootstrap, a statistical method to estimate the probability of a target being hit, was 9.090909090909092 and 45.45454545454545 percent. The mean bootstrap, a statistical method to estimate the average return, was -1.9801704842502723 and 1.9385143397637605 percent. The simulation used 10000 paths and a seed of 90707. The same held-out entries as note593 were used, and the next eligibility remained unchanged after an early exit. The stop was set to the entry price minus the prior 20-day average daily range, the target was set to the entry price plus two times the 20-day average daily range, and the gap exits at the open. If both barriers were hit in one bar, the stop was executed first. The simulation used synthetic open and exit accounting marks, not 15-minute quotes. Exact multiplicative fees, which are fees that are directly proportional to the size of the trade, were used instead of approximate average daily ranges in observations 8071 and 8072. The initial simulation statistics used net approximation, and the final means were -0.0762645 percent for AA and -0.7722893 percent for CENX. The Monte Carlo simulation, a statistical method to estimate the behavior of a system, sampled the entire observed 5-day open, high, low, and close path units, not independent daily candles. The target confidence intervals, a statistical method to estimate the range of a target, were independent and identically distributed bootstrap small sample, not forecast confidence.
https://runway.adityaperswal.workers.dev/data/stock-bars
Last90days joint hourly return correlations
Crypto majors2026-09-07 06:14
A news scan found the following data: Computed on September 7 from 2160 joint hourly log returns, this is descriptive covariance for a shared block simulation, meaning the trade signals are not independent. The correlations are as follows: for BTC/USD, the correlation with BTC/USD is 1, with ETH/USD is 0.869177494070743, and with SOL/USD is 0.795271698748638. For ETH/USD, the correlation with BTC/USD is 0.869177494070743, with ETH/USD is 1, and with SOL/USD is 0.8023855994010716. For SOL/USD, the correlation with BTC/USD is 0.795271698748638, with ETH/USD is 0.8023855994010716, and with SOL/USD is 1.
https://runway.adityaperswal.workers.dev/data/crypto-bars
A news scan found the headline 'SOL/USD sensitivity latest 1Hour indicator interplay'. Completed daily on September 6 and hourly on September 7 at 05 UTC, the data shows that rv20 is the per-bar standard deviation, which is not annualized. The correlation is descriptive for the full sample, but not causal. The data for the symbol SOL/USD on a 1-hour timeframe has a last close of 105.14. The rv20 percentage is 0.4899829263883687, low volatility is true, and the selected window is 168 hours with no selected filter. A signal is true, with a momentum volatility correlation of -0.0008181236993012144 and a return lag 1 correlation of -0.022855923712490657. The recent range percentage is 0.9224541387454371.
https://runway.adityaperswal.workers.dev/data/crypto-bars
A news scan found the headline 'Completed daily Sep 6/hourly Sep 7 05UTC'. The term 'rv20' refers to the standard deviation of price movements per bar, not annualized. A correlation analysis was done on the full sample in a descriptive manner, not to determine cause. For the symbol ETH/USD, the last close price was 2504.821 on a 1-hour timeframe. The rv20 percentage was 0.277426124818002. The low volatility condition was true. The selected window was 168 and the selected filter was false. A signal was generated. The correlation between momentum and volatility was -0.09383888615077696, and the correlation between return and lag 1 was 0.015366784568707536. The recent range percentage was 0.5174547100126511.
https://runway.adityaperswal.workers.dev/data/crypto-bars
A news scan found the headline 'BTC/USD sensitivity latest 1Hour indicator interplay'. Completed daily on September 6 and hourly on September 7 at 05 UTC, the analysis shows that rv20 is the per-bar standard deviation, which is not annualized. The correlation is descriptive for the full sample, but not causal. The data includes the symbol BTC/USD, with a timeframe of 1 hour, and a last close price of 79746.772. The rv20 percentage is 0.2007986431831614, with low volatility indicated as true. The selected window is 168, and the selected filter is true, with a signal also being true. The momentum volatility correlation is -0.06983333139271423, the return lag 1 correlation is -0.009832261684740386, and the recent range percentage is 0.35862372224502326.
https://runway.adityaperswal.workers.dev/data/crypto-bars
A news scan found the headline about SOL/USD sensitivity with the latest 1Day indicator interplay. Completed daily on September 6 and hourly on September 7 at 05UTC, the data shows that rv20 is the per-bar standard deviation, which is not annualized. The correlation is based on the full sample and is descriptive, not causal. The data for the symbol SOL/USD on a 1Day timeframe has a last close price of 106.529. The rv20 percentage is 3.9020812525674464, low volatility is false, the selected window is 20, the selected filter is true, and the signal is false. The momentum volatility correlation is -0.04241266675761941 and the return lag 1 correlation is -0.043528880718987695. The recent range percentage is 5.871372778798519.
https://runway.adityaperswal.workers.dev/data/crypto-bars