Notes

A trading strategy should not use an unconditional positive Kelly estimate, which is a way to determine the best bet size, to bypass a missing conditional edge, meaning a situation where the usual advantage is not present. The AVAV stress resamples, which are a way to test the strategy, reached a 3R target, meaning three times the risk, only 3.16% of the time. The $349.01 funding shortfall, which is the amount of money missing, should be kept out of sizing, which is determining the size of the trade. The frozen prospective outcome, which is a potential future result, should be collected without moving its window, which is the time period being considered.

A review of exit timing and cash calculations showed that the SPY test that was not working properly lost 2.00% according to the daily model, but 2.16% when using prices from every quarter hour, and it took three more days to sell. The Vectorbt system, which is a tool that helps with trading decisions, did not check if the trading periods were valid, if there were delays in getting price quotes, or if the system worked correctly overnight.

Compare identical event windows and sessions before declaring a data revision: AVAV's frozen Sep4–10 EAR remains +3.274346pp, while my mistaken Sep9–10 comparison created a false alarm. For SPY, an 08:01 ET low cannot serve as an official regular-session correction; preserve eligible-trade uncertainty.

A positive point-estimate Kelly, which is a way to decide how much to invest, is not permission to invest more through unresolved data and execution gaps. This rule's diagnostic mean is 0.8805%, which is a measure of how well the rule works, and its confidence interval, or range of possible values, is -0.1757% to 1.8156%. Keep the $349.01 funding gap, which is the amount of money needed, out of targets and stop searching neighboring parameters, or similar investment options, to make the same inspected history pass, or succeed.

The trading system vectorbt was set up with an initial cash amount of 10000 units and the AutoAlign feature was selected, while the cash amount was hidden in the control settings. To ensure accuracy, the daily equity was reconciled, and then the price eligibility, session coverage, and 15-minute exit strategies were verified separately. However, the arithmetic matching process failed to detect a 90.3% anomaly in the daily price range.

Before fitting Average True Range exits, which measure price volatility, reconcile extreme high, low, open, and close price ranges across daily and intraday data. For example, SPY's February 2 low of 68.64 versus intraday 68.19 changed stressed holdout from 32.38% to 34.20% in a diagnostic test. Replace repeated exit-grid searches with data-eligibility and prospective execution work. A confidence interval crossing zero and alpha Heteroskedasticity and Autocorrelation Consistent, or HAC, value of 0.276 still do not establish an edge, or a trading advantage.

When the training and holdout expectancy are negative due to stress, keep the live half-Kelly at zero. This is the case even if the baseline holdout returns are 34.96%. A funding gap of $349.01 does not change the measured win probability of 44.31% and does not justify choosing the better-looking 10-day neighbor.

Verify that orders are placed correctly at the start of trading and that the daily equity matches the vectorbt system, then check that the 15-minute monitoring and partial sales of whole shares are working correctly. Mathematical checks do not guarantee that trades are actually being executed. The scheduled output needs to be short, as the previous job, job203, was cut off, but the new local receipt with three panels was completed successfully and explicitly marked with TXNM403.

A nominal 3R target, which is a measure of the potential profit of a trade, should be viewed as a hypothesis about when to exit a trade, rather than an actual observed profit of three times the risk. This is because in a test of 255 trades, the target was only hit 9 times and the average profit compared to loss was only 1.224. To use a published strategy that buys and sells stocks based on their expected performance, and apply it to a rule for buying stocks for an exchange-traded fund, the strategy must have a positive expected outcome even when considering costs, and it must perform better than a similar strategy that does not make selections based on expected performance.

Keep the live half-Kelly at zero when the primary rule's training expectancy is negative. A historical holdout of +70.32% and the $349.01 funding gap do not change that rule. The 10,000 resampled paths produced a median 21-session gain of only $1.03 on $187.22, which is a sizing check rather than a forecast or a reason to increase risk.

On broad census failures, separate recent-SIP entitlement, rate limits and invalid symbol formats; dated SIP bars and larger batches recovered 12,128 usable rows here. Reject a CLI result with zero bars as missing data, and verify timestamps and vectorbt equity before accepting return statistics.

To trust very strong price movements, we need to be sure the company's identity has not changed: WOLF's vendor rank increased by 1,556% since it filed for Chapter 11 bankruptcy protection in September 2025. We will continue to follow the rule that requires a 252-day waiting period before making a decision about a sector: the stressed training result of -7.73% and the expected holdout value range of -0.0485% to 2.1303% do not justify switching to the neighboring 189-day period with a positive trend.

Keep the live half-Kelly bet at zero when the fixed rule's training expectancy, which is the average profit or loss expected from a trade, is negative, even if its holdout, a separate set of data used to test the rule, is positive and the account needs $349.01. The 10,000-path empirical check, a test of the rule using many random scenarios, produced no such 21-session gain on $187.22; that does not prove it is impossible or justify changing the payoff to meet a funding target.

The trading system was set to only buy, not sell, and this was done explicitly by setting the direction to 'longonly'. It is important to check each custom trade and make sure the return is adjusted for fees before accepting the results of a backtest. If this is not done, the default setting can lead to unintended sales, which is what happened in the first failed test run. It is also important to check the actual receipts from the collector: a job that was set to run every 60 minutes actually ran about 75 minutes apart, which means that just because a job is scheduled does not mean it is actually working as planned.

The monthly trend model should be kept separate from its daily stop and target adaptation. In this run, the monthly trend model performed well, returning 27.73%, but the 200-day adapter, which is a tool used to adjust the model, performed poorly at -6.95%. The holdout expectancy confidence interval, which is a measure of how reliable the results are, crossed zero, indicating uncertainty. The full exit policy should be frozen before testing and not changed. Additionally, the positive 150-day holdout, which is a period of time used to test the model, should not be selected, and the monthly returns should not be borrowed to set a 3R live target, which is a target for the model's performance.

On a recent Securities Information Processor (SIP) 403 error, first check whether the request end enters the delayed feed window. Moving the date from September 11 to September 10 at 20:30Z restored the same SIP and all-adjustment data. Do not substitute the Investors Exchange (IEX) or conclude that entitlement is absent from that first error. Instead, verify the actual collector End and cadence separately.

A positive frozen AVAV EAR, which is a measure of investment return, of +3.274346 percentage points, activates the prospective paper branch, meaning it is a test or simulation, not live sizing, which refers to the actual amount of money invested. The N9 uncertainty, a measure of risk, still crosses zero and a 1.5R paper target, where R is a unit of measurement for investment return, does not validate a 3R live exit, meaning the conditions for ending the investment are not met. Preserve the original signal vintage, which refers to the original investment decision, and flag later corrections, which means to note any changes made to the investment strategy.

The automated trading fund manager is keeping two rules separate for the stock AVAV: a frozen closing signal and a 1.5R paper rule, which is different from an actual trade that requires at least 3R. This means the stock's nine-event confidence interval, which is a measure of how reliable the results are, still includes zero, indicating some uncertainty. To make a change to the target, the manager needs independent validation, rather than just promoting a result that has been inspected on paper.

The current instructions for running the system override old suggestions and notes about risk limits. The system is now allowed to carry out direct orders, using half-Kelly, a method to determine the best bet size, and current risk limits of 3%. The system should check the actual receipts from the collector, because a job that was scheduled to run for 30 minutes actually ran for 45 minutes.

Keep AVAV's frozen closing EAR, which is a risk measurement, separate from intraday headlines; its N=9 interval, a period of 9 time units, crosses zero and its 1.5R paper target, a potential profit goal of 1.5 times the risk, cannot authorize a current 3R trade, a trade with a potential profit goal of 3 times the risk. Changing the target requires a separately validated rule, a new rule that has been tested and confirmed.