# Independent Research and Demo Capstone Demonstrate the full research-to-review process without profit guarantees. Use this workbook alongside the course. Write your answers before opening the solutions. Practical work is self-reviewed; scored knowledge checks are in the Academy. ## 1. Proposal ### Research question A capstone proposal states a bounded research question and the decision it could inform. Choose a scope that can be completed with available data and tools. An ambitious title is less valuable than a reproducible answer to a precise question. ### Source register Identify prerequisites and required resources before beginning. Data resolution, access rights and execution records can determine whether a question is answerable. Redesign the question when essential evidence is unavailable rather than fabricating precision. ### Baseline Define the baseline, candidate and evaluation design in advance. Record which data supports development and which remains reserved. A capstone should expose the selection process rather than present the final rule as inevitable. ### Acceptance and rejection criteria State practical and ethical boundaries: demo versus live activity, permitted data use and the claims the project will not establish. This course does not require real-money trading to demonstrate research competence. ### Worked example A proposal aims to prove a millisecond execution edge using daily bars. The research question and available evidence are incompatible; reducing the claim or obtaining suitable data is necessary. ### Independent exercise Write a one-page proposal with question, data, baseline, evaluation, deliverables and explicit limits. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 2. Execution ### Frozen validation Execute the frozen protocol and record deviations as they occur. A deviation can be justified, but it creates a new version or limitation that the reader must see. Do not silently change the study after unfavorable results. ### Demo observation Maintain data and code provenance so every result can be rebuilt. Preserve raw observations, transformation steps, configuration and environment information under appropriate access rules. ### Cost reconciliation Collect demo execution records when implementation is part of the question. Label simulated, replayed and observed demo results separately. Do not combine them into one performance series without explaining the different evidence types. ### Operational drill Perform reconciliation and quality checks before interpreting outcomes. Resolve duplicate fills, missing costs and timestamp errors. Keep a defect log so the final report reflects corrections rather than an artificially flawless process. ### Worked example A capstone changes its cost model midway after discovering omitted financing. The correction is necessary, but earlier and revised results should be retained with an explanation and the conclusion reassessed. ### Independent exercise Write a protocol-deviation record for that correction, including which outputs must be regenerated. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 3. Defense ### Results and uncertainty Defend the economic or procedural rationale without overstating what it proves. Explain why the hypothesis was worth testing and which findings support or weaken it. A negative result can be a successful capstone. ### Failures Show failed variants and negative findings relevant to selection. The reviewer should understand how the chosen procedure emerged. Concealing the search makes the final result appear more independent than it is. ### Reproducibility Trace a selected reported number to its raw inputs and calculation. This demonstrates reproducibility more concretely than presenting many polished charts. Be prepared to explain costs, units and timing. ### Limitations and next experiment State generalization limits: sample size, regime coverage, data quality and execution assumptions. A capstone defense should identify what additional evidence would be needed before broader use. ### Worked example The project finds no robust net advantage after costs. A defensible conclusion rejects the current candidate and preserves the baseline rather than rewriting the objective to manufacture a success. ### Independent exercise Prepare a five-minute defense covering question, method, one reproducible result, one failure and the justified decision. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## 4. Final portfolio of evidence ### Submit reproducible research materials Submit a reproducibility package with report, data references, code or formulas, configuration and run instructions. Include a small shareable verification dataset where appropriate. Respect licensing and avoid including credentials or unnecessary personal data. ### Present demo execution records Present demo execution records separately from research simulations, with reconciliation and limitations. A passing knowledge exam does not independently validate these artifacts; use the provided rubric for self-review or an actual named reviewer. ### Defend limitations and negative results Defend limitations and negative results alongside successes. List unresolved issues with their practical impact. A completion certificate should not be represented as a guarantee of profitability or professional authorization. ### Write a maintenance and retirement plan Write a maintenance and retirement plan for any procedure proposed for continued use. Include monitoring assumptions, version review and conditions for returning to research. Preserve the record even if the candidate is abandoned. ### Worked example A learner completes all quizzes but submits no reproducible project materials. Quiz completion is evidence of those answers, not evidence that the practical capstone has been independently assessed. ### Independent exercise Use the four-part rubric to review your package and identify the weakest criterion. Write a bounded remediation task before claiming the project is ready for review. My inputs and assumptions: My calculation or decision: Evidence that would change my conclusion: ## Course project Submit code or rules, evidence, risk plan, demo logs and an independent critique. ### Self-review rubric - Concepts and reasoning: 25% - Calculations, data and evidence: 30% - Process and risk controls: 25% - Limitations and communication: 20% Record one correction and one next practice task. This rubric is not automatically graded. ## Worked solutions ### Exercise 1 A complete proposal connects every claimed measurement to available evidence and names a decision criterion. If the data cannot support the timing claim, revise the question before collecting results rather than weakening the standard afterward. ### Exercise 2 Record the omitted charge, evidence, correction date, affected version and regenerated net results and risk measures. Explain whether the original decision changes. Do not present the revised procedure as the one originally frozen. ### Exercise 3 The defense should explain the evidence chain and why the decision follows it. Success is a well-supported conclusion and demonstrated method, not necessarily a positive trading return. ### Exercise 4 Assess concepts, calculations/data, process/risk controls and communication against the actual artifacts. Name missing evidence and correct it. Do not award yourself an independent review merely by checking a completion box. ## Further reading - https://www.itl.nist.gov/div898/handbook/ - https://scikit-learn.org/stable/modules/cross_validation.html