What Sprint 5 Is About
Sprint 4 was your Software Development Sprint — you built the pipeline that collects data, runs it through LLMs, and seals a prediction. If you achieved that, well done. Sprint 5 raises the bar: this is the Integration and Intelligence Sprint, where your pipeline stops being a collection of scripts and starts behaving like a coherent, self-contained system.
The single question Sprint 5 asks is this: can your system run end-to-end with minimal human intervention? The Scrum team's job this week is to close the gaps — between agents, between scripts, between the human score and the final prediction file — so that by Sunday night, the only human decision is the Wild Card override.
Sprint 5 in One Sentence
Connect what you built in Sprint 4 into one coherent pipeline — from Friday market close to Sunday's sealed prediction — and make it visible in a dashboard or structured output your whole team can read.
Capability Tiers — Where Should Your Team Be?
Level 1
Manual
Data collected by hand. Predictions typed. No scripts. No automation.
Sprint 1–3 BaselineLevel 2
Structured
Scripts fetch data. Output saved to GitHub. Some automation in place but still manually triggered.
Sprint 4 → 5 FloorLevel 3
Automated Pipeline
GitHub Actions runs fetch automatically. LLM API calls scripted. Delta scoring runs after actuals are known. Prediction file structured and sealed by tag.
Sprint 5 TargetLevel 4
Intelligent System
Full pipeline plus a live dashboard or structured HTML report your team can read. Prescriptive output, not just data. Human Score integrated into output automatically.
Sprint 5–6 StretchSprint 5 Expectation
Teams that completed a working fetch_market_data.py and GitHub Actions trigger in Sprint 4 should be reaching Level 3 this sprint. Teams still at Level 1 need to close that gap this week — the window is narrowing as A1 approaches.
Sprint 5 Deliverables — What Goes into vW25
Your vW25 release tag must contain the following. Seal it by Sunday 21 June 23:59 SGT. Late tags cannot receive a delta score.
| Deliverable | Owner | Level | Description |
|---|---|---|---|
Prediction fileprediction_W25.json |
R7 R2 | REQ | SPX, NDX, IWM directional prediction + confidence + Human Score Wild Card. Structured JSON or Markdown. Sealed inside vW25 tag. |
| Agent outputs Almanac · Macro · Technical |
R3 R4 R5 | REQ | Each agent writes independently before LLM synthesis. All three must be present and committed before the tag is cut. |
| LLM synthesis ChatGPT · Gemini · DeepSeek |
R8 | REQ | Identical prompt sent to all three mandatory LLMs. Responses saved and committed. Comparison table showing where LLMs agree and diverge. |
| Automated fetch log | R9 | REQ L3 | Evidence that data was fetched automatically (GitHub Actions run log, auto-committed data file with timestamp, or equivalent). Teams at Level 2 must show manual fetch with a commit timestamp before Friday midnight SGT. |
| Pipeline diagram or README update | R1 R9 | REQ | A diagram (FigJam, draw.io, or Mermaid in README) showing end-to-end data flow from market close → prediction seal. R1 signs off the README update. |
| Delta score report — vW24 | R10 R6 | REQ | R10 calculates direction accuracy and error % for last week's vW24 prediction vs Friday's actual close. Committed as delta_W24.md or equivalent. R6 provides actuals data. |
| Dashboard or structured report | R9 R1 | STRETCH L4 | A simple HTML page, Streamlit app, or GitHub Pages view showing the current week's prediction, agent summaries, and LLM comparison in a readable format. Not required to be live — a screenshot in the repo counts. |
| Sprint Retrospective notes | R2 | SCRUM | R2 Scrum Master documents what the team completed, what was blocked, and what changes in the next sprint. Goes into the Weekly Workbook and FigJam board. |
Role Focus — Sprint 5 Priorities
R1 · Product Owner
System Coherence
Define what "done" means for the integrated pipeline. Sign off the README. Ensure all agent outputs connect to the prediction file.
R2 · Scrum Master
Integration Unblocking
Chase every role daily. Run a mid-week check by Wednesday. Retrospective is mandatory this sprint — A1 workbook needs it.
R3 · Almanac Agent
Seasonal Context
Historical patterns for the week of 16–20 June. Document sources. Output must be committed before R8 runs LLMs.
R4 · Macro Agent
Fed + Data Watch
Macro events for w/c 16 June: Fed speakers, CPI revisions, any major data releases. Structured Markdown output.
R5 · Technical Agent
Trend + Momentum
SPX/NDX/IWM technical read from Friday 13 June close. Key levels, trend direction, momentum signal.
R6 · Data / Actuals
vW24 Actuals
Pull confirmed Friday 13 June closes. Feed to R10 for delta scoring. Also prepare data structure for current week's collection.
R7 · Human Score
Wild Card Override
Review all three LLM outputs + three agent outputs. Apply override if consensus is wrong. Write Wild Card insight — this is the primary marks differentiator.
R8 · LLM Operator
Synthesis + Comparison
Send identical prompt to ChatGPT, Gemini, DeepSeek. Save all responses. Build comparison table. Claude optional if access available.
R9 · GitHub Lead
Pipeline Owner
Get GitHub Actions fetch running automatically if not yet done. Ensure all branches merged before Sunday. Cut the vW25 tag.
R10 · Calibration
Delta Score vW24
Calculate how accurate last week's prediction was. Direction correct? Error %? Commit delta_W24.md with retrospective notes.
Five Things That Will Sink Your Sprint
01
Waiting until Sunday to merge
If all branches land Sunday afternoon, there is no time to fix merge conflicts. R2 must chase commits by Thursday. R9 merges by Saturday.
02
Agent outputs after LLM synthesis
R8 cannot run LLM synthesis until R3, R4, and R5 have committed their outputs. The dependency is strict. R2 enforces the order.
03
No delta score for vW24
R10 must commit delta_W24.md this sprint. A prediction sealed last week with no retrospective score is an incomplete Scrum cycle.
04
Wild Card with no reasoning
R7's Human Score is the primary differentiator. "I agree with the LLMs" is not a Wild Card. Write a specific, sourced, non-obvious insight.
05
Missing the vW25 tag
A tag not cut by Sunday 23:59 SGT cannot receive a delta score next sprint. The tag is the immutable seal. There are no extensions.
Sprint 5 Timeline
Monday 15 June 2026
Sprint 5 Kickoff — after Sprint 4 presentations
Teams receive Sprint 5 brief. R1 and R2 plan the week's backlog.
Mon–Wed 15–17 June
Agent work week — R3, R4, R5 research and write
Almanac, Macro, and Technical agents commit initial outputs. R9 ensures pipeline is running.
Thursday 18 June
R8 LLM synthesis — runs all three mandatory LLMs
Identical prompt sent to ChatGPT, Gemini, DeepSeek. Comparison table committed. R10 begins delta_W24 calculation.
Friday 20 June — US market close
R6 collects actuals · Automated fetch runs
Friday close data for SPX, NDX, IWM collected. GitHub Actions pipeline should trigger automatically at 22:00 SGT.
Saturday 21 June
R7 Human Score + Wild Card — final review
R7 reads all outputs, applies override if warranted, writes Wild Card insight. R2 chases any missing role submissions.
Sunday 21 June 2026 · 23:59 SGT ⚑
vW25 Release Tag Deadline
R9 cuts the vW25 tag. All prediction files, agent outputs, LLM synthesis, and delta score must be committed before the tag is created. No extensions.
Monday 22 June 2026
Sprint 5 Presentations · Sprint 6 Kickoff
Every role (R1–R10) speaks. Demo of the integrated pipeline. R10 presents delta score for vW25 prediction.