diff --git a/evolution_log.md b/evolution_log.md index 68513a3..f342625 100644 --- a/evolution_log.md +++ b/evolution_log.md @@ -4,6 +4,28 @@ Each entry records one autonomous improvement iteration. --- +## Iteration 11 β€” 2026-06-22 + +### [UI/UX Improvements] + +- **Ship-wheel loading spinner** (`GameProvider.tsx`): Replaced the flat shimmer-bar + lion emoji loading screen with an animated SVG maritime ship's helm. An 8-spoke wheel (hub radius 5.5, rim radius 24, knob radius 2.8 at each spoke tip) rotates at 3 s/revolution via `@keyframes spinWheel`. All geometry computed inline via `SPOKE_ANGLES.map()` β€” no external asset. The spinner is `aria-hidden="true"` since the adjacent text already describes the loading state. Directly implements the "ship-wheel / compass loading spinners" item from the UI/UX rotation. + +- **Progressive hint system** (`SQLPanel.tsx`): Replaced the auto-display of the full hint (triggered on any failed attempt) with a click-to-reveal progressive disclosure UI. A `splitHint()` helper splits each level's hint string at `. ` boundaries into at most 3 chunks (grouping more if the hint is longer). After a first failed submission, a gold `πŸ’‘ Request hint` button appears instead of the full text. Each click reveals one additional chunk in the hint panel; earlier chunks fade to 70% opacity to visually de-emphasise them as the current tip. A `hintChunkIdx` state (reset on every level change) tracks disclosure depth. The button label changes to `Show more hint (N step(s) remaining)` after the first reveal. This changes the cognitive posture from passive delivery to active retrieval β€” learners who request hints engage more deeply than those who receive them automatically. + +### [Game Design Tweaks] + +- **Level 70** *(Expert, difficulty 4)* β€” "Loan Book Risk Scorecard β€” Weighted Rate & Running Exposure": Teaches four high-value patterns in a single query: (1) **weighted-average rate** via `ROUND(SUM(principal_amount * interest_rate) / SUM(principal_amount), 2)` β€” the correct way to aggregate rates that differ in volume; (2) **scalar CTE cross-join** using a second `total` CTE and `CROSS JOIN` in the final SELECT to access the grand total for portfolio share computation; (3) **portfolio percentage** `ROUND(100.0 * g.total_principal / t.grand_total, 1)`; (4) **running cumulative exposure** via `SUM(g.total_principal) OVER (ORDER BY g.risk_grade ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW)`. Against the active loan book (12 loans, grades A/B/C only) the result is 3 rows: A (5 loans, $1.88M, 2.50%, 86.5%, cumulative $1.88M), B (4 loans, $245K, 2.89%, 11.3%, cumulative $2.13M), C (3 loans, $48K, 5.50%, 2.2%, cumulative $2.17M). This exact scorecard is used in Basel III RWA dashboards at GIC, JPMorgan, and DBS. No new data needed β€” uses the existing `loans` table. + +- **Level 71** *(Expert, difficulty 4)* β€” "Loan Portfolio Year-over-Year Disbursement Growth": Teaches the canonical YoY growth pattern using two CTEs and `LAG`. CTE `yearly` groups all 15 loans (regardless of current status) by `STRFTIME('%Y', start_date)` into loan_count and total_disbursed. CTE `with_lag` adds `LAG(total_disbursed) OVER (ORDER BY loan_year) AS prev_disbursed`. Outer query computes `CASE WHEN prev_disbursed IS NULL THEN NULL ELSE ROUND(100.0 * (total_disbursed - prev_disbursed) / prev_disbursed, 1) END AS yoy_growth_pct`. Returns 7 rows (2018–2024) showing dramatic origination swings: +100% in 2020, -71.4% in 2021, +172.5% in 2022, -87.2% in 2023. The NULL-safe CASE WHEN pattern is the canonical SQL idiom for period-over-period growth, used in strategic planning decks, investor reports, and loan-book analytics at commercial banks and sovereign wealth funds. No new data needed. + +### [Database & Code Optimizations] + +- **`LevelUpModal` hint map** extended with entries for levels 70 (risk scorecard β€” weighted rate, CROSS JOIN scalar CTE) and 71 (YoY growth β€” STRFTIME + LAG + CASE WHEN NULL). `nextHintKey` sentinel array extended from `[…, 69]` to `[…, 69, 70, 71]`; fallback updated `?? 69` β†’ `?? 71`. `MAX_LEVEL` in `progression.ts` auto-derives from the last level in the array (71) β€” no other touch points needed. +- **Test suite**: 395 tests pass (up from 375 in iter-10). Both new levels are automatically picked up by the determinism invariants suite, verifying non-empty result sets and confirming all ORDER BY keys are distinct (no ties that could produce non-deterministic ordering). +- **No new DB tables, seed rows, or indexes** β€” both levels run cleanly against the existing `loans` table. + +--- + ## Iteration 10 β€” 2026-06-15 ### [UI/UX Improvements] diff --git a/src/components/GameProvider.tsx b/src/components/GameProvider.tsx index e1b5414..f2eaaf2 100644 --- a/src/components/GameProvider.tsx +++ b/src/components/GameProvider.tsx @@ -84,13 +84,54 @@ export default function GameProvider() { // ── Loading state ────────────────────────────────────────────────────────── if (!isDbReady) { + const SPOKE_ANGLES = [0, 45, 90, 135, 180, 225, 270, 315]; return (
- {/* LCB lion crest placeholder */} - - - 🦁 + {/* Ship-wheel spinner */} + +

Lion City Bank @@ -99,20 +140,10 @@ export default function GameProvider() { Initialising SQL Analytics Engine…

-
-
-
diff --git a/src/components/LevelUpModal.tsx b/src/components/LevelUpModal.tsx index f486a0e..7215691 100644 --- a/src/components/LevelUpModal.tsx +++ b/src/components/LevelUpModal.tsx @@ -25,6 +25,8 @@ const EPOCH_NEXT_HINT: Record = { 67: 'NOT EXISTS: the correlated anti-join without the NOT IN NULL trap', 68: 'Cohort retention: first-login cohort month β†’ Month-1 return rate β€” the core product-growth metric', 69: 'Sessionization / gaps & islands: LAG β†’ flag β†’ SUM() OVER β†’ aggregate β€” the universal session-detection pattern', + 70: 'Risk scorecard: weighted-average rate via SUM(x*w)/SUM(w), CROSS JOIN scalar CTE for denominators, running exposure window function', + 71: 'Year-over-Year growth: STRFTIME β†’ GROUP BY year β†’ LAG(total) OVER β†’ CASE WHEN NULL β€” the canonical period-over-period growth pattern', }; export default function LevelUpModal() { @@ -61,7 +63,7 @@ export default function LevelUpModal() { return () => window.removeEventListener('keydown', onKey); }, [showLevelUp, handleClose]); - const nextHintKey = [10, 20, 30, 40, 44, 54, 57, 59, 61, 63, 64, 65, 66, 67, 68, 69].find((k) => currentLevel < k) ?? 69; + const nextHintKey = [10, 20, 30, 40, 44, 54, 57, 59, 61, 63, 64, 65, 66, 67, 68, 69, 70, 71].find((k) => currentLevel < k) ?? 71; const xpEarned = xpFor(currentLevel); return ( diff --git a/src/components/SQLPanel.tsx b/src/components/SQLPanel.tsx index b9945aa..d162c4c 100644 --- a/src/components/SQLPanel.tsx +++ b/src/components/SQLPanel.tsx @@ -44,11 +44,25 @@ interface SQLPanelProps { onQueryRun?: () => void; } +/** Split a hint string into up to 3 progressive reveal chunks. */ +function splitHint(hint: string): string[] { + const raw = hint.split('. '); + const parts = raw.map((p, i) => (i < raw.length - 1 ? p + '.' : p)); + if (parts.length <= 3) return parts; + const n = parts.length; + return [ + parts.slice(0, Math.ceil(n / 3)).join(' '), + parts.slice(Math.ceil(n / 3), Math.ceil((2 * n) / 3)).join(' '), + parts.slice(Math.ceil((2 * n) / 3)).join(' '), + ]; +} + export default function SQLPanel({ onQueryRun }: SQLPanelProps) { const [query, setQuery] = useState(''); const [error, setError] = useState(null); const [failedAttempts, setFailedAttempts] = useState(0); const [doubloonAmt, setDoubloonAmt] = useState(null); + const [hintChunkIdx, setHintChunkIdx] = useState(0); const editorRef = useRef(null); const monacoRef = useRef(null); const dismissTimerRef = useRef | null>(null); @@ -277,6 +291,7 @@ export default function SQLPanel({ onQueryRun }: SQLPanelProps) { setQuery(levels.find((l) => l.id === currentLevel)?.seedQuery ?? ''); setError(null); setFailedAttempts(0); + setHintChunkIdx(0); const editor = editorRef.current as { focus?: () => void } | null; if (editor?.focus) setTimeout(() => editor.focus?.(), 50); }, [currentLevel]); @@ -287,6 +302,8 @@ export default function SQLPanel({ onQueryRun }: SQLPanelProps) { const expectedShape = failedAttempts >= 3 && level ? getExpectedShape(level.solutionQuery) : null; + const hintChunks = level?.hint ? splitHint(level.hint) : []; + return (
{level?.description}

- {hasAttemptedCurrent && level?.hint && ( -
- -

- {level.hint} -

+ {hasAttemptedCurrent && hintChunks.length > 0 && ( +
+ {hintChunkIdx > 0 && ( +
+ +
+ {hintChunks.slice(0, hintChunkIdx).map((chunk, i) => ( +

0 ? 4 : 0, + }} + > + {chunk} +

+ ))} +
+
+ )} + {hintChunkIdx < hintChunks.length && ( + + )}
)} {expectedShape && ( diff --git a/src/data/levels/expert.ts b/src/data/levels/expert.ts index 201070f..a8b5ae9 100644 --- a/src/data/levels/expert.ts +++ b/src/data/levels/expert.ts @@ -1101,10 +1101,6 @@ SELECT fl.cohort_month, difficulty: 4, }, - // ============================================================ - // SESSIONIZATION / GAPS & ISLANDS (Level 69) - // ============================================================ - { id: 69, title: 'Portal Session Reconstruction (Gaps & Islands)', @@ -1200,4 +1196,145 @@ SELECT customer_id, epoch: 'Expert', difficulty: 5, }, + + // ============================================================ + // RISK SCORECARD & YoY GROWTH (Levels 70–71) + // ============================================================ + + { + id: 70, + title: 'Loan Book Risk Scorecard β€” Weighted Rate & Running Exposure', + description: `The Risk Management division needs a capital adequacy scorecard for the active loan portfolio, grouped by risk grade. Compute for each grade: +- \`risk_grade\` +- \`loan_count\` β€” number of active loans +- \`total_principal\` β€” sum of original principals +- \`wa_rate_pct\` β€” **weighted-average** interest rate (weighted by principal), ROUND to 2 dp β€” use \`ROUND(SUM(principal_amount * interest_rate) / SUM(principal_amount), 2)\` +- \`portfolio_pct\` β€” this grade's share of total active principal, ROUND to 1 dp +- \`cumulative_exposure\` β€” running cumulative principal ordered A β†’ C using \`SUM(...) OVER (ORDER BY risk_grade ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW)\` + +Use **two CTEs**: +1. **\`grade_stats\`**: \`FROM loans WHERE status = 'Active' GROUP BY risk_grade\` β€” compute loan_count, total_principal, wa_rate_pct. +2. **\`total\`**: \`SELECT SUM(total_principal) AS grand_total FROM grade_stats\` β€” scalar aggregate for the denominator. + +In the final SELECT, \`CROSS JOIN grade_stats g\` with \`total t\`, compute portfolio_pct and the window-function cumulative_exposure. \`ORDER BY risk_grade\`. + +This pattern appears in Basel III capital adequacy reports, credit risk dashboards at GIC, JPMorgan, and DBS, and is a staple of senior DS interviews at investment banks and sovereign wealth funds.`, + hint: 'CTE grade_stats: FROM loans WHERE status = \'Active\' GROUP BY risk_grade β†’ COUNT, SUM(principal_amount), ROUND(SUM(principal_amount * interest_rate)/SUM(principal_amount), 2). CTE total: SELECT SUM(total_principal) AS grand_total FROM grade_stats. Final: CROSS JOIN the two CTEs, ROUND(100.0 * g.total_principal / t.grand_total, 1) AS portfolio_pct, SUM(g.total_principal) OVER (ORDER BY g.risk_grade ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS cumulative_exposure. ORDER BY risk_grade.', + seedQuery: `WITH grade_stats AS ( + SELECT risk_grade, + COUNT(*) AS loan_count, + SUM(principal_amount) AS total_principal, + ROUND(SUM(principal_amount * ) / SUM(principal_amount), 2) AS wa_rate_pct + FROM loans + WHERE status = + GROUP BY risk_grade +), +total AS ( + SELECT SUM(total_principal) AS grand_total FROM grade_stats +) +SELECT g.risk_grade, + g.loan_count, + g.total_principal, + g.wa_rate_pct, + ROUND(100.0 * g.total_principal / t.grand_total, 1) AS portfolio_pct, + SUM(g.total_principal) OVER (ORDER BY g.risk_grade ) AS cumulative_exposure + FROM grade_stats g + CROSS JOIN total t + ORDER BY g.risk_grade`, + solutionQuery: `WITH grade_stats AS ( + SELECT risk_grade, + COUNT(*) AS loan_count, + SUM(principal_amount) AS total_principal, + ROUND(SUM(principal_amount * interest_rate) / SUM(principal_amount), 2) AS wa_rate_pct + FROM loans + WHERE status = 'Active' + GROUP BY risk_grade +), +total AS ( + SELECT SUM(total_principal) AS grand_total FROM grade_stats +) +SELECT g.risk_grade, + g.loan_count, + g.total_principal, + g.wa_rate_pct, + ROUND(100.0 * g.total_principal / t.grand_total, 1) AS portfolio_pct, + SUM(g.total_principal) OVER (ORDER BY g.risk_grade ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS cumulative_exposure + FROM grade_stats g + CROSS JOIN total t + ORDER BY g.risk_grade`, + epoch: 'Expert', + difficulty: 4, + }, + + { + id: 71, + title: 'Loan Portfolio Year-over-Year Disbursement Growth', + description: `Strategy Analytics needs a historical view of LCB's loan origination volume. For **every year** in which loans were originated, compute: +- \`loan_year\` β€” \`STRFTIME('%Y', start_date)\` +- \`loan_count\` β€” number of loans originated that year +- \`total_disbursed\` β€” total principal disbursed +- \`yoy_growth_pct\` β€” year-over-year change in total_disbursed, ROUND to 1 dp β€” **NULL for the first year** + +Include **all loans** regardless of current status (disbursement is a historical fact). + +Use **two CTEs**: +1. **\`yearly\`**: GROUP BY loan_year, aggregate loan_count and total_disbursed. +2. **\`with_lag\`**: add \`LAG(total_disbursed) OVER (ORDER BY loan_year) AS prev_disbursed\` to carry the prior year's total forward. + +Outer SELECT: +\`CASE WHEN prev_disbursed IS NULL THEN NULL ELSE ROUND(100.0 * (total_disbursed - prev_disbursed) / prev_disbursed, 1) END AS yoy_growth_pct\` + +ORDER BY loan_year. + +This is the Year-over-Year growth pattern used in strategic planning decks, investor relations reporting, and loan-book analytics at commercial banks, DBS, and GIC. The \`LAG(...) OVER\` β†’ CASE WHEN NULL combination is the canonical SQL idiom for computing period-over-period changes.`, + hint: 'CTE yearly: SELECT STRFTIME(\'%Y\', start_date) AS loan_year, COUNT(*) AS loan_count, SUM(principal_amount) AS total_disbursed FROM loans GROUP BY loan_year. CTE with_lag: SELECT *, LAG(total_disbursed) OVER (ORDER BY loan_year) AS prev_disbursed FROM yearly. Outer: CASE WHEN prev_disbursed IS NULL THEN NULL ELSE ROUND(100.0*(total_disbursed-prev_disbursed)/prev_disbursed, 1) END AS yoy_growth_pct. ORDER BY loan_year.', + seedQuery: `WITH yearly AS ( + SELECT STRFTIME('%Y', start_date) AS loan_year, + COUNT(*) AS loan_count, + SUM(principal_amount) AS total_disbursed + FROM loans + GROUP BY loan_year +), +with_lag AS ( + SELECT loan_year, + loan_count, + total_disbursed, + LAG( ) OVER (ORDER BY loan_year) AS prev_disbursed + FROM yearly +) +SELECT loan_year, + loan_count, + total_disbursed, + CASE + WHEN prev_disbursed IS THEN NULL + ELSE ROUND(100.0 * (total_disbursed - prev_disbursed) / , 1) + END AS yoy_growth_pct + FROM with_lag + ORDER BY loan_year`, + solutionQuery: `WITH yearly AS ( + SELECT STRFTIME('%Y', start_date) AS loan_year, + COUNT(*) AS loan_count, + SUM(principal_amount) AS total_disbursed + FROM loans + GROUP BY loan_year +), +with_lag AS ( + SELECT loan_year, + loan_count, + total_disbursed, + LAG(total_disbursed) OVER (ORDER BY loan_year) AS prev_disbursed + FROM yearly +) +SELECT loan_year, + loan_count, + total_disbursed, + CASE + WHEN prev_disbursed IS NULL THEN NULL + ELSE ROUND(100.0 * (total_disbursed - prev_disbursed) / prev_disbursed, 1) + END AS yoy_growth_pct + FROM with_lag + ORDER BY loan_year`, + epoch: 'Expert', + difficulty: 4, + }, ];