Idiot Finance

The problem

Idiot Finance lets a learner explore preset and custom ETF portfolios with virtual money. The application makes allocation, review, transaction state, and historical research understandable while keeping calculations and failure handling explicit.

An investing simulation loses credibility if duplicate clicks change the balance twice, edited weights rewrite an existing position, or missing prices quietly become a convincing chart. The interface and financial state need to agree.

Preset and custom portfolios share a simulation engine. A browser-local practice account and a read-only historical research path have different purposes and calculation rules.

How it works

1. 01

   Preset / custom definition

2. 02

   Review + state revisions

3. 03

   Read-only market / synthetic adapter

4. 04

   Precise simulation engine

5. 05

   Versioned browser state + recovery

From input to output

1. 01Define a preset or a valid custom allocation, with weights and versions made explicit.
2. 02Capture a reviewable virtual transaction before fetching eligible market observations.
3. 03Recheck account and definition revisions, calculate the complete basket, and persist one account update.
4. 04Keep historical price simulations read-only, expose missing coverage, and validate backups before replacing saved state.

Data sources

• Deterministic synthetic prices in the isolated offline demonstration
• Read-only Alpaca market bars and calendar endpoints in the configured market-data path
• Versioned portfolio definitions and browser-local simulation records

Technical approach

• Integer cents, scaled prices and units, BigInt arithmetic, and explicit rounding for virtual transactions
• Immutable definition snapshots and request IDs for reproducible account actions
• Web Locks and revision checks for browser-local coordination
• Separate historical price-only calculations with explicit source coverage
• Validated migration and backup restore with a fresh storage generation

What you can inspect

• An existing position keeps the definition used when it was created; editing a draft creates a separate version.
• An accepted duplicate command returns its prior receipt rather than creating a second virtual transaction.
• Incomplete historical coverage withholds unsupported full-period metrics, and research does not spend practice cash.

The deliverable is an investing simulation whose account changes and chart limitations can be inspected. The engineering evidence is in the arithmetic, state transitions, and rejected operations.

Where this is useful

• Shows how financial domain rules become precise application behavior.
• The same state and validation patterns are relevant to internal tools that must preserve a consistent record through retries and recovery.

Checks & engineering decisions

Verification evidence

• Reviewed unit-test sources exercise allocation rounding, no-money-creation rules, command identity, definition snapshots, and backup validation.
• Existing release artifacts separately record unit/API checks, browser scenarios, synthetic fixtures, and limited provider observations. Historical results are not represented as newly executed portfolio tests.
• The displayed account and prices are synthetic. They demonstrate UI and state behavior, not investment performance.

Key decisions

• Keep virtual execution arithmetic separate from read-only historical research calculations.
• Commit the whole virtual basket or reject it; a partial account update is not a completed transaction.
• Pin a position's definition so later draft edits cannot rewrite its history.
• Keep stale and missing market coverage visible, and withhold metrics that lack sufficient data.
• Validate saved state before restore and use a new storage generation to reject stale in-flight actions.

type Command = {
  requestId: string;
  roundId: string;
  expectedRevision: number;
  portfolioId: string;
  type: "buy" | "sell";
  amountCents: number;
  definitionVersion: number;
  positionId: string | null;
  draftId?: string;
  draftRevision?: number;
};

Role & status

Simulation application · Case-study demonstration Source review: 2026-10-14.

Limitations

• This is a virtual-investing application; it does not place real securities orders or provide personalized investment advice.
• Guest account state is local to the browser and is not secure custody or a multi-user account service.
• Historical research is price-only; corporate-action support is bounded and does not establish complete total-return modeling.
• The portfolio uses an illustrative synthetic-data view. No public product endpoint or public source repository is linked.

Explore the evidence

Source review: portfolio engine and store, architecture, research backtest, backup policy, and release verification artifacts. Existing release evidence is historical; no market-data calls or project tests were run for the portfolio source review.