Read the algorithm.
Then watch it run.

An article can prove that a definition is correct. It cannot show you what the algorithm does on a Tuesday when two constituents are missing a price. These labs take the published implementations, point them at real market data, and let you inspect every input, output, and edge case the contract had to survive.

1 live lab28 algorithms running505 market sessions

Why this page exists

The gap between
knowing and shipping.

Most financial engineering fails in the space between a correct formula and a running system. These labs exist to close that space in public: same definitions, same tested code, now under real data with the failure modes visible.

01

Formulas hide their edges

A closed-form definition says nothing about missing prices, mid-window constituent changes, or sessions where a ratio is undefined. A lab has to answer all three, on screen.

02

Code you can trace

Every algorithm card names the exact import path from the published package, so the thing you are watching is the thing you would install — not a demo reimplementation.

03

Study becomes practice

Read the article for the contract, open the lab to see the contract hold, then lift the same call into your own service. The loop is the point.

Live now

The portfolio.

One lab today, covering an entire published domain end to end. More follow as their domains complete.

Live

D04 / Market Breadth and Internals

Market Breadth Lab

28 breadth algorithms · real S&P 500 data · point-in-time

Every published algorithm in Market Breadth and Internals runs here against the same point-in-time universe — the full family tree from raw advance/decline counts through the McClellan oscillators to index concentration.

Algorithms
28
Universe
S&P 500 (505 members at window start)
Window
2024-07-19 → 2026-07-24
Sessions
505 trading sessions
Point-in-time
63 constituent changes replayed
Coverage
99.8% latest coverage
Charting
TradingView lightweight charts
Open the Market Breadth LabRead the domain first
  • Every algorithm card shows the exact import statement from the published package
  • Toggle each algorithm between chart, table, sample input, and sample output
  • An “Explain” panel on every algorithm, plus a guided tour and a “Why this matters” primer
  • A 📖 link on every card that opens the canonical article on this site
  • Standalone repository links for algorithms released on their own
  • Point-in-time universe: constituent changes are replayed, not back-filled
  • Incomplete sessions are flagged rather than silently filled — the contract refuses a definitive answer
  • Light, dark, and follow-system themes

What is inside

Five families.
Every algorithm linked.

The lab is organised in the same family tree as the library. Each algorithm below has a canonical article here and a running implementation there.

D04-F01

Participation

5 algorithms

Advance/decline breadth — the raw partition of a universe into winners and losers.

D04-F02

McClellan

6 algorithms

The 19/39 EMA family, ratio-adjusted variants, and their summation indices.

D04-F03

Extremes & Trend

6 algorithms

New highs versus new lows, and the share of members above their moving averages.

D04-F04

Thrust & Pressure

6 algorithms

Volume-weighted pressure, thrust conditions, and the divergence detector.

D04-F05

Concentration

5 algorithms

How much of the index is really just a handful of names.

How to learn with it

Four passes
and it is yours.

The labs are not a showcase to admire — they are the practice half of the platform. Work the loop once per algorithm and you leave with something you can put in production.

The same call, in your own serviceimport { calculateNetAdvances } from "fintech-algorithms/market-breadth-and-internals/advance-decline-breadth/net-advances";
  1. Read the contractOpen the canonical article. Learn the definition, the input universe, the edge states, and what the output is allowed to mean.
  2. Watch it computeOpen the lab and switch that algorithm between chart, table, sample input, and sample output. The numbers you see came from the tested implementation.
  3. Break the happy pathFind the flagged sessions — incomplete coverage, undefined ratios, replayed constituent changes — and check how the contract refused to guess.
  4. Install and rebuildCopy the import, call it on your own data, and compare against the lab. If your numbers differ, one of you has an assumption written down and the other does not.

What comes next

A lab per finished domain.

A domain earns a lab once its algorithms are complete and tested. That is why there is one here and not forty-eight.

D04 · LiveMarket Breadth Lab28 algorithms running
In buildNext labs follow the atlasSee which domains are complete
Open sourceEvery algorithm is publishedBrowse the index on GitHub

The labs are engineering demonstrations of algorithm behaviour on historical data. They are not investment advice, trading signals, or a forecast of future prices.