5,000+ peer-reviewed papers · now in Claude

5,000 academic strategies, every one runnable

Peer-reviewed papers shipped as runnable Python with clean data. Backtest, fork, and deploy from a single workbench.

Start free →

5,000+

peer-reviewed papers

+147 this quarter

1.04 TB

cleaned market data

updated monthly

60+

video lessons

18 hr · zero → live

Value and Size Effect: Now You See It, Now You Don't

February 2002 · Annaert, Van Holle, Crombez, Spinel
Yearly
Equities
Results from the backtest

Backtest period

1990 – 2026


Sharpe ratio

1.51


Annual return

9.09%


Annual volatility

5.87%


Maximum drawdown

9.81%


Net worth · log scalepublished19902000201020202026
Code
import pwb_toolbox.backtest as pwb_bt
import pwb_toolbox.datasets as pwb_ds


class ValueSizeEurope(pwb_bt.BaseStrategy):
    """Annual Size and Book-to-Market sorts, long/short deciles."""

    params = (
        ("quantiles", 10),
        ("min_universe", 60),
        ("weighting", "equal"),
        ("gross_exposure", 0.9),
    )

    def __init__(self):
        super().__init__()
        self.last_rebalanced_year = None

    def next(self):
        # Form at July (year t), hold until next July (year t+1).
        dt = self.datas[0].datetime.date(0)
        if dt.month == 7 and self.last_rebalanced_year != dt.year:
            self.last_rebalanced_year = dt.year
            self.rebalance(dt)
The method

Research that survives its own trial count.

Finding a strategy that backtests well is the easy half. The hard half is telling an edge apart from the residue of having looked a hundred times. This is the AI loop the platform is built around — and the one it enforces.

1

Search the prior work

Search the mechanism, not the name you gave it. In-sample against out-of-sample Sharpe shows whether the edge survived publication.

2

Replicate, then vary

Match the published figures first. Then one change at a time, each variant declaring its parent.

3

Reject on the numbers

The deflated Sharpe prices in how many times you looked. Below 0.50, reject rather than improve.

Then plug your favorite LLM into the MCP server and let AI do the work you cannot: stay honest at scale. It runs the hypothesis, counts every trial, registers the strategy, and starts the out-of-sample clock. Your only remaining job is to wait. Read the full procedure.

The Stack

Everything you need, nothing you don't.

Three primitives that compose into anything from your first momentum backtest to a multi-strategy live book.

STRATEGIES

Executable papers

Every paper paired with clean Python, the dataset it needs, and a one-click backtest.

  • 5,000+ papers, all peer-reviewed
  • Fork to your workspace, modify freely
  • Deploy to live in one click
Open strategies →
DATASETS

Clean market data

Equities, crypto, futures, FX. Adjusted, deduped, gap-filled. Optimized for backtesting.

  • 1.04 TB, growing daily
  • Parquet, queryable from Python
  • Bias-free survivorship-adjusted
Open datasets →
COURSE

Algo-trading curriculum

Sixty-plus lessons. Videos and notebooks that take you from zero to a live book.

  • 18 hours of video, fully indexed
  • Notebooks with real datasets
  • Pairs trading · momentum · arb
Open course →

Works where you work

The whole catalog, inside Claude.

Add one connector and your assistant researches with you — against the live catalog, not against what a model happens to remember. Works with Claude, Claude Code, Cursor and any other MCP client.

“Find me a momentum strategy with a Sharpe above 1 that trades commodities.”


What it can reach
  • 5,000+ papers, ranked and searchable
  • Every strategy, with its backtest assumptions
  • The dataset catalog and the course
From the community

Built with quants, for quants.

I taught myself Python after running risk for a US bulge bracket for 25 years. Still need a community to develop industrial-grade code. This is it.

AH

Ali Hobballah

Former Head of Trading

The work is exceptional. It has accelerated my learning in financial markets considerably. I encourage everyone serious about quant to support and follow it.

CF

Capital Flows

Macro Research, Substack

Well-researched, insightful, and clear to follow. The combination of paper plus runnable code is what makes it stick. The Discord is a real bonus.

JG

James Gowen, Jr.

Independent Trader

Pricing

From hobbyist to hedge-fund.

One subscription, full archive. No per-seat. No per-strategy. Cancel anytime.

Reader

Premium blog & weekly newsletter

$10/mo
Backtester

Full code, all data, deploy live

$50/mo · $450/yr
Institution

Seats, SLA, dedicated support

Custom

Start with a paper, not a blank file.

Free account, no card. Browse the catalog, run a backtest, and see whether the edge survived its own publication.