Kwant Klubben
SDU · Odense Paper only · no pooled money

We have fun
with fat tails.

A student quant-finance club that treats backtests like crime scenes. We pre-register hypotheses, log every trial, and KILL the strategies that can't survive an honest validation battery — because a clean negative result is still a finished project.

Join the klub → Read the manifesto
Empirical return distribution · NVO daily
Return distribution with fat tails
−3σthe tails the Gaussian swears can't happen+3σ
House rules

How we keep ourselves honest.

01
SPEC before code
Pre-registering the hypothesis is the cheapest anti-overfitting device there is.
02
Log every trial
No entry in trials.csv? Then the result doesn’t exist. The honest trial count is what makes the Deflated Sharpe mean anything.
03
Validate before you believe
Walk-forward, permutation test, Monte Carlo, DSR. Most things die here. That’s the point.
04
Failures stay in the repo
A strategy that honestly failed is just as valuable a library entry as one that survived.
Strategies spec'd 0 across FX, equities, options
Honestly killed 0 −64% survival rate by design
Survived validation 0 paper-trading
Members 0 IBKR paper · USD 100k each
The library

Every project, honest outcome and all.

The repo is the portfolio. SPEC → code → trials.csv → honesty card → write-up.

NVO / LLY pairs

Live paper

Market-neutral pharma duopoly spread with a pre-registered z-band and deflated-Sharpe validation.

equitiesstat-arbcointegration
Net Sharpe0.71

Options pricer + Greeks

Survived

Closed-form BS pricer cross-checked against Monte Carlo, then run on a live NVO/LLY chain to map the skew the model can’t explain.

optionsblack-scholesmonte-carlo
IV error1.2%

XS-momentum (12-1)

Killed

Construction valid (UMD ρ=0.88) but the gross premium is gone. The anomaly decayed, not our data. Two kill conditions hit.

equitiesfactor
Perm p0.87

TSMOM multi-asset

Killed

Net SR 0.52, DSR 0.93 — but permutation p just missed the pre-registered 0.10 bar. Timing added no Sharpe over the static book. Killed per spec.

multi-assettrendvol-target
Perm p0.112

Stylized facts & fractals

Survived

Return distributions vs normal across timescales: QQ-plots, tail indices, and the shuffle test that proves clustering is real and fat tails are forever.

researchhursthill
Hurst0.58

Markov regime model

In review

Vol/trend state classification on weekly returns, conditioning an existing strategy and permutation-testing whether it honestly improves anything.

regimeHMM
States3

Bring a hypothesis.
We’ll help you kill it.

No finance background required — just curiosity and a tolerance for being wrong in public. Apply below, and if it's a fit we'll get you into the first meeting.

Apply to join
Setup check
$ git clone kwantklubben
$ pip install -r requirements.txt
$ python -m honesty.test_honesty
All sanity checks passed