Quantitative evidence instrument / public access

Don’t trust the line.
Test its neighborhood.

EPSILON turns one market claim into a field of nearby experiments. It reveals whether the conclusion survives when cost, slippage, window, or universe changes.

Run an evidence field →See the method

Controlled synthetic marketsNo login requiredNo real capitalOpen source

Live product language / 00

The result is not the object. The evidence field is.

Illustrative controlled experiment

Evidence plate / epsilon.evidence.v1

Recent momentum remains directionally positive under nearby execution and sampling assumptions.

Direction preserved

4/4

Falsified if

The return direction reverses when one plausible assumption changes while the others remain fixed.

Perturbation field / normalized equity

Nearby assumptions, visible consequences.

ε = one changed input per trace

Baseline · +6.20%Fee ×5 · +4.32%Slippage ×5 · +2.71%Window +30d · +5.18%Narrow universe · +5.83%

Every trace is normalized to its own starting equity. Colour denotes a documented assumption change—not confidence, quality, or predicted performance.

Exact baseline and perturbation outcomes
Run	Changed input	Return	Sharpe	Drawdown
Baseline	None	6.20%	0.910	-3.80%
Fee ×5	0.0001 → 0.0005	4.32%	0.680	-4.30%
Slippage ×5	$0.01 → $0.05	2.71%	0.430	-5.20%
Window +30d	2024-01-02 → 2024-02-01	5.18%	0.790	-3.50%
Narrow universe	AAPL,MSFT,GOOGL → AAPL,MSFT	5.83%	0.860	-3.90%

Method / ε

A mathematical idea made into a research behavior.

In analysis, ε describes a nearby change: an error, tolerance, perturbation, or residual. EPSILON asks the practical version—what happens to your conclusion in that neighborhood?

Observe

Select a market subject and write the claim before seeing the answer.

→02

Define

Attach a rejection rule, universe, window, and execution assumptions.

→03

Perturb

Change one nearby assumption at a time and compute the consequences.

→04

Challenge

Publish the evidence object so another person can inspect and fork it.

→

Evidence discipline / 01

Monochrome until challenged.

EPSILON does not use colour to decorate performance. Colour appears only when an assumption changes, so every visual difference has a research meaning.

ε is operational

Every coloured trace corresponds to one exact parameter change.

Failure remains visible

A reversal, failed run, or unknown source is preserved instead of tuned away.

Evidence can travel

The claim, conditions, outcomes, and provenance stay together in one portable object.

From private result to public reasoning

Publish something other people can disagree with precisely.

An Evidence Plate packages the original claim, exact changes, result traces, limitations, and a Fork link. The social unit is not a screenshot of returns—it is a challenge another person can reproduce.

Create your first plate →Inspect public impact

Begin / no account required

Make the assumptions move before the conclusion does.

Open EPSILON →

Don’t trust the line.Test its neighborhood.