BONDSTATS
RESEARCH INSTRUMENT / CF-01METHODS · V3

BondStats / Experimental finance

The Counterfactual Ledger.

A record of what changes when one intervention is tested against the same simulated financial conditions. Not a prediction. A comparison you can reproduce.

RESEARCH EDITION CONTROL / INTERVENTION / PROVENANCE

ANALYTICAL INSTRUMENT · PAIRED PATH METHODNo external requests · Deterministic rerunsIndependent research · BondStats Ltd
The counterfactual desk

Hold the world constant.
Change the decision.

Select a financial mechanism, define a baseline and an alternative, then examine the measurable difference under exactly matched synthetic conditions. Every result is a model outcome, not evidence about a live institution.

EXPERIMENT RECORD / CONTROLLED COUNTERFACTUALPREPARING
Research protocol / Control versus intervention

Synchronous execution and market depth

Does staggering correlated orders reduce execution pressure?

SYNTHETIC · NOT EMPIRICAL
Control: mean impact—Unmodified baseline policy
Intervention: mean impact—Same stochastic conditions
Relative change—Reduction relative to baseline
MODEL TEST PENDINGIllustrative model · no live market feed
Paired path / 24 matched simulated intervalsSynthetic impact proxy
Run the experiment to see paired paths
Unmodified controlTest intervention
Pre-registered claim

Staggered execution reduces mean price impact by at least 10%.

Run an experiment to evaluate the claim within the specified synthetic model. A model-level outcome does not constitute external validation.

Intervention trade-off—
Interpretation will appear here together with the control/intervention trade-off.
Robustness / One-factor stress
DepthControlTreatmentReduction
Run to examine five parameter settings.

Only the first protocol parameter varies. All other assumptions and the random seed are held constant.

PROVENANCE / SESSION RECORD

Recent matched runs.

0 in this session

Stored only in this open page. Select a record to restore its inputs; export a JSON packet for durable reproducibility.

Research packets contain the current model inputs and fingerprint.

Ready to run. The ledger uses no external API calls.

Train the trainer / Read the experiment

What would actually falsify the claim?

A useful experiment must be able to fail its own proposed explanation. Inspect the hypothesis threshold, the outcome and the cost of intervention.

RESEARCH EXERCISE

Before adjusting the intervention, predict how an aggressive control might change both the primary outcome and its operational trade-off.

Choose an interpretation to reveal the methodological point.

Transparent method / Current protocol

From assumption to result

  1. Formulate a falsifiable claim.
  2. Hold random inputs constant between policies.
  3. Compare outcomes and opportunity costs.
  4. Change assumptions before generalising.
Formula and operational definitions

The method for the selected protocol appears here.

What is not demonstrated?

No empirical calibration, predictive performance or institutional certification is implied.

Scientific integrity

A model result is not
a real-world finding.

The ledger is an independent methods environment. It does not describe a real institution or reproduce a published scientific result. Each synthetic comparison is designed to make assumptions inspectable and falsification possible.

Reproducibility

Same seed. Same result.

Matched trials use a deterministic pseudo-random generator. The downloaded research packet includes the selected protocol, all parameters, the seed, model version, reported metrics and sensitivity series.

Scientific honesty

Simulations are not observations.

The models are deliberately simplified and uncalibrated. A hypothesis can be supported *within a model* while remaining untested against real markets, banking records or operational incidents.

Research connection

Built to extend AFP.

The Autonomous Finance Project examines systemic questions about automated finance. This independent laboratory records controlled intervention comparisons and repeatable falsification exercises, rather than duplicating AFP’s models.

Documentation and data rights

No external data, research paper, banking telemetry or API is embedded in these experiments. All formulas and generated paths belong to a local synthetic model. Findings do not establish observed causal effects, regulatory compliance, institutional safety or investment suitability. General site terms: Legal Notice. Open the companion Financial Agent Mandate Protocol to investigate the rules that might constrain an autonomous financial agent.