RESEARCH
We publish the methods behind our models — including a sovereign index in a market where we sell nothing, so the work can be judged on its own terms. Every figure is reproducible from official sources, and every open limitation is named.
22 sovereigns scored on their capacity to keep financing themselves under stress. Debt-to-GDP carries 6% of the composite; debt servicing is the test.
MethodologyFive principles governing every model we build or publish, each stated with the artifact that evidences it — versioning, frozen parameters, reproducibility, published diagnostics, disclosed limitations.
Case studyOne factor model, one day, six covariance versions. Two failed backtests, the product flaw they exposed, and the problem the successful test then revealed.
RegisterEvery correction to a published figure, dated and described. Empty is not the same as absent — the register exists before it is needed.
Most risk vendors describe their models in marketing terms and keep the methods private. That leaves a buyer's model validation team — the function that actually decides whether a model can be used — with nothing to inspect until after a contract is signed.
We take the opposite approach. The methods are published, the diagnostics are published including the ones that disagree with us, and the limitations are stated on the artifact rather than buried. TSRI-C exists partly to make that checkable in a setting where we have no commercial stake in the answer.
We email when a new release or research note is published — nothing else, no fixed schedule, and separately from any product communication. Write to research@tensilerisk.com to be added, or to request a methodology paper, an audit file, or a walkthrough of how a model was built.