Research · Papers, background notes, scoreboard
The method, published.
Working papers on whether stated confidence survives contact with recorded outcomes. Out-of-sample scoring, provenance on every figure, code published with every paper.
Papers
- Analysis·Market risk
Tail Risk Backtesting: When the Test Improves and the Model Does Not
I changed 3 things about my own backtest at once, the expected-shortfall diagnostic improved, and my changelog credited the model. Running the experiment that separates them shows the reference distribution did about 11 times the work. Re-scored against 6 nulls with the conditional ES statistic held fixed, the same 5 models go from 0 of 10 model-and-level cells passing to 10 of 10.
- Analysis·Manufacturing
Tool Change Intervals, Audited Against Measured Wear
A 16-insert audit of a fixed change interval against wear that was actually measured: the rule misses by about a quarter of the change point, the range a shop would quote around it covers 68% rather than the promised 90%, and pulling early enough to be safe is paid for in tool life.
- Analysis·Energy
German Grid Forecasts: Where the Uncertainty Bands Fail
A 13-month audit of German day-ahead forecasts against what happened: systematic midday bias, uncertainty bands that fail out-of-sample (including conformal ones), and what that costs: ≈€45,500 per 100 MW of avoidable imbalance exposure at reBAP, concentrated in exactly the hours where failing is expensive.
Background notes
Tutorials on the ideas the papers lean on. Each is versioned, and a paper pins the version it was written against. The background index lists them with their formats.
Calibration
A plain-language tutorial: what it means for a stated probability to be honest, how to check when events are common, and the ladder of remedies when they are rare.
Clustered Data
A plain-language tutorial: what changes when your observations arrive in groups rather than one at a time, why the sample is smaller than it looks, and why the mistake this causes is invisible to the check most people run.
Contagion
A plain-language tutorial: the two reasons failures arrive together (shared exposure and true propagation), why real disasters are usually both, and why the fix for one makes the other worse.
Extreme Value Theory
A plain-language tutorial: how to estimate losses worse than anything you have ever seen, what gives you the right to try, and what the answer costs. One dataset, carried from first look to final estimate.