Hacken vs Kalmar: Goal-Model Assumptions and Checks
Review the Hacken–Kalmar archive, including an incomplete 1X2 total and why a Poisson illustration must not be presented as verified software output.
The earlier Hacken–Kalmar article combined software claims with a separate calculation from assumed goal rates. These are different kinds of evidence and should be labelled accordingly.
The archived inputs and result figures
The text quoted goal rates of 3.38 for Hacken and 1.26 for Kalmar, totalling 4.64. It also quoted result estimates of 76.60%, 12.24% and 10.32%. Those result estimates sum to 99.16%, leaving a 0.84-point gap.
The article did not preserve a dated output that resolves that gap. It should not be silently described as ordinary two-decimal rounding.
A reproducible illustration is not a verified forecast
Under an assumed independent Poisson model, a score probability is calculated by multiplying the home-goals probability by the away-goals probability. This is a mathematical illustration using the quoted rates; it does not establish the method used by the paid software or verify the rates themselves.
The previous article's exact-score table was a secondary calculation, not read from the display. It should therefore not be advertised as product output. Nor are these rates shot-based expected-goals statistics merely because they represent expected counts.
Related probabilities need to agree
The archived team-scoring values were 96.58% and 72.39%, while BTTS was quoted at 67.70%. For the same two events, the lower bound for BTTS is 68.97%. The quoted value is below it, revealing a further inconsistency.
The original text also described a probability near 50% as meaningless. That is not correct: it communicates uncertainty between two outcomes. What is missing is evidence about the estimate's accuracy.
What this page now establishes
It establishes the arithmetic discrepancies in the previously published text. It does not establish a current Hacken–Kalmar forecast, an actual result, product accuracy or a predictive advantage.
A valid case study needs the original workbook output, sample window, generation time and market definitions. See the store for exact product demonstrations and the separate website methodology for how free forecasts are produced.
Correction, 19 September 2026: Removed unsupported fixture context and software-method claims; identified incomplete and incompatible probabilities and distinguished secondary calculations from source output.
Revised with AI assistance using the available article evidence. Predictions are estimates, not guarantees. 18+. Support for gambling-related harm.