Method: checks, established data and the atom bridge

The value of a theory lies in agreeing with nature when compared with it. The model is tested by five rules.

1. Checks first. Is the model internally consistent? Do the numbers agree, does incoming energy equal outgoing energy, does every row of the working table follow the rule? Possible readings of a point that looks inconsistent are written down as candidates.
2. Established data only. The model is tested only against directly measured, law-level data: numbers measured by several independent methods, with small errors, unchanged for years. Numbers inferred within the standard cosmological model (such as the dark-energy and dark-matter fractions) are kept separate; comparisons with them are for information only.
3. Rule before calculation. How the model gives a number is written down before the calculation; then come the calculation and the comparison with measurement. A result that matches exactly or within the measurement error counts as matching and becomes an anchor of the model. The chance probability (how many other expressions built from the model's numbers come equally close?) is reported alongside as information. A measurement used to set the model's clock is an input; the tests are the numbers the model gives afterwards.
4. Multiple candidates, progress by elimination. If the algorithm can answer a question in more than one way, all readings are written down as candidates before calculation. Candidates are eliminated by measurement and by whether they agree with structures developed later; an eliminated candidate stays on record with its reason. What fails is that reading; as the model as a whole advances, the candidates decrease.
5. Atom bridge: prediction and proof at every step. The atom is the centre of progress: the work proceeds step by step inwards (nucleus, quarks) and outwards (molecules, crystals, living structures, celestial bodies). At every step the prediction of the previous step is tested by measurement and the prediction for the next step is written before calculation; no open point is left without an owner. As anchors accumulate, questions that cannot be answered today are read again (atom bridge).

Verdict colours

• Consistent: an internal check of the model is correct.
• Matches: the model's number equals an established measurement exactly or within the error; the mapping becomes an anchor.
• Close: the difference is between 2σ and 5σ.
• Eliminated: the model's number is clearly far from the measurement; that reading is eliminated and other readings are tested.
• For information: a comparison with an inference of the standard model; no verdict is given.
• Silent: the model does not yet give a number on this point.