Alignment-tendency distillation (ours, not the SRD's)

The SRD gives each race an alignment prose tendency. We distill it to a (moral, ethic) bias pair on the engine’s axes — moral +1 good / −1 evil, ethic +1 lawful / −1 chaotic, each in −1..+1, where the nine alignment names derive at ±1/3 thresholds. The bias is a generation-time mean shift, not a cap.

Magnitude scale:

Bias	Reading of the prose
0.0	no tendency stated
±0.2	hedged pull (“not strongly inclined toward…”)
±0.3	clear lean (“more often X than not”, “tend toward X”, “many end up there”)
±0.4	strong lean (“most often X”, “share the X bent”, “inherit a tendency toward X”)
±0.5	emphatic majority (“most are X”, “lean strongly toward X”) — pushes the mean past the ±1/3 label threshold

Per race (SRD prose quoted in each races/*.srd.ron above the value):

Race	moral	ethic	Reasoning
Dwarf	+0.3	+0.5	“most dwarves are lawful”; “tend toward good as well”
Elf	+0.3	−0.5	“lean strongly toward… chaos”; “more often good than not”
Halfling	+0.5	+0.5	“most halflings are lawful good”
Human	0.0	0.0	“no particular alignment”
Dragonborn	+0.4	0.0	“most dragonborn are good”, but explicitly polarized (“tend to extremes”) — the bimodal spread can’t live in a scalar mean
Gnome	+0.4	0.0	“most often good”; law/chaos split by profession
Half-elf	0.0	−0.4	“share the chaotic bent”; no moral tendency stated
Half-orc	−0.2	−0.4	“tendency toward chaos”; “not strongly inclined toward good”; the “usually evil” clause is conditioned on orc upbringing
Tiefling	−0.3	−0.4	“many of them end up [evil]” though not innately; “inclines many… toward a chaotic alignment”