⚔️ Toram Refining Guide

Verified against confirmed image data. LUK column = 0 TEC contribution (LUK does not affect success rate, only degrade). Rec columns show best char per slot count based on expected ore cost.

Each pair of columns (TEC / LUK) is for one slot count. Outlined border = winner for that slot. ∞ = divergent: expected levels lost per fail exceeds gain rate; pity required to stabilise.

Each cell is the expected ore cost for that single transition only (e.g. the +B→+A row is the cost to go from +B to +A, not from +0 to +A). Values are not cumulative.

How the value is calculated: each attempt costs 1 ore. On success (probability p) you advance. On fail, you may degrade and lose levels, which costs extra ore to recover. The formula is:
  E = (1/p) / (1 − (1−p) × P(degrade) × avg_levels_lost)
But this is only an approximation — the tables use the exact Markov chain, solving all 14 levels simultaneously so that the true recovery cost of each lost level is accounted for.

Note: these values assume the same char is used throughout, including recovery from degrades. See Table 2c for the optimal two-char strategy where TEC handles recovery at mid-levels.

Same structure as Table 2 — each cell is the per-transition cost only, not cumulative. Outlined border = winner for that slot.

Pity adds a flat % bonus to success rate (+1% per 150 points). At high levels where both TEC and LUK are floored at 1%, pity is the only way to reduce cost. Notice how the +A→+S and +B→+A rows shrink dramatically as pity increases, while low-level rows (where TEC already has high success) barely change.

Same format as Table 2b — each cell is the per-transition cost only, not cumulative. Outlined border = winner for that slot.

What makes this different: Table 2b assumes the same char handles recovery after a degrade. This table uses the locally optimal char for each recovery level. For example, if LUK is attacking +B→+A and degrades to +C, the cost to recover +C→+B is calculated using TEC (which has 11% at that step) rather than LUK (which has 1%). Each level's recovery cost is set to whichever char — TEC or LUK — has the lower expected cost there, solved simultaneously across all levels.

Same format as Table 2c — each cell is the per-transition cost only, not cumulative. Outlined border = winner for that slot.

The LUK column adds an Anti-Degradation item (75% extra prevention), reducing the LUK degrade chance from 12.5% to 3.125%. Like Table 2c, recovery after a degrade uses the locally optimal char at each recovery level — so if LUK+AD degrades and TEC is cheaper for the recovery climb, TEC's cost is used.

Same format as 2c and 2d — each cell is the per-transition cost only, not cumulative. Outlined border = winner for that slot.

Models the practical strategy where you only use Anti-Deg items for the two most expensive steps (+B→+A and +A→+S), saving anti-deg costs at lower levels. Below +B, plain LUK is used. Optimal cross-char recovery applies throughout.

This table shows only the +A→+S transition cost (one step), broken down by pity tier. It is equivalent to the +A→+S row in Table 2b, but presented as a standalone summary. Values are not cumulative from +0.

At +A/+S both TEC and LUK have the same success rate (1% + pity), so success rate is not the differentiator here — only the degrade chance is. Outlined border = winner per slot. ∞ = practically divergent (cost >5000 ore). The exact model shows that without pity, both chars diverge — failing at +A drops you to +B, which is also ~1% with the same degrade risk, trapping you in a loop. Pity is required to escape this.

Use this to pick ore. Higher grade = better success rate at the same refine level, at higher cost per ore.