#Tuning pass 1 — LandingChancePPM sweep and Pareto table

Outcome: no LandingChancePPM satisfies both acceptance targets. STOPPED before touching a second parameter family, as instructed.

Approved scope of this pass (user, 2026-08-09, DRAFT MATH gate): sweep only LandingChancePPM. Base reels, trigger weights, base paytable, jackpot values, coin values, prize-type weights and the RNG mapping are untouched, and remain untouched.

#Manifest

#Why conditional sampling rather than whole rounds

The acceptance band for full-grid frequency is 0.05%–0.25% of triggered features. At a 1-in-155 trigger, resolving a 0.1% conditional rate from whole rounds would need tens of millions of rounds. Conditioning on the trigger spends every sample inside the feature, which is where LandingChancePPM acts. Finalists would still be confirmed on ≥2,000,000 whole rounds — there are no finalists.

#Fixed points (EXACT, and structurally independent of LandingChancePPM)

LandingChancePPM is read only inside the mechanic's step loop, after a trigger, so it cannot move either of these. Both were re-enumerated during the sweep and are unchanged:

#Derived acceptance targets

required feature RTP  = 96.0% − 75.1160%           = 20.8840%
required cond mean    = 0.208840 / 0.00643486      = 32.4545×
total RTP band        = [95.7%, 96.3%]
full-grid target      = 0.10% of triggers, band [0.05%, 0.25%]

#Coarse sweep

#Refinement

Max award is exactly 500.0000× at every candidate, and never above it. Containment holds throughout; reachability holds because a full grid still occurs.

#Pareto table — the two targets are disjoint

Both acceptance criteria are monotone in LandingChancePPM, and they move in opposite directions. Their feasible intervals do not overlap.

full-grid window :  37,000 ─────────── 51,200
RTP window       :                                  64,700 ── 66,800
                                       └── gap ≈ 13,500 PPM ──┘

Consequences at each frontier:

#Why they conflict

The conditional mean is, to a very good approximation, locked cells × E[landed prize], plus a small grand term. With the approved prize tables:

E[landed prize] = 3065 / 1026 = 2.9873×
cells needed for a 32.4545× mean = 32.4545 / 2.9873 = 10.86 of 15

A grid that routinely reaches 10.9 of 15 cells cannot make a full grid rare. That is the whole conflict in one sentence. The measurements bear it out: 10.73 cells at 90,000 PPM comes with a 3.45% full-grid rate, and the full-grid band tolerates at most ≈8.6 cells.

Equivalently, at the full-grid-compliant 45,000 PPM the feature locks 8.4949 cells, so reaching the required 32.4545× would need

E[prize] = 32.4545 / 8.4949 = 3.8205×      (+27.9% on the current 2.9873×)

Note also that the ordinary settlement mean at 45,000 PPM is 25.348×, already below the required 32.4545× before full grids are considered at all. No landing chance fixes that, because landing chance buys cells and the shortfall is in value per cell.

#Options for the next decision (not acted on)

If Option A is chosen, two sub-routes reach E[prize] = 3.8205×:

• A1 — scale regular coin values. E[regular] must rise from 2.64× to 3.4950× (+32.4%), leaving prize-type weights alone.
• A2 — reweight the prize-type pool toward Mini/Minor/Major, leaving coin values alone. The jackpot term must rise from 425 to ≈1,280 weight-value units, which roughly triples Mini/Minor/ Major weights and materially changes how the feature feels rather than only what it pays.

A1 and A2 are not equivalent in player experience even where they match on RTP, so this is a design choice rather than an arithmetic one.

#Discipline notes

• One parameter family moved. Nothing else was touched, and no config file was modified — the sweep overrode LandingChancePPM in memory only.
• Common random numbers throughout, so differences between rows are the parameter and not noise.
• No finalist run was performed, because no candidate passed screening. Independent confirmation seeds and the ≥2,000,000-round confirmation are still owed once a candidate exists.
• No new Draft revision was created and draft-v0.1 is unchanged, because no landing probability was selected.
• Every conditional figure here is ESTIMATED. Base RTP and trigger probability are exact.
• DRAFT / NOT FOR CERTIFICATION.