Model royalbid-pam-0.2-mcf
Plusgrade-Informed Upgrade Auction Model
Plusgrade’s actual RoyalUp allocation algorithm is proprietary and not public. RoyalBid simulates the economic behavior of a revenue-optimized upgrade marketplace using publicly documented concepts — premium inventory optimization, competing offers, dynamic pricing, reserve pricing, released inventory and downstream upgrades.
- 01
Hidden Inventory Estimator
Publicly visible cabins are thinned by a binomial allocatable share, plus a cancellation/reappearance draw from unlisted cabins. A sold-out category is not treated as 0%.
- 02
Retail Displacement
Each allocatable cabin may be held for normal retail sale. That probability falls as sailing approaches.
- 03
Competitor Bid Estimator
For every origin → target edge, the number of rival offers ~ NegativeBinomial(μ, k). Normalized bids ~ Beta(α, β), shifted by demand and days-to-sailing bucket.
- 04
Dynamic Reserve Estimator
minimum + inventory protection + demand premium + retail opportunity cost − spoilage discount, with a shared market shock per simulation. Never presented as the real internal reserve.
- 05
Released Cabin Value & Cut Line
From the optimal allocation, residual shortest paths give the exact total bid your offer must beat for each target — including the value your released cabin creates downstream.
- 06
Revenue Optimization Solver
Each scenario is solved exactly as a min-cost flow of free cabins across the whole upgrade graph: one upgrade per reservation, inventory incl. released cabins, reserve thresholds. Every accepted upgrade releases its origin cabin, so cascades (e.g. $1,500 + $800 + $500) are valued automatically.
- 07
Monte Carlo Simulation
10,000 scenarios by default. Each records the per-person cut line you would have needed. The bid curve is the empirical distribution of those cut lines.
- 08
Sweet Spot
Maximizes P(win) × (your upgrade value − total bid), then Bargain, Win-Focused and Maximum Rational bids are derived from marginal cost per probability point.
Calibration status
Passengers report real RoyalUp wins and losses after each sailing. Each result is scored against what the simulator predicted at that bid; the average miss shifts future win chances (shrunk toward zero until about 30 results exist), uncertainty bands widen or narrow with the error score, and confidence rises to MODERATE at 30 results and HIGH at 100 well-calibrated results. Probability ranges are deliberately wide. Once real win/loss reports exist, parameters are updated by Bayesian updating and checked with calibration curves and backtests.