Announcing RivalRyze Points v2

Smarter scoring, fairer leaderboards, cross-platform integrity

TL;DR
We’ve rebuilt our points system from the ground up. The goal is simple: reward genuine difficulty and effort while staying consistent across Steam and PlayStation. Points v2 uses platform rarity, game shape, release timing, set size, metadata priors, and cross-platform sanity checks to produce totals that feel right — even when a game’s raw rarity data is noisy.

Why a new system?

Firstly, because the current system sucks. Designing a points model with a relatively small user base is tricky. Most “global difficulty” signals (like earned rates and trophy tiers) live with the platforms, not with us—so v2 emphasizes robust platform data, curve-based weighting, and defensive anti-cheat heuristics. As RivalRyze grows, we’ll layer in truly community-driven learning (per-game calibrations, personal skill normalization, etc.).

How points are calculated

• Difficulty is anchored to how rare achievements actually are globally, with a curve that grows steeply for truly rare unlocks.
• We shape totals using real-world context: small sets, very new releases, massive player counts, or shallow rarity tails are treated differently.
• Cross-platform “auto-pop” detection prevents duplicate trophy/achievement pops from inflating your points.
• Leaderboards got a front-end refresh with daily/weekly/monthly/all-time views, per-platform and overall tracks, and cleaner ranking.
We’re early; the user base is growing, but we won’t pretend we can learn your personal skill curve yet. This version is deliberately “game-first”: it reads the world from platform data (Steam, PSN) and trusted metadata (IGDB/Steam tags), then calibrates points with guardrails to avoid obvious pitfalls in small or immature data. As RivalRyze scales, we’ll fold in behavioral signals and personalization — but v2 already fixes the biggest fairness issues you felt in v1.

How the math works

1. Base difficulty from rarity
   Every achievement gets an effective rarity p_eff (0–1) that emphasizes the rare tail without overreacting to a single outlier. Intuition: we blend the minimum observed rarity with a weighted average of the rare end of the distribution and work in log space to stay stable.

Base points = B + M · (1 / p_eff − 1)^β, capped
• B (baseline), M (scale), β (steepness), and a hard cap define the overall shape.
• Visual: “Base difficulty vs global earned rate” — points climb rapidly as rarity goes below ~5–10%.

2. Allocate within a game using rarity weights
   We distribute a game’s total “difficulty budget” to each achievement using a gamma weight on rarity:
   weight ∝ (100 / rarity%)^γ, normalized across the set
   • γ (gamma) controls how much extra bite the rarest items get.
   • We monotone-enforce points by rarity so rarer never pays less than easier in the same game.
   • Visual: “Per-achievement weighting by rarity%” — a smooth, monotone curve; the rare end dominates as γ rises.

3. Modifiers that reflect real-world context
   These are transparent guardrails that keep totals fair across very different games.

   • Average/dispersion modifier (m_avg)

   • Achievement-count modifier (m_ach)

   • Release-age modifier (m_rel)
   • Kind/genre/tag prior (m_kind)
   • Visual: “m_kind vs prior score at different metadata coverage levels.”

   • Rarity-maturity damp (m_mat)
   • Visual: “m_mat vs age (days)” under a shallow-tail scenario.

   
   

Cross-platform auto-pop detection (cheaters hate this part)

Some games copy progress between platforms or allow instant unlocks after account linking. We don’t want those to dominate competition, so we flag bursts that look like auto-pop and exclude them from points and points-based challenges.

We look for three things:

1. Tight unlock clusters
   A dense sequence of unlocks with very small gaps (e.g., ≤ 4–5 seconds between adjacent items) and a minimal cluster size (e.g., 7+ unlocks) within a short window (e.g., ≤ 5–7 minutes).

2. Earlier unlocks on another platform
   The same achievements already unlocked earlier on a different platform (we require a meaningful lead time, e.g., ≥ 24 hours, to avoid clock-skew or simultaneous releases).

3. Enough overlap to be confident
   The fraction of the cluster that matches earlier other-platform unlocks must exceed a ratio threshold (e.g., 60%), or pass an absolute count fallback for very large clusters.

Leaderboards Update

• New periods: Daily, Weekly, Monthly, and All-Time—each recalculated on schedule.

• Platforms: Steam, PSN, and Overall (combined).

• Rank logic: stored, versioned, and consistent; ties are stable and deterministic.

• Integrity: points and achievement counts exclude auto-popped unlocks.

• Performance: we added an optimized view and lightweight aggregation so ranks update fast—even for big communities.

What’s tunable (and why that’s good)

We built this to be adjustable without risky rewrites:

• Rarity curve: the shape and caps (how much rarer items are boosted, and how they saturate).

• Base points: calibration by achievement/tier/family.

• Windowing: period boundaries, schedule, and re-aggregation cadence.

• Auto-pop: cluster size, gap threshold, earlier-platform window, and overlap ratio.

• Normalization: per-game or per-franchise adjustments to prevent outliers.

These “knobs” let us respond quickly to real-world data and community feedback while keeping the core model stable.

How you can help us tune it

Send feedback—especially if you notice edge cases:

• Did something get flagged as auto-pop that shouldn’t?

• Is a particular game’s points feeling off relative to effort?

• Are the leaderboards behaving the way you expect per period and platform?

Your reports help us refine thresholds and curves quickly.

A note on transparency

We’ve shared the principles and shapes of the math and the kinds of thresholds we use, but we’re intentionally not publishing all parameters and internal tables. That’s to keep the system fair and resilient. If you’re curious about your own data or think we missed something, reach out—we’re happy to investigate.

This is a big step toward fair, fun, and truly cross-platform competition. Thanks for rolling with us—please keep the feedback coming, and we’ll keep tightening the system week by week.

— The RivalRyze Team 🚀