PR Rating: Formula & Server-wide Baseline Values

Our PR (beta) doesn't reuse the wows-numbers formula — it's built from scratch on our own server-wide sample. The formula, the weights and every baseline are on this page. Feel free to check the math.

Sample: 489,712 active accounts · 974 ships · 1,896,141,472 random battles.

• Current data: version 4, sampled 2026-10-15
• PR baseline: sampled 2026-10-01, recalculated about every 4 weeks
• Next update: sampling starts 2026-10-22 08:00 (JST), published the same day

We are collecting better approaches and feedback on the current method — give feedback

The formula

1. 1.For each ship, divide your per-battle damage, frags, spotting, tanking (potential damage), capture points, survival rate and win rate by that ship's average on your server (big table below). That gives seven ratios, clamped to 0–3. The win-rate deviation is stretched ×2.5 first: win rates sit so close to 50% that they'd disappear next to the other stats.
2. 2.Weight the seven ratios by ship class and add them up. Ship score = 1000 + (overall ratio − 1) × 2000. 1000 is the server average, and every 20 points is 1% of overall performance.
3. 3.The account rating is the battle-weighted average of your ship scores. A ship needs 10+ battles to count, the account 30+ in total.
4. 4."Better than X%" places your rating among 490k accounts; the colors follow the same percentiles (table below).

How baselines are built

A baseline is simply the server-wide per-battle average on that ship, weighted by battles. Ships with few samples get pulled toward their tier-and-class average (E = (B·m + K·m̄) / (B + K), K = 5000 battles), so new and rare ships never start with absurd baselines. Ships with no sample aren't rated. Each server is computed separately — Asia differs from the merged values by 4–9% on average. Your score uses your server's baselines; the big table below shows the merged ones.

Modes: the random-battle family (all/solo/div-2/div-3) shares one baseline set; operations, ranked and co-op are out of scope.

Stat weights (per class)

The weights aren't hand-picked — they come from a regression: we take your six stats on one ship and try to predict your win rate on your other ships. Stats that predict better get more weight. Why other ships? Because predicting the same ship's win rate would cheat: winning battles already comes with more kills and more survival, so the numbers are a result of winning, not the cause. The win-rate stat itself is fixed at 20% and stays out of the regression. Two floors were set by hand: 8% tanking for battleships, 15% spotting for destroyers. The regression alone gives them about 2% — not because tanking or spotting is useless, but because damage and frags already carry most of that information. Weights are public though, and players will play to them; we're not going to tell battleships to stop tanking. The floors cost a little accuracy, and we publish that too: on a 20% holdout (240,821 accounts), these weights predict per-class win rate with R² = 0.74; without the floors it's 0.752, and a hand-tuned baseline only got 0.66. Every future change follows the same rule: it ships only if it predicts the holdout better.

Color bands

Bands are percentile-based; the score column is the rating at that percentile, calibrated on the 490k-account distribution.

Known limitations (feedback wanted)

• Weights measure predictive power, not tactical value. A small weight on tanking or capping doesn't mean the play is useless — it means the number adds little once your damage and frags are known;
• The sample skews toward clan players — solo-player sampling is still in progress, so baselines may run slightly high (your PR may be slightly underrated);
• Division modes include carry effects yet share the solo baseline;
• No player-side small-sample shrinkage: ship scores at 10–20 battles are volatile.

Changelog

• v1 (2026-07-15): PR rating (beta) launched. Seven weighted stats, regression-fitted weights with validation published (floors: 8% battleship tanking, 15% destroyer spotting); per-server baselines; 1000 = server average.

Data update log

We sample every Thursday: combat stats update weekly, the PR baseline about every 4 weeks. The last 12 updates are below.

• Version 4 (sampled 2026-10-15): 495,170 accounts. Compared with the previous version, a typical ship moved 0.01 pp in win rate and 0.0% in average damage. 2 ships newly reached the display threshold.
• Version 3 (sampled 2026-10-08): 492,511 accounts. Compared with the previous version, a typical ship moved 0.01 pp in win rate and 0.0% in average damage.
• PR baseline updateVersion 2 (sampled 2026-10-01): 489,712 accounts. Compared with the previous version, a typical ship moved 0.03–0.05 pp in win rate and 0.2% in average damage. 23 ships newly reached the display threshold. PR baseline updated too: baselines moved 0.1% on a typical ship; the median player's score changed by −2.

Server-wide baselines (all ships)

The denominators from step 1. Merged across servers here — scoring uses your server's values. Updated with the PR baseline (about every 4 weeks).

← Player Stats