Calculation guide

How NextMove computes its scores

This page explains the non-obvious calculations in the product so users can understand what is official, what is estimated, and what is only meant to guide training.

Ratings

Check what is official and what is estimated

The anchor ratings come from analysed game ratings in your saved reports, while blended estimates are convenience summaries only.

Peers

Understand percentile comparisons carefully

Peer profiles compare you against similar saved reports inside the product, not against the full player pool on a chess site.

Training

Separate review retention from playing strength

Your spaced-review rating reflects how stable your trained mistakes are inside the card system. It is not your playing Elo.

Scheduling

See why cards return when they do

The interval and goal rules explain why some cards come back quickly, why others wait longer, and how the daily streak is awarded.

Playing strength

Mode-specific rating anchors

NextMove keeps Bullet, Blitz, Rapid, and Daily separate. For each saved report in a specific mode, it builds a mode anchor from the analysed player ratings in that report.

Per-mode anchor

Use the median analysed player rating when there are at least 5 rated games in that mode; otherwise fall back to the most recent analysed rating.

Primary mode

The primary mode is the saved mode report with the largest rated-game sample for the same username and platform.

Provisional label

A mode is marked provisional when it has fewer than 8 rated analysed games.

Playing strength

Blended cross-mode estimate

The blended estimate is only a convenience summary. It is intentionally shown as a range, not as an exact Elo, because different time controls do not measure the same kind of strength.

Mode weights

Primary mode: 60%
Second mode: 25%
Third mode: 15%

Blend center

Take the weighted average of the top three mode anchors, then round the center to the nearest 5 points.

Blend range

The visible range widens when your mode anchors are far apart or when the total rated sample is small. The endpoints are rounded to the nearest 25 points.

Peer comparison

Percentile profile

Peer comparison is mode-specific. NextMove tries to compare you against players on the same platform, in the same mode, and around the same rating anchor.

Cohort selection

Prefer same platform + same mode + players within about 150 points of your primary-mode anchor. If that cohort is too small, the comparison widens step by step.

Percentile

For each metric, percentile is based on where your value sits inside the cohort. Lower is better for error rates; higher is better for first-mistake move.

Important limit

These percentiles are product diagnostics, not official chess-platform percentiles.

Training retention

Spaced-review rating

The training rating measures how stable your reviewed mistakes are inside the drill-card system. It is not a playing-strength rating.

Skill score

Skill score runs from 0 to 100 and combines card strength with review accuracy. Strength contributes 65% and accuracy contributes 35%.

Skill rating

The displayed training rating is computed as 600 + skillScore × 14.

Mastery

A card counts as mastered when it reaches strength 4.

Momentum

Trend direction

Momentum is based on saved daily snapshots for the primary mode of the current username and platform.

Inputs

NextMove compares your latest saved day against the previous saved day using major errors per game and average first major mistake move.

Momentum score

Errors-per-game improvement is weighted more heavily than first-mistake timing. Large positive change becomes Improving, large negative change becomes Sliding, and small change becomes Stable.