ChessPivot

How the scores in this comparison are awarded

A grid of 56 criteria across 11 categories. Every level of every criterion states a fact that anyone can check on a public page, and every score cites the page it was read from.

categories

11

checkable criteria

56

services scored

147

Back to the comparison
01

The principles

  1. 01

    One level, one fact

    Every criterion has four levels, 0 to 3, and each level states something observable: a published figure, a feature visible on a page, a line in a policy. If the fact is there, the level is earned; if not, it is not. Nothing is left to taste.

  2. 02

    The evidence travels with the score

    Every score cites the page read and the fact found there. A service page therefore lays out all 56 pieces of evidence, so a reader — or the service itself — can redo every check.

  3. 03

    Undocumented means uncredited

    A service that has a feature but documents it nowhere is scored as if it did not. This is the one rule that stops scoring from resting on what we happen to know. It is hard on quiet services — which is exactly what the correction procedure is there to fix.

  4. 04

    Primary sources come first

    Official site, pricing page, FAQ, app store listing, privacy policy, code repository. An outside review can confirm a fact; it never replaces it.

  5. 05

    The same grid for everyone

    A playing server, a video course, an analysis engine and an online coach all face the same 56 criteria. A narrow tool collects many zeros: that is not a verdict on its quality, it measures what it brings to the question asked — getting better. The category tables give specialists their due.

02

The arithmetic

A category score is the mean of its criteria on a 0-10 scale: a service scoring 3 everywhere gets 10, one scoring 0 everywhere gets 0. Within a category, every criterion counts the same.

The overall score is the weighted mean of the eleven categories. The weights are an editorial choice, owned and arguable — which is why the ranking is also published unweighted and category by category, and why you can change the weights yourself.

That score does not name "the best service". It names the one that covers most of the question, under those weights. A reader who mainly wants to play should read the Play category; a coach, the Coaching one; a tournament player, Openings and Diagnosis.

  • APersonalized diagnosis20 %8 criteria
  • BOpenings18 %6 criteria
  • CTargeted training16 %9 criteria
  • DPrice and value14 %4 criteria
  • ERigor and honesty8 %4 criteria
  • FContent and method8 %5 criteria
  • GPlaying and practice5 %4 criteria
  • HExperience and accessibility5 %5 criteria
  • IPreparation and study2 %4 criteria
  • JHuman coaching and community2 %4 criteria
  • KPersonal data and interoperability2 %3 criteria

Weights add up to: 100 %

03

The grid: 11 categories, 56 criteria

APersonalized diagnosis20 %

A service can only target training if it knows what the player keeps missing, and shows it to them. This is where individual progress starts, and it is the area covered least often.

A1 Importing your games (Chess.com, Lichess, PGN files)

0
No import: the service cannot reach any game played elsewhere.
1
Paste a PGN/FEN, one game at a time.
2
Connects to ONE platform (Chess.com or Lichess) OR imports PGN files in bulk.
3
Connects to Chess.com AND Lichess, plus PGN file import.

Evidence expected Import page, FAQ, screenshot of the connection screen.

A2 Automatic engine analysis of every game

0
No engine evaluation.
1
Evaluation of one position on demand, with no pass over the whole game.
2
Automatic analysis of every imported game; engine and depth undocumented OR capped by a quota.
3
Automatic analysis of every game, engine AND depth published, with no blocking quota on the paid plan.

Evidence expected Product page, technical FAQ, changelog.

A3 Recurring mistakes identified across all your games

0
Nothing aggregated: each game is analyzed on its own.
1
Aggregated statistics (average accuracy, results by opening) with no mistake named.
2
Mistakes named and counted across all games (e.g. "undefended piece hanging: 14 times").
3
Mistakes named, counted AND ranked by frequency and by cost (severity measured with the engine).

Evidence expected Screenshot of the report, public list of the patterns detected.

A4 Types of mistakes covered (tactics, strategy, endgame, time…)

0
No typology published.
1
A single area (e.g. tactics only).
2
2 to 3 areas among: tactics, strategy/structure, opening, endgame, king safety, time management.
3
≥ 4 of those areas, including endgame or time, with the list of patterns published.

Evidence expected Public list of areas / patterns.

A5 Written explanation of each mistake

0
Numeric evaluation only.
1
A generic label ("blunder", "inaccuracy") with no explanation.
2
Explanatory text for each mistake, in general terms ("the center is weakened").
3
Explanatory text for each mistake that cites the concrete moves (line, threat, square) it claims.

Evidence expected Public sample comment, method page.

A6 Stated verification of the diagnoses (procedure, rates)

0
No information about reliability.
1
Users can report a problem ("this diagnosis is wrong"), with no figure published.
2
Verification procedure described publicly (set of controlled cases, human review).
3
Procedure described AND rates published (false positives, recall) OR reference cases open to inspection.

Evidence expected Method page, engineering blog, public repository.

A7 Tracking each weakness over time

0
No history.
1
Rating (ELO) curve only.
2
Play indicators (accuracy, patterns) tracked period by period.
3
Each named weakness tracked period by period (1/3/6/12 months), with the direction of the trend.

Evidence expected Dashboard screenshot.

A8 Comparison with players of the same rating

0
No comparison.
1
Your position in a global leaderboard.
2
One overall indicator compared with your rating band (percentile).
3
Each area / weakness compared with your rating band.

Evidence expected Screenshot, documentation.

BOpenings18 %

The repertoire is what players work on most, and what the market sells most. We measure how far it is personalized, how firmly it is anchored in real games, and how reliable the lines are.

B1 Building a personal repertoire

0
None.
1
Fixed repertoires (courses) that cannot be edited.
2
Free-form building (PGN import, tree editor).
3
Free-form building plus suggestions based on the player’s own games or level.

Evidence expected Product page.

B2 Training the repertoire (drills, games)

0
None.
1
Passive replay of the lines.
2
Interactive drill (play the right move).
3
Drill + spaced repetition + full games against an opponent that plays the repertoire.

Evidence expected Product page.

B3 Spotting where you left your repertoire in your games

0
None.
1
Result statistics by opening.
2
The service shows where the player left the repertoire in their games.
3
Deviations detected AND turned into exercises/reviews.

Evidence expected Product page, screenshot.

B4 Verification of the lines taught (engine, practice)

0
Not documented.
1
Lines from a titled author, with no engine check announced.
2
Lines checked with an engine OR sourced from a game database.
3
Lines checked with an engine AND sourced, with a published threshold or protocol.

Evidence expected Method page, about page.

B5 Opening explorer and statistics

0
None.
1
Statistics on your own games only.
2
Explorer over an external database (masters or online).
3
Multi-database explorer (masters + online + your own games) with evaluations.

Evidence expected Product page.

B6 Explanation of opening plans and ideas

0
None.
1
Names and moves only.
2
Plans and ideas explained in writing.
3
Plans explained + traps + model games, in text AND visual form (video or diagrams).

Evidence expected Public example.

CTargeted training16 %

What the player does every day. We measure how far it is personalized, how well it covers the phases of the game, and the mechanisms that make learning last.

C1 Exercises drawn from your own games

0
None.
1
Replay the move you missed in a game (one position, no bank).
2
Exercises built from your own mistakes, presented as a series or a personal bank.
3
Exercises from your own mistakes, sorted by the weakness detected, with a success rate per weakness.

Evidence expected Product page, screenshot.

C2 Bank of tactical exercises (size, themes, rating)

0
None.
1
< 5,000 exercises OR no sorting by theme.
2
≥ 5,000 exercises sorted by theme, with a difficulty rating.
3
≥ 100,000 exercises, themes, adaptive rating, public quality control (votes, reports).

Evidence expected Figure published by the vendor (page, store).

C3 Endgame training

0
Nothing.
1
A few endgame lessons or puzzles, with no dedicated module.
2
Dedicated module: theoretical positions played against the engine OR an interactive course.
3
Dedicated module + practical endgames taken from real games (including the player’s own) + endgame tablebases.

Evidence expected Product page.

C4 Strategic and positional training

0
Nothing.
1
Passive content (video/text) with no exercises.
2
Positional exercises (plans, structures, "guess the move" on master games).
3
Positional exercises sorted by strategic theme (≥ 3 themes: structures, plans, pieces) with the plan explained.

Evidence expected Product page.

C5 Training defense and alertness

0
Nothing.
1
Live blunder alert during a game.
2
Exercises dedicated to one sub-type (hanging pieces OR opponent threats OR the saving move).
3
Exercises dedicated to ≥ 3 defensive sub-types, each with its own difficulty rating.

Evidence expected Product page.

C6 Training calculation and visualization

0
Nothing.
1
One coordinates/board-vision exercise.
2
Blindfold mode OR exercises on calculating sequences.
3
A progressive calculation/visualization program with several levels.

Evidence expected Product page.

C7 Difficulty matched to your real level

0
Fixed level, or chosen by hand.
1
Levels by rating band.
2
Adaptive rating (Glicko/ELO) on the exercises.
3
Adaptive rating with a documented target success rate (e.g. 70-85%) per exercise family.

Evidence expected Documentation, FAQ.

C8 Spaced repetition (named algorithm)

0
None.
1
Replaying the exercises you failed by hand.
2
Spaced repetition, algorithm not named.
3
Spaced repetition with a named algorithm (SM-2, FSRS…) and visible review due dates.

Evidence expected Documentation (name of the algorithm).

C9 Personalized training plan

0
None.
1
Generic advice by level.
2
A daily work queue (what to do today), the same for everyone.
3
A plan personalized for each player, with an adjustable time budget and explicit priorities.

Evidence expected Screenshot, product page.

DPrice and value14 %

What a complete path really costs, and how it is sold.

D1 What the free plan actually lets you do

0
Nothing useful without paying (demo/trial only).
1
Very limited free features (< 5 exercises/day or 1 analysis/day).
2
Real daily use is possible for free, within quotas.
3
The core of the service is free and unlimited.

Evidence expected Pricing page.

D2 Annual price of the full plan

0
> €150/year.
1
€76-150/year.
2
€31-75/year.
3
≤ €30/year, or free.

Evidence expected Pricing page (annual price, or 12 × the monthly price if there is no annual plan).

D3 Commercial transparency (price shown, trial, cancellation)

0
Price not shown before signing up, OR renewal traps reported.
1
Price shown, with no trial and no refund.
2
Price shown + trial or refund + cancellation online.
3
Price shown + trial with no credit card + one-click cancellation + no surprise auto-renewal.

Evidence expected Pricing page, terms of service.

D4 Number of paid tiers (how simple the offer is)

0
A subscription AND one-off purchases both needed to unlock everything.
1
Three paid tiers or more.
2
One or two paid tiers, the top tier unlocking everything.
3
A single paid tier, or a service that is entirely free.

Evidence expected Pricing page.

ERigor and honesty8 %

What a service claims, and what it proves.

E1 Published evidence of effectiveness

0
No data.
1
Testimonials or a figure with no method behind it ("3× faster").
2
Data published with the method described (sample, duration).
3
Independent or peer-reviewed study.

Evidence expected Page, publication.

E2 Chess verification of the content (positions, lines)

0
Errors reported but left uncorrected OR no information at all.
1
Corrected once reported.
2
Engine checking or review by a titled player announced.
3
Verification protocol published (legal positions, moves checked, thresholds).

Evidence expected Method page.

E3 Transparency of the method (algorithms, code)

0
Black box.
1
General description ("AI", "engine").
2
Algorithms and settings named (engine, depth, SRS).
3
Source code public OR full technical documentation.

Evidence expected Documentation, repository.

E4 Marketing claims that can be checked (no overselling)

0
An unsourced numeric promise of rating gain.
1
Superlatives that cannot be checked ("world no. 1").
2
Factual wording.
3
Factual wording + limits stated explicitly.

Evidence expected Home page, advertising.

FContent and method8 %

The courses and the structure of progress. We measure how content is organized by level, who the authors are, the formats, and the volume.

F1 Courses organized by rating level

0
No courses.
1
Courses with no level indicated.
2
Courses labeled by level (beginner/intermediate/advanced).
3
A complete path for each rating band (e.g. 0-800, 800-1200…) up to ≥ 2000.

Evidence expected Catalog.

F2 Course authors named and titled

0
Authors not identified.
1
Authors named, without a title.
2
At least one titled author (FM/IM/GM) named.
3
A team of several named titled authors (≥ 3).

Evidence expected Authors / about page.

F3 Content formats (text, video, interactive)

0
No content.
1
One format (text OR video).
2
Two formats.
3
Text + video + interactive exercises tied to the content.

Evidence expected Catalog.

F4 Published method of progression

0
None.
1
Generic advice.
2
Curriculum published with milestones.
3
Curriculum published + adapted to the player’s diagnosis.

Evidence expected Method page.

F5 How often the content is updated

0
Content undated, or untouched for > 3 years.
1
Updated within the last 12 months.
2
Updated within the last 3 months.
3
Documented publishing at least monthly (changelog, blog, dated catalog).

Evidence expected Visible dates (blog, changelog, store).

GPlaying and practice5 %

Playing is still essential. We measure the size of the pool, how realistic the artificial opponents are, and whether you can replay one specific position.

G1 Playing humans online (pool, anti-cheating)

0
Not possible.
1
Small pool (> 1 min wait in blitz, or < 1,000 players online).
2
Medium pool with instant pairing.
3
Very large pool (≥ 10,000 online), every time control, documented anti-cheating measures.

Evidence expected Counter of players online, fair-play page.

G2 Realistic artificial opponents

0
None.
1
An engine throttled into levels.
2
Opponents with a personality or a style.
3
Models trained on human play, calibrated by rating band, with realistic thinking time.

Evidence expected Product page, technical documentation.

G3 Replaying one specific position against the machine

0
Not possible.
1
Play from a position you enter (FEN).
2
Replay a position from your own games against the engine.
3
Automatically replay the critical positions of your games (the turning point) against a calibrated opponent.

Evidence expected Product page.

G4 Tournaments and organized competition

0
None.
1
Occasional tournaments.
2
Daily arenas/tournaments.
3
Daily tournaments + leagues/teams + events with titled players.

Evidence expected Tournaments page.

HExperience and accessibility5 %

A tool nobody uses helps nobody improve. We measure the platforms, perceived quality that can be measured (store ratings), languages, accessibility, and offline use.

H1 Platforms available (web, iOS, Android, desktop)

0
One non-mobile platform (desktop only).
1
Responsive web only OR mobile only.
2
Web + one native app (iOS or Android).
3
Web + iOS + Android (+ desktop).

Evidence expected App stores, download page.

H2 User ratings and reviews (stores, press)

0
Store rating < 3.5/5 OR no rating and no third-party review.
1
Store rating 3.5-4.2 OR mixed third-party reviews.
2
Store rating ≥ 4.3 with ≥ 100 reviews OR consistently positive third-party reviews.
3
Store rating ≥ 4.5 with ≥ 1,000 reviews.

Evidence expected Store listing (rating, number of reviews), reviews cited.

H3 Interface languages available

0
A single interface language.
1
2 to 4 interface languages.
2
5 interface languages or more.
3
5 languages or more, with the teaching content translated (move notation localized).

Evidence expected Language selector, store listing (languages listed).

H4 Accessibility (keyboard, screen reader)

0
No information.
1
Size/contrast settings.
2
Keyboard navigation AND screen-reader support announced.
3
Accessible mode documented (guide) and tested.

Evidence expected Accessibility page, FAQ.

H5 Offline use and known incidents

0
Depends entirely on the server, with incidents reported in the last 12 months.
1
Depends on the server, with no notable incident reported.
2
Key features available offline (mobile or desktop).
3
Everything works offline (local software) OR a public status page showing uptime.

Evidence expected Status page, store reviews, documentation.

IPreparation and study2 %

Everything around training: preparing for an opponent, annotating and sharing a game, entering a game played over the board, finding a game again in your history.

I1 Preparing for a named opponent

0
None.
1
Browsing the games or the public profile of a named player.
2
Report on a named player: repertoire by color and results.
3
Report + weaknesses you can exploit + suggested preparation lines.

Evidence expected Product page, screenshot.

I2 Annotating and sharing games (studies)

0
None.
1
Text comments on a game.
2
Variations + comments saved and shareable by link.
3
Collaborative studies (several authors, several chapters) or a public annotation database.

Evidence expected Product page.

I3 Entering a game played at a club (photo, scoresheet)

0
Not possible.
1
Entering the moves on a virtual board.
2
Recognition of a position from a photo or a diagram.
3
Recognition of a handwritten scoresheet (OCR) into PGN.

Evidence expected Product page, store.

I4 Searching your games (filters, position)

0
None.
1
Chronological list.
2
Filters (opening, result, time control, opponent, color).
3
Search by position, by pattern, or in plain language.

Evidence expected Product page, screenshot.

JHuman coaching and community2 %

A human coach is still what players cite most above 1800. We measure access to coaches, how alive the community is, and the tools available to trainers.

J1 Access to human coaches

0
None.
1
A list of coaches with no vetting.
2
Vetted coaches (title or rating checked) OR regular group lessons.
3
Vetted coaches + individual follow-up (feedback on the player’s games) included or bookable.

Evidence expected Coaches page, terms.

J2 Size and activity of the community

0
None.
1
Comments or a quiet forum (< 1 visible message a day).
2
Active forum/Discord (messages every day) OR a club/team.
3
A community of ≥ 5,000 visible members with daily activity and regular events.

Evidence expected Public counter, visible activity.

J3 Tools for trainers, clubs and schools

0
None.
1
Several accounts under one subscription.
2
Coach area: list of students, assignments, activity tracking.
3
Coach area with a diagnosis for each student (weaknesses, progress) and a group view.

Evidence expected Dedicated page.

J4 Engagement mechanics (streaks, virtual currency, reminders)

0
Paid engagement mechanics (virtual currency, boosters, weekly subscription).
1
Streaks/leaderboards with no paid mechanic, but with pushy reminders.
2
Streaks/leaderboards with no commercial pressure.
3
Visible progress without a compulsory leaderboard, and it can be switched off.

Evidence expected Screen, terms of service, store.

KPersonal data and interoperability2 %

What players can get back of their own data, what is done with it, and whether the service can talk to other tools.

K1 Exporting your data (games, repertoire)

0
No export.
1
Export on request (by email).
2
PGN export of your games or your repertoire.
3
Full self-service export (games, repertoire, personal data).

Evidence expected FAQ, account page.

K2 Protection of personal data (GDPR, hosting)

0
No policy, or data resold / advertising tracking.
1
Policy in place, hosting not specified, third-party trackers.
2
GDPR-compliant policy, no resale, hosting specified.
3
GDPR + EU hosting + no advertising trackers.

Evidence expected Privacy policy.

K3 Public API and source code

0
Closed.
1
Standard import/export (PGN) only.
2
Documented public API.
3
Open source code.

Evidence expected Documentation, repository.

04

The limits, owned

What the grid does not measure

Actual effectiveness. No criterion measures how many rating points a player gains, because no service publishes data that would allow it. The grid measures what is provided and proven, not what works.

What rests on a threshold

The numeric thresholds (exercise counts, review counts, yearly price, community size) are arbitrary by construction. They are published so they can be argued with: change them and the scores move predictably.

What could not be read

Some pages block reading, or only show prices after signing up. Those cases are flagged on the service page and score the lower level until corrected.

What goes stale

Prices and features change. Every service page carries its reading date and is re-read within twelve months at the latest.

05

Suggest a correction

Anyone — a reader, a coach, a club, a scored service — can suggest a correction: fix a score by citing the public URL that establishes the fact, add a missing service, propose a criterion, or flag a mistake. Every request is looked at; if the URL establishes the fact, the score changes and the correction is dated and published. A feature that exists but is documented nowhere cannot be credited: document it, then write to us.

Email contact@chesspivot.com
06

Who publishes this comparison

ChessPivot, which is one of the services compared — read from its public pages only, by the same process as the rest. That is exactly why the grid, the levels, the weights and the evidence are all public: start by checking ours.