How to Interpret CCRL Ratings Compared With Other Chess Engine Rating Lists

CCRL ratings are best read as one credible, method-specific estimate of engine strength — not a universal ranking. The site itself recommends consulting other rating lists for broader context, because its numbers come from a defined hardware and time-control setup (benchmark hardware Intel i7-4770k, Stockfish 10 used to calibrate time controls, generic books capped at 12 moves, ponder off). If you want a single definitive ordering of engines, no list gives you that; if you want a well-documented sample of results you can inspect game by game, CCRL is designed for exactly that.

What CCRL actually publishes

CCRL (Computer Chess Rating Lists) was formed in 2006 by Graham Banks, Ray Banks, Sarah Bird, Kirill Kryukov, and Charles Smith. It is run by computer chess enthusiasts who test engines as a hobby and combine results as a group.

The available lists and statistics include:

Each list is updated independently, and links sit below the Best Versions lists. That independence matters when comparing: a strong 40/15 result does not automatically transfer to the Blitz or Chess960 tables.

The methodology points that shape the numbers

• Fixed benchmark hardware. Testing uses an Intel i7-4770k as the benchmark machine, with Stockfish 10 calibrating time controls. Ratings from a different machine or a different calibration engine are not directly interchangeable with these.
• Controlled playing conditions. Games use generic books with a maximum of 12 moves and ponder off. Engines that rely heavily on deep opening preparation or on pondering will not show their full practical behaviour here.
• Early ratings fluctuate. CCRL states that early ratings may move around, which is expected when an engine has few games behind it.
• ELO is saved only at 100+ games. An engine below that threshold may appear without a stored rating, so absence of a number is not evidence of weakness.
• LOS statistics. The Likelihood of Superiority figure shows the probability of an engine outperforming the one ranked immediately below it — a useful check on whether a gap between two neighbours is meaningful or just noise.
• Testers choose their own methods. While general testing guidelines are followed, individual testers decide what and how to test; some use systematic methods, others prefer tournaments. This is a genuine source of variation across the lists.

What CCRL explicitly does not do

CCRL does not judge engine originality. Inclusion or exclusion from the lists does not imply endorsement or rejection of an engine. Viewers are expected to form their own opinions. If your question is about whether an engine is a derivative or a clone, CCRL is not the source that answers it — and you should not read a listing as a verdict on that question.

Comparing CCRL with other rating lists

Use the same dimensions across lists rather than comparing headline numbers directly:

1. Hardware. Check what machine the other list benchmarks on. Different CPUs mean different effective search speeds, which shifts results.
2. Time control. CCRL separates 40/15, Blitz, and 40/2. Confirm the other list's time control matches before placing two numbers side by side.
3. Opening book policy. CCRL uses generic books up to 12 moves. A list using engine-specific or deeper books will produce different outcomes.
4. Ponder setting. CCRL tests with ponder off. Lists that allow pondering measure something different.
5. Game count behind a rating. CCRL stores ELO at 100+ games. A rating built on far fewer games deserves less confidence regardless of which list publishes it.
6. Variant coverage. Chess960, Chess324, and DFRC results only compare meaningfully with lists covering the same variants.

A practical rule: treat a rating as a statement about performance under a described setup. Two lists agreeing on an engine's rough tier is meaningful; two lists disagreeing by a small margin usually reflects methodology, not a real strength difference.

Getting from a rating to a judgement

CCRL provides the raw material to check claims yourself rather than trusting a single figure:

• Download games by engine, by month, or by ECO code.
• Click an engine name to see its opponents, homepage links, and comparison tools.
• Use the openings report for ECO statistics, draw rates, and White win percentages.
• Sort any list by column to reorder by rating, games, or other fields.

If you want to verify whether engine A is genuinely stronger than engine B, pull their head-to-head games and the LOS figure between them, then check whether the same ordering holds in a second list with comparable methodology. That combination — direct games plus a second source — is more informative than either list alone.

Joining as a tester

If you want to contribute results rather than just read them, sign up on the CCRL forum and contact a member. For security reasons, CCRL only accepts testers who are known to the community or vouched for by trusted members.