How to read the rating list

Below is the explanation of rating list columns, from left to right. (Though you can also construct custom lists with some columns missing)

Logo

Just a small picture representing an engine.

You can click on it to open a selection of all versions of this engine, participating in KCEC testing. Usually there will be at most 2 such versions because the older versions are removed.

Rank

A number that shows engine's position in the list. 1 is best, 2 is second best, etc. With ranks you can easily see the top 10 engines, top 20, etc. Some features of KCEC ranks:

1. Two (or more) engines with the same rating will share the rank, like "15–16".

2. Only one version of any engine may have a rank. This is not the issue with the list on the main page, as it only includes one version per engine. The complete list (and custom lists) however may include multiple versions of the same engine. In such case only one of those versions will get a rank - the latest one. If they have the same date, then the stronger one will get a rank.

3. Only engines that played at least 100 games can have rank.

Engine

First line: Engine name and version (including any compile or setting details). Color: Green = free engine, Orange = free and open source. You can click on it to see various details for this engine, including list of all opponents with scores and performances.

Second line: List of authors, and release year.

Protocol

Protocol used to communicate with the engine. "UCI" - Universal Chess Interface, "WB" - Winboard protocol (version 1 or 2), "CB" - ChessBase protocol.

Some engines support more than one protocol, in that case the list mentions the one that was actually used in KCEC testing. The order of preference is UCI, WB, CB.

So, "UCI" means that an engine supports UCI protocol and it was used in testing. "WB" means that an engine does not support UCI protocol (or has broken support of it), so Winboard protocol was used instead (with WB2UCI adapter). "CB" means that no Winboard or UCI version of that engine is available, only the one with ChessBase proprietary binary protocol.

Sometimes there are explamation marks around the protocol name. This means that the engine has some serious problems with protocol support, stability, or time losses. Check the engine details page (by clicking on engine name) for more information about those problems.

Country

Flag of a country, represented by an engine. Usually it's a country where the engine author is from. You can click on a flag to see all engines from that country.

Continent

Picture of a continent represented by an engine. Clicking on it will show all engines from the same continent.

Rating

Elo rating of each engine, computed with Bayeselo program based on the complete database of KCEC games.

All lists on this site use the same ratings. When you select a custom set of engines for comparison, the custom rating list will still use the ratings based on complete database, same like the complete list.

KCEC ratings are calibrated to CCRL 40/4 rating list from 2008-07-25 using all shared engines. (By default Bayeselo computes ratings centered around 0, the whole rating list has to be shifted up to match more familiar rating scale).

Rating error bars

95% confidence intervals of the rating. If you see a rating of 2845 and intervals of +15 −15, this means that we can be 95% sure that the true rating is within [2830,2860] range. Computed by Bayeselo.

Av.Op. - Average opponent

Approximate average opponent's rating is represented graphically by a number of arrows. Down arrows mean that the average opponent is weaker, up arrows - stronger. Number of arrows shows how much weaker or stronger. Every arrow counts for 20 Elo points. No arrows means that the average opponent is within (-20,20) range. Rounding towards 0 is applied, that is: +19.9 is represented as no arrows, +20 - as one arrow, +39 - still as one arrow, etc.

To see the actual average opponent's rating as a number, click on the engine name (to open engine details page) and scroll down - it is shown below the list of opponents.

Performance slope

Shows how much engine's performance changes depending on opponent's strength. Positive slope means that this engine plays better with strong opponents, and worse with weaker opponents. The number itself is how much performance points this engine gains (on average) per 1 added point of opponent's strength.

Below is the 95% confidence interval of the slope.

Av.Df. - Average difference from opponent

Approximate average distance between an engine and its opponents is represented as a number of small circles. Each circle counts for 20 Elo points. Rounding towards 0 is applied, that is: 19.9 is represented as no circles, +20 - as one circle +39 - as one circle, etc.

To see the actual average difference as a number, click on the engine name (to open engine details page) and scroll down - it is shown below the list of opponents.

Drawness

Shows how many draws an engine made compared to the expected number of draws. Expectation is computed by the last formula from the Draw rate page (depending on rating difference and average rating) for each match this engine played. The actual total number of draws the engine made is shown in percents of the expectected total number of draws.

Second line is the 95% confidence interval.

Games

Number of games played by an engine. It also serves as a link to download the archive of engine's games.

LOS

LOS means Likelihood of Superiority. These numbers show how sure we can be that an engine is stronger or weaker than its neighbors in the list.

Missing columns

Often found in other rating lists, these columns are removed from the KCEC rating lists:

Draws - Draw percentage among all games played by an engine. As the Draw rate page shows, draw rate greatly depends on rating difference and average rating in a match. So this number reflects mostly the testing process rather than the engine's own quality. It is still available at the engine's details page, below the list of opponents.