Al Piskun: add archive link

2026-07-20T09:46:39Z

add archive link

← Older revision		Revision as of 09:46, 20 July 2026
Line 35:		Line 35:
	}}		}}

	'''Chess@Home''' was a short-lived [[BOINC]] volunteer computing project that set out to build a large, machine-evaluated database of chess positions as the first step toward training a classifier able to estimate which side stood better in a given position. The project was run from chess.qugate.org and was maintained by Michal Stanislaw Wojcik, a Polish BOINC community member, with early technical support from the Poland BOINC Foundation.<ref name="about">{{Cite web \|title=About Chess@Home \|url=http://chess.qugate.org/ \|archive-url=https://web.archive.org/web/20140208000329/http://chess.qugate.org/ \|archive-date=8 February 2014 \|website=Chess@Home \|access-date=19 July 2026 \|url-status=dead}}</ref>		'''[https://web.archive.org/http://chess.qugate.org/ Chess@Home]''' was a short-lived [[BOINC]] volunteer computing project that set out to build a large, machine-evaluated database of chess positions as the first step toward training a classifier able to estimate which side stood better in a given position. The project was run from chess.qugate.org and was maintained by Michal Stanislaw Wojcik, a Polish BOINC community member, with early technical support from the Poland BOINC Foundation.<ref name="about">{{Cite web \|title=About Chess@Home \|url=http://chess.qugate.org/ \|archive-url=https://web.archive.org/web/20140208000329/http://chess.qugate.org/ \|archive-date=8 February 2014 \|website=Chess@Home \|access-date=19 July 2026 \|url-status=dead}}</ref>

	== Overview ==		== Overview ==

Al Piskun: first light

2026-07-20T09:28:24Z

first light

New page

{{Infobox software
| name = Chess@Home
| logo =
| logo caption =
| screenshot = Stockfish icon (2010-2020).png
| caption = The icon of Stockfish, the open-source chess engine Chess@Home used to evaluate positions
| description = Chess@Home was a short-lived BOINC volunteer computing project that used Stockfish to evaluate tens of millions of chess positions, building a labeled dataset intended for training a machine-learned classifier of winning positions.
| status = Completed
| category = Computer science, Machine learning
| compute = CPU
| dependencies = Stockfish (bundled)
| developer = Michal Stanislaw Wojcik
| author = Michal Stanislaw Wojcik
| sponsor = Poland BOINC Foundation
| maintainer = Michal Stanislaw Wojcik
| released = {{Start date and age|2014|01|10}}
| completed = Yes
| discontinued = c. March 2014
| repository =
| programming language =
| operating system = Windows
| size =
| stats as of = {{Start date and age|2014|02|08}}
| average performance =
| active users =
| total users =
| active hosts =
| total hosts =
| rac =
| credit per day =
| gpu performance =
| cpu performance =
| website = {{URL|http://chess.qugate.org/}}
| license =
}}

'''Chess@Home''' was a short-lived [[BOINC]] volunteer computing project that set out to build a large, machine-evaluated database of chess positions as the first step toward training a classifier able to estimate which side stood better in a given position. The project was run from chess.qugate.org and was maintained by Michal Stanislaw Wojcik, a Polish BOINC community member, with early technical support from the Poland BOINC Foundation.<ref name="about">{{Cite web |title=About Chess@Home |url=http://chess.qugate.org/ |archive-url=https://web.archive.org/web/20140208000329/http://chess.qugate.org/ |archive-date=8 February 2014 |website=Chess@Home |access-date=19 July 2026 |url-status=dead}}</ref>

== Overview ==
The project's stated goal was to construct a chess winning-position classifier, or score estimator, built on simple positional features such as material count, the number of squares controlled by each side, and the number of checks available in a position. The first phase aimed to establish how well relatively cheap evaluation methods and standard data mining techniques, including logistic regression, k-nearest neighbors, support vector machines, and decision trees, could perform at this task. A longer-term goal was to apply more advanced machine learning methods to approximate different levels of human chess proficiency, by tuning a search heuristic between more and less thoroughly trained states.<ref name="about" />

To build the underlying dataset, the project first assembled a large collection of positions taken from games played by titled players, and planned to add further positions from weaker amateur games sourced from the [[wikipedia:Free Internet Chess Server|Free Internet Chess Server]] games archive, in order to represent a wider range of playing strength.<ref name="about" /> Each position was to be evaluated by [[wikipedia:Stockfish (chess)|Stockfish]], an open-source chess engine, with the resulting centipawn score (or mate distance, where a forced mate was found) and best move recorded for later analysis.<ref name="about" />

== Scale of the task ==
The project's own estimate, published on its about page, illustrates why the work needed distributed computing rather than a single machine. Evaluating one position to a fixed node budget of five million nodes took roughly five to six seconds on a single core running at approximately 2.5 GHz. Applied across an initial collection of 25 million positions, that works out to:

<math>6\ \text{s} \times 25{,}000{,}000 \approx 1.5 \times 10^{8}\ \text{s} \approx 41{,}666\ \text{hours} \approx 1{,}736\ \text{days}</math>

on one CPU core alone, and the project expected its full dataset, once amateur games were added, to grow to hundreds of millions of positions.<ref name="about" /> By the time of the 8 February 2014 snapshot of the project's homepage, volunteers had already evaluated 30,301,792 positions.<ref name="about" />

== Technical details ==
The BOINC client application distributed to volunteers was named chessathome, for example ''chessathome_1_windows_intelx86.exe'' on Windows, and served as a wrapper that launched the bundled Stockfish engine to perform the actual position evaluation; Stockfish's search was the sole consumer of CPU time on the volunteer's machine.<ref name="about" /> Completed tasks were credited at 9 BOINC credits each, corresponding to roughly ten minutes of work on a 2.5 GHz core.<ref name="about" />

== History ==
Chess@Home was announced to the BOINC community on 10 January 2014, described in the announcement simply as a new project related to the game of chess.<ref name="announce">{{Cite web |title=Thread 'Chess@Home' |url=https://boinc.n-helix.com/forum_thread.php?id=8842 |website=BOINC combined message boards |access-date=19 July 2026}}</ref> Its administrator described himself as new to running a BOINC project and credited Krzysztof Piszczek of the Poland BOINC Foundation as a mentor who helped set up the project's server.<ref name="about" />

The project's active life was brief. By 17 March 2014, a volunteer reported that the project had been offline for more than a week; another community member confirmed that the project had finished, adding that the administrator himself was not certain whether further work would be issued in the future.<ref name="announce" /> No further tasks were subsequently issued, and the project is understood within the BOINC community to have concluded at that point.

=== Relationship to Chess960@home ===
Despite the similar name, Chess@Home was not connected to the earlier, unrelated project [[Chess960@home]]. When Chess@Home was announced, a member of the BOINCstats community asked directly whether the two were related; the response confirmed they were separate projects, and that permission to use a similarly chess-themed name had been sought beforehand.<ref name="boincstats">{{Cite web |title=Forum::New projects::Chess@Home |url=https://www.boincstats.com/forum/11/9016,1 |website=BOINCstats/BAM! |access-date=19 July 2026}}</ref>

== See also ==
* [[Chess960@home]]
* [[wikipedia:Stockfish (chess)|Stockfish]]
* [[wikipedia:BOINC|BOINC]]

== References ==
{{Reflist}}

Chess@Home - Revision history

Al Piskun: add archive link

Al Piskun: first light