Note 2: This project is in the design stage. If you see something is wrong of missing - don't be surprised. But please post any improvement ideas in the linked threads.
It's year 2011, and we still don't have any usable 7-piece chess endgame tables, despite the obvious huge interest in the community. I propose to establish a bounty to stimulate and reward the developers working in this field. Part of the bounty will be offered for the generator and probing code, part for the actual generated tables, and part for support and infrastructure work. This document will contain up-to-date summary of the project.
Creating a practical generator for 7-piece endgame tables is an immensely challenging task. Creating even remotely useful 6-piece generator takes months of work for a highly skilled and motivated developer. 7-piece endgames is a lot harder than that. A number of people worked in that direction, but their generators are mostly kept private. The available ones are not open source and have very limited functionality. [TODO: Expand with examples].
Up to now the progress in this field was driven by enthusiastic hobbyists. Every now and then someone will try to write a generator (which may take years), mostly out of their own curiosity. If successful, then some tables are generated and distributed, often with the help from the community. This model worked well for 5-piece endgames, with over a dozen competing formats. It also produced some useful 6-piece tables, most notably the Nalimov format. Somehow this model did not produce a useful 7-piece EGTB generator so far, probably because it's so hard to do. So perhaps it makes sense to change the model and try encouraging the developers with some real funding.
It's apparent that the community has a lot of interest in creation of practical 7-piece endgame tables. How large is that interest - remains to be seen by the progress of this bounty project. Up to now there was no way for the community to directly motivate the developers in organized fashion. Now this bounty project is providing such way.
As a small historical reference, the problem of availability of the 6-piece endgame tables in Nalimov format (about 1.2 TB) was solved by the community with amazing efficiency in 2005-2006 (See the project page). Now the whole set is available from multiple sources online. This time, with 7-piece tables, the problem is not only sharing the existing tables, but also generating them, and creating a useful generator first. Quite a bit bigger task. Is the community strong enough to solve it? With your help, I'm sure it is!
What this project is going to achieve? By the time when we call the project complete, 7-piece-EGTB-assisted endgame analysis should become an easy and enjoyable experience, available to everyone free of charge. Roughly speaking, this means the following:
- An efficient and feature-rich generator for 3-to-7-piece endgame tables is available, free and open source.
- The access code is available and integrated into various chess software.
- All popular tables are generated and available to download.
- Obscure / less popular tables can be generated by those who need them (for example 6-vs-1 tables).
- A unified web-based query interface provides access to all generated tables.
Also, some talented developers will make some real money. And, of course, computer chess community will have accomplished a mega-project.
The rough idea is:
Phase 0. The community drafts the plan and requirements. (This is where we are now).
Phase 1. Advertisement, accumulation of funds, improving the requirements. Ends when the generator is submitted. Part of the bounty that was allocated for the generated is paid to the successful developer (or a team).
Phase 2. Generation and verification of the tables. Creation of the infrastructure for efficient distribution of the tables. Creation of the unified query interface. Key tasks are funded from the remaining part of the bounty.
Phase 3. Project complete, everyone is happy.
Phase 1 will most likely result in one of these scenarios:
Scenario 1. Someone submits the solution qualifying to all requirements. Bounty is paid to the developers. We proceed to phase 2 - generating and distributing the tables.
Scenario 2. Someone releases a more or less useful 7-piece EGTB generator, which does not satisfy the bounty requirements. This results in community losing interest to this bounty and pulling back the funds (Yes, all donations can be taken back at any time, until the moment when a solution is submitted). Alternatively, we may decide to lower the requirements, to still reward the developers.
Scenario 3. Nothing happens.
The first scenario is more likely to occur if the community succeeds to assemble an attractive prize for the developers to win. So the outcome of this project will mostly be decided by the amount of interest the community has for 7-piece EGTB.
This bounty project:
- This project at FOSS Factory
- 7-man EGTB Bounty Reborn - General Discussion
- 7-man EGTB Bounty Reborn - Metric Discussion
- 7-man EGTB Bounty Reborn - File Format Discussion
- 7-man EGTB Bounty Reborn - Infrastructure Discussion
Discussions in other forums