multiPV historical chess games analysis concept

multiPV historical chess games analysis concept
--------------------------------------------

1) online (web) database similar to http://www.chessgames.com/perl/chessgame?gid=1008361 interface
2) moves are in 2 column table similar to this http://goo.gl/jHvl34
3) cells (moves) are coloured depending how good was move (green for 1st line, pink supotimal, red for blunder etc)
4) current eval (and D were eval was obtained) for each black & white move
5) if move is subotimal - additional columns with list of all better PV's - eval and 5-6 plies PV-line
6) all actual moves and PV moves are clickable and shown on the main board (nice to implement arrows as sesse.net)
7) games are only obtained from chessgames.com (each analyze has direct link to source at chessgames.com)
8) site admin (or user) may request additional Depth if they feel so (4c type of task), but only for limited number of additionl plies (site calculate MN cost estimation for this analize based on previous run and allow tasks not more than 50 000 MN estimation)
9) after game - separate table with summary of worst moves by both sides (sorted by eval drop versus 1st line) and total eval drop by each side. After game was win (>4 eval for winning side for the rest of the game), eval drop aren't counted (or there is separate table of such moves)

How analyze is done
1) web site has queue of games to be analized (site admin or visitors add links at chessgames.com they want to analyze. moderation, voting etc can be implemented)
2) there is windows/linux downloadable kit (like fishtest, but GUI), site admin (or visitors donating their CPU) run this tool, it is UCI-wrapper running latest Stockfish and doing actual work.
3) UCI-wrapper connects to site and check if queue is not empty, if there is task - assign it and download. If task was not finished for some period of time (e.g. 24 hours) it is unassigned and available again in queue.
4) there are 5 possible kind of tasks
	a) shallow multiPV (~50 mN per move, ~30 min per game on laptop) for all legal moves
	b) deep multiPV (500 ... 50 000 MN per move) for heuristic number of lines dependent of 4a result (actual move +- some eval margin)
	c) deepen multiPV for particular move by couple D (is available only for finished b-type task)
	d) regular eval to highest D (e.g. final eval)
	e) request +N extra lines for multiPV (worse than was actually played) if kibitzer is curious "what if play this way here"
5) User GUI-UCI-Wrapper has several user options:
	* hide to system tray
	* hash size (detect available ram for default)
	* process priority (IDLE = default)
	* number of threads (all CPU threads available - default, detect HyperThreading)
	* schedule when allow analysis (sleep all other time)
	* allow manually put current task in ignore list (user don't want to contribute to this game)
	* type of tasks to allow (default - all 3 types)
	* Syzygy path, if user can provide
	* user preferences on type of games to be analized (years period, player strength, human or comp etc),
	  white/black or priority list
6) Completed job is uploaded partially after each move is done. Interrupted jobs can run from this point on any other machine (after task is implicit abandoned or unassigned by 24hr timeout)

Known problems:
1) Open book detection - don't waste time on positions in open book main lines. Analyze only if it is side lines of the book
2) Heuristics for 4b type of analysis

*******************************
LIST: Historical chess games analysis multiPV   ---->   http://pastebin.com/9psux9XU
*******************************