2 * Manual page - full usage documentation
3 * GTP interface documentation
6 * Further optimize board implementation, profiling fun
7 * Clean up GTP interface, allow custom GTP commands for modules
8 * gogui-friendly GTP interface
9 * Implement parameter setup over GTP (less important)
10 * Fix the build system to allow fully parallel build
11 However, revamp to something like cmake (or, ugh, autotools)
12 is not guaranteed to be appreciated.
15 * Try to disable the bsize pattern feature
16 It just fudges the pattern evaluation since for most tactical
17 patterns fourth line vs. fifth line just doesn't matter. Maybe
18 its max should be 2 and maybe it should just be gone, needs
19 regenerating the pattern database and benchmarking.
20 * Develop dedicated playout handling for few common tactical situations
21 Monkey jump (and its followup sequences!), bent four in the
23 * Fix seki recognition to be stable
24 Try to research and fix most cases harmed by selfatarirate=100
25 or find another way to fix seki.
26 * Try to avoid using a hash table for 3x3 patterns
27 Instead autogenerate procedural matching code; may be more
28 efficient (the near-guaranteed L1 cache miss is fairly expensive).
29 * Optimizing our tree implementation for cache-efficiency
30 Statistics of all children of a parent node shall be contained
31 in an array of the parent node so that move evaluation during
32 the descent can access them sequentially in memory, instead
33 of walking a linked list. Pasky already tried once but it's
34 somewhat arduous and dull work.
37 * Automated building of opening book
38 * Expanding and tagging the regression suite
39 Even better, create a nice UI for our users to contribute and
41 * Implement Pachi support to fishtest
42 http://tests.stockfishchess.org/tests would allow crowdsourcing
43 Pachi parameter tuning.
44 * Split playout aspects to custom-stackable pieces?
45 * Port to Intel Phi (if we get the hardware :)
47 Some heuristics to test:
48 * Local trees (work in progress, no luck so far)
49 * Liberty maps (work in progress)
50 * Implement a tsumego solver and apply it once per playout (stv insp.,
51 see Eric van der Werf's PhD thesis?)
52 * MM local-based patterns in playouts (work in progress, no luck so far)
53 * Balanced local-based patterns?
54 * Killer moves (redundant to RAVE?)
55 * Reverse status learning
56 Run on game corpus. Start at final position, watch development
57 of status of all stones. The moment the final status and expected
58 status changes, analyze, especially if move choice differs. Use
59 learnt status-fixing moves in simulations somehow.
60 Tried to do this on Pachi-played games; no measurable effect
61 (maybe too small sample).