Widest Path / MasterGo
Here is the widest path for the complete MasterGo database. The widest path consists of the most common starting move, followed by the most common follow-up to that move, etcetera. Note that the MasterGo database contains mostly games from the 1990s. Note also that MasterGo looks at all games in which the given position was present, irrespective of the order of moves to get there, and that it when necessary will turn or flip the board to get there. - Andre Engels
Tapir: It is still the same path 'til in the 2010 MasterGo database (>50k games). is played more often as an approach (d in the 2nd diagram).
- Total database: 18000 games, black wins 56%.
- Black 1: 12290 games, black wins 56%. Black a is the second most common (5193), all other starting moves taken together occur in only 517 games.
- White 2: 4881 games, black wins 55%. White b (4166 games) is a close runner-up, white 9 (2436) also occurs quite often.
- Black 3: 1878 games, black wins 57%. b (1489 games) and 9 (1023 games) are the runners-up.
- White 4: 1155 games (out of 1916), black wins 53%. c (472) is the runner-up.
- Black 5: 639 games (out of 2715), black wins 49%. Much choice here, in total 16 different moves are present in the database. low Chinese fuseki at 10 (480) and shimari at d (414) are the other most common moves.
- White 6: 300 games (out of 640), black wins 50%. Clearly more common than e (122) and g (122).
- Black 7: 162 games (out of 309), black wins 50%. Clearly the direction of play (see Kobayashi formation); runner-up is f with 74 games.
- White 8: 116 games, black wins 49%.
- Black 9: 53 games (out of 117). g (51 games) is a close runner-up.
- White 10: 33 games, black wins 48%. Only common alternative is h (19 games).
- Black 11: 31 games, black wins 52%.
- White 12: 28 games, black wins 46%.
- Black 13: 15 games, black wins 60%. Black 15 (10 games) is an important alternative.
- White 14: 8 games, black wins 62%. White a (7 games) occurs almost as often.
- Black 15: idem
- White 16: 7 games, black wins 71%.
- Black 17: 5 games, black wins 80%.
- White 18: 2 games, black wins 100%. b, c and d all occur once.
The smallest path ends at black 19. Follow-ups are white e (played by Yu Chae-hyeong against An Tal-hun in the Korean promotion tournament 1998) and white f (Nakazawa Ayako against Chinen Kaori in the 1997 women's Honinbo semifinals).
I did the same with my program, but using amateur games as well and noticed that the path converges and diverges several times, and even at move 39 there are still variations. A Fuseki library of 450,000 games is used.
It becomes obvious that including strong amateur play in the "widest path" leads to a much thicker Fuseki tree. Even ten moves past the terminal node of the pro-based widest path, amateur Fuseki - if you can call it fuseki at this stage - is much more varied, as expected. At this position there were still 6 variations over 39 games.