Tuesday, November 22, 2016

Round 10 - Playoffs 2 of 2016 US Tak Open

With the second round over, the first set of 8 people has been eliminated, bringing us down to the top 24 players.  Interestingly 3 of the 16 games ended in a tie-breaker. While this may not seem like a lot, it really shows how closely matched some of these players are and will be important to look at as part of the post-tournament analysis.

In addition, next week will feature more exciting matches. I personally am looking forward to the match between NohatCoder and SultanPepper. Whoever you may be rooting for, the tension is palatable.

One of the games I particularly enjoyed this past week was the first game between Gray_Mouser and SuperJujuAwesome:

I was particularly engaged at the end of the match, right before black's final play:

[... 21. c6']
During the game, black made a crucial mistake, playing d2<, missing white's ability to play 2d11<. Practically, this demonstrates that even those 'advanced' still miss taks all the time. But, suppose black had spotted this? Given more time and a more in-depth dissection, what moves would make the most sense?

In dissecting the position, white has clear tempo. With a tak threat, black is forced to respond. More so, white actually has three possible completions: c1, 1d1<, and 2d11<. The first two rely solely on the c-file, whereas the third also takes advantage of the connection from c3-b3.

Abstractly speaking, there are three main ways to defend against Tak: cutting the source, capturing connections, or 'stemming' the flow (interfering with Tak completion moves). To see this more explicitly, let's explore each of these approaches in the context of this position.

To cut the source here, black would need to capture c4,c5, or c6. Capturing any of these would prevent white from immediately completing tak. The downside, however, is that white can then press his flat lead, and increase position in the board. For example, b6> could easily be followed up with d5'!, allowing white to keep control and tempo. Similarly, b5> could incentivize filling at b5 - claiming the vacated spot with white's own flat, improving control in the b-file.

[{21. ... c6> 22. d5'}]

Alternatively, black can 'stem' the flow. Usually, this is most viable with a single completion, but here playing Sc1 would block all of the mentioned moves. This, though, is a very defensive move, to which white would still be able to further press tempo with a strong a1'! Note that in addition to the north-south threat, this also supports white's west-east threat (f5-c5;c5-c2,c2-a2)

[{21. ... Sc1 22. a1'}]

The last option to consider, then, is cutting the connection point: capturing c3. By capturing this, both taks are cut off at their central point. The interesting bit is that there are two fairly reasonable ways of doing this: 1d3< and 3d3<. The first one moves only the cap onto c3, creating a strong block, but also building a potential west-east road threat, across ranks 3 and 4. The second one, by contrast, leaves d3 open, but gives black the option of spreading north or south, capturing white's main forces:

[{21. ... 3d3<}]
In addition, there's even a hint of a possible north-south road utilizing black's b-file pieces, along with the new control along the c-file.

In return, white now has to manage black's threat to cover pieces, while maintaining a presence and taking advantage of his diminishing flat lead. Without an immediate tempo, white has more space to play positionally, probably playing d3, filling the vacated space, or e4, gaining control over the east edge and building towards a potential road along the e-file (and threatening to use the capstone on b3 to complete it).

Overall, I hope this fleshes what I see underlying my predominantly intuitive processes of dissecting a position. Some of these calculations are quick - spotting Taks can be fairly quick. Identifying blocking moves in of itself takes skill. And then quickly running through each, and even looking out for the potential follow-ups, either playing it out or using heuristics.

Let me know how you think about Taks. What types of Tak are easiest or hardest for you to spot? What types of responses do you tend towards?


  1. Taks like this one, where my opponent can throw a small stack while giving up some of my pieces, are exactly the kind I tend to miss most often. My brain doesn't really parse a flat with one prisoner as a threat, I think.

  2. I've found that after working on ptn.ninja for the past several months (and not actually playing Tak), I'm missing very obvious Taks when I do play. I think I've trained myself to rely on the road indicators, and I fail to see capstones as parts of roads. Then again, it could just be that I haven't played in months.