q5go-2.1.3: Autoplay and some other minor updates

[And]

launched with katago! very large request, change to a larger value of 25 (use for analysis) computer go - change an engine configuration - board size

EDIT or ... I do not understand, is it now not necessary?

[And]

I checked it works great! thanks a lot!!!

[bernds]

launched with katago! very large request, change to a larger value of 25 (use for analysis) computer go - change an engine configuration - board size

EDIT or ... I do not understand, is it now not necessary?

I'm not entirely sure what you are asking. The idea behind setting a board size for an analysis engine was that LZ only supports 19x19, and by recording that in the settings, we can have consistency and safety checks in the batch analysis window: you have to select a board size, the list of available engines will reflect that choice, and it won't let you add files that can't be analyzed by the engine.
With KataGo now supporting more board sizes, maybe I can come up with something else. In the meantime, as a workaround, you can just duplicate an existing engine configuration and change the board size to whatever you like.

[And]

I was primarily interested in winrate. he is working now. the analysis may be interesting, for example, if I download sgf 29x29.
that is, maybe useful

[Dragon]

bernds, Thank you so much! I like more than Sabaki!

[bernds]

I've finally completed (for now) the new IGS multi-observation window. This allows you to have multiple observed games open in a single window, with clickable previews to switch between them.

To keep the screenshot small, everything is a little squashed - it looks better at normal window sizes.

multiobserve.png (579.46 KiB) Viewed 17607 times

The old behaviour (one window per game) is also available, and you can switch between the two styles at any time.

[And]

huge gratitude for removing the limitation on board size during analysis !!!
is it possible to change the direction of analysis in the settings? For some reason, my analysis goes from end to beginning

[bernds]

is it possible to change the direction of analysis in the settings? For some reason, my analysis goes from end to beginning

That is intentional. At least in theory, by analyzing the later moves first, the engine could reuse the evaluation when examining an earlier move, leading to better results.

[And]

if it were possible to open the sgf, turn on autoreplay, and run the analysis - that would be fantastic!
maybe it's not difficult to do it?

[Dragon]

the program is getting better and better! many thanks!

[Dragon]

if it were possible to open the sgf, turn on autoreplay, and run the analysis - that would be fantastic!
maybe it's not difficult to do it?

yes, that would be very good!

[Dragon]

I hope for the release of KataGo or another program for playing on 37x37. could you make such an opportunity when choosing a game against the computer? now up to 36x36.

[bernds]

I had some trouble getting Kombilo to run the other day, since Python 2 seems to be getting less supported by my Linux distro. So I implemented pattern search for q5go.

It's not quite as feature-full as Kombilo yet - it's missing a statistics display, for example, and it only searches primary variations at the moment. But it seems to work pretty well, and is quite fast and comfortable to use: in a normal board window, press R and leftclick-drag a rectangle, then press S to bring up the pattern search window and initiate a search. After double clicking on a game in the game list, the found pattern is highlighted and you can press N to go to the next move in this area. Shift- or Control-click to place stones of a specific color, right-click and drag a new rectangle and press S or A to search again (in the current or full game list). Results can be displayed on the board as rank (A-z as in Kombilo), total number of continuations, or percentages.

At this point I probably also need to add some code to generate databases without Kombilo. I'll probably use a separate file format so as to not interfere with Kombilo (and because I don't need/know how to generate all of the data).

I'd be happy if anyone could give this a spin to see if/how well it works for them. With this in place it's probably time to make another release soon.

Screenshot:

patsearch.jpg (389.6 KiB) Viewed 16777 times

[bernds]

There have been a few updates to the pattern search. It turns out there is a problem with the files Kombilo generates for pattern search: they are not system independent. On Windows, for example, I had a database created by a 32 bit Kombilo, and it caused errors when trying to load it with a 64 bit q5go.
As a result, it really became necessary to add my own file format. The q5go.db files can be created alongside a Kombilo database and will be used preferentially if present. They have a few other advantages, for example being smaller and supporting games with different board sizes in the same directory.

There are also some new features - for example, double clicking a game after a search now shows not only the found pattern, but also the first few moves of continuation within the search region.

As always, up-to-date automated builds for MacOS and Windows can be had from the Github Actions page, or by clicking on the green checkmark on the main page. I'd appreciate if I could find one or two testers to make sure it doesn't just work on my machine.

[ez4u]

...

As always, up-to-date automated builds for MacOS and Windows can be had from the Github Actions page, or by clicking on the green checkmark on the main page. I'd appreciate if I could find one or two testers to make sure it doesn't just work on my machine.

When I try either the green checkmark or the actions page, I do not find compiled versions to download. I can see a 76 MB "artifact" listed that is named q5Go-win64_msvc2017_64, but it is not a downloadable file. Is there something further that I have to do or are the beta testers supposed to download and compile the code itself?