Update 0.33.0 - New AI, Practice Board AI, Puzzle Settings, and Imported Content

New Features

[p]New AI - KataGo Human[/p]
- [p]There is now a new AI in the game, KataGo-Human, which was trained on human games and should provide an opponent that has more human-like characteristics than the super-human level of play provided by the ordinary KataGo opponent. KataGo is doing the heavy lifting here, so many thanks to lightvector and his continued support of the Go community.[/p]
- [p]When using the human trained KataGo, you can select which era of Go the AI will play as. This includes the eras before and after AlphaGo Zero had an impact on human play and strategies.[/p]
  - [p]If selecting the modern AI or pre-AI eras, a rank selection is provided to select which rank (20 Kyu - 9 Dan) the AI will play similar to.[/p]
  - [p]If selecting to play against the historical pros era, a year selection is provided to select which year (1800-2023) the AI will play as, which is modeled after professional/master level players from that time.[/p]
- [p]Players can also choose whether or not to use the “Refined Move Selection” mode when using the KataGo Human opponent. When selected, additional processing will be performed during the move selection process which may provide a more realistic opponent, but may also make the opponent take longer to respond. For those that are familiar with KataGo’s human model setup, when this checkbox is checked, both a human and non-human model will be used with additional parameters to control which moves will be played. When it is not checked, only the human model will be used.[/p]
- [p]This new AI is available in both quick play and the campaign mode.[/p]
[p]New Puzzle Settings[/p]
- [p]In the settings menu, there is a new tab labeled "Puzzle" which is used to control several new settings listed below.[/p]
  - [p]Load Next Puzzle - How the next puzzle will be loaded after being completed.[/p]
    - [p]Manual - The default value, and how things previously worked, which is you click the button to load the next puzzle.[/p]
    - [p]Auto - After completing a puzzle, the next puzzle is loaded automatically after the specified delay in seconds has elapsed.[/p]
  - [p]Allow Hints - Controls whether the "Show Correct Move" button will be displayed if available.[/p]
  - [p]Attempts Limit - How many tries you get to solve a puzzle. An attempt is used when the player reaches an incorrect ending in the puzzle's move tree, when the player enters a move that isn't part of the puzzle, or if the player undoes moves and begins to play moves down a different path through the tree. By default there is no attempt limit, but a value of 1 to 20 can be specified.[/p]
  - [p]Time Limit - How long the player has to solve the puzzle. By default there is no limit, but a time limit of 5 seconds to 20 minutes can be specified.[/p]
  - [p]Learning Center - Settings that are specific to the learning center's workshop puzzles or imported puzzles.[/p]
    - [p]Randomize Order - After loading a puzzle, clicking the load next puzzle button will load a random puzzle from the list instead of by sequential order.[/p]
    - [p]Randomize Color - The puzzle's stone colors will be swapped or left as-is on a random frequency.[/p]
    - [p]Randomize Orientation - The puzzle's layout will be randomized by flipping horizontally, vertically, and/or diagonally.[/p]
[p]New Puzzle Session Stats[/p]
- [p]When solving puzzles in the campaign fortification process, workshop, or imported content, stats about how many puzzles have been solved or unsolved for that session will be displayed.[/p]
- [p]Expanding the solved or unsolved header will display a list of each previous puzzle, allowing the user to click them to return to the particular puzzle.[/p]
[p]AI Analysis on the Practice Board[/p]
- [p]AI analysis (find moves, territory, score and win rate graphs) is now available when using the practice board, similar to how it is available when reviewing a normal match.[/p]
- [p]This was previously not enabled because the practice board allows for things that are not allowed during a real match, such as placing multiple stones of the same color, changing the color of stones, deleting stones, etc. that are usually used to setup board positions on the practice board. During these instances, the AI will be asked for analysis based on the board state itself rather than the sequence of moves to get there. When using the alternating stone mode, the analysis will still however be requested based on the sequence of moves. All of that being said, there may be some rare occurrences of differences in AI analysis such as super ko evaluation when mid-match setup nodes are being used to add or remove multiple stones at once, since a full move history is unavailable.[/p]
[p]Learning Center Imported Content[/p]
- [p]The Learning Center now has a new menu item for imported content, which allows you to select folders of content on your computer to be imported into the game, such as puzzles, matches, etc. similar to the workshop content. The difference is that instead of downloading from the workshop, the files are instead loaded from a folder on your PC.[/p]
- [p]The content within this section will also be similarly tracked for completion, allowing you to for instance see which puzzles you've already solved. Once progress has been tracked for an imported item, the ability to reset the progress is also available.[/p]
[p]Recent Matches Tracking[/p]
- [p]Matches played from the Quick Play menu will now be stored automatically into a new section of the menu called Recent Matches.[/p]
- [p]By default, this will store the past 20 matches and will remove the oldest match to make space for a new one once the 20 match limit is reached.[/p]
- [p]For matches that you want to be saved long term, players can open the recent match for review and click the Save Match button at the top right to save it to the Saved Matches folder.[/p]
- [p]There is a new setting in the Playback Settings menu called Save Recent Matches which will control the number of matches that will be included in the Recent Matches folder. This can be switched to 0 to not save any recent matches or set to a maximum of 100. For those worried about disk space, the match files are very small, and a total of 100 matches will likely equate less than half a megabyte of space.[/p]
- [p]Both matches that are finished and matches that are in progress will be automatically saved to the Recent Matches folder when leaving the Go board screen or exiting the game. Because of this, the prompt for an unsaved match that used to be displayed has been removed if the Recent Matches feature is enabled.[/p]
- [p]Similar to other matches saved, these recent matches are also included in Steam Cloud so they can be accessed from multiple PCs.[/p]
[p]Selectable Neural Network Size[/p]
- [p]Within AI Settings, there is a new field for Neural Network which controls the size of the network that will be used by the analysis AI and KataGo opponent.[/p]
- [p]The larger networks are stronger but also slower, so users can now decide which works best for their hardware, strength, and speed requirements. In general, choose a network that is as large as possible while still maintaining a reasonable speed when reviewing and providing analysis. Most users should stick to large if using a GPU Analysis AI (OpenCL) or use medium for a CPU enabled AI (Eigen or Eigen AVX2).[/p]
- [p]Existing networks have been updated and two new networks have been added.[/p]
- [p]Users can also select Custom, which will allow the user to import a different network file. Many network files for KataGo can be found here.[/p]
- [p]For users using the OpenCL analysis AI, the tuning process will need to run again for the updated KataGo and network. This can take several minutes, but once completed it won't need to be retuned when launching the game again.[/p]
[p]There is a new option for KataGo Version found in the AI Settings called KataGo Custom. This is for advanced users that want to import a custom version of KataGo which will be used for analysis, reviews, and the KataGo and KataGo-Human opponents. For example, users with modern NVIDIA GPUs that have installed and configured TensorRT on their machine could import the TensortRT build of KataGo which has the potential of better performance if the PC can support it.[/p]
[p]When reviewing an automated review for moves that can be improved, once the moves have either been found, or the Show Correct Move button has been used, the markers on the board that show each next move in the tree will now include the move rank (1, 2, 3, etc.) from the AI review. This allows users to quickly view the rank of each move on the board and also differentiate between the moves that were included in the automated review and moves that were manually added afterwards.[/p]
[p]A new tuning progress user interface element has been added which will be displayed for users using the OpenCL version of KataGo while KataGo is tuning its configuration to meet your PC's specifications. Previously a message was displayed for this when it started and ended, but no progress information was available between then. Due to the KataGo updates and network file updates mentioned in the above new features, this tuning process will run for the initial launch after this update is downloaded. Once completed, the tuning process will not need to be run again.[/p]
[p]A new "Loading..." message has been added above the AI player card while the AI is being loaded for a match. Some AI opponents will load almost instantly and won't see this message, but for others like the casual AI or the new KataGo human AI, this message will be present until setup has been completed and the AI is able to start searching for moves. Additionally, the player clock won't start counting down for the AI until it has been fully loaded and it is ready to search for moves.[/p]

Updates

[p]The underlying engine that The Conquest of Go is built on has been upgraded to a later version to resolve some engine-level issues. This has the chance of affecting anything within the game, so if there are any bugs that have arisen, please don't hesitate to report them here.[/p]
[p]The difficulty presets for quick play matches and creating campaigns have been reconfigured. The casual difficulty setting has replaced the old beginner preset, and the novice preset has been changed to use the new KataGo-Human opponent. This means that the old presets that used GNUGo with handicap stones have been eliminated, but the intermediate difficulty with the GNUGo opponent still remains. This gives a more straight forward difficulty level progression without the need for handicap stones. Of course players can still achieve the settings of the old presets by adjusting the settings and creating their own custom difficulty level, if desired.[/p]
[p]After a quick play match and choosing rematch, if the match settings remain the same, the game will no longer need to go through a loading screen to start the next match.[/p]
[p]As a result of the improvements made to the AI being available on the practice board, the AI analysis when reviewing normal matches/SGF files is also now improved as it is now able to handle SGF setup nodes (adding multiple stones or removing stones) that occur mid-match, which could possibly be included in SGF files that have been imported.[/p]
[p]Offline quick play and campaign matches can now specify that there is no time limit for a match by dragging the Main Time slider all the way to the right, which will set the time to "No Limit". The previous method of dragging the Main Time and Byoyomi Time all the way to the left to specify No Limit still exists, but some users found this difficult to discover. The new method is closer to the thought process of "infinite time" rather than the old method which is the thought process of "no time limit".[/p]
[p]When doing fortification puzzles during the campaign, the Battle Summary panel will now take up less space, allowing additional space to be given to the panel showing puzzle information.[/p]
[p]In the Saved Matches section, in addition to being able to delete files, you can now also delete folders.[/p]
[p]Within the Settings menu, the names of each tab no longer include the word "Settings", which previously was redundant and took up more space.[/p]
[p]During the campaign tutorial, the "Next" button in the Go scene that is used to progress the tutorial now has an animation when ready to be clicked to better help players know what to click next during the tutorial.[/p]
[p]Running background AIs on macOS silicon to use silicon versions if possible before using Intel versions has been improved.[/p]
[p]The default KataGo AI has been changed from KataGo Eigen to KataGo Eigen AVX2 for better performance. The AVX2 instruction set should be available in most CPUs that were created in the past 10 years. For older CPUs that do not support this, the game should fallback to the original KataGo Eigen which has wider support. As mentioned before, if you have a GPU, the KataGo OpenCL version will likely give you much better performance, so feel free to try that by going to Settings > AI Settings > KataGo Version.[/p]
[p]The text for the Analysis AI setting found in Settings > AI Settings has been changed to "KataGo Version" as this setting not only applies to the analysis functionality, but to KataGo usage as a whole (opponents, analysis, review).[/p]

Fixes

[p]Fixed an issue where automated reviews were able to be started on SGF files and quick play matches even if the match didn't contain any moves.[/p]
[p]Fixed an issue with the move rank/order in automated reviews, where it was occasionally skipping numbers in the rank field.[/p]
[p]Fixed an issue on macOS where the mouse cursor button detection was not reliable when playing in windowed mode.[/p]
[p]Fixed tracking puzzle completion for workshop items on macOS.[/p]
[p]Fixed the game periodically hanging on the loading screen for macOS silicon users until refocusing the window.[/p]
[p]Removed additional "Correct" text from Steam Workshop puzzle solutions when the localized "Correct" text has already been displayed.[/p]