Recommend
4 
 Thumb up
 Hide
43 Posts
1 , 2  Next »   | 

Stephen Tavener» Forums » General Creator Forum

Subject: Ai Ai Bug Reports rss

Your Tags: Add tags
Popular Tags: AiAi [+] [View All]
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
Please post any bugs here.

Ai Ai links all in one place:
Suggestions
Announcements
FAQ
Ai Ai Home Page
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Selamat pagi,
Sweden
flag msg tools
Help, I'm being held prisoner in an overtext typing facility! I don't have much time, they could find out at any m
badge
I'm that weirdo whose number of badges sold prior to yesterday Bail Organa is keeping track of
Avatar
mbmbmb
Delta: Once in the late game, when new pieces could no longer be added, the game got hung up on both players alternately shifting pieces to the same spot, almost clearing the board of pieces before declaring a computer victory.

Several games bug out if you set the board size to something other than the default size, so that you can't tell the true spatial relationships and/or win conditions and there may be ghost pieces showing up. Delta and Lot are examples of this. Changing the board size should probably be disallowed if the engine isn't set to scale the game mechanisms accordingly.

Inphexion: Once the game declared that three of my pieces had captured a fourth of my pieces and told me to relocate it, effectively converting it to an enemy piece.

Catchup: I can't seem to tell what actual rules are being used. At times I place only two pieces despite the computer having the biggest group. At other times the computer plays three pieces despite having the biggest group. At one point I thought I had more or less worked out that the rule was that you play three pieces whenever your opponent grows their biggest group, regardless of comparative size, but firstly this doesn't always hold true, and secondly I can't seem to find any such incarnation of the game.

Also, while not technically a bug, the AI is just rubbish at some of the games (notably Gonnect and Y), seemingly playing random moves and letting you win. Is this just because of choosing the wrong AI, and if so, would there be a way to specify a default AI for each game based on what's been tested to work? I've just used GUCT for everything because that's the default setting and I wouldn't know what works better for any given game.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
Hi! Thanks for your interest. In general, when you find a bug, it would be very useful if you could save the position file and email it to me. That way I can reproduce the bug and fix it faster.

Kaffedrake wrote:
Delta: Once in the late game, when new pieces could no longer be added, the game got hung up on both players alternately shifting pieces to the same spot, almost clearing the board of pieces before declaring a computer victory.

Bug - I'll look into it. If you do have a .mgs file, that will make my life easier.

Quote:
Changing the board size should probably be disallowed if the engine isn't set to scale the game mechanisms accordingly.

True. It's on my to-do list, as part of a bigger task (the game file should tell the GUI which parameters can be tweaked, and by how much).

Quote:
Inphexion: Once the game declared that three of my pieces had captured a fourth of my pieces and told me to relocate it, effectively converting it to an enemy piece.

Bug. Again, a .mgs file would be useful.

Quote:
Catchup: I can't seem to tell what actual rules are being used. At times I place only two pieces despite the computer having the biggest group.

There should be three files, with three different rulesets. I'll update the .mgl files to be more descriptive.

Quote:
Also, while not technically a bug, the AI is just rubbish at some of the games (notably Gonnect and Y), seemingly playing random moves and letting you win. Is this just because of choosing the wrong AI

It's mostly because some games are just more resistant to AI than others. You can increase the thinking time to make the AI stronger. Unless I have encoded heuristics for the game (only true of Vault so far, but more are coming), then GUCT is usually the strongest player.

Quote:
If so, would there be a way to specify a default AI for each game based on what's been tested to work? I've just used GUCT for everything because that's the default setting and I wouldn't know what works better for any given game.

I am working on something along these lines at present.

Anyhow, thank for helping make Ai Ai better - and do send me the .mgs files if you have them.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
Kaffedrake wrote:
Delta: Once in the late game, when new pieces could no longer be added, the game got hung up on both players alternately shifting pieces to the same spot, almost clearing the board of pieces before declaring a computer victory.

Fixed.

Quote:
Catchup: I can't seem to tell what actual rules are being used. At times I place only two pieces despite the computer having the biggest group.

Each game file now links to the explicit rule set on BGG. Any discrepancies between the two are a bug!

Quote:
If so, would there be a way to specify a default AI for each game based on what's been tested to work?

That can now be done in the .mgl file. In most cases, though, it hasn't I am working on stronger play, though, so keep watching.

Some user feedback on which games play well/best settings would be good? Please be my guest...
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Selamat pagi,
Sweden
flag msg tools
Help, I'm being held prisoner in an overtext typing facility! I don't have much time, they could find out at any m
badge
I'm that weirdo whose number of badges sold prior to yesterday Bail Organa is keeping track of
Avatar
mbmbmb
mrraow wrote:
Anyhow, thank for helping make Ai Ai better - and do send me the .mgs files if you have them.


Thank you for making it in the first place!

Another general bug: if you set the win animation to "none", the display still switches to some previous position for a second when the game ends. This happens even if you set the duration to 0 seconds.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Selamat pagi,
Sweden
flag msg tools
Help, I'm being held prisoner in an overtext typing facility! I don't have much time, they could find out at any m
badge
I'm that weirdo whose number of badges sold prior to yesterday Bail Organa is keeping track of
Avatar
mbmbmb
Two more things:

1. Are the game end conditions properly implemented in hex Murus Gallicus? In 3P at least nothing happens if you get a piece to the other side. Then again I have no idea if hex MG is an official variant at all or how it should be played. (As an aside, the AI seems to play MG exclusively defensively trying to win by move starvation, claiming advantage right up to where your attacking forces start poking holes in its defenses.)

2. In games with more than two players, the AI will still call any win prognosis significantly below 50% "unfavourable". It should recalibrate the draw expectation to 33% or 25% as appropriate (and perhaps adjust as players drop out of contention).
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
Kaffedrake wrote:
1. Are the game end conditions properly implemented in hex Murus Gallicus? In 3P at least nothing happens if you get a piece to the other side. Then again I have no idea if hex MG is an official variant at all or how it should be played. (As an aside, the AI seems to play MG exclusively defensively trying to win by move starvation, claiming advantage right up to where your attacking forces start poking holes in its defenses.)

The 2p versions have the reach-the-other side goals (except for MuriusGallicusEscape.mgl), the multiplayer games don't. From the rules:
Escape
Murus Gallicus becomes a more challenging game when the winning conditions are reduced to stalemate and resignation only. All other rules apply. The strategy here is to divide, isolate, and conquer.
Multi-player Escape
The escape variant can be played with four players around a square game board, for example a checkerboard. During the initial setup, players do not put towers on the four corner cells. When a player is stalemated, he or she is eliminated and his or her stones remain on the board. The escape variant rules also apply well to regular hexagonal game boards, making three-player games feasible.

Quote:
2. In games with more than two players, the AI will still call any win prognosis significantly below 50% "unfavourable". It should recalibrate the draw expectation to 33% or 25% as appropriate (and perhaps adjust as players drop out of contention).

I'll have to think about that one. Myself, I'd call a 1-in-3 chance of winning unfavourable!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Craig Duncan
United States
Ithaca
New York
flag msg tools
designer
Avatar
mbmbmbmbmb
This is probably a user problem (i.e. my mistake) but I have been unable to run Ai Ai on Windows 7. I've doublechecked I have Java 8 installed. I see the aiai.jar file on the unzipped file, but when I click on it it just brings up a list of folders (ai ai, dictionary, drawable, etc.). Java does not run.

I've checked that my computer associates .jar files with Java as the default program to open such files with.

Any advice?
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
Curious. I have windows 7 myself, and no problems. I'd guess you have an unzip application which is also linked to .jar files. What happens if you right click on the file?

Failing that, you can certainly run it from a command prompt.

Change folder to aiai directory, then

(Path to java) -jar aiai.jar
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Craig Duncan
United States
Ithaca
New York
flag msg tools
designer
Avatar
mbmbmbmbmb
mrraow wrote:
Curious. I have windows 7 myself, and no problems. I'd guess you have an unzip application which is also linked to .jar files. What happens if you right click on the file?


Right-clicking pulls up a menu with options such as Open, Open Inside, Open Outside, Edit, and then things like Copy, Rename, etc.

I'll try the commmand prompt when I get a chance.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Craig Duncan
United States
Ithaca
New York
flag msg tools
designer
Avatar
mbmbmbmbmb
Ok, I think I found my initial mistake. I was clicking on the Ai Ai file icon from within the .zip file without extracting the Ai Ai folder first.

When I click on the extracted aiai.jar file, Java now launches. But I get an "Application Error" message that reads: "Could not parse launch file. Error at line 1."
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
cdunc123 wrote:
"Could not parse launch file. Error at line 1."

... probably because there is no launch file.

Again, this is odd - it looks like your machine is trying to run the application using java webstart, which almost certainly won't work (a) because there is no .jnlp file to tell it how, and (b) because I rely on being able to access local files (the mgl files).

I suggest you try the following:

1) Make sure the files are unzipped into a folder somewhere you can find them with paths intact. Your base folder should contain three things:

- aiai.jar
- default.mgc
- mgl folder (containing game definitions)

2) Start a command prompt (start menu, type cmd in the search box)

3) Change directory to the Ai Ai folder

4) Type the following:

java -jar aiai.jar

This should work if you have one copy of java correctly installed; if not, you may need a fill path to the java executable.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Craig Duncan
United States
Ithaca
New York
flag msg tools
designer
Avatar
mbmbmbmbmb
Hi Stephen

Yes, I think for some reason my office computer was trying to run the program with Java Webstart. Weird.

In any case, it worked fine from my home computer. Thanks for the help, and for Ai Ai. It looks to be a good way to try out some games new to me.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Benjamin T. Rancourt
United States
North Carolina
flag msg tools
mbmbmb
Hey Stephen,

It appears that AiAi, at least in GUCT mode, is not actually evaluating moves in Catchup. Through the whole game, it rates every position at exactly 50%. As a result, it plays randomly. GUCT is supposed to be doing a series of play outs, correct? If it was actually doing that, it wouldn't give 50/50 every time, even it it's not good at Catchup. Correct?

I tried giving it more thinking time, which made it much better at Ecalper, but this didn't work for Catchup at all. Still random play.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
Quote:
I tried giving it more thinking time, which made it much better at Ecalper, but this didn't work for Catchup at all. Still random play.

Which ruleset was this? Ketchup #3, #4, or Catchup (#6)

 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
LetsBeGood wrote:
It appears that AiAi, at least in GUCT mode, is not actually evaluating moves in Catchup.

Fixed; it wasn't noticing the end of the game in the random playouts - one optimisation too far!
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Benjamin T. Rancourt
United States
North Carolina
flag msg tools
mbmbmb
mrraow wrote:
LetsBeGood wrote:
It appears that AiAi, at least in GUCT mode, is not actually evaluating moves in Catchup.

Fixed; it wasn't noticing the end of the game in the random playouts - one optimisation too far!
Great, thanks!

I have downloaded the new version of AiAi, but so far I have not been able to use it because it won't display the game board. When I open AiAi, it usually starts with a completely black or white window. Hovering the mouse over buttons, the game description, or the menu makes them become visible, but the board has never become visible for me.

That said, I'm very happy with the larger text and icons for high resolution displays. It will be so much nicer to use, once I can use it.

Also, I noticed a Distrify game bug: when I am one turn away from winning, AiAi declares "Draw in 0 moves" and ends the game. (Since there wasn't anything about Distrify in the latest announcement, I assume this has not changed.)

Thanks again for all of your work on this.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
LetsBeGood wrote:
I have downloaded the new version of AiAi, but so far I have not been able to use it because it won't display the game board.

Curious. Did you download over the top of the old version? If so, look for a file called [username].mgc and delete it, then try again.

If you run ai ai from the command line, you may well see an exception... that would also be useful information.

If not, I'll have to dig further What operating system do you have? Also, what's your screen resolution?

Quote:
Also, I noticed a Distrify game bug: when I am one turn away from winning, AiAi declares "Draw in 0 moves" and ends the game. (Since there wasn't anything about Distrify in the latest announcement, I assume this has not changed.)

Thanks. I'll see if I can reproduce the problem tomorrow. If not a save game file would be useful, so I can replay the game ans see what happens.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
If you have been having problems with games not loading,or the AI not working, please download the latest version. Ai Ai was running out of memory on a lot of machines (anything with less than 4Gb ram would struggle!), and I have spent some time reducing the memory requirements.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Craig Duncan
United States
Ithaca
New York
flag msg tools
designer
Avatar
mbmbmbmbmb
I think I am still having a memory problem.

When I click in the jar file I get the following message:



Clicking OK yields:



I tried a command prompt and got this:



Any advice? I am running Windows 10 version 1607
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
Hi! The second error message is one I have only seen from people who try to run Ai Ai without extracting the zip file. Try the following:

1. Please make sure you have unzipped the "ai ai.zip" file you downloaded somewhere to your hard disc, with the folder structure intact.

2. Navigate to the folder containing "ai ai.jar" using explorer

3. Hold down CTRL-SHIFT and right-click in the folder

You should now have a pop-up menu with an option that says "Open Command Window Here"

4. Click on it.

5. Type "java -version" and return

This should return the java version, which should be 1.8.something

6. Now try running:
java -jar "ai ai.jar" -Xmx1024m
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Craig Duncan
United States
Ithaca
New York
flag msg tools
designer
Avatar
mbmbmbmbmb
It took me a while to find time to try this, but I just now tried it.

I believe I unzipped the folder with the file structure intact, and went ot the folder containing the jar file. Here is a screenshot:



But the ctrl-shift right click trick didn't work for me. Doing this trick at a random spot in the folder just gives me the normal "View / Sort By / etc." menu that one gets without any ctrl-shift right click.

Doing the trick on the jar file icon itself just gives me the usual "Open / Cut / Copy" menu -- the same one that appears without the ctrl-shift.

So, ctrl-shift doesn't see to be doing anything for me.

Sorry to be so ham-handed at computering!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Stephen Tavener
United Kingdom
London
England
flag msg tools
designer
The overtext below is true.
badge
The overtext above is false.
Avatar
mbmbmbmbmb
cdunc123 wrote:
Sorry to be so ham-handed at computering!

No worries. On day I'll get around to converting to JavaFX, which will allow me to add a proper installer.

Until then, ...

Please download the following:
https://drive.google.com/open?id=0B4VtOLJ4uh1UT2l6a2VHRlVVR1...

... and copy it to the same folder as "ai ai.jar".

You should just be able to double-click on it to launch ai ai with the correct settings.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Craig Duncan
United States
Ithaca
New York
flag msg tools
designer
Avatar
mbmbmbmbmb
That worked! Thanks, Steven. Oddly, Ai Ai started up on the Santorini page:



But I can use the menu bar at top to choose other games.

In short, all good!
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Selamat pagi,
Sweden
flag msg tools
Help, I'm being held prisoner in an overtext typing facility! I don't have much time, they could find out at any m
badge
I'm that weirdo whose number of badges sold prior to yesterday Bail Organa is keeping track of
Avatar
mbmbmb
For Tak:

When moving a stack, you cannot drop off less than 2 pieces in a square.

You have omitted the opening moves with reversed colours.
2 
 Thumb up
5.00
 tip
 Hide
  • [+] Dice rolls
1 , 2  Next »   | 
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.