| Author | Message | ||
| Mats W (Kålroten)
|
Crusade-promotion. Please improve this game by, perhaps, making the board bigger and by introducing large empty areas (perhaps also areas that cannot be trod on). You could add "bounce-movement" capability (as in my "Bouncers"). Then you actually retain the pieces' inability to move by themselves. If they cannot capture they can only move by bouncing, i.e. they must have a piece behind the back in the alignment direction (so stalemate is still a possibility). Moreover, perhaps you ought to add a variant where the player himself decides where to drop the pieces (as in Stratego). My point is that the game now is overly tactical. My impression is that it's a continuous tactical slaughter and on account of this one can hardly beat the computer (which makes me pissed!). I reminds me too much of Reversi or Blobs, which are games which I find incomprehensible. I think this game could evolve into a really good battle game if it becomes less tactical and more strategical. | ||
| Mats W (Kålroten)
|
I am being really over-ambitious now, but I've tried combining this promotion-idea with bounce-moves, like I proposed. It is an interesting battle game, but whether it's any good I don't know. I need some input. It's a challenge to make a good battle game for Zillions. There are certain advantages with bounce-movement, and approach-capture, which makes these methods suitable for battle games. Bounce-movement allows a natural promotion of the pieces' power. Approach-capture makes possible blockade-moves, that is, putting a piece in front of an enemy piece, preventing its motion in that direction. One can also withdraw a piece from an enemy piece, thereby disallowing its movement in the opposite direction. It creates a multi-faceted game with many positional factors. But maybe it's cumbersome, I don't know. It's preliminary and can be downloaded here: BounceBattle. | ||
| Keith Carter (Keithc)
|
Hello Kalroten. Thank you for the feedback. One of the neat things about Zillions is the ease with which you can modify a game to try variations or to introduce an additional concept. Go for it. I can see that our gaming tastes differ. I enjoy the Reversi family of games and Blobs and Blobz are the two games I play most often against my wife. FWIW if I added a two space jump move to Crusade it would be a complete reversal of Blobs with the same sets of moves being used to empty the board rather than fill it. So if Blobs does not work for you then Crusade probably won't either. The Crusade games skip a development phase and use a continual tactical slaughter by design to create a game that I can start and finish during a 15 minute work break. I find it easier to beat the AI in Crusade-promotion than Crusade since winning the first area contested means having promoted pieces to help win the second area contested and so on and so on. For a quick test of how a bigger board with large empty areas would introduce a more strategic element you could keep the same 9x9 board and reduce the number of pieces dropped from 81 to 51. To a significant extent a more open board version has already been tried with the Conversion Chess games. I limit my games released for Zillions to games Zillions plays well. Zillions does not play well a variant where the sequence of dropping pieces goes beyond the number of moves Zillions is looking ahead. A Stratego variant of Crusade would probably be a good game but it would not be a good Zillions game. (Karl, please jump in if you think I am wrong). The second and third variants in Crusade-promotion really do ramp up the tactical difficulty to the AI's advantage. They were included because there is a Crusade player out there who requested harder variants. I dropped a fourth variant that filled the board with major chess pieces. Keith | ||
| Mats W (Kålroten)
|
> I limit my games released for Zillions to games Zillions plays well. That's not a valid criterium! Good games are good, regardless if Zillions plays them well or not. Mats | ||
| Keith Carter (Keithc)
|
>That's not a valid criterium! Good games are good, regardless if Zillions plays them well or not. Good games that are no good with Zillions should be released on Zillions? I use zillions to play games against the AI. Why should I spend hours developing a game in zillions format that is AI dead? By your sole criterium of game goodness a good real time strategy game should be released to the Zillons site because good games are good regardless of whether Zillions plays them well or not. How about a good multiplayer diplomacy game or a good role playing game, or a good bidding based card game? There is nothing about a game being good that logically requires its release in a zillions format. Zillions, while a remarkable piece of software and broadly based, is not an all encompasing game site. There are other more appropriate sites to release good game designs that do not work well with Zillions. | ||
| Mats W (Kålroten)
|
My point is that you can actually test a game, whether you like it or not, etc., even if Zillions plays it badly. This game could then become popular, or be implemented, outside a Zillions context. Moreover, even if a game functions badly in Zillions, the ideas behind the game could inspire new constructs of Zillions games that Zillions plays well. A good example is Thrall, which Zillions plays badly, but the idea behind it is inspiring. It is reminiscent of two gladiators fighting in an arena. It could be implemented as a physical boardgame and sold in game stores. | ||
| Greg Schmidt (Gschmidt)
|
If you guys don't mind, I'd like to share some of my thoughts on the topic of whether or not it makes sense to implement a game in Zillions based on the level of play the AI is capable of. o I think the threshold for competitive AI play is often subjective. Unless the game play is either random or suicidal, there can be a huge range of subjectivity involved in assessing what a decent AI is. o There are some Zillions games for which I now regularly beat the AI, but I'm glad that there was some level of AI there in the first place in order for me to begin to hone my skills. o Why bother implementing chess in Zillions when there are better dedicated chess programs available? I'm not offering an opinion here, but merely pointing out that the same argument for weak AI could apply as well to other "established" games implemented in Zillions. In other words, weak as compared to what - the best currently AI available for that particular game? That's a pretty tall order. o I've played some games where the AI is too strong. For some simple games, Zillions is able to solve the entire game. When the AI is too good, that is not much fun either, but I still wouldn't spite those games. o Some games, such as TwixT were specifically implemented in Zillions with the caveat that play should be between humans. In that case, Zillions is the vehicle for quickly implementing and managing games between two humans. I wouldn't spite those games either. o My kid, who's just learning about strategy games, sometimes plays Zillions games. He loses interest if he's constantly being trounced. In that case, it may actually be an advantage that the AI plays poorly.  o Even if the AI plays badly, the game itself may have merit and making it available via Zillions introduces it to the public. Zillions can be considered as a medium for exchanging game ideas. o The AI for a game could be improved over time. I can think of a variety of reasons: __ 1- You, or someone else, may later discover improvements to the zrf. __ 2- A new release of Zillions could lead to improvements. __ 3- Someone chooses to write a custom engine for the game. Given that one believes they have a solid game idea, my preference would be for them to publish it via Zillions anyway even if the AI itself is not considered to be up to par. Just state in your documentation that the game is "best played by humans" - as others have done. | ||
| Keith Carter (Keithc)
|
One of the reasons I limit my Zillions of games designs to games Zillions plays well is so that I can test the game. I come from the Steve Jackson Games school of game development: play test, play test, playtest. I play a game dozens of times, make adjustments, play dozens of more times etc. until I think I have the game situated as best I can and then I play it scores of times more before I release it. While I am able to get some human vs human testing in, I can't get in nearly enough for my approach to game design. A game that Zillions plays well is much easier for me to test. | ||
| Keith Carter (Keithc)
|
Theoretically they can. Games that Zillions plays well can too. Maybe even more likely so since you can play Zillions to experience the concept and have existing code as a starting point for the new constructs. I would rather put my limited time resources into AI functional games. Do you know of any examples where a game that Zillions plays badly inspired new constructs that Zillions plays well? | ||
| Keith Carter (Keithc)
|
Greg Schmidt Your ideas are welcome. Discussions have been so thin the last few years on this board it is nice to have multiple threads with active discussions. You make so many points its going to take some time to respond. I will try to break it down into smaller pieces. <o My kid, who's just learning about strategy games, sometimes plays Zillions games. He loses interest if he's constantly being trounced. In that case, it may actually be an advantage that the AI plays poorly. I have not run into a problem with the Zillions AI being too strong. I can dumb it down by reducing its thinking time, fix the number of moves it looks ahead to as little as 1, increase the variety of its moves and limit its play strength. | ||
| Keith Carter (Keithc)
|
Sure there is some value there. The game Cannon is a good example. Zillions plays Cannon fairly well. It is a very good trainer for beginning players to help crystalize in their minds how the elements described by the rules work together. However, there are things about the game that Zillions does not get for various reasons and it falls by the wayside as an opponent. Wouldn't it have even more value if the AI were stronger and it was able act as a skill honer longer? If it were my game, and I only had enough time to develop it along one line (which is my situation) I would post and support it at the much more active (player and site owner) Superdupergames.org. More bang for the development buck. | ||
| Keith Carter (Keithc)
|
Whoops. Closing the bracket made the quote drop. Here is the post in complete form. <There are some Zillions games for which I now regularly beat the AI, but I'm glad that there was some level of AI there in the first place in order for me to begin to hone my skills. Sure there is some value there. The game Cannon is a good example. Zillions plays Cannon fairly well. It is a very good trainer for beginning players to help crystalize in their minds how the elements described by the rules work together. However, there are things about the game that Zillions does not get for various reasons and it falls by the wayside as an opponent. Wouldn't it have even more value if the AI were stronger and it was able act as a skill honer longer? If it were my game, and I only had enough time to develop it along one line (which is my situation) I would post and support it at the much more active (player and site owner) Superdupergames.org. More bang for the development buck. | ||
| Keith Carter (Keithc)
|
Lost the quote on some of my other posts too. I am reposting the complete post. <Moreover, even if a game functions badly in Zillions, the ideas behind the game could inspire new constructs of Zillions games that Zillions plays well. Mats W Theoretically they can. Games that Zillions plays well can too. Maybe even more likely so since you can play Zillions to experience the concept and have existing code as a starting point for the new constructs. I would rather put my limited time resources into AI functional games. Do you know of any examples where a game that Zillions plays badly inspired new constructs that Zillions plays well? | ||
| Keith Carter (Keithc)
|
Another dropped quote so I am reposting. <My point is that you can actually test a game, whether you like it or not, etc., even if Zillions plays it badly. Mats W One of the reasons I limit my Zillions of games designs to games Zillions plays well is so that I can test the game. I come from the Steve Jackson Games school of game development: play test, play test, playtest. I play a game dozens of times, make adjustments, play dozens of more times etc. until I think I have the game situated as best I can and then I play it scores of times more before I release it. While I am able to get some human vs human testing in, I can't get in nearly enough for my approach to game design. A game that Zillions plays well is much easier for me to test. | ||
| Keith Carter (Keithc)
|
<o Why bother implementing chess in Zillions when there are better dedicated chess programs available? I'm not offering an opinion here, but merely pointing out that the same argument for weak AI could apply as well to other "established" games implemented in Zillions. In other words, weak as compared to what - the best currently AI available for that particular game? That's a pretty tall order. Greg Schmidt I believe that Chess, along with the many other known and public domain games, is included as a selling point for the mass market. The Zillions AI plays chess well enough to provide an adequate opponent for the mass market. It also provides an in place zrf script as a starting point for chess variants by budding scripters. I released a chess variant in Zillions format because Zillions was the best program I knew of for it. If I had known of another better program I would have used that instead. Zillions? Maybe later if I felt there was the right combination of time available and strength of AI play. | ||
| Keith Carter (Keithc)
|
<o Some games, such as TwixT were specifically implemented in Zillions with the caveat that play should be between humans. In that case, Zillions is the vehicle for quickly implementing and managing games between two humans. I wouldn't spite those games either. Greg Schmidt I am not intending to spite those games. I have for years played games using Zillions as a pbem engine. But human vs human games represent less than 1% of my Zillions gaming. Most of the people I knowplay almost all of their games against the Zillions AI. Again, I get more bang for the development buck by only developing games for Zillions that Zillions plays well. Other sites, like, again, Superdupergames.org are a better choice for managing games between two humans. | ||
| Karl Scherer (Karl)
|
Just a small point that stayed unmentioned: With ever increasing computer speed and efficiency Zillions AI gets automatically more "intelligent". Hence using the status quo as a reference of the capability of the Zillions AI is of very limited use. | ||
| Keith Carter (Keithc)
|
<o Even if the AI plays badly, the game itself may have merit and making it available via Zillions introduces it to the public. Zillions can be considered as a medium for exchanging game ideas. It does introduce it to the public. Though as a medium for exchanging game ideas Zillions is pretty limited. There has been more discussion on the discussion boards recently but overall Zillions is fairly inactive as a medium for exchanging game ideas. Small number of active participants and not many discussions going. For a game that Zillions does not play well Zillions is not a competative medium for exchanging ideas given my limited gaming time resources. | ||
| Keith Carter (Keithc)
|
<Just a small point that stayed unmentioned: With ever increasing computer speed and efficiency Zillions AI gets automatically more "intelligent". Hence using the status quo as a reference of the capability of the Zillions AI is of very limited use. Karl Scherer Hey Karl, Nice to see you jump in. Thats a kind of sort of thing. For most games that I have seen Zillions not "get" the ability to look even further ahead over time won't make a significant difference. Zillions will look another ply or two ahead (or more)and still not get the game. It is not a lack of speed or efficiency that is holding the AI back in the opening and mid-game. Increased computer speed and efficiency would make a difference where it allows Zillions to see the end of the game even with games that the AI plays poorly. | ||
| Derek Nalls (Omegaman)
|
Three problemmatical factors of importance to our craft deserve mention: 1. The higher the branching factor of a game, the less effectively it can be played on a state-of-the-art computer running any program (including ZOG) even with extremely long time controls (due to a combinatorial explosion taking place past its deepest, completable ply). Computer playtesting is critical to proving the stability and overall quality of a game. 2. The higher the branching factor of a game, the less impact future advances in computer technology (esp. CPU speed) will have. For example, Hex Chess SS has an est. branching factor of nearly 300. Thus, a 9000-fold increase in PC's owned worldwide would only yield the benefit of a minimal quality play, 8-ply search completion within a reasonable time where currently, nothing deeper than an unacceptably low quality play, 6-ply is practical (throughout the game). For reference, a 10-ply search is needed for solid play and a 12-ply search is needed for incisive play. 3. The higher the branching factor of a game, the better its potential quality in oft-measurable terms such as non-triviality, tactics, strategy, theoretical depth, fairness, balance, equality, stability, symmetry, etc. Conclusions- At least with board games of the chess variant type, the incisiveness with which the Zillions program plays a game generally varies inversely with its potential quality and importance. Unfortunately, games with extremely high branching factors are much riskier creations since, by being out-of-reach for modern computer playtesting technology, they cannot be absolutely proven to be well-designed or even, free from fatal or disastrous flaws. My opinion is that serious chess variant inventors should never publish anything with a branching factor less than 100-200 [By comparison, standard FIDE chess is 35.] yet must do everything within their restricted ability to assure the quality of their gameworks in the absence of the most useful, conclusive test (i.e., incisive computer AI programming and play). In a word, when working at such a high level of complexity, it is relatively easy to invent junk which is extremely difficult to detect as being junk. Ultimately, whichever opposite direction we choose creatively, it places ALL of our "very best works" (borne of much work and insight) in great jeopardy of someday being correctly, widely identified by wiser, future generations as nothing more than "time-wasting junk". | ||
| Greg Schmidt (Gschmidt)
|
Keith Carter wrote: [I am not intending to spite those games.] Sorry, I did not mean to imply that you were. No offense was intended on my part and I hope no offense was taken. I was mainly trying to say that if someone developed a decent game for which Zillions played poorly, I personally would not think less of that game. As Mats pointed out and I must agree, Thrall seems to fit that category well. Thrall appears to be a great new game, even if Zillions plays it marginally. And yes, I was disappointed that the AI ran out of gas quickly, but after playing the game, felt very glad to have been introduced to it. And BTW, Mats produced a very nice alternate board for the game as well as raising the yet to be accepted challenge "can anyone improve the AI"? I think another example is the game "Random Links", an interesting connection game, but one which Zillions just doesn't seem to "get" as connection games are problematic for Zillions which wants to deal mainly with positional and piece advantage. [While I am able to get some human vs human testing in, I can't get in nearly enough for my approach to game design. A game that Zillions plays well is much easier for me to test.] There's no question that there are definite advantages to having Zillions plays your game well. That makes me wonder, how common it is for game developers to tweak their rules, just so that Zillions will play a better game. I know I've been guilty of that. But now I digress... [Do you know of any examples where a game that Zillions plays badly inspired new constructs that Zillions plays well?] Well sort of, but not really. There was a recent discusssion of connection games and potentially a Zillions plug-in geared towards them. Theoretically, one could write a completely general Zillions plug-in that is designed specifically for connection games. Practically, I fear that no one will bother to write one since it would require duplicating most of the Zillions engine in order to do that. What's really needed is a simple way to replace only the evaluation function and I hope that someday we'll see a version of Zillions that addresses this issue. And now at the risk of further rambling, I will add that I am frequently amazed by the level of ingenuity demonstrated by the Zillions developers. I think that some games have inspired new insights in how to manipulate the AI to play these games. Games that may have previously been considered unplayable. It looks like the whole notion of "bonus systems" has been developed to transform a game into one which the AI can deal with. So while I can't really say that I'm aware of new Zillions constructs, I can say that there have been new techniques that developers have layered on Zillions in order to circumvent these limitations. In my own case, and with some help from this discussion board, I was able to implement a novel game where having one's own pieces captured is actually rewarded. I think of it as running my game through a transform into the "Zillions domain" (Z-Transform?). Not the ideal way to approach the problem, but it frequently works. It reminds me a bit of the advance from structured programming to object oriented programming. There was no real technological breakthrough, it was more of a mental breakthrough, the so called "paradigm shift" that lead to new ways of thinking about a problem. I see the same thing going on here. [I have not run into a problem with the Zillions AI being too strong. I can dumb it down by reducing its thinking time, fix the number of moves it looks ahead to as little as 1, increase the variety of its moves and limit its play strength.] Take a look at some of the games by Markus Salo. He is after simplicity and he has done a wonderful job inventing new minimalistic games. However, for these types of games, Zillions is essentially able to solve the entire game tree and thus the AI is invincible. [Wouldn't it have even more value if the AI were stronger and it was able act as a skill honer longer?] Yes it would, I would just hope that good games with weak AI (such as the examples above) will still get published. Derek Nalls wrote: [Ultimately, whichever opposite direction we choose creatively, it places ALL of our "very best works" (borne of much work and insight) in great jeopardy of someday being correctly, widely identified by wiser, future generations as nothing more than "time-wasting junk".] You could be right about that, but for me in the here and now I ultimately have to ask the basic question "Do people enjoy playing the game?". And finally, Keith Carter wrote: [Your ideas are welcome. Discussions have been so thin the last few years on this board it is nice to have multiple threads with active discussions.] Thanks Keith, and I agree with your observations. It's nice to see the recent flurry of insightful discussion on this board by many talented individuals, something I appreciate as well. | ||
| Mats W (Kålroten)
|
>Games that may have previously been considered >unplayable. It looks like the whole notion of >"bonus systems" has been developed to transform a >game into one which the AI can deal with.....(Greg) It would be handy with a FAQ file for Zillion programmers about programming techniques. I had to invent this very bonus technique myself for Bario. If I had known about this in advance this project would have caused me less headache (but I'm just about ready now). The problem in Bario is that the engine refruses to transform his Quantum pieces because they are so hideously strong where they sit. This is a correct evaluation, but he can't win if won't do anything. So I present the engine invisible "tidbits" on squares outside the board, as a reward for transforming Quantums. It works fine, and the engine has started playing chess, although very reluctantly. | ||
| Keith Carter (Keithc)
|
<<[I have not run into a problem with the Zillions AI being too strong. I can dumb it down by reducing its thinking time, fix the number of moves it looks ahead to as little as 1, increase the variety of its moves and limit its play strength.] Keith <Take a look at some of the games by Markus Salo. He is after simplicity and he has done a wonderful job inventing new minimalistic games. However, for these types of games, Zillions is essentially able to solve the entire game tree and thus the AI is invincible. Greg I have played some of Markus Salo's games. If you set the Zillions computer player to a fixed look ahead of only 2 or 3 moves then Zillions will not solve the entire game. The game I had in mind when I mentioned being able to dumb down the AI was 4x4 tic-tac-toe (which I did for my kids). That game is even simpler than Markus' games. | ||
| Keith Carter (Keithc)
|
<There's no question that there are definite advantages to having Zillions plays your game well. That makes me wonder, how common it is for game developers to tweak their rules, just so that Zillions will play a better game. I know I've been guilty of that. But now I digress... Greg In my case it is more extreme than that. I design a game by combining game elements that I already know Zillions handles well. I don't have any game designs that Zillions does not play well to release to Zillions because I don't try to design such games. It is a matter of resources. I work 66 hours a week. I only have a handful of hours each week for games and an approach to game development that requires scores if not hundreds of hours of time. Also any game development time takes away from game playing time. So I go with what will give me my best return on my time investment and that, for the last 7 years or so, has been games that Zillions plays well. For those of you with more time go ahead and release stuff to Zillions it does not play well as an exchange of ideas. Mats W (Kålroten) - go ahead and make and post improvements to Crusade-promotion that you wish to see if you have the inclination and time to do so. |