Welcome, Guest. Please Login or Register.
Feb 23rd, 2020, 2:40pm

Home Home Help Help Search Search Members Members Login Login Register Register
Arimaa Forum « Christian Freeling on inventing games (part 2) »


   Arimaa Forum
   Arimaa
   Off Topic Discussion
(Moderators: christianF, supersamu)
   Christian Freeling on inventing games (part 2)
« No topic | Next topic »
Pages: 1 ... 7 8 9 10  Reply Reply Notify of replies Notify of replies Send Topic Send Topic Print Print
   Author  Topic: Christian Freeling on inventing games (part 2)  (Read 153552 times)
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #120 on: Aug 30th, 2017, 3:56am »
Quote Quote Modify Modify

 
Experience suggests that the community usually prefers to see a more objective indication of a game's qualities than the inventor's assurance. So forget that I saw Starweb as being a strategy game right from the start and consider this. Michael Howe raised the question at BGG whether a 'minority strategy' would be viable:
 
Quote:
So I think an interesting question then becomes: can a human expert beat another human expert by only taking 8 corners?  
Do you think it would be a viable strategy between top players?

The latter is important because in beating Cameron's AI it even proved a viable strategy when playing second with a 7-11 minority. Doubtlessly that would change against a stronger AI.  
The viability of a minority strategy is important because it largely increases the diversity of strategic planning. It means that it's not all 'corners first and see what happens'. Note also that the pie rule in itself already begs the question whether or not to take a corner.  
Starweb strategy is all its own and just as different from Hex or Havannah as these two differ themselves. Tactics on the other hand are strangely familiar, like the taste of a new kind of citrus fruit: it's different from the known ones but it's citrus, no doubt.  
 
I found worthy opponent in Tony van der Valk. His current ratings at LG for Hex and Havannah are 2144 and 2162 respectively and he dedicates a significant part of his life to promoting Hex at hexboard.com. In terms of familiarity with hex connection games his status is beyond argument. If your browser supports java, here's the 'minority strategy' test game:
 
Christian Freeling - Tony van der Valk

I opened at Q14 and Tony declined a swap and took White. This is the position after Black sacrificed a corner with L17 and White took it at M19. Black's stone cuts through White's '6-group' and this will be the opening theme.
White did also decline a corner with Q12 and Black replied with R14, connecting and defending the top left group.
 
Black's strategy is to unite the six corners on the left. White goes into the centre by threathening to connect the two '3-groups' on the right.

White's threat has given him a 4-group on the right, provisionally isolated by the defending black group that has connected with the bottom 2-group. White now aims to connect the top and bottom 2-groups and black's big one on the left is not yet safe.

So here White has provisionally connected his two 2-groups, but with two 4-groups and two singles his score adds to 22. Black with a 6-group and a 2-group would get to 24, so a cut attempt was required. Because of the leftmost white stone a connection there had to be secured at the cost of a white escape to the centre, threathening to cut off the D1 corner. And that's a 5-points reduction of White's score, quite enough to tip the balance.  
Obviously Black's reply will be at K9 (a 'cup' in Havannah) and White can push either up or down the long file. A 'tight' push leads white to a black 'triangle', either at M10 or H5, so upward Black must immediately deviate diagonally to prevent a cut. Downward it's more complicated. Ton told me that on at least three occasions he had played the wrong move and this adventure may hold one of them.

Here Black moves into a white triangle. Allowing to connect to the D1 stone means nothing has changed and it's still 24-22 for Black. So White connects his bottom 2-group to the G1 single and allows a cut in his central 4-group. But now that D1 is cut off, Black has a 5-group, a 2-group and a single. That's 19 points. However ...
 
Black's cut allows him to connect the 5-group to A7 at the cost of having his 2-group cut. Starweb connections are not equivalent! In the final diagram Black has the 6-group that he aimed for and two singles. That's 23 points. White has a single, a 2-group, a 3-group and a 4-group. That's 20 points.
This is one game, and I'm quite sure it was played 'reasonably well' on both sides. It doesn't prove Starweb is a strategy game, but the success of the here employed minority strategy would at least suggest it is.
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #121 on: Nov 26th, 2017, 12:32pm »
Quote Quote Modify Modify

 
On 'inside out' inventing. A little essay I wrote, here's the declaration of intent:
Quote:
Inventing abstract strategy games hardly ever makes it to someone's profession. Public interest is usually modest and among publishers it is even less, so why bother. However, the emergence of the world wide web has made that there are far more inventors now than there were a few decades ago. The numerous new platforms where abstract games can be played, allow them to be presented to the general public more easily and in consequence inventing them has been evolving. I fear that this is for the larger part fuelled by a vision of publishing a game for fortune and fame. That's pretty naive and totally legal. But please don't make it bad games.
For a small minority inventing abstract strategy games is art for art's sake, a quest sometimes bordering on obsession. Insofar as it may be called an 'method', and insofar as fuelling obsessions may be called 'beneficial', they may benefit from 'inside out' inventing. It may render a game in seconds, and the result is usually 'modification proof' with hardly any issues to be resolved. It may also encounter some scepticism by the majority of inventors because of that.
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #122 on: Dec 12th, 2017, 6:38am »
Quote Quote Modify Modify

Here's another thing. In writing the essay mentioned above I had to revisit the primal vision that led to Mu: the core behaviour of the board. That vision came in the context of a weirdly flawed game called Atlantis that employed an 'explosion' mechanism.
 
Mu straightened out the concept but the mechanism retained the inherent chain reactions that make it feel like something between a strategy game and a pinball machine.  
 
So when I revisited that primal vision I had a fleeting moment where I thought "isn't there another mechanism to clear the top layer, create the wall and grow reinforcements in the process"? And it stuck.
 
The game is called Stiles and I will not publish the formal rules till in 2018 (because it's nice to have one in 2018 and because I've other things to do) but the rules now published at BGG are basically complete. But they are written 'inside out', starting from the core behaviour, as they developed in my mind.
« Last Edit: Dec 14th, 2017, 4:55am by christianF » IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #123 on: Dec 23rd, 2017, 6:35am »
Quote Quote Modify Modify

 
I've renamed Stiles as Storisende. I had a game called Storisende (it's here) but ditched it about a year ago along with a couple of others. Nice name though and appropriate for a last game. Because whether or not another one drops out of the blue, this should remain the last one.
 
I'll officially publish it in January but here are the provisional rules, not yet public. They're no different from those in the Stiles thread, but now hopefully more clear and concise (without overdoing the latter).
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #124 on: Feb 11th, 2018, 11:07am »
Quote Quote Modify Modify

 
Recently Markus Hagenauer launced the
BEST COMBINATORIAL 2-PLAYER GAME OF 2017 - NOMINATIONS page at the abstract strategy forum at BGG and I entered Starweb.
 
In the context of the contest Stephen Tavener has set out to program as many of the entries as possible in his java based generic AI program, and Starweb is now included. The program plays a fairly mean game and you will need a good strategy and clever tactics to beat it. Enjoy!
 Smiley
 
Download Stephen's AI program
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #125 on: May 4th, 2018, 7:49am »
Quote Quote Modify Modify

Stephen Tavener's AI AI now includes four of my games in two more or less related pairs, Havannah & Starweb and Symple & Sygo.
 
The last couple of days I had to necessarily brush up my Sygo play because I got beaten four times in a row on an 11x11 board. Come on! Then I had a narrow loss and a narrow win on 15x15. I should maybe mention at this point that AI AI's thinking time was set to 1 minute per placement. Sygo may well have positions that allow some ten placements per turn. So I had to sit it out a bit, sometimes.
 
Anyway, then came 19x19 and here are images of a position and the subsequent position in the first game:
 


 
AI AI played black and his last placement is indicated. The group involved is dead of course, as is the black group in the top left corner. As are, alas, the two white groups top right.
The bottom diagram shows the position after I grew six groups, capturing the black group bottom right in the process. But I lost the game because I lost the centre.
 
The next game I had my revenge and more or less obliterated my opponent. This version is extremely aware of life and death and creating eye space, but as it turned out a little nearsighted. So sneakily circumfencing a target and then trapping it in an ambush turned out to be quite successful. I'm extremely happy with the implementation!
 
Time perhaps to remind the community that Sygo is a flip capture variant that:
 
- Has no cycles.
- Has an embedded and totally fair balancing mechanism, so it needs no komi.
- Is the only flip capture Go variant that needs no additional mechanism or rule to ensure life.
 
Now understandably Sygo was never played all that much but Stephen's AI AI plays a very good game, so if you want to learn to play it well, I can't at the moment imagine a better way to do that.
 
Download Stephen's AI AI
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #126 on: May 12th, 2018, 12:00pm »
Quote Quote Modify Modify

 
 
I'm very happy to announce the release of Storisende by Stephen Tavener  in his AI AI:
 
Download AI AI
 
Storisende has not yet been relased at mindsports.nl and has not yet been added to the BGGdatabase. I'm glad to reassure everyone that it has been played Smiley  
You can find the rules here (for now).
I'm especially glad about the graphics Stephen provided. The "only couples can breed" angle that is at the the core of the game, actually materialised!
The program has 5 playing levels, a time setting of 1/10 of a second to an hour, and is better on smaller boards. I suggest convex boards for beginners.
I play max strength and 1 minute per move on 14 modules in a convex layout and now I can win thanks to a couple of new tactics I learned from the AI.
 

 
A hint: Storisende is a territory game, but only if you manage to survive in the first place. So look at this.
 

 
Material is still equal, but the distribution is very different. White controls the Wall and Black gets suffocated.  
 

 
In a territorial context the game was already over, but the image is to show that growth (or options thereto) and extermination play a key role and that controlling the Wall beats 'controlling' territory. Black is stuck in a couple of patches, White has free range.
 
 
*****

 
 
I'm aware of the fact that my 'last game' has become something of a running gag at BGG and indeed, I can't quite exclude a game coming out of the blue like Starweb last year, but lately I prefer thinking about inventing rather than engaging in it.
 
Dieter Stein made Polar and then Urbino. Polar is a game based on core behaviour that almost inherently points to its rules. The invention process is inside out. Urbino is assembled and the core behaviour of Polar is one of its parts. The invention process is outside in, not starting with core behaviour but with external requirements not only for a good game, which is still a first requirement for Dieter, I'm sure, but also for a good product.
 
I notice that a natural game with high organicity, like Polar, is considered to be less likely to be succesful as a product than a far less organic but more interesting looking game like Urbino. To overexpose what I'm saying: If we didn't know that Go was such a great game, it would look rather dull, especially if compared to Chess.
 
I'm quite sure that for the average onlooker the initial behaviour of Urbino is far more interesting than that of Polar and that this makes it a better product. It's what the buyers want so it makes perfect sense. But it often obscures discussions about what makes a good game. I trust Dieter to have made Urbino a good game, but I could immediately see that Polar was a very good game. But a game whose behaviour you slowly have to learn. A game that might eventually turn into a 'sport weapon'. But not a game that behaves interestingly from scratch.
 
This, in short, is what I'm focussing on lately, and one of the reasons is practical: inside out inventing is far less bothered by rule changes, modifications, bugs, 'special' rules and the like than outside in. Starweb emerged in seconds and the one rule that solved both decisiveness and symmetric play followed the next day. Storisende is based on two kinds of core behaviour, the organic emergence of the Wall and the Focus way of moving stacks, with capture by replacement so that they all remain one colour. So when a birdie whispered "only doubles can breed" the whole thing became self explanatory and I've played it in my head for weeks on end because Ed was otherwise occupied.
But nothing changed of course and nothing has changed since. It's a great organic game incorporating about every aspect of abstract strategy. Its goal is territorial but its sub-goals include elimination, connection, race, blockade, general fights, local fights, invasions, careful manoeuvring and a host of specific tactics, some of which I'm sure are yet to be discovered. Although I've done my best to make it recreational, it turned out to be a game of deep strategy. That's why I can beat the AI, long term planning and an increasing awareness of the role and use of tactics, some of which I was taught by the AI itself. Thing is: you have to slowly learn its behaviour before it gives something in return. I'm extremely happy with it because I couldn't have wished for a better game to be my last.
 
P.S. Being able to beat the AI is relative. On a 7-modules convex board (max strength, one minute) I've lost more games than I've won.
« Last Edit: May 14th, 2018, 4:57am by christianF » IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #127 on: May 14th, 2018, 5:00am »
Quote Quote Modify Modify

 
I've finally played a fairly balanced game against the AI (maximum stregth, one minute thinking time) and thus I've finally seen it to the end. Storisende is a territory game but its goal can only be reached by winning the fight on the Wall, and this is basically an elimination game. The interaction between the two goals, one formal and one inherent, is so far as I know new in the realm of abstract strategy games. That's why I wrestled so long with visualising the endgame when I wasn't able to play yet. I knew all the time how it should end, but its nice to see it acted out.  
The final count according to the AI was Black 9, White 8. That's because Stephen went from the premiss that 'areas of territory never merge' which is totally true, but only during the game. After the game tiled cells, if any, count as territory as if the tiles weren't there. I've clarified that in the rules now and Stephen will adapt the AI shortly.
 

 
Considering the above it will be clear that Black will not move the man on the tiled cell at the top. As it is it claims 5 cells, should it move then it would result in two 2-cell territories with only one of them claimed. For the same reason the black man in the center claims 3 cells: the tile is considered to be absent.
 
That being said it will be clear that White has 16 cells and Black has 8. The black man on the Wall is able to claim the last unclaimed cell. As it is that's totally unnecessary, but suppose the score were 12-12 instead of 16-8.
 
In a strategic sense I went for a slight majority on the Wall. I always do but this time it worked out. Fights on the Wall have a a lot of 'trapping tactics' to support them. Exchanging is one thing, but isolating a man in one of the numerous niches, attacking it while simultaneously covering its escape cells is better and requires very delicate manoeuvring. This allowed me eventually to eliminate White on the Wall and use the survivors to claim enough territory. And if it had been 12-12 with one black man still on the Wall and one territory cell still unclaimed, now that would have been the way a balanced game should end (of course after the last man on the Wall has claimed the last cell).
But this was near enough Smiley
« Last Edit: May 14th, 2018, 5:02am by christianF » IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #128 on: Jun 20th, 2018, 11:19am »
Quote Quote Modify Modify

Storisende has been published at mindsports: Storisende.
 
I've also written a little essay: Organicity in Abstract Strategy Games  
that was previously published at Nick Bentley Games.
 
Also, Moving Forward Looking Back now includes Storisende, although for a different reason.
IP Logged
aaaa
Forum Guru
*****



Arimaa player #958

   


Posts: 759
Re: Christian Freeling on inventing games (part 2)
« Reply #129 on: Jun 21st, 2018, 10:24am »
Quote Quote Modify Modify

What's your take on this attempt at a playable go variant on a hex board?
https://nickbentley.games/2018/03/14/blooms-rules/ & https://nickbentley.games/2018/05/31/life-in-blooms/
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #130 on: Jun 21st, 2018, 12:05pm »
Quote Quote Modify Modify

on Jun 21st, 2018, 10:24am, aaaa wrote:
What's your take on this attempt at a playable go variant on a hex board?
https://nickbentley.games/2018/03/14/blooms-rules/ & https://nickbentley.games/2018/05/31/life-in-blooms/

 
I think Nick would agree that Blooms is quite exemplary for an inside out invention with a matching organic behaviour. The basic idea was to reduce the large number of liberties on a hex grid and Nick's solution is brilliantly simple. Characteristically the implications of Blooms' behaviour are, while still under investigation, surprising already, as the article shows. A great find!
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #131 on: Aug 7th, 2018, 12:17pm »
Quote Quote Modify Modify

 
Summer in the Netherlands started in May and barring the first three weeks of June it has been unrelenting. I enjoyed it but it has been going on a bit so I'm glad we're bound to go normal. Meanwhile my main occupation has been to keep an eye on the snakes, sunbathing in the garden. Big beach umbrella, lots of coffee and marihuana and good music (optional). No games except for playing Dameo and Storisende (now playable at mindsports) online.
 
However, I'd like to introduce you to a provisional Grand Chess program made by Eric Joyce. It looks great and it's provisional because he plans to implement considerable improvements. At the moment the program is rather slow. Here's a quote from the accompanying mail:
Quote:
A future milestone for this project (and one for which I can maybe earn some school credit) will be to set up a machine learning routine that can cultivate skill at Grand Chess independent of my programmed rules. I plan to attempt this using neural networks.
 
...
 
You had mentioned sharing a link to this implementation on Board Game Geek. There is now something to show, though I would issue the obligatory caveats that seem to accompany every web project: "work in progress," "not done," "under construction," etc. I've done a bit of test-playing before writing you this email, but there are bound to be more bugs to find. Any and all bug reports will be most welcomed, from you or from anyone else. If something does go wrong with the program, a screen shot would help me to reconstruct the position and determine the problem's cause.
 
I had hoped to reach this milestone quite a bit sooner, since I've already written an 8 x 8 Chess engine. The better part of my work on Grand Chess was in fact the discovery of bugs in the 8 x 8 code. It simply took the occasion of a variant on that code to flush those bugs out.

Enjoy Smiley
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #132 on: Oct 1st, 2018, 10:33am »
Quote Quote Modify Modify

Luctor et Submergo - playing Emergo against Stephen Tavener's AiAi
 
Stephen Tavener has implemented Emergo in his AiAi. It's the square version on a 7x7 and a 9x9 board (if played on the dark squares of a checkered board, the actual boards are 'orthogonalised').  
 

 
I've always maintained that you don't need a neural network to make an Emergo AI with superhuman capabilities, and indeed, Stephen didn't. In the last three days I've played more than a hundred 7x7 games against it, giving the AI 5 seconds for a move and needing on avarage about a minute myself. I won barely ten. On two occasions when I thought victory was mine, it managed a combination that resulted in self blocking (and thus a draw).
 
The program is ruthless but it is a great teacher. It shows the efficiency of the game by not wasting any time in wiping you off the board (while you actually remain on the board).
 
For beginners it looks like chaos, but there's order in it. It requires a form of visualisation that is very familiar for Draughts players. Although Emergo knows a placement stage (exellently handled by the AI) it is by all means a movement game. I'd almost say the ultimate movement game. Here you are, and a few plies down you're in a totally different position.
 
I highly recommend the progam. You won't believe the tricks it can show on a 25 cells board with 2x8 men. If you don't think it is possible to have a game that in human terms is inexhaustible, with such limited material, then you might want to reconsider.
 
You can download the program here.
 
It would be nice for interested posters to announce their first victory against the AI in this very thread. For me it will count as a proof that you actually understand the game.
 
P.S. As always, 'legal' is implemented perfectly.
IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #133 on: Jan 12th, 2019, 6:07am »
Quote Quote Modify Modify

Chess+
Nick Bentley and yours truly designed a new entrance to "The old Lady's Castle", with much the same arguments that Fischer gave for Chess960: better balance and a strongly reduced impact of accumulated opening theory. However, Chess960 is unsatisfactory in terms of architecture. I'll leave that without comment at this time.
 
Nick and I feel Chess+ is better, simpler and more integrated, because the entering protocol and the normal move protocol are interwoven. But feelings are for hippies. This is quite new and whether it delivers what it is intended to deliver remains to be seen. There is a chance Stephen Tavener will include it in his AiAi engine, and if and when that happens, test procedures over large numbers of Ai vs Ai games will be possible. 'Nuff said, here are the rules:
Quote:
Rules
- Unless otherwise indicated, the rules of Chess apply.
- The material is the same as in Chess.
 
Initial position
All pawns are placed on the same initial squares as in Chess: white pawns on the second row and black pawns on the seventh. Both White and Black have their pieces 'in hand' beside the board.
 
Entering and moving
'Entering a piece' means placing a 'piece in hand' on a vacant square behind an unmoved pawn of the player's own colour, with the option of moving that pawn one or two squares straight forwards or capturing with it, in the same turn. The conditions for making a move while a player still has a piece or pieces in hand are:
 
- If a player's own king is not yet on the board then the player on his or her turn must enter a piece.
- If a player's own king is on the board then a player on his or her turn must enter a piece OR make a regular move with any piece or pawn.
 
Kings
From the entering protocol it follows that players must enter their king on or before the eighth turn. A king may not be entered 'in check'. A player unable to enter his or her king if required loses the game.
 
- There is no castling.
- En Passant is as in Chess.
 

P.S. If you wonder about bishops, so did we. Smiley
 
Edit:
There's a thread running at BGG about the protocol and in it is a reference to this article by Frederic Friedel. In the article the author mentions a problem of Chess960 players who, confronted with a set-up neither did choose, sometimes start a game in a state bordering on catatonia.
 
May I add this consideration regarding the difference of both protocols. Chess960 confronts players with an opening arrangement that they didn't choose and that didn't evolve. And that's a big deal! A Chess+ opening evolves in every single game, so the typical cases of catatonia that are mentioned in the article will never occur: players grow into the game and an initial position doesn't have to be decided upon. We consider that another significant advantage.
 
 
« Last Edit: Jan 15th, 2019, 6:13am by christianF » IP Logged
christianF
Forum Moderator
Forum Guru
*****



Arimaa player #4019

   


Gender: male
Posts: 802
Re: Christian Freeling on inventing games (part 2)
« Reply #134 on: Feb 13th, 2019, 9:01am »
Quote Quote Modify Modify

And now for something completely different, Stephen Tavener's AiAi now supports online play, either live or turnbased, without a browser.
 
The program is java based and free. It features a lot of games including a couple of mine, in particular Storisende, a game I'm quite taken to, not because I haven't lost a game in it (I played maybe six or seven games) but because of its merger of a formal territorial goal and an embedded existential one, that dominate strategic thinking at different stages. I'm willing to play against anyone who enjoys a game with a wide array of strategies and tactics. But be warned, it's not a short game.
IP Logged
Pages: 1 ... 7 8 9 10  Reply Reply Notify of replies Notify of replies Send Topic Send Topic Print Print

« No topic | Next topic »

Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1!
YaBB © 2000-2003. All Rights Reserved.