Author |
Topic: Essay by Christian Freeling on inventing games (Read 522362 times) |
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender: 
Posts: 804
|
 |
Re: Essay by Christian Freeling on inventing games
« Reply #870 on: Jan 8th, 2012, 3:35am » |
|
on Jan 7th, 2012, 11:06pm, omar wrote:| Any news about when the Havannah challenge match is going to be? Have any dates been set yet? |
|
Neither the organizers nor I found it advisble that I should be part of the organizing committee, so frankly I don't know. But you can contact Ton van der Valk about it. He's the head of the committee. You can reach him via: HexBoard.com.
|
|
 IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re:
« Reply #872 on: Jan 15th, 2012, 6:54am » |
|
There was a question by Megajester concerning applet display at mindsports.
If you're using Explorer 9 you should put it in 'compatibility mode', presumably in the prefs (I never use Explorer, so that's a guess).
|
| « Last Edit: Jan 15th, 2012, 7:00am by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #873 on: Jan 15th, 2012, 9:17am » |
|
Ah, the quiet, the rest, and still a fair couple of views. Now that I'm retired (you'd better believe it) I'd like to return to a subject that this thread was about in the first place. It has everything documented and the period is mirrored in Late arrivals and final whispers, with 16 games listed, two of which matter (in so far as abstract games matter). Of these two Sygo is a fast, modern and streamlined Go variant based on othellonian capture. It's a great game, trust me, but there's nothing mysterious about its strategy or tactics: you go deeper, but in not entirely unfamiliar waters. The other one is Symple, and that's the one I'd like to talk about in this context:
on Mar 8th, 2009, 6:30pm, Fritzlein wrote:I'm intrigued by Freeling's claim that he (unlike normal people) can tell from the rules of an abstract strategy game whether or not the game will be good. He explicitly says that he doesn't need to be able to play at a grandmaster level to know what it will feel like to play at a grandmaster level. He begs us to take his word on four or five of his games that haven't yet been proven to be excellent games, and offers us Havannah as evidence because he knew it was a great game decades before a serious gaming community embraced Havannah and uncovered the glory that he knew all along would be waiting.
I have argued in other threads in this forum precisely that one can't tell a great game just from its rules. You must play to know. Arimaa is fabulous because of its emergent complexity, and by definition, emergent complexity can't be obvious from the start. If you can see something on the surface, it is not emergent. I can't believe that anyone, even a "game whisperer" could have foretold the intricacies of the camel hostage strategy from the bare rules of the game. The way we play and talk about Arimaa today would be impossible without the accumulated experience of the community. |
|
And believe me, I can well understand these arguments. But before I go into them, better than I did at the time I hope, I'd like to give a recap of Symple's discovery/invention that I posted at BGG in the AI Game Design: The Shibumi Challenge thread that was started by Cameron Browne as an experiment in automated game design. It raises such questions as "Did Ludi invent or simply discover Yavalath and its mechanism?", where Yavalath is a game the rules of which were the result of a game-creating/-finding program called Ludi. Or, "Do the rules make the game or vice versa"? Those are good questions. Here's the recap:
*****
Finding Symple
In 1983 Craige Schensted aka Ea Ea published Star. Thematically it combined 'static connection' (as in Hex) with 'dynamic connection' (as in LOA and more recently Ketchup and Yodd). It had groups called 'stars' that could grow and gather points by static connection, while being penalized a fixed number of points for being stars in the first place. So connect two of them and save a penalty, that's the 'dynamic connection' part. It had its charm and promise, but something wasn't quite right. The inventor sensed that and eventually came up with *Star, 'the game that all the others wanted to be'. Whether that be true or not, it was an improvement, and the phrasing suggests that *Star was seen as the 'quintessential' implementation.
A uniform game is organic if its mechanics and its theme align naturally, quintessential if every rule is 'necessary and sufficient' for its implementation. Occam's Razor is the key instrument to strip any preconceived 'assumptions' during the process of invention or discovery.
The one 'preconceived assumption' with regard to the quintessential implementation of a game is that it exists. You don't have to invent it, just find it.
This notion was what Ea Ea and Benedikt Rosenau and I had in common, but neither Benedikt nor I thought it was *Star. Neither was it Superstar or YvY, my own 'not quite' games on the theme.
Where is it?
October 2010 Benedikt mailed:
Quote:You are among the most cluesome abstract gamers/designers I know. I have been thinking a lot about a certain class of games recently and I want to share my thoughts with you, hoping for feedback.
There is the family that got started with Star, moved on to Superstar, *Star, and YvY. The games of this family share a pattern, namely:
a) you score by taking certain fields and
b) imposing a tax: the more groups one has in the end, the more is subtracted from the score.
I have three issues with these games ... |
|
I replied that these issues, whatever issues, didn't interest me that moment because I was caught up in Draughts' evolution, but Benedikt insisted:
Quote:| In other words, I am at the limit of design without heavy playtesting. I cannot achieve what I want. A telling experience. |
|
Tell me about it. The only thing that kept his quest at the back of my mind was the notion that we had missed something:
where was the quintessential 'group penalty' game?
What to leave out?
You're unlikely to find a quintessential implementation by adding something to anything. We found Emergo, by removing the 'checkers' part (initial position, forward orientation and promotion) of column checkers, watching the columns interact of their own accord. Thus it became the only column checkers game that did not emerge as the 'columnification' of an existing game. Less is better.
I looked at 'group penalty' from that angle: what to leave out. Not the 'dynamic connection' part, because it is at the basis of the group penalty idea: connect two groups and get one penalty less.
What about the static part? The "a) you score by taking certain fields" part? It began to morph:
"you score by taking certain fields ... by growing".
Yes, sure, implicitly ...
"you score ... by growing".
How about that. It's less and it's inherently logical! Could that be the key removal? Why should "certain fields" be taken? Isn't the most natural score of a group its size? Of course it is.
It implied a thematic shift from being a blend of static- and dynamic connection to a blend of territory and dynamic connection. Territory measured as groupsize.
This seemed a good step, but it didn't make a game yet. I was thinking hexagonally and 1-move turns: start growing groups and connecting them while preventing the opponent to do so, slow, same growth rate and regarding strategy ... is there any? Boring.
Occam's Irony
The problem with a 'vision' is that it's just an instant where everything fits. I had been wrapping my mind around the new basic concept for a few days, when one night while drifting off to sleep I got one fleeting glimpse of what was to become the Symple move protocol. I thought ... "so simple, what's wrong?" and submerged. The next day I remembered and without more ado mailed Benedikt:
Quote:You asked for it, so don't complain if this works 
Take a hexhexboard, two players, first move swappable.
* On his turn a player has two options, and he may use either or both or neither.
* Option one: Put a stone on a vacant cell, thereby creating a new group.
* Option two: Grow every existing group by one stone.
* Option one, if used, precedes option two.
* The game ends when the board is full.
* The count is the number of stones minus two points for every group. |
|
After Benedikt's suggestion not to use both options in the same turn, now the core of the main strategic dilemma, the game soon turned square, and as a bonus a highly sophisticated turn-order balancing mechanism was found to be embedded in the move protocol. Where could it go wrong?
Well, it already had. Occam's Razor, for te occasion turned out to be double edged. It made me remove the "taking certain fields" condition, but enthousiasm about that made me blind for an unfounded assumption I had unwittingly made. And when the game turned out to lack drama, and I was the first to acknowledge that, I held it for an inherent property. Quintessential games "are what they are" and all that. But it was a bug posing as a feature and I had failed to listen carefully. Sh!t happens.
Do the rules make the game or vice versa?
The vision I had has two interesting aspects to it. The first is that I saw the 'organism'. Do the rules make the game or vice versa? Both are possible, obviously. Ludi rendered Yavalath by trying random rule combinations, and Yavalath clearly has an 'organic' quality about it. Symple was conceived as an organism, and only the next day did I formalize what I'd seen. Reverse process, similar result.
The second aspect concerns my assumption. I saw groups working together to secure vacant territory too small to invade. Remember a new group gets a penalty, and depending on its height ('-2' had soon become '-2n') invasions may be unprospective. Implicitly I saw territory in a Go-like perspective. That was wrong. It led to the 'pass' as a legal move, it gave each his territory, area counting, it was mostly strategy with modest tactical means and it had 'no particular preference for drama'. Nothing was actually wrong with it, just that I was slightly disappointed with the result. A game would develop "like a big ship slowly heeling to one side", as I put it at the time.
Now where did I go wrong? I made an assumption that is inherent in Go. Go has capture and is implicitly first and foremost about vacant territory, area counting notwithstanding. That's how it slipped into Symple and required 'regulation'. But Symple is not Go and it is not inherently about vacant territory at all. Symple counts 'Points minus Penalty' and measures territory as groupsize, not as vacant points. Vacant points are, much like Othello, only meant to occupy. That's what I should have considered at the time - some form of compulsory movement. Fortunately Luis Bolaños Mures gave that push, and it shows how subtle one can screw up.
Here are some example games. The mcts_bot is programmed by Marcel Vlastuin.
We now think the Devil invented Symple and made us find it.
You may sometimes sense his presence if you play!
christian
Star *Star Superstar YvY Symple
|
| « Last Edit: Jan 15th, 2012, 5:14pm by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #874 on: Jan 16th, 2012, 8:34am » |
|
Invention or discovery?
Craige Schensted aka Ea Ea not only published Star but also, according to wiki, has 'invented' the connection game Y, together with Charles Titus. They may have been preceded by Claude Shannon who in any case found it independently. That's two parties 'inventing' the same game.
Piet Hein, according to wiki, has 'devised' the connection game Hex. According to the article the game was 'independetly re-invented' by John Forbes Nash. That's again two parties 'inventing' the same game.
Both games are arguably quintessential connection games. Isn't it fairer to say that each of the games was independently 'found' or 'discovered' by two different parties? I'm not trumpeting the Truth here, just painting the landscape. I'm pretty sure both 'inventions' started with suddenly 'seeing' the game. After that, how much is there to formalize? Boardsize for one, but considering that part of the 'invention' ... well. There's the corners to consider, and as it turns out, making them belong to both sides allows for the most concise phrasing of the rules that should govern the idea.
Being placement games without cycles or capture, it should also be clear that the first player has 'the initiative' and that in order to be fair there should be some balancing mechanism to compensate the second player. Enter the Pie rule, in itself a dubious case of 'invention versus discovery'. If these games were to miraculously disappear from the face of the earth and the minds of its inhabitants, would they emerge any differently?
Don't get me wrong, for the fast majority of games I'd go with 'inventions', but there is a small minority that is so elementary and self-explanatory, that some distinction between inventions and discoveries would in my view serve clarity.
|
| « Last Edit: Jan 16th, 2012, 8:36am by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #875 on: Jan 16th, 2012, 10:57am » |
|
Predictability
on Mar 8th, 2009, 6:30pm, Fritzlein wrote:| Arimaa is fabulous because of its emergent complexity, and by definition, emergent complexity can't be obvious from the start. If you can see something on the surface, it is not emergent. I can't believe that anyone could have foretold the intricacies of the camel hostage strategy from the bare rules of the game. |
|
To paint the landscape, again, Fritzlein appears to consider Arimaa, if not the measure of game inventing, at least as a game that is great by all measures. I'm not disputing that. But predictability, or indeed lack thereof, isn't the same in every game. What about Hex, Y, Havannah or indeed Go itself?
Nowhere to go but deeper
To start with the latter, so much is known now about the game, and nomenclature is so elaborate, that you can pick any Joseki and say that you "can't believe that anyone could have foretold the intricacies of it". That's not an entirely fair argument is it? It's right in a suggestive way: indeed, the intricacies of a particular Joseki that has been shaped and molded by experience could not have been predicted. But what about the emergence of such joseki? Once the basic behaviour of the game is clear, you might ask "where can it go?" (no pun intended). In Go the situation might not be immediately clear, because the basic concept involves capture, cycles and different ways of counting territory - aspects with consequences that are not entirely self-explanatory. But these being settled by concensus, where can the game go but deeper? And doesn't that include the natural emergence of cornerfights? What else can there be than ongoing refinement of strategy and tactics?
In Hex there's no capture, or cycles, or counting and no draw. Same question.
The answer is two-fold: either such a game is solved, as small small versions of all of the above games (excluding Arimaa (?)) are, or one must go deeper. In human terms, in doing so one will invariably find an ongoing refinement of tactics and strategy. That's as predictable as getting wet in the rain. No special powers needed. In the case of Hex one might also strongly suspect that strategy will go deeper, but not much wider, for lack of a clear dilemma that would allow different approaches and different styles of play. In Go, though more difficult to assess, standing at its cradle one might yet have suspected that its strategy would be not only deep but also wide as an ocean. That would have required some 'special powers' maybe. Such as 'seeing' the game in the first place.
Assembling games
There's one reservation I've always made: organic games can be very predictable in certain aspects of their behaviour because they're mostly uniform. Predictions regarding the behaviour of Chess variants or similar multi-piece games are usually not very reliable because they can be made to 'work' with relatively little effort. These games are 'assembled' from various pieces and rules, and never 'discovered'. The more experience one has in putting the right parts together, the sooner it will lead to presentable games. It may be equally hard however to discover their qualities as their bugs, if any. They need elaborate testing to get out 'of the woods'. Even Fritzlein for instance, feels Arimaa may not be quite beyond suspicion yet, if I remember correctly. Draws and cycles can be a b!tch.
But I'm writing this in the context of predicting Symple's behaviour and character in so far as I've come to understand it. Since draws and cycles don't play a role there they won't bother us.
The effort would be meaningless if I weren't convinced of Symple's enduring significance as a strategy game that is both deep and wide, and as a touchstone for AI development.
|
| « Last Edit: Jan 17th, 2012, 11:55am by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #876 on: Jan 18th, 2012, 7:50am » |
|
Predicting Symple
Symple is finite, decisive and has a highly sophisticated turn order balancing mechanism. The answer to question "Where can it go but deeper?" would appear to be "nowhere", and while Fritzlein's argument ...
Quote:| ... and by definition, emergent complexity can't be obvious from the start. If you can see something on the surface, it is not emergent. |
|
... may have some truth to it, it may also be the result of only looking at the surface. By seeing something under the surface, no special powers needed, emergent complexity can be obvious from the start.
The thematic dilemma
Mindsports summarized the difference between strategy games and tactical games thus:
Quote:"Strategy games have strategies varied enough to allow different styles of play, tactics varied enough to induce their own terminology, and a structure that allows advantageous sub-goals to be achieved as calculable signposts along the way.
Tactical games have strategies that are either fairly obvious (however deep), like Pente, or fairly obscure, like Othello." |
|
Different styles of play are rooted in the strategic dilemmas a game presents and the different approaches made possible by them. For them to emerge, the dilemmas must be present in the first place. A game without them is not a strategy game.
Few games, if any, have one thematically embedded like Symple. The counting in Symple has two legs:1. The stone count
2. The group count The first one is territory based and it has a fixed value in the sense that every stone always counts as one point.
The second one is connection based and it has a variable value in the sense that every group always counts as 'minus P' points, where P is an even number.
The best way to serve 1. is to grow fast. A quest for fast growth implies needing many groups which leads to a high penalty.
The best way to serve 2. is to connect as much as possible, saving penalties but reducing growth which leads to less territory points.
The requirements of style
Sometimes one has to grow or place a single, just to follow suit, and much depends on timing. But there is no strict division between a placement and a growing stage, and whether a player is more inclined to do one or the other, or whether he prefers fast growth and a high penalty to reducing the growth rate and lowering the penalty, these considerations will eventually give rise to different styles of play. That's the subjective part of it.
There's also an objective part: the higher the value of P, the higher the relative importance of connections above growth. Extreme values may throw the relation between the two off balance, though the mathematical implications may be interesting, but there a substabtial range that gives natural gameplay under various heights of the penalty, with different degrees of emphasis on growth versus connection.
Marcel Vlastuin's MacBook Pro runs the program on one core (2,6 GHz) and at the time of this base-15/P6 game took some 3 hours for an opening move, about 10 minutes for an endgame move. The game is a fairly close call where I ended up with one group after white_23. Black at the time had 7 groups, yet he creates a new one with black_23. I'm 25 points ahead at that time, but the slow growth rate will get me and after black_30 I'm foced to invade. Remember I'm a beginner, so bad judgement inevitably plays a role. But even so, the contours of the different ways to approach the game are already clearly visible.
|
| « Last Edit: Jan 18th, 2012, 8:11am by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #877 on: Jan 18th, 2012, 9:13am » |
|
Invasions & Shape
To get away from the heavy heavy stuff, let's illustrate some basics about invasions and shape.
Inevitably groups grow into gridlock, having less and less external liberties, because those are the ones you'd like to grow first, and relying more and more (for better or worse) on internal liberties for growth. Here's a very simple example, no external liberties involved.
Both have one group with six liberties, it's black's turn and he's a point down, so filling in his own group loses the game.
He therefore invades at D7. White has no other option than to attack the invader (a counter invasion allows black to grow both groups, choking the invader in the process). Black is now P+1 down.
3. Black grows both groups (E7 F3), white blocks at C7 (black P down).
4. Black grows both groups (E6 D2), white fills in his last liberty (black P-1 down).
5. Black grows B1, white must invade (black 1 up).
6. Black grows F2, white must invade (black P+1 up).
Of course white's answer after the black invasion should have been to resign. It shows how important shape is in the endgame: you'd want as many internal liberties as possible, but each one with as little as possible neighbors. Again, everything in Symple appears to be double edged.
There's an obvious consequence to the above: once established as inevitable, an invasion should take place sooner rather than later, to maximize the invader's impact.
Now imagine this situation embedded in an endgame (and if it's not close, there's no endgame), surrounded by other groups, possibly still with external liberties, and spiced by threatening connections and invasions. Add the fact that local answers often have an immediate 'global' impact (because of compulsory growth of all possible groups), then you may get some idea of the intricacies of endgames.
|
| « Last Edit: Jan 18th, 2012, 12:16pm by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #878 on: Jan 19th, 2012, 8:28am » |
|
The lay of the liberties
Of course there are variations on the above theme.
Each of the next diagrams is the same in that white is one point up and it's black's turn. Can he win by invading?
It's all in "the lay of the liberties", their number and their distribution inside a group, outside a group and over the various groups.
Crunch time
Crunch time comes when most or all external liberties have been taken. Now the speed at which the internal liberties are filled if both start 'plugging' matters. Since forced invasions may be an inevitable result, the outcome of a 'plugging race' is crucial. Basically the groups with the highest number of liberties matter: the rest of them are plugged in the process. If you pit a group with six liberies against two with four liberties, the former wins because the latter are plugged simultaneously. Connect the two and suddenly the six group is pitted against a group with eight liberties.
So to keep the option on a group with much breathing space, you must plan it early on. As it happens I've got an example:
CF - Symplebot.
At white_17 I included J6 to prepare a connection between the two groups on the right and the center. Black answers with an interesting 'invasion' at I2. Don't let the 8 points penalty fool you: the connection is 'for the taking' for black. The invasion does cost him a growing round but allows extra future growth into the white bottom area, and it also saves internal 'breathing space' for crunch time.
But it also allows the connection white aimed at: after white_18 there are five white groups, four of them virtually connected, and nine black groups of which two pairs are virtually connected. So white is P (8 in this game) ahead in future 'reduced penalties' and after black_18 he's 22 points ahead in the count. That's a 30 point buffer against black's faster growth rate. Given enough breathing space that should be enough. But I'm still an absolute beginner, so we'll have to wait and see.
Next day and a few moves onwards
Black_19 is an invasion at the only external liberty of the white bottom group that white could have taken without connecting. I can connect at H5, simultaneously isolating the black invader, but I decided to grab four more external libeties keeping the option to connect. Black follows suit, but must fill two important internal liberties in the process (A2 and F14 are both in groups that have many liberties to spare at crunch time - and it's about crunch time).
White_21 gives me three more external points. Black cannot take both E5 and H5 so the connection is secure. However, if black takes H5, he will connect and save a penalty. But to do so he must grow and thus further plug his bottomleft and topleft groups.
Since I have more liberties, black, for fear (virtually speaking) of losing the plugging race, decides to invade at L10. I'm not sure about the value of that - it saves internal liberties, but at the price of a penalty, and it doesn't threat an immidiate cut.
White_22 therefore allows me to take another external one at O15 and take the connection at H5, which saves a double penalty since I connect and isolate at the same time (as opposed to connecting at E5).
Black_22 catches up in growth. Note that black has an additional 8 points secured because two bottom groups are virtually connected. His topleft group has 7 liberties and is 'invadable'.
The big white group has 7 internal liberties that are not so easily invaded, and a couple of external ones (including the central ones, of which either G7 or G8 may be considered to render another internal one).
Local events with global impact
Given white's lead, he would probably win despite a growing rate of one stone a turn. But invading, if at all, must be done now, and I'd like to see it work out. Never mind that black can cut at L8 now, saddling me up with another 8 points penalty. So white_23 invades at B15.
Note how local events have an immediate global impact. Next to saving liberties, the very point of the invasion is that if black responds to the invader, he must grow his other groups too. Thus his breathing space is dwindling. Meanwhile white still has G7 or G8 to create one more internal liberty. As far as I can see black has not even a remote chance to catch up.
Edit: Friday evening (Saturday morning actually)
And indeed he doesn't. Note that white's advantage in this endgame is the result of the connection he prepared at white_17 by including J6. This is the kind of thin tactical line that can easily be missed by MCTS. The importance wasn't the eight points saved, nor that it slowed down white's growth rate. The importance was to establish a large group with a generous amount of internal liberties to prepare for crunch time.
|
| « Last Edit: Jan 20th, 2012, 5:54pm by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #879 on: Jan 21st, 2012, 7:19am » |
|
The future of Symple
Regarding the future of Symple I can safely say that it will survive me. Most game invented these days probably will  .
I've enjoyed the quest for it tremendously especially the vicious attacks by Mental Mark, for whom the game for some reason seems to be important. He's now actually paying posters to support him.
I also appreciate that Symple's co-inventor has warned me that compusory movement is 'ugly' and that I'm in severe danger of being subjected to 'ridicule'. That may be. I give my games and insights, and I don't take anything away from anybody.
I'm a game inventor, I listen to games, that's how Symple revealed its character. If I were to listen to people, I might as well have become a psychologist.
I'm also well aware that elegant tactical games are fun without raising a treshold. Symple is a strategy game: it will only reveal its secrets to those willing to learn that there's 'more to it than meets the eye'. It's a game you cannot try without being tried by it.
Once again I want to thank Omar for starting what turned out to be my 'blog', more or less. I really appreciate that.
Viewers who are interested can play or watch Symple at mindsports.
I'll keep you posted about Symple's adventures in the 2013 CodeCup Challenge.
|
| « Last Edit: Jan 22nd, 2012, 3:51pm by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #880 on: Jan 24th, 2012, 10:22am » |
|
Invitations for invasions at crunchtime
Here's an example taken from actual play, to illustrate the difference that an invasion at the right moment may make.
|
| « Last Edit: Jan 26th, 2012, 7:43am by christianF » |
IP Logged |
|
|
|
SpeedRazor
Forum Guru
Skepticism is the Immune System of Science
Gender:
Posts: 72
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #881 on: Jan 25th, 2012, 3:10pm » |
|
Wow, I just noticed that this thread has crossed 100k views!
I think that it almost needs an [ongoing] Table of Contents - (LOL), as it has touched on so many interesting and diverse topics over the years by the Original Poster, Omar; the Author, Christian; the various Contributors, Us; and even a Heckler or two
All in all, a worthy endeavor I should say...
|
| « Last Edit: Jan 25th, 2012, 6:23pm by SpeedRazor » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #882 on: Jan 27th, 2012, 8:03am » |
|
on Jan 25th, 2012, 3:10pm, SpeedRazor wrote:| All in all, a worthy endeavor I should say... |
|
Thanks SpeedRazor 
I'm no expert, but I play against MCTS bots regularly. By the nature of the approach, the method quickly points in a crude general direction, so 'fairly strong' is usually a realistic target in the short run.
There are also specific weaknesses. Thin but long tactical lines easily drown in the multitude of play outs, and also the programs tend to be 'near sighted' in that circumferential strategical moves aren't easily spotted as such.
I'm fairly good at Havannah, so there I'm better equiped to exploit these weaknesses than in Symple. My approach in Symple is based on the premiss that the emphasis should be on keeping planned connections secure and isolating the bots groups. When approaching crunchtime, having at least one group with a fair number and a good distribution of internal liberties is important.
At the same time, actually connecting one's groups leaves the remaining group with limited growth potential, and dangerously inflexible when it comes to countering invasions, whether by growth or by placement.
So you get screwed sometimes. Note that it's my turn and I'm 11 points ahead.
To clarify, the white topright and centergroup are not yet connected. I can't 'neutralize' by invading at N7 because O7 is also a connectionpoint.
That's another 10 points plus for white.
The combined white group has 7 internal liberties against the black one's 2. Black is bound for two or three forced invasions.
That's another 20 or 30 minus for black. No need to establish the exact count, screwed. 
|
| « Last Edit: Jan 27th, 2012, 8:29am by christianF » |
IP Logged |
|
|
|
christianF
Forum Moderator
Forum Guru
Arimaa player #4019
Gender:
Posts: 804
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #883 on: Jan 29th, 2012, 10:23am » |
|
Smooth Calculator
I'd like to express my appreciation of Marcel Vlastuin's mcts-based Symple_bot. There's nothing like an emotionless calculator when it comes to revealing tactics, or flaws in one's thinking about tactics.
Symplicated?
Symple is a newly discovered organism that strategically speaking is unusually wide. "Better a bad plan than no plan" seems to actually summarize a player's choices. So I made plans, giving somewhat more weight to connectivity than to growth, because the latter is more direct, while the former involves long term planning. And we're experimenting with different penalties in an upwards direction, which would appear to support such a strategy.
I don't give much a priori thoughts to tactics. If the system is sound you'll usually not be able to imagine the situations you actually encounter. Here's an opinion that I found annoyingly ignorant with regard to the behaviour of games like Symple, that "have nowhere to go but deeper".
Quote:| "This filling in phase of Symple is substantially different because there are no decisions to make. It is a sequence of forced moves. You just fill in, until you resume making decisions again." |
|
Annoyingly because a) it's never that simple and b) the opinion is based on looking at a middlegame position of Symple old style (without compulsory movement) and then suddenly switching to compulsory movement. That is wrong because Symple now develops a very different kind of middlegame in preparation of crunchtime. Distribution of internal liberties, both within a group and over the several groups, is a leading strategical issue. You simply don't get 'old style' middlegames where distribution didn't matter all that much, if at all.
Plugged in the process?
Remember this?
on Jan 19th, 2012, 8:28am, christianF wrote:| Basically the groups with the highest number of liberties matter: the rest of them are plugged in the process. If you pit a group with six liberies against two with four liberties, the former wins because the latter are plugged simultaneously. |
|
I'm glad to admit after my last game against Symple_bot, that it's a bit more complicated than that. It always is, isn't it?
I followed my 'connectivity' strategy, aiming at connecting late and creating a group with enough breathing space to survive crunchtime. I didn't do so bad either, not even in retrospect, because I lost by a mere 5 points on the last forced invasion. But for a long time I seemed to have the better of the game, and a white_22 I was 58 points ahead, with three goups (but black has two or three secure options to connect against white's one more or less secure connection point at J7). I'm low on internal liberties, but there are still quite a few external ones. Black at that moment has 10 groups and is rather low on internal liberties too.
White_22 and _23 are 'penalty exchanges'. I isolate two groups at the cost of creating two, plugging a couple of potential liberties in the process. I still thought things were looking pretty good.
I'll leave Symple_bot's handling of the subsequent finer points for you. I may have made mistakes in the endgame, but I can't really find any and I'm still recuperating. 
|
| « Last Edit: Jan 29th, 2012, 4:09pm by christianF » |
IP Logged |
|
|
|
NickBentley
Forum Senior Member
Arimaa player #6704
Gender:
Posts: 44
|
|
Re: Essay by Christian Freeling on inventing games
« Reply #884 on: Jan 29th, 2012, 6:44pm » |
|
This stuff is great Christian. I wish more designers would describe the evolution of tactics and strategy following the invention of their games. I can't try out every game in the world, and reading stuff like this helps me to decide which to play. I want to have some sense of whether I'd like the tactics/strategy before I play, and I'm not much of a "game whisperer" so I need help.
|
|
IP Logged |
|
|
|
|
Home
Help
Search
Members
Login
Register
Notify of replies
Send Topic
Print
)
