Moving Forward Looking Back

Questions, discussion, and ideas and suggestions can be posted here.
Post Reply
User avatar
Site Admin
Posts: 27
Joined: Sun Sep 17, 2017 2:32 pm
Location: Kansas City

Moving Forward Looking Back

Post by Relayer » Sat Oct 07, 2017 9:09 am

I noticed this article posted in March, 2017, on Christian Freeling's website Mindsports and found it so compelling that I wanted to share the relevant parts for chess here. Mr. Freeling is the creator of Grand Chess.

"Like Johan Cruijff and David Bowie I was born in 1947 so it may be an appropriate moment to look back a bit and speculate on what is yet to come. I think it's fair to say that my generation has gone through some significant changes and that humanity is now in a transitional period the likes of which it has never seen before and the end of which isn't in sight. 'Interesting times' so to say.

The changes didn't fail to affect the small niche labelled "abstract strategy games" in a fundamental way. I happened to be there most of the time, from the frustration of early programmers when their brainchild failed to find a mate in a king+rook versus king endgame, to the point that AI has become vastly superior in Chess and Draughts, and will soon have the same status in Go and Shogi. Improving these programs is an ongoing challenge, but programming new games is somewhat less so nowadays. Commercial abstracts come with apps and an AI opponent to go with it, but that's more a routine than a challenge. And if new games do have leverage enough to actually challenge programmers, then the programs will need less time to become 'vastly superior' because humans learn slowly and will need time to catch up. This is the new reality. It means that insofar abstract games are played as sports, the champions will have new and superior tutors and championships will be played out between the best pupils.


In About Grand Chess I said, about a decade ago:

We live in a time not unlike the renaissance. In those days the printed word accelerated events and led to almost aggressive innovation, now the internet does the same. At the current rate of development, Chess will show serious signs of fatigue within a decade. More study, better programs, more knowledge, less fun, less adventure, more grandmasters, more draws and no more heroes.
I'll allow for one more player to define an era like Fischer or Kasparov.

And sure enough there's Magnus Carlsen and his era was still extending in 2017. World Championship matches are mondial events that are in a constant state of transition. Live streaming, AI supported live analysis by renowned grandmasters, live chatrooms and accessibility through a multitude of media - in the near future we may all be able to attend and discuss the games in a virtual environment. But things other than coverage change too.

As I write, the last world championship was held in New York in November 2016. Carlsen's opponent was the equally young Russian Sergey Karjakin. After twelve regular games the score was equal with two decisions and 10 draws and a series of 4 tiebreaker games was needed to decide the match. Carlsen won.

I won't say that this result is predictive, but if one would consider Chess as a metropolis then it might be argued that AI can turn on the lights in uncharted alleys in ways that never before in its history have been possible. And that light is now used by all top players.

The 'best pupil' era

Exploring new alleys in well charted areas has always been the pride and joy of Chess progress. Till a few decades ago this herculian task was driven by the efforts of masters and grandmasters. That in a way is still the case, but computer programs have evolved into indispensable employees that search the branches departing from the chosen alleys with a hard artificial light and at an incredible speed. Flawed lines are detected fast even if the flaws become apparent farther down the line than humans might go in their analysis. These programs don't care about style, they don't lie and they are available to all.

It may be difficult to establish how much of a Chess player's success is due to talent and how much is sheer preparation. Talent is an evasive property to define and preparation goes as far as it goes. In the old days, when I was young, grandmasters had different styles, resulting from different temperaments and different ways of approaching the game. Computerprograms have neither temperament nor styles resulting from it. They may have different ways of approaching the game, but not all that different and in any case better than the player using them. At the preparation end things go much faster because of this, and nothing much can be kept a secret for very long. Beyond the preparation horizon talent is still a significant factor, but thorough and unrelenting preparation pushes that boundary ever further. That means that once a game takes to uncharted territory, differences in talent may often turn out not to be enough to force a decision.

The games in world championship matches or tournaments will increasingly be channeled into the narrow alleys carved out by stronger AI. It may not go so far as to reduce the players to a medium for contests between programs, but games will increasingly be 'flavoured' by it. Understanding what's actually going on on the board will increasingly be the prerogative of top players who are aware of the latest developments. The rest of us will see great games, with a constant eye on what the AI evaluation makes of it to guide our judgement.
With quantum computers on the horizon I deem nothing impossible in the future, but as yet Chess is lightyears away from being 'solved'. So there's no proof for or against the assumption that it is a determined draw. But you can bet your boots on it. It means that under AI driven preparation the number of draws will slowly but surely increase. Carlsen is the first exponent of a generation that grew up with AI and he became the one where talent and timing coincided to make him define an era. But he's bound to be the last. Chess world championship matches will increasingly be decided by preparation. Preparation to the point where the boundaries between it and 'talent' fade and all have the same equipment to make the best pupils rule.

What has Grand Chess to do with this?

Insofar Grand Chess represents 'Chess with a complete set of pieces', the same as it always has. Including the Queen as a composite piece naturally implies a question about the other two composite pieces. I've written about that in How I invented ... Grand Chess.


Look at the diagram and you see that each side has one king, three composite pieces, six single pieces and ten pawns. That's the triangular series. It breathes balance.

Grand Chess' particular arrangement is arbitrary but it emerged naturally and thirty years of play by a range of qualified players, grandmaster Larry Kaufman among them, have not revealed any opening unbalance. Kaufman applied 8 criteria to chess variants and rated Grand Chess second after Shogi and higher than Chess, Chu Shogi and XiangQi.

But there's more arbitrariness, especially with regard to promotion. And it's all Philidor's fault.

Philidor's issues

In the mid eighteen's century André Danican Philidor was both the world's strongest and its most famous Chess player. His Analyse du jeu des Échecs was considered a standard chess manual till well into the nineteenth century. It abided by the current rules, yet Philidor had some issues with these. He doubted the initial double pawn move and ridiculed en passant capture. That's no issue anymore and Grand Chess abides by consensus. He had some issue with castling too, but that's no issue anymore either and in Grand Chess it never was in the first place. But his thought regarding promotion are still topical. So far as I know Philidor was never concerned with Chess' lack of structural completeness, but promotion to any chosen piece, which was to all intents and purposes promotion to a queen, certainly violated his sense of material completeness. Concerning his fellow countrymen and Chess players he remarks:

I presume they have been led away (like myself, formerly) by a bad custom established in all probability by the person who first brought Chess into France; I'm inclined to believe it must have been some player at Draughts, who knowing little more than the moves of the pieces, imagined one might make as many queens in the game of Chess, as at Draughts. I would only ask what a fine sight it is to see upon a Chess board, two pawns on the same square, to distinguish a second queen; and if by chance a third should be made (as I have often seen it at Paris) then it is still a finer sight while the bottom of one pawn is almost sufficient to cover a square on the board. Is not therefore this method most ridiculous, especially as it is practised in no country where the game of Chess is known?

Since he blames those darn Draughts players for deforming Chess, I don't think Philidor would have liked using topsy turvy rooks any better. Was he right? Obviously not, but he stuck a chord regarding Grand Chess. I'm used to looking at games not from a historical but from a more general point of view. I hold structural completeness in high regard, but like Philidor I also value material completeness. And where Chess would certainly become less of a 'sport weapon' if promotion were restricted to pieces captured by the opponent, in Grand Chess things are quite different. The price of material completeness isn't quite that high because there are three composite pieces. Of course you must have one available for promotion and even if that will usually be the case, it yet poses a restriction that sometimes may require the exchange of a composite piece before an intended promotion can be effectuated. To be on the safe side regarding the consequences, which would include the inherent emergence of some interesting purpose driven tactics, I gave pawns the option to promote upon reaching the eighth or nineth rank and the obligation to do so upon reaching the tenth. Compared to Chess it means that a pawn starts one rank closer to promotion.

Back to the future

Chess has no problems, but something has been lost forever. We will keep seeing players with different temperaments but no longer will they be able to afford having these affect their approach too much. A player like Mikhail Tal, who's attacking, combinatorial style was pervaded with improvisation and unpredictability, wouldn't stand much of a chance in the new reality. He would have to face walls of preparation and early deviation into 'uncharted territory' would reveal that this territory wasn't quite as uncharted as he imagined. His namesake and contemporary Mikhail Botvinnik on the other hand would probably have felt quite at home in it.

If we compare Grand Chess with a metropolis, it would relate to Chess like New York to Volendam. But an old Volendam with a well lit center and well lit suburbs, where top players know their way around, and where more sparsely lit alleys are only to be found beyond the limits of the city proper. Grand Chess on the other hand is a very young New York, where streetlights are sparse, even in the central areas. There's no 'preparation horizon' because none was needed yet. Main alleys radiating from the center still have to be carved out and computer programs to guide the process still have to materialize. It's an environment where players like Tal can again thrive and where talent and daring are ready to take on thorough preparation. It's bringing the past back to the future.

With the current knowledge of Chess programming, if programmers care to make one, a Grand Chess program that outplays 99,99% of human players would soon be a reality. It would be a tremendous tool to carve out main alleys of opening play. But even under constant improvement it would not allow a preparation horizon with a ply depth equal to what is currently the case in Chess. Grand Chess is simply too wide to allow humans to prepare that deep, ever. The ratio of more or less charted areas to the great uncharted ones will always be significantly lower than in Chess. Thus adventurous, daring and talent driven play will stand every chance against study and preparation. As I said before, preparation goes as far as it goes, and in Grand Chess that will only ever be so far. And yes, I love tautologies.

There's no special insight required to see that Grand Chess is a well considered implementation of a structurally complete Chess, better than the one of Capablanca and a number of subsequent variants including Gothic Chess (among others), Embassy Chess and Janus Chess. All three employ 8x10 boards with the rooks once again tucked tight in the corner and a castling rule to 'solve the problem' - it's not a bug, it's a feature. Gothic Chess uses the same set as Grand Chess, Embassy Chess the same set in the same configuration and Janus Chess features two cardinals (Archbishops) but no marshal (Chancellor) - worth mentioning only because it's so weird.

Speaking of which, inventors being what they are, the emergence of Capablanca random Chess was of course all but inevitable. Bad examples tend to be followed.

The notion of a structurally complete Chess is firmly rooted in the Chess community's collective consciousness and in the past 30+ years Grand Chess has given it a focus that is beyond disappearing and clearly growing. I have neither the means nor the skills nor indeed the inclination to influence this growth. I'm quite happy to know that players will enjoy the game for much longer than I'll be able to watch it."

Post Reply