Why Chess 960 is Bad

It should be called Chess8

@RamblinDave said in #5:

I dunno, maybe someone could come up with a scheme where we randomly select a 960-ish starting position and run tournaments in it, at a variety of levels, for the next year? Then the theory gradually develops over the course of the year, before being reset again. Although maybe with modern engines people would just figure out most of the best lines by running a supercomputer for a few days, and then we'd suddenly have reams of theory again.

Great minds think alike, apparently:
"I have always liked the idea of choosing a few decent positions. And, I don’t think you need more than 15 to 20, out of the 960 possible random chess positions, many of which violate our sense for normal chess geometry. Any change of the position is a challenge, but 10 to 15 to 20 positions can be chosen, and I believe that in the future, every year, we should start with a new position. Again, it should just be one position. I feel an insult if players should start with something that is totally ridiculous, and you have three minutes to prepare... No, I mean, come on, chess is also about some research. You don’t want to have the same extensive thing, fine. But, you have one year of playing one position, which means that players can actually get adjusted and they could do a little bit of research. So at least you have five, six opening moves that are theory now and then you go on to another position. "
https://en.chessbase.com/post/bisik-bisik-with-garry-kasparov-part-3

Fischerandom is the most significative variant in chess history. It is not bad; it ́ is the future of chess.

Oh thanks @Splorer for your comment and a chance to briefly clarify the rules of Chess 960 because most players haven't played it yet, and I think they are really missing out on the best chess experience.

Again, these are Bobby Fischer's thoughts and until Lichess, were not mine.

The major rule about piece placement is to have the king placed between the rooks.
So the two corners are out of the mix.

Next rule is that bishops are on opposite colours. I'm Canadian so we spell "colors", differently, btw.

Castling rules don't change> but some things are a little confusing at first.
And when playing online with LIchess there could be things that differ from other chess sites.

Admittedly there are quite a lot of rules about castling to know, and you might need to refresh yourself before first playing 960 because this was my pinch point.

During the last couple of weeks I played 960 OTB for the first time. It was very fun to play someone at my level of standard chess and be the first to explain the rules to them.

We just kind of stumbled through it with a simple concept of placing the king first followed by the rooks, bishops, knights and by process of elimination, the queen was placed on the last remaining square. It does take a little longer, but not much.

The 8 starting positions of "Alpha-Hotel", would only be for official online games where it's impossible to cheat with illegal moves or starting configurations. Eight positions is plenty, and they could also periodically change the starting lineup. I'm just talking out loud and dreaming at this point.

My intention is to just start a conversation about the evolution of Chess 960, Fischer Random, Shuffle Chess, New Chess or whatever we call it.

How do we make it better and easier to digest for newcomers that might be reluctant to try something different.
I've already heard all the complaints so can anyone suggest some positive ideas?

Hey @ungewichtet I can always count on you to have an interesting take on my posts. Personally Chess 960 is fine as it is. However I'm greedy and its about getting the reluctant to try it because whatever the name, it is the evolution of the game and I want to see it become much more popular so I can watch more talented players play it.

Those most reluctant of course are the big wigs or authorities with the big ratings. They have a vested interest in keeping it just as it is and maintaining the status quo. It's not their cup of tea for a good reason. They have spent a lot of time on opening theory to get to those big ratings. For players at our level, openings hardly matter.

There really is no down sides, except that people hate change unless compelled. We all like to talk a big change game, but when push comes to shove, we really don't. Ain't human nature a hoot ?!

I like Garry Kasparov’s idea of choosing 1 position for a “season”, and letting that be the position that is played until a season ends (after which point, another position is chosen). Kasparov sees a season as lasting one year, but arguably it could be a shorter timeframe for online play. Something from 1-3 months or so.

Imagine if lichess implemented chess960 seasons (with a unique trophy for some achievement in a chess960 season to promote play). That would be so cool.

Lichess is a chess platform that could easily promote new ideas and change within the game, given the large playerbase. It’s surprising that they do not utilize this to promote new ideas for the game.

Most chess players don’t care.

also interesting would be to know the difference in legal positions. what kind of structure in middle game is reachable at certain depth from different initial conditions.

can all identified structures from real opening theory in standard be reached from all 960 or subsets (and which are or not).

The limited time or depth might make one more aware for the tempo cost of a given initial position. shedding light on standard chess hidden constraint to geometrically (i.e. structure) minded plans. This might be a way to learn real theory about openings (constraints from initial position).

How far do we need to go, for the middle-game chess to be equivalent. can middle game chess be improved from this new angle that sheds light of the consequence of initial position mutations early in the game. The mobility rules (besides required adaptation of castling) are exactly the same. and even castling is a compatible generalization of the standard initial position one. So one could learn something. I don't have the time for it. but... machines do.

In computer chess, does anybody know of a reinforcement learning NN approach like alpha-zero or lc0 trained with 960 initial conditions? and then how does it fare on standard chess compared to one trained only on one initial position (such as standard one). The basics physics dynamic is the same. no matter what the initial condition is. Plus they would have had to solve how to train like that. Currently my understanding is that like PGN without FENs, they have some initial position effect in having more terminal outcome in raw exploration phase early than further, and that would happen relatively too with tightened exploration (along the successive reinforcement training batches). So that in training, They never trained in RL from a diverse set of initial positions on the same engine. How to use the same learning NN to learn from a set of initial conditions. latest I heard was using opening books, but I am not sure of the details, and not sure it is about the same NN.

Find hooks for people who love the mysteries of opening decision consequences down the game.