Richard, we think you're the top, the bee's knees, the cat's pajamas. But here in 13 Hexland we have a bot infestation problem. For a while it was kinda fascinating, in a horrifying sort of way, to watch them pick up our best ideas and smash them against rocks.
The worst part is, now the bot maintainers have turned into Jack Nicholson from A Few Good Men. They won't let us download their bots, because we can't handle the truth. We need a way to salvage our self respect here, for a few months at least.
Yes, you have already provided 19x19, and we're very grateful,v, but a stepping stone would make the transition so much smoother.
And there is no humiliation in losing against someone else who has been playing hex for 10 or 15 years... and the humiliation is all on him if he loses.
Maybe we should just take more advantage of variant that is already available - 19x19? I don't think we can expect adding 15x15, but 19x19 does not seem to be popular in our community.
By the way - does anybody have any information, if Maciej has any interest in returning to hex? It would be great to watch his games against Arek, Daniel or lazy.
I take advantage of this discussion in which there is a strong and justified request for unlur, et where Y is also mentioned, to draw attention to yXy, a game designed some years ago. Coming from a previous idea by Steven Meyer, it just merges hex and Y in a single game. yXy has been implemented at the play- by- email site.
I dont think that 19x19 is too long, the average professional go game lasts about 250 moves. The average high level Hex game on 13x13 here on little golem lasts about 25 moves, which translates to about 15% of the board being occupied by the end of the game. So extrapolating from this, the board size for which the average Hex game would last closest to 250 moves would be 41x41.
Some Go players do think that the 19x19 games last too long, but serious players seem to mostly not have this opinion. Which makes me think that a good estimate of where hex will end up is somewhere around 41x41 as the board size for serious players.
So for example, to have an hex game that lasts around 240 moves, you would need 27x27 (that is, you would need to double the area of 19x19). Hex may end up earlier, of course, but then a Go game can also end up earlier if you lose a big group. Professional players of Go play at a MUCH MUCH higher level than top Hex players here so you can't directly compare.
For example, I've taken the last 10 games played by mootwo against other top bots, the average game length is 30.8% of the board. If you include the next 10 you probably reach very close to 33%. If instead we take the match AlphaGo vs Lee Sedol, 5 games, average game length is 206. Why it's so short? Because one side was much stronger than the other. And btw mootwo is also quite a lot stronger than the other bots so it's not entirely fair to use its games to compute the statistics for Hex.
I've taken the games of Gerhard Knop (he seems to be the highest rated active Go player here) and his games in the last championship lasted an average of 234 (64.8% filled). For sec, the player who has won the last championship here, his average game length in the championship was only 187 (51.8% of the board).
For mootwo, his latest hex 13x13 championship, average game length is 29.5 (17.4% of the board filled). Is this enough evidence that only 1/6 of the board gets filled? Well, if weak players are playing against strong players then yes only 1/6 of the board gets filled! We simply have a lot of room to improve.
And of course 1/3 and 2/3 are only easy to remember numbers. Maybe Hex games is at 30% instead of 33% and Go is at 70% instead of 66%. Still not that much different. Then of course we've to consider the difficulty of choosing what to play. The reason why large hex boards aren't popular is because we really have very little idea on what to play at the beginning.
To get an idea of relative difficulty of Hex and Go we can see largest boards that have been rigorously solved. For go, that's 5x6 (solved in 2009). For hex, it's 9x9 (solved in 2012). The area of 9x9 is 2.7 times larger than the area of 5x6. Assuming the same proportions hold then 19x19 Go is somewhere between 31x13 and 32x32 Hex. If we assume 6x6 is equivalent to 9x9 then we would be somewhere between 28x28 and 29x29. Clearly this is not a 41x41 board. The assumption of proportional scaling is of course very very questionable.
I think there is one key difference that has not been brought up. Go is a point game, while Hex is a "victory condition" game.
In Go, larger boards are ok, because we can mostly analyze smaller areas in term of "locally correct play". Of course that is never completely true, but it's close enough if you throw in some general idea of the overall board (ladder breakers, direction of play etc.) . I can get 20 points in one corner, and that is a good move even if it's not clear what it does to the overall board "texture".
Hex is an "all or nothing" game, and the first moves become more and more obscure as we increase the size. Having said that, I think that serious professional play could be done on 19x19 also in Hex.
Postal play (like little Golem) requires less moves. I'm amazed that people play large board Go here.
Richard, I agree, the initial moves could be harder in hex than in go, especially if the area of the hex board is double than that of the go board. Then in the later phases it's probably the opposite because in hex the balance between the players is very delicate and there are only very few choices that keep the game properly balanced.
I'm not sure if postal games are shorter. For sure it's interesting that we've 37x37 Go available here. In fact I'm watching a few games right now from the infinity tournament. These games usually end with a lot of space still unoccupied. This is probably because the level of play is really too low relative to the possibilities offered by such a large board.
Anyway, as feature request, I think we could agree on 27x27 and then 38x38. In this way we keep doubling the area and it's a nice geometrical progression. I'm not sure about the merit of the sizes that are between 13x13 and 19x19 such as 15x15 or 16x16. The more sizes we've available, the more fragmented the community will become...
If we want to follow formula ceil ( sqrt ( ( N^2 ) * 2 ) ) then we get 9x9, 13x13, 19x19, 27x27, 39x39. Without the ceil then we would get 9x9, 13x13, 18x18, 25x25, 35x35.
Little Golem