1900

Congratulations to blue blimp and stepanzo, on becoming the first players to achieve the 1900 rating at Catchup and Morelli respectively.

As far as Morelli is concerned, any of you Grand Panjandrums of abstract strategy who haven't taken the time yet to try the game, are warmly encouraged to wander over and see how you fare against the current King of the Hill.

Catchup 5   Top 20 link

http://www.littlegolem.net/jsp/info/player_list.jsp?gtvar=catchup_NEWSTANDARD&countryid=&filter=&Send=submit

as at 09:22 BST

List of users   Page: 1 2 3 4 5 6

Full nameCountrycatchup Size 5blueblimpCanada1932Jos DekkerGermany1807gamesorryChina1724gridgoUnited Kingdom1693metzgerismUnited States1686Richard MalaschitzSlovakia1663hyperpapeUnited States1657livingtechUnited States1647Ray GarrisonUnited States1640HjalltiBelgium1639ululu1634ddoggGermany1627vstjrtPoland1626TheBurglar FreeHosting NetUnited Kingdom1613jugularUnited Kingdom1611orggGermany1611Scott ArnoldUnited States1600bennokSwitzerland1594Mirko Rahn---1592mungoSweden1591

no inversion of text/background colours if copy all

List of users   Page: 1 2 3 4 5 6

Full nameCountrycatchup Size 5blueblimpCanada1932Jos DekkerGermany1807gamesorryChina1724gridgoUnited Kingdom1693metzgerismUnited States1686Richard MalaschitzSlovakia1663hyperpapeUnited States1657livingtechUnited States1647Ray GarrisonUnited States1640HjalltiBelgium1639ululu1634ddoggGermany1627vstjrtPoland1626TheBurglar FreeHosting NetUnited Kingdom1613jugularUnited Kingdom1611orggGermany1611Scott ArnoldUnited States1600bennokSwitzerland1594Mirko Rahn---1592mungoSweden1591

ouch ? how to copy with correct colours

I have the iOS app to thank. It taught me the game, and the AI is good for quick practice. https://itunes.apple.com/us/app/catchup-abstract-strategy/id718429890?mt=8

Is the Ai implementing new rules?

Do you know if Martin Grider used some special heuristics or just brute force?

The app currently uses the old rules. I don't know what the future plans are.

I read somewhere that the app's AI uses a monte carlo algorithm (probably Monte Carlo Tree Search, MCTS), but I don't know details beyond that. I wrote the dumbest-possible MCTS AI for the game (just random moves for the playouts), and it performed a lot worse than the app's AI unless the dumb AI had ~100x as much thinking time, so I assume there are some additional improvements.

Thanks for the answer.

Did anyone solve catchup endgames:

Say player1 has a longest chain A with a possible hexagons to add on his side, player 2 has longest chain B with b hexagons on his side and they share x hexagons so like a|A|x|B|b.

Are there simple cases (second longest chain too short to matter) where it is best for player 1 to play in x instead of playing in b? I guess the new rule changes that too.

I did some analysis for those situations (and some more) under the old rules: http://boardgamegeek.com/article/18510014#18510014. The analysis doesn't directly specify a winner for your situation but it does give a method to calculate the winner. The new rules will indeed change the analysis a little. Aside: under the new rules, there are simple endgame positions where optimal play requires passing (which isn't currently possible on LG): http://boardgamegeek.com/article/19064720#19064720.

blueblimp, what do you mean with "not possible on LG"? The rules do not allow passing, as far as I know.

If you mean placing 1 stone instead of 2, or 1 or 2 stones when you are allowed 3, see the discussion in this thread: Catchup: Wrong result recorded, game rules not 100% correct.

I have given a description there of how to do this at littlegolem. It's not obvious and a bit clumsy and I hope Richard implements this through normal GUI - even if it is rarely needed.

There are also links about games (with previous rules), where the only winning move was like that. Example: game 1658767, situation after move 25

In the same game, Blue missed a chance to win the game by placing only 1 stone at the top: game 1658767, situation after move 22 - if my analysis was correct.

Sorry, by "passing" I just meant playing fewer stones than the maximum. You're right that it's not allowed to place 0 stones on a turn.

Thanks for linking that thread. I'd read on BGG that it was possible to pass using the URL, but I didn't know the specifics of how to do it. The linked game is also very interesting. I speculated that in the old rules, the only situations where passing was needed would be tactical situations, but the example game requires passing even just for optimal territory filling. Nice.

I agree with your analysis that Blue could have won 1658767 after move 22. It seems that the game reduces to a single-heap normal play subtraction game ( http://en.wikipedia.org/wiki/Nim#The_subtraction_game_S.281.2C_2.2C_._._..2C_k.29). Because the number of useless remaining hexes is 7, which isn't a multiple of 3, Blue can repeatedly force that number to be a multiple of 3, forcing Orange to be the first to make a move in the center. After Orange's stone at D9, Orange can make the number 8 (by passing) or 7 (by playing any useless hex) and neither is a multiple of 3.

Neat stuff. I hadn't seen such a nice reduction in a Catchup game before.

It is actually possible to play fewer than the maximal number of stones on LG.

Here's the paydirt from an earlier thread:

If anyone wants to place less than 2 or 3 moves, the procedure is: click on the place you want the stone to be placed and click on another random empty cell. In the next page do not click Send move but manually edit the url.

If it ends with "...move=eaff" and you want to move only into the E1 (ea) cell, edit the url to "...move=eaea" and click enter.