Arimaa Forum - Print Page


    
      
        Arimaa Forum
        (http://arimaa.com/arimaa/forum/cgi/YaBB.cgi)
      

        Team Games >> 2009 One vs TheMob >> 2015 Move 1
        
(Message started by: deep_blue on Jun 20^th, 2015, 4:24pm)

Title: 2015 Move 1
Post by deep_blue on Jun 20^th, 2015, 4:24pm

Hey mobsters,
lightvector told me he would play silver with sharp. So even if the game did not yet start we can start discussing our setup and then necessarily also our strategy.

P.S. Once we get our own board on the forum we can move this thread there.

Title: Re: Mob game 1g
Post by supersamu on Jun 20^th, 2015, 5:34pm

Looking at the last post on page 4 of this thread: http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=talk;action=display;num=1207682686 , we can see that the most popular setup was:

RHCMECHR
RRRDDRRR

(standard 99of9 with cats behind the traps, followed by 99of9 with dogs behind the traps).
Maybe woh can update these statistics, but I doubt that the top two setups have been overtaken.
I think both of these are good choices for 2 reasons:

- We as a mob are not only composed of players of different styles, but also of different strengths. So choosing an unconventional opening might dissuade weaker players or players unfamiliar with that particular setup/opening to contribute, because they don't understand the plan we want to follow with an unbalanced/unusual setup.

- I would like us to beat sharp with very straightforward play that doesn't rely on a perceived weakness of sharp's opening, and choosing the most common setup, which is more of a reactionary one, would help prove that point.

So I propose that we choose one of the 2 99of9 setups. I personally have no preference there.

For anyone interested, these are the previous games of the mob:
http://arimaa.com/arimaa/gameroom/pastgames.cgi?id=6169

Title: Re: Mob game 1g
Post by arimaa_master on Jun 20^th, 2015, 5:46pm

My suggestion is one of the two proposed:

on 06/20/15 at 17:34:03, supersamu wrote:

RHDMEDHR
RRRCCRRR

(standard 99of9 with dogs behind the traps).

Title: Re: Mob game 1g
Post by PerkofBR on Jun 20^th, 2015, 6:03pm

I start all my gold games ( and most of silver ones too, even tho its considered bad :P) with the regular 99of9 setup, the one with cats behind the traps.

Cats are usually homeguards, so its best to place then behind the traps.

Title: Re: Mob game 1g
Post by half_integer on Jun 20^th, 2015, 6:11pm

I also typically use the 99of9 with the dogs behind the traps. Players who plan to use the dogs in the middle once the E, M, and H are tied up would prefer that they are at d1 and e1.

Since I feel that my opening is a weakness, I actually look forward to any discussion on the impact of the setup on the opening, and plans for the opening.

Title: Re: Mob game 1g
Post by deep_blue on Jun 20^th, 2015, 7:15pm

IMO cats behind traps have the idea to aggressively attack with dogs later while dogs behind traps are used to play more likely a home game.
So while suggesting setups we should also discuss how aggressive we want to play.

Title: Re: Mob game 1g
Post by browni3141 on Jun 20^th, 2015, 9:20pm

I am open to a variety of Hb2 Ed2 Me2 Hg2 structures, but dogs behind the traps must be near the bottom of my list. Development of the c2 dog can be a little bit awkward as moving it temporarily removes a defender of the trap and a piece on b3/d3/c4 is required for forward development.

Additionally, forward development of the cats in the center is a little bit awkward as they are vulnerable in the opening in cases where a dog wouldn't be. Also, since dogs desire to advance more than cats, the wings may be left a little bit weak by the late game. This can also be a problem in setups with both cats on one wing.

I will suggest some setups which I prefer and a bit of reasoning behind them.
This one is not exactly the same structure, but it is similar:
RHCEMRCH RRRDDRRR
Horses often move to the a/h files early anyway so this makes sense, and f3 is not weakened too much because this is done on the camel wing. Silver could respond by making a horse advance on this wing undesirable, though, making the horse a little bit misplaced.
A gold setup I frequently use:
HMDEDCHR RRCRRRRR
http://arimaa.com/arimaa/forum/cgi/YaBB.cgi?board=talk;action=display;num=1344359232;start=
My top two would probably be either my setup or 99of9 cats behind the traps.

I disagree that we need to talk about how aggressive we want to play. We should play the moves which we think are best, which usually won't leave us with an option of being aggressive or defensive.

Title: Re: Mob game 1g
Post by ikalyoncu on Jun 20^th, 2015, 10:53pm

I prefer 99of9 with cats behind the traps

Title: Re: Mob game 1g
Post by Hufflepup on Jun 21^st, 2015, 3:06am

I favour a 99of9 type setup as well. Too often with unbalanced setups I find that the other side reponds in such a way that I want to switch wings with a piece which just costs time.

When attacking I've found that centralised dogs are easier to deploy than dogs behind the trap, however dogs behind the traps have worked better for me in home games.

As I'm guessing we will want to at least attempt some sort of attack in the opening I am currenly favouring 99 of9 with dogs in the center.

Title: Re: Mob game 1g
Post by half_integer on Jun 21^st, 2015, 7:01am

Browni's arguments in favor of cats behind the traps make sense. My only rationale for the dogs there is in case there is the opportunity to frame a dog, the other sides of the frame are more likely to be in place. However, I would accept it as fact if a stronger player stated that a dog frame in the opening doesn't happen at this level of play.

Some setups with the dogs behind the traps and the cats on c1 and f1 could also be considered - the dogs are then the same distance from d2 and e2 to move up the center, and the cats are only one step away from replacing them.

I quickly flip-flopped on Browni's setup. My first reaction was that using a setup which not everyone understands would make it more difficult to discuss a plan in the opening. But, I quickly was attracted to the idea of learning the benefits of such a setup by actually playing it, with expert guidance. Do other mid-rated players feel the same way? I, for one, would like to better understand some of the non-symmetric setups and their rationales.

So, my vote is currently for Browni's setup for the reasons just given - unless other advanced setups are proposed.

Title: Re: Mob game 1g
Post by Samraku on Jun 21^st, 2015, 7:35am

on 06/21/15 at 07:01:27, half_integer wrote:

Yes, I think using Browni's setup could be educational.

Title: Re: Mob game 1g
Post by deep_blue on Jun 21^st, 2015, 12:25pm

Half_integer made an interesting point there. I for one would like to learn too how to play such a setup. Still there is the danger though that we reach positions that most players are not comfortable with.
I guess EMH usually gives up a camel hostage to then swarm the trap. Even if that objectively may be sound I am not so comfortable with such positions.
So all in all I am not yet sure if I would want to play such an unbalanced setup. Browni, feel free to further advertise your setup. ;)

Title: Re: Mob game 1g
Post by Knedlik on Jun 21^st, 2015, 4:17pm

I was actually looking forward to learning how to best make use of the 99of9 setup as gold. Since even browni has used it in most of his games in the last WC, it clearly is one of the best setups as gold and more of us might be able to contribute in the opening as we are more likely to be familiar with the position. So for the moment, my vote would be in favour of the 99of9 setup (with cats behind traps).

Title: Re: Mob game 1g
Post by Algorias on Jun 22^nd, 2015, 2:52am

on 06/21/15 at 07:01:27, half_integer wrote:

I quickly flip-flopped on Browni's setup. My first reaction was that using a setup which not everyone understands would make it more difficult to discuss a plan in the opening. But, I quickly was attracted to the idea of learning the benefits of such a setup by actually playing it, with expert guidance. Do other mid-rated players feel the same way? I, for one, would like to better understand some of the non-symmetric setups and their rationales.

I'm of two minds about this. One one hand, I would find such a setup more interesting to play. However, the mob needs every advantage it can get, and playing a move for any other reason than believing it is the strongest available doesn't sit well with me.

I must add though that unfamiliarity of a move to some players shouldn't be a reason not to pick it. We have time on our side, and therefore the tactical blunder rate should be close to 0.

The only thing I'd be worried about is someone proposing an unorthodox move, it getting accepted, and then that person disappearing for a few weeks, right when their experience is most needed by the hivemind.

Title: Re: Mob game 1g
Post by PerkofBR on Jun 23^rd, 2015, 12:36am

I think we should play agressively against Sharp.

If we play a defensive game, like Fritz and Deep played in the screening, the game might take too long, and perhaps have the same fate as the last mob vs gang. In one year, we can only able to play 20 - 30 moves, so if the game happens to have 60+ moves, its gonna to be a problem.

Also, a agressive game is more fun :)

Title: Re: Mob game 1g
Post by browni3141 on Jun 23^rd, 2015, 10:51am

I don't think style of play is really a choice. We should try to play the best moves. I generally play aggressively because I think that it is the best course of action. You will also see me play less aggressive lines of play when I think they are called for.
My point is that I think discussing how aggressive we should be is counter-productive.

Title: Re: Mob game 1g
Post by deep_blue on Jun 23^rd, 2015, 2:03pm

Lightvector mentioned that he can't guarantee that he doesn't accidentially reads our discussions when done in the chat. So how about another chat like this: http://chatwing.com/noreply-818

Title: Re: Mob game 1g
Post by half_integer on Jun 24^th, 2015, 6:30am

This discussion seems to be winding down, and I have only seen 4 real suggestions: 99of9 with either dogs or cats at the traps, and the two setups proposed by browni. So maybe it is time for a poll. I would suggest that we first decide between generic 99of9 and asymmetric, and then have a second poll to select the exact setup from the winning class.

Title: Re: Mob game 1g
Post by aaaa on Jun 24^th, 2015, 9:15am

on 06/20/15 at 17:58:16, half_integer wrote:

Harvestsnow's post prompts discussion of another related voting topic: In a vote where there are more than two choices, it may be the case that no choice receives a majority of the votes (a tie is a special case of this). I suggest that we should not merely take the move with the highest number of votes, but ensure that each move made has the approval of the majority of voters.

The simplest way to do this is, whenever there is a tie or lack of a majority, the two most popular options are kept and a poll between them alone is taken.

on 06/24/15 at 06:30:58, half_integer wrote:

These posts suggest that you're not aware that the whole idea of using a Condorcet voting method (like Schulze) is that it allows people to submit full and honest rankings of all the candidates in order of their preferences, with relatively little fear of having to be tactical about it. I would advise, after each vote, to only possibly consider having another round of discussion and voting in case there is no Condorcet winner.

Title: Re: Mob game 1g
Post by deep_blue on Jun 25^th, 2015, 4:10pm

Are there any other suggestions for setup? I will start the setup vote after the time control vote finished.

Title: Re: Mob game 1g
Post by harvestsnow on Jun 25^th, 2015, 8:23pm

I don't see how ordering preferences between "more/less than 7 days" and "more/less than 21 days of reserve" makes sense. These options are not exclusive. Is the preference system supposed to work in these conditions?

Title: Re: Mob game 1g
Post by deep_blue on Jun 26^th, 2015, 4:24am

Well, when you like two of those you simply give good numbers for two of them and low numbers for rest.
If you think 7d is too short you DON'T vote for uncapped reserve or so even if you like I guess (or you vote it as 3rd choice then) because that one includes 7d. If more than 7d would win we would then need to discuss how much time instead.

Title: Re: Mob game 1g
Post by deep_blue on Jun 26^th, 2015, 5:33am

I suggest that for now we move the discussions to the old One vs. Mob forum boards.

Title: Re: Mob game 1g
Post by arimaa_master on Jun 26^th, 2015, 5:53am

on 06/26/15 at 05:33:57, deep_blue wrote:

I suggest that for now we move the discussions to the old One vs. Mob forum boards.

Title: Re: **2015 move 1**
Post by clyring on Jun 26^th, 2015, 6:35am

I have no objections to this thread being moved there or to threads for subsequent moves being started there. However, if we relocate our discussions to one of the older mob sections we should very clearly label all mob-sharp threads as '2015 move X' to limit possible confusion with the older discussions.

EDIT:

on 06/25/15 at 16:10:27, deep_blue wrote:

Are there any other suggestions for setup? I will start the setup vote after the time control vote finished.

I reject starting voting on the setup before the game actually starts.

Title: Re: **2015 move 1**
Post by deep_blue on Jun 26^th, 2015, 7:18am

on 06/26/15 at 06:35:42, clyring wrote:

Good point, I will keep that in mind.

Quote:

EDIT:
I reject starting voting on the setup before the game actually starts.

Agreed, but after the time control poll is over we WILL start the game.

Title: 2Re: Mob game 1g
Post by clyring on Jun 29^th, 2015, 7:21am

Now that the game has actually started there is some urgency to this discussion. I note that the talk so far has been rather thin. While I understand this for the 99of9 variants, I think we should further discuss our options, especially browni's more unusual proposal RHCEMRCH/RRRDDRRR, which has not seen much testing in human practice as far as I am aware.

I'm going to label our four main proposals so far as follows:
(A) RHCEMCHR/RRRDDRRR (99of9, 99of9 cats forward)
(B) RHDEMDHR/RRRCCRRR (99of9 dogs forward)
(C) HMDEDCHR/RRCRRRRR (browni setup)
(D) RHCEMRCH/RRRDDRRR (h-horse setup)

Under which circumstances are we likely to not want to advance a horse in the east using setup (D)?

Silver setting up with mh in the east. This seems the clearest case of our horse not wanting to directly attack, but then we have still gained time relative to the 99of9 if our response in either case is to slide our camel and horse over to match.
Silver setting up with m in the west and hh in the east. This would certainly limit our horse's ability to achieve an attacking position in the east, but seems better than we could hope for with the 99of9 against an ehh setup as our camel is already on the right wing.

More generally the horse on the h-file only ever seems misplaced if our camel is not coming east soon. We could just set up our camel on the g-file to start with (D*), which would lead to something similar to (C), leading me to the conclusion that the two setups are quite closely related and the critical tries against each are those in which the camel might wish to leave its native wing early, rather than those in which the horse might think twice about advancing early. Generally with the elephant and camel a net three steps further from the preferred flank in (D) than (C) it will be much harder to apply immediate pressure on the preferred flank but easier to pursue an alternative plan in the former setup.

There are two broad categories of responses to the browni proposals that I consider interesting:

Variants on ehh/m with the camel as a primary defender of the eastern wing. Due to the limited ability of the two horses to quickly generate threats of their own on the opposite wing, against (C) browni has previously (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=332901) placed one of the horses centrally to aid in the initial defense of c6 which seems to come under too much pressure early on otherwise. Against (D) the central horse probably isn't immediately necessary but has its own implications more similar to an ehh/m setup against a 99of9 setup. (D*) mentioned above to illustrate the relationship between (C) and (D) I don't think is as strong as either (C) or (D) against this flavor of setup as gold is significantly less well equipped to either apply pressure or switch gears.
Setups intending a counterattack on the other wing with mh. I admit I haven't seen many other players take this approach but it is my preferred response to (C) and similar setups, potentially leading to nearly symmetrical arrangements of the camels and horses, but with both elephants on gold's camel wing, which only needs to be worth two steps. However, due to the centralized camel, I think this idea probably holds less substance against (D). Unfortunately I don't have a particularly nice example game. The three I could quickly dig up: 1 (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=303488) 2 (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=310880) 3 (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=322240)

Let the dissent begin! :)

Title: Re: 2015 Move 1
Post by deep_blue on Jun 29^th, 2015, 9:30am

I wonder why even people like browni who strongly favor that setup still used 99of9 in some of the world championship games. I still think a standard setup has the advantage of more players being comfortable with it but then again some of the browni setup arguments sound convincing so I am open for that too.

Title: Re: 2Re: Mob game 1g
Post by deep_blue on Jun 29^th, 2015, 6:25pm

on 06/29/15 at 07:21:22, clyring wrote:

(D) RHCEMRCH/RRRDDRRR (h-horse setup)

Under which circumstances are we likely to not want to advance a horse in the east using setup (D)?

Silver setting up with m in the west and hh in the east. This would certainly limit our horse's ability to achieve an attacking position in the east, but seems better than we could hope for with the 99of9 against an ehh setup as our camel is already on the right wing.

What about m in the east and hh in the west? This looks better than a ehh attack vs. the 99of9 IMO. (because of the slightly misplaced H). So the only way to play for a better position is playing MH vs. m then but does this really give fast attacking chances (since getting M to g3 also costs time)?

EDIT: Btw, even browni said he was not sure if he would vote for his setup.

Title: Re: 2015 Move 1
Post by Algorias on Jun 30^th, 2015, 5:00am

I wonder whether we are attempting to play the best game possible against a versatile and strong opponent, or trying to exploit sharp's specific weaknesses. From the discussion here, the former seems to be the case. I think we should seriously consider the alternative, however.

In my opinion, sharp is weakest in nonstandard situations. For example, after an even camel trade, it stays strong tactically but flounders about strategically. Consider c&g game 3 of the 2015 challenge match:

http://arimaa.com/arimaa/gameroom/opengamewin.cgi?client=1&role=v&side=w&gameid=422656

Here, sharp is lured into a strange position with an early advanced elephant (move 4s), followed by a short maneuver resulting in the exchange of camels. At this point, sharp starts playing significantly worse, bleeding 4 minor pieces over the next 25 or so moves. This was one of 2 the games won by the human side, and it was a brutal bot-bashing that makes sharp look very stupid, specially in the context of it just having swept the challenge match.

This might have of course been fixed by lightvector in the meantime, and maybe he would consider some move in that sequence a significant enough blunder to intervene (he has a lot of leeway on how to interpret the agreed upon rules). A deliberate baiting strategy of this kind is therefore somewhat risky. We also don't know to what extend sharp's 2 minute play predicts its one week play.

With that said, I still propose that we work out a starting setup geared specifically to achieve an early camel trade, be that 99of9 as in the example game or something different, and add it to the poll options.

Title: Re: 2Re: Mob game 1g
Post by clyring on Jun 30^th, 2015, 6:45am

on 06/29/15 at 18:25:54, deep_blue wrote:

I anticipated and mentioned this in my post:

on 06/29/15 at 07:21:22, clyring wrote:

Generally with the elephant and camel a net three steps further from the preferred flank in (D) than (C) it will be much harder to apply immediate pressure on the preferred flank but easier to pursue an alternative plan in the former setup.

There are two broad categories of responses to the browni proposals that I consider interesting:

Variants on ehh/m with the camel as a primary defender of the eastern wing.

This situation against the h-horse setup is very similar to 99of9 vs ehh/eastern m setups. I don't think the h-horse changes too much or is misplaced: Note that g3 often remains vacant for a few moves anyway in 99of9 vs ehh/eastern m setups and that there is no horse available to quickly threaten to occupy g3. Indeed the horse might even be better placed on the h-file if we want to use it to push back or pull silver rabbits or give it up as a hostage.

Title: Re: 2015 Move 1
Post by Algorias on Jun 30^th, 2015, 7:05am

Whoops, I hadn't realized the poll already went out. Well, regardless of the initial setup chosen, my general question still stands.

Title: Re: 2015 Move 1
Post by ikalyoncu on Jun 30^th, 2015, 7:25am

Hey mobsters, I have a couple of questions as I am quite new to this mob vs somebody type of play.

1. Is there a permanent link to this game? If so, could you please share. I assumed it will pop-up under "Live Games" section once the game has started, but I see that this is not the case.

2. Will the bot be pondering about the game all the time, or will lightvector control when/how much the bot is active?

Title: Re: 2015 Move 1
Post by clyring on Jun 30^th, 2015, 11:17am

on 06/30/15 at 05:00:13, Algorias wrote:

Considering lightvector has reserved the right to intervene whenever and however he sees fit, I think the benefits of trying to exploit sharp's weaknesses are limited.

on 06/30/15 at 07:25:34, ikalyoncu wrote:

The game will be accessible through this page (http://arimaa.com/arimaa/gameroom/opengamewin.cgi?gameid=429383&role=v) until it ends, after which it will be available at this page (http://arimaa.com/arimaa/gameroom/comments.cgi?tgid=429383) as long as we continue to use this site.
It doesn't show up under live games because it is, in fact, not a live game, but a postal game.

on 06/30/15 at 07:25:34, ikalyoncu wrote:

2. Will the bot be pondering about the game all the time, or will lightvector control when/how much the bot is active?

I believe lightvector has full discretion to do anything and everything relating to this game except read the Mob forum or solicit outside help.

Title: Re: 2015 Move 1
Post by deep_blue on Jun 30^th, 2015, 1:34pm

on 06/30/15 at 05:00:13, Algorias wrote:

In my opinion, sharp is weakest in nonstandard situations. For example, after an even camel trade, it stays strong tactically but flounders about strategically. Consider c&g game 3 of the 2015 challenge match

I strongly disagree with that idea. One game tells NOTHING, the lemmings might just occur in that special position and nowhere else. Also I became convinced now that we should beat sharp "normally" to show that one can still do that.

Title: Re: 2015 Move 1
Post by browni3141 on Jun 30^th, 2015, 8:02pm

on 06/30/15 at 13:34:18, deep_blue wrote:

Also I became convinced now that we should beat sharp "normally" to show that one can still do that.

Were my two wins not normal?

Regarding trying to use exploitative strategies, I am against it even if lightvector couldn't intervene. Bot-bashing teaches little about the game and perhaps even harms general playing ability, and I would like for this game to be instructive.

The reason I say I may not vote for my setup is because there are so many gold setups which I can't distinguish between easily. I personally don't have a strong opinion about what we should do. 99of9 with dogs behind the traps is the only suggested setup I would be unhappy with.

Title: Re: 2015 Move 1
Post by Manuel on Jul 3^rd, 2015, 1:51am

What is currently holding up making the first move? I understand the poll has gone out 3 days ago, but there is no move made yet...
Do we get to see the outcome of the poll?

Title: Re: 2015 Move 1
Post by deep_blue on Jul 3^rd, 2015, 8:30am

Poll ended. Winner was 99of9 cats forward and I played that move.
Link to result:
http://civs.cs.cornell.edu/cgi-bin/results.pl?id=E_8295027e28c4f8c5

In case that one gets lost:
1. 99of9, cats forward: RHCMECHR/RRRDDRRR
2. browni setup: HMDEDCHR/RRCRRRRR
3. 99of9, dog forward on M side: RHDMECHR/RRRCDRRR
4. 99of9, dog forward on E side: RHCMEDHR/RRRDCRRR
5. browni's second suggestion: RHCEMRCH/RRRDDRRR
6. 99of9, dogs forward: RHDMEDHR/RRRCCRRR

	1.	2.	3.	4.	5.	6.
1.	-	13	19	19	16	17
2.	5	-	14	15	14	14
3.	1	6	-	9	12	9
4.	1	5	5	-	9	8
5.	3	4	8	8	-	11
6.	3	6	8	8	8	-

Of 22 voters, 20 voted. The right to vote (those 22) had:
Algorias, arimaa_master, aurelian, browni3141, chessandgo, clyring, CraggyCornmeal, DeathCure, deep_blue, half_integer, Harren, harvestsnow, Hufflepup, ikalyoncu, Knedlik, PerkofBR, phairland, RonWeasley, Samraku, SilverMitt, Skarn, supersamu.