|
||
|
Title: Challenge Strategies Post by deep_blue on Mar 13th, 2015, 8:41am Since the screening is supposed to be a possibility for humans to find the bot's weak spots there should also be a forum thread to collect those working strategies. I think noone would disagree that sharp wins the Screening so here's my experience of my first sharp game: 1. By shuffling in the opening like I did one can gain a reasonable amount of reserve which can help later. 2. Sharp seems to overevaluate hostages so one should give it the possibility to give material to gain a hostage. 3. Sharp seems to sometimes attack too aggressively like it did against me after taking the camel hostage. This can lead to material loss for it. (not sure though that 3. always works) |
||
|
Title: Re: Challenge Strategies Post by browni3141 on Mar 13th, 2015, 2:47pm on 03/13/15 at 08:41:23, deep_blue wrote:
Sharp can be doubly exploited here. It will let you gradually develop while it does relatively little, which is even better than mutual shuffling. This might even be good general Arimaa strategy anyway in a lot of positions. |
||
|
Title: Re: Challenge Strategies Post by deep_blue on Mar 31st, 2015, 9:21am After my second Screening game I am fairly sure now that Sharp can't be beaten on Score, it will eventually start to advance. But when shuffling in the beginning one can exploit that advancing see this game where I got a very nice position and only lost due to a blunder (terrible last step on 46s): http://arimaa.com/arimaa/gameroom/comments.cgi?gid=333472 |
||
|
Title: Re: Challenge Strategies Post by Belteshazzar on Apr 1st, 2015, 6:07am on 03/13/15 at 08:41:23, deep_blue wrote:
I noted here (http://arimaa.com/arimaa/gameroom/comments.cgi?gid=327465) that Sharp was actually willing to trade a horse plus a cat for the camel. Someone could try to make that trade in the opening, giving up their camel for a horse and cat, and then at least have an edge in quantity of material. This could provide new insight as to the relative value of pieces. If nothing else, it might be a good experiment for the third game if your match is already decided. |
||
|
Title: Re: Challenge Strategies Post by chessandgo on Apr 2nd, 2015, 3:48am Sharp seems to play really well after taking a caMel hostage, blitzing the (diagonally) opposing enemy trap with camel and horse, and playing great technique when the enemy elephant leaves for a trade to keep shared control of the trap for as long as possible. Unless someone has an example of recurrent misplay by sharp, I'm not thrilled by playing a type of position that I know sharp plays well. Maybe try something more unusual, like camel vs camel opening trade? |
||
|
Title: Re: Challenge Strategies Post by deep_blue on Apr 2nd, 2015, 4:12am Chessandgo, there is a counter example for its hostage play. Just look at my Screening win, that should be enough counter example: http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331427 |
||
|
Title: Re: Challenge Strategies Post by chessandgo on Apr 2nd, 2015, 11:13am I'm not sure what you mean by counter example. If you mean that sometimes sharp takes a bad shotage, then the terrible 38s in your game is indeed proof of that. If you mean a counter, period, I'm clueless about what it is. Your game does make one want to play a game with camels on the same side and bait sharp into giving a piece for the hostage. Since camels on the same side should make the game less tactical (at least initially), I think I'll mirror sharp if I have silver on my first game. If I have gold, setup 99of9, and sharps sets up EHH, I have no idea. |
||
|
Title: Re: Challenge Strategies Post by harvestsnow on Apr 5th, 2015, 3:30pm EHH vs. ehh used to work well against sharp. It would let a rabbit be pulled in the HH side while not being able to make progress in the camel side. A good example of this type of game from last year's screening: http://arimaa.com/arimaa/games/jsShowGame.cgi?gid=296721&s=w (arimaa_master vs sharp) This game is particularly interesting from a weakness study perspective in that sharp willingly shifted its second horse (instead of the camel) to the frame side. So there actually was a slope towards the weakness area, regardless of the setup. It seams to me that sharp's evaluation of balance has improved since last year. We saw good rebalancing moves, like 15g in http://arimaa.com/arimaa/games/jsShowGame.cgi?gid=333270&s=w , and also moves that seemed to have the purpose of imbalancing the opponent. I expect it to be harder to exploit this way, although this particular strategy hasn't been attempted in the screening. I fear that in the first game I mentionned, sharp2015 would quickly take advantage of the crazy MHH setup, block the lateral moves and send a horse to the weak side. An interesting point is that sharp won most of its games with hostage or frame strategies, despite being known for its aggressive style. Sharp seems to wait for the ennemy camel to lose mobility before trying to take control of an ennemy trap (game 331726 (http://arimaa.com/arimaa/games/jsShowGame.cgi?gid=331726&s=w) being a possible exception). The stalling games suggest that it should be possible to develop slowly into a good position while accumulating reserve, and without giving anything away. Easier said than done obviously. I wanted to classify the screening and WCC games according to their durable positionnal features and strategic themes. I'm not sure I will actually finish it, so here is what I have so far. It's a bit raw and possibly Camel frames: http://arimaa.com/arimaa/gameroom/comments.cgi?gid=332283 99of9 http://arimaa.com/arimaa/gameroom/comments.cgi?gid=327729 722caasi http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331377 kzb52 http://arimaa.com/arimaa/gameroom/comments.cgi?gid=332780 Hufflepup Camel hostage: http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331427 deep_blue http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331791 SilverMitt http://arimaa.com/arimaa/gameroom/comments.cgi?gid=332499 omar http://arimaa.com/arimaa/gameroom/comments.cgi?gid=332693 supersamu http://arimaa.com/arimaa/gameroom/comments.cgi?gid=333299 SilverMitt Maybe one of the most promising games? Sharp's camel hostage is less intimidating after it has lost its horses. I like silver's position here after 27s ed4e Re3e ee4s cg4w. http://arimaa.com/arimaa/gameroom/comments.cgi?gid=327504 bot_Z Horse hostaged by central elephant (fence?): http://arimaa.com/arimaa/gameroom/comments.cgi?gid=327631 Harren http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331765 arimaa_master http://arimaa.com/arimaa/gameroom/comments.cgi?gid=333206 arimaa_master Dog frame/hostage: http://arimaa.com/arimaa/gameroom/comments.cgi?gid=332973 Boo Notable here is the conversion of the frame into a hostage in order to free the holding pieces. Horse frame: http://arimaa.com/arimaa/gameroom/comments.cgi?gid=333270 Boo rabbit frame http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331865 aaaa http://arimaa.com/arimaa/gameroom/comments.cgi?gid=332510 Fritzlein horse infiltration: http://arimaa.com/arimaa/gameroom/comments.cgi?gid=333397 aaaa (EMH deadlock around a cat frame in the twin trap) http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331726 omar |
||
|
Title: Re: Challenge Strategies Post by Fritzlein on Apr 5th, 2015, 10:08pm If I had known you were going to spend so much time in preparation, harvestsnow, I might have downgraded my estimate of sharp's chances against you. Bravo! |
||
|
Title: Re: Challenge Strategies Post by chessandgo on Apr 8th, 2015, 12:57pm Thanks a lot for the long post Harvey! Not much in the good news department, unsurprisingly. |
||
|
Title: Re: Challenge Strategies Post by Belteshazzar on Apr 12th, 2015, 2:05am People in the chat room now think that you should just play your best game, instead of employing an anti-sharp strategy. |
||
|
Title: Re: Challenge Strategies Post by mattj256 on Apr 17th, 2015, 5:40am Hi all. I just finished going through all the screening games looking for strengths, weaknesses, and patterns in Sharp's behavior. (These notes mostly don't include the Challenge games although I did watch all of them.) This is somewhat disorganized and I don't have the energy to write it more neatly but I hope someone else will. Things Sharp does well ================== A piece is "close to" the enemy piece of the next-lower rank. E.g. friendly E next to enemy M, friendly M next to enemy H, etc. Sharp always desires this, but there's a bonus if the weaker piece is also frozen or immobilized: Elephant taking camel hostage Horse taking dog hostage etc. Trap control Restrict mobility of enemy pieces. (Mobility is defined as the number of adjacent squares that a piece can move with a single step, push, or pull, without another piece coming over to unfreeze it.) Knows when to put its elephant on a trap square. Bots in general are good at this. (Sharp's elephant usually only moves onto the trap square if there is some immediate tactical advantage.) 332283 34s - human already losing 333318 29g - prevent H from being pushed onto c3. Captures rabbit two turns later. 331726 23g - E is tactically safe but I don't understand what Sharp is looking for here. 332693 3s - frames H on the following move. 331765 45g - goals two moves later. There was one in browni's first game, don't have the energy to look it up. That pattern that nobody has a name for: two strong friendly pieces with a weak enemy piece in between: 333270 29g 331865 4s 333206 26s 331791 36s Avoiding premature rabbit advances. (Advances rabbits in the opening without provocation, but won't advance past the 4th rank unless the rabbit is tactically safe. Tends to not initiate goal attacks unless it has shared or exclusive control of an enemy trap.) - 333270 34g - advances rabbit to c6, then goals three moves later. Superhuman goal defense. - 327786 86g - Fritzlein is ahead by HDDCR and can't push a rabbit to goal without making more captures. - 331427 75g - Sharp is strategically lost. Deep_blue wins nine moves later. - 327631 28g - Sharp defends all goal threats and wins on time. (Sharp had goal-in-one on the board?) - 331726 48s - Omar is materially way behind so he tries for a goal attack. Sharp completely shuts it down and wins six moves later. - 331791 39g - Sharp shuts down opponent's goal attack, then goals on 44s. These are some short Sharp goal attacks: - 333270 34g - advances rabbit to c6, then goals three moves later. - 332973 27s - advances rabbit to b4, goals two moves later. Very accurate tactics 332499 29g - Sharp allows opponent to break a frame and rotate out opponent's elephant. (Frame started 20s.) Sharp also declines to capture C on 30s. Sharp goals 32s. Sharp mistakes ============ Allowing two-for-one or three-for-one replies. (Maybe this is not a mistake?) useless two steps in the move: - 334567 9s - browni3141 1st challenge game - 327729 21s In the opening, allows opponent to shuffle pieces and gain reserve time. (Many have said this already.) Allows its elephant to be blockaded. (Is this a mistake?) After it uses Elephant to take Camel hostage, allows the trap to be stuffed and/or blockaded. 327786 - Elephant pinned to cat, fights fiercely to break the frame but doesn't abandon the frame until immediate material loss is threatened on a different trap. Willing to give up a large space advantage to opponents. (Mistake?) 331726 23g - E is tactically safe but I don't understand what Sharp is looking for here. The next few moves Sharp shuffles his E and H around without gaining any immediate advantage. Sharp's Elephant pinned 327786 17g - abandons rabbit only to avoid immediate material loss in c3. 327786 19g-31g - abandons cat only to avoid immediate material loss in f3. 332488 13s - abandons H, and over the next few moves wins MH for HDR 333397 28g - abandons C to capture D 331765 27g - abandons R to threaten H in c3. This draws enemy E to c3. Then sets up a frame in f6 followed by material gain in f6. Result: wins HD for HR. 333206 48s - Sharp is able to break the frame. Sharp style ======== Tries to get its high-ranked pieces close to the next-lower enemy piece. Doesn't commit to offense until the enemy elephant is strategically committed. Will frame enemy M, but only if it can free its own M or take an enemy H hostage: 331377 16g 332283 10s 327729 12s 332780 10s 327680 30s (Opponent walked into this one.) 333482 8s (This one is an exception: Lion freed his H and then timed out.) Will abandon framed enemy M: 327729 32s [Edit: Position where Sharp has NO interest in the M frame even though one is available: 422564 17s and onwards ] If opponent is totally passive and plays for score, after around move 30 Sharp starts to be more aggressive: - 333472 33g - 331427 32s Good things to do. (Much easier said than done!) ====================================== Maintain exclusive control of one home trap. 333206 - arimaa_master maintains exclusive control of one home trap for almost the whole game. When he loses control of both home traps he loses the game. If Sharp takes your M hostage, leave immediately with E and make a strong counterthreat: 331427 45g (deep_blue gets HD for M) 327786 9s - This is a very good setup for the human, and I believe fritzlein was the only one who has been able to achieve it: first Sharp's E, then opponent E, then opponent M, all in a straight line. This completely shuts down Sharp's camel-pulling. 327786 17s - sets up C frame. 327786 36s: the f4 horse isn't captured right away, but the threat forces Sharp's E to abandon the western wing which leads to a capture in c6 two turns later. |
||
|
Title: Re: Challenge Strategies Post by mattj256 on Apr 17th, 2015, 10:26am This is a shorter (and hopefully more readable) version of my previous post. 1. Sharp is REALLY strong on goal defense, and I don't think people give Lightvector/Sharp enough credit for that part of its game. When a person is behind in material against a strong player it's normal to launch a speculative goal attack. I haven't seen any evidence that this works against Sharp. When people win against Sharp it's because they've won the game strategically and have a huge material lead. (Or they win on score, which is harder than it sounds.) Examples of Sharp's VERY tenacious goal defense: http://arimaa.com/arimaa/gameroom/comments.cgi?gid=327786 (Fritzlein vs. Sharp) 75s Sharp's position is strategically lost and it still lasts almost 20 moves before allowing goal. http://arimaa.com/arimaa/gameroom/comments.cgi?gid=331427 75g - Sharp is strategically lost. Deep_blue wins nine moves later. 2. If you try to make no progress and win on score, after around move 30 Sharp changes its strategy a bit. Sharp first gets the opponent elephant to move to one wing, then advances with horse(s) on the opposite wing. [Edit: browni3141 says Sharp just advances the horse but doesn't otherwise change its basic strategy.] 3. Sharp is very opportunistic. It likes to frame your camel if that's available, but only if it can rotate out its camel or freeze your horse. (In the first case its camel is the strongest free piece. In the second case it has two free horses against your one free horse.) It's not going to frame your camel just for the sake of framing the camel. [Edit: SilverMitt: "I think it takes any frame without too much positional evaluation."] 4. If Sharp's elephant is pinned, Sharp doesn't generally abandon the frame unless there's an immediate material threat in a different trap and the only way to meet that threat is to move the elephant. 5. Sharp will speculatively advance rabbits up to the third or fourth rank. Sharp generally doesn't step the rabbit past the fourth rank unless it can guarantee long-term shared or exclusive control of the enemy trap. Usually Sharp will get shared or exclusive control of the enemy trap first, and advance the rabbit only when it can calculate a forced goal or forced win of material. 6. Sharp is very aware that its elephant should be near the enemy camel, its camel should be near the enemy horse, and so on. Even if it can't take the hostage it still tries to follow this pattern. Sharp tries to take hostages with the "right" piece: horse hostaging dog, dog hostaging cat, etc. Humans usually don't worry so much about horse hostaging cat, but Sharp does. On the flip side, I don't think Sharp is too concerned with keeping its forces balanced. Perhaps this can be used to advantage? 7. When people win against Sharp, they capture at least one of Sharp's pieces in their home traps. (That's my impression based on my memory, and I hope it is correct.) I think the common wisdom is that it's important to take over an enemy trap, because a home game is too slow. When playing Sharp I don't think that holds. Sharp is more opportunistic than aggressive. If you try to take over one of its traps, Sharp will try to hostage/frame/flip pieces to maintain control of the trap and threaten your pieces. Likewise, if you leave a home trap lightly defended Sharp will contest the trap as long as it works out tactically. But it doesn't seem to me that Sharp is very inclined to bully its way into full control of one of your traps. Sharp is more likely to make a threat against one trap, and when you respond to that threat Sharp attacks a different vulnerable piece somewhere else. When Fritzlein was able to win against Sharp, Fritzlein didn't advance anything past the fifth rank until move 21s, and by that time he already had a clear advantage. Another thing -- easier said than done -- is to maintain exclusive (not shared) control over at least one home trap. [Edit: browni3141 says just play your best game and don't worry about anti-Sharp strategies or imitating anyone else's style.] If I were rated as high as Fritzlein maybe I would be able to beat Sharp too. This is the best I can do, as I won't be winning any games against Sharp any time soon. I hope this is of help to the defenders. Matthew |
||
|
Title: Re: Challenge Strategies Post by SilverMitt on Apr 17th, 2015, 11:23am Let's see if I can organize my thoughts into anything coherent... As I see it, there are 3 types of tactics: captures, goal attacks, and hostages/frames. While sharp is naturally quite good at the first two, which are more immediate, it is quite poor at the third due to its more long-term aspect. One still has to be careful about giving up a strong or unbreakable frame, of course. In particular, while, when pulling rabbits, you have to be quite certain they can't be used as a goal threat, and dogs and above can become strong assets near an away trap, cats are almost universally extremely vulnerable in the opening, as they are frozen even by dogs, and sharp can be extremely loose with its cats. Additionally, sharp is incredibly speculative in its attacks on away traps, but also quite good at claiming space and bringing reinforcements so as to have the resources to contest it properly. Allow an opponent to get one strong piece planted near your trap and it's quite difficult to reclaim control over it afterwards. In any case, it should be possible to tell, in many cases, what sharp is likely to do next and prepare for it. If you are not ready, reacting afterwards may be too late. |
||
|
Title: Re: Challenge Strategies Post by browni3141 on Apr 17th, 2015, 7:10pm on 04/17/15 at 10:26:37, mattj256 wrote:
This is what I want to do for myself. I don't necessarily advise others to follow. |
||
|
Title: Re: Challenge Strategies Post by Belteshazzar on Apr 18th, 2015, 10:57am on 04/17/15 at 19:10:27, browni3141 wrote:
Now you're the only one who matters, although it would be encouraging if chessandgo or harvestsnow (or clyring?) could win in the third round. |
||
|
Arimaa Forum » Powered by YaBB 1 Gold - SP 1.3.1! YaBB © 2000-2003. All Rights Reserved. |