Expert Intuition


#1

So, I read this short article about Magnus Carlsen, who besides having a pretty badass name, is also one of the youngest grandmasters of chess ever. That article, and the article it’s based on, are also about Magnus’ chess intuition, where he’s gotten to such a high level in chess skill that he works off of a feel of the pieces as often as working deliberately off of conscious reasoning. The reason I thought this might be a fun topic for FGD is pretty obvious; do you find in yourself or expert players around you a strong feel for any game, or a feel for the “flow” of certain games?

In my own experience, I know that some of the strongest, most experienced players in my area are less interested in frame data and strategies formulated by others and are more interested in experiencing and exploring the game themselves, to find what works for them. I personally love frame data, but as I play more and more, I do recognize how one can feel out the different priorities of certain moves, and the strategies a character should be utilizing all by oneself. I have a really hard time getting a feel for Street Fighter games; sometimes I just don’t know what I should be doing at all. It feels very stiff to me. The less stiff a game is and the more freedom of movement or moves I have though, the more I not only enjoy the game, but can feel the flow of the game better, and become better at predicting what comes next. In Tekken, with more and more experience, I can feel out when the next low is coming without having to do so on reaction. With Virtua Fighter, getting a feel for how I can move around and a feel for what moves the enemy will throw out next becomes more intuitive with every session. ArcSysGames make sense, and I always feel in complete control of my character’s movement.
The one game I can’t feel at all though is Street Fighter. I don’t know quite what it is, but it is far less intuitive to me.

So, your thoughts? Can you feel an opponent out, without too much overt analysis? What say you?


#2

I sort of know what you mean. It took me a long-ass time to get a “feel” for The Last Blade 2. To this day I’m pretty “wooden” at the game, haha.

But the games I play often I have a good feel for, so I think it’s something you end up developing with time invested. So play more SF until you drop your stiffness at it!


#3

Im trying to get used to the Arcana Heart system. Things like the advanced combos sort of throw me off. I guess im too used to the mvc series. As for the SF games, I dont know, some of the things in that game are technical for me to grasp and some of the commands are a bitch to pull off (same with supers in the Guilty Gear series…points to Dizzy’s supers).


#4

I believe that this “feel” is the result of your brain automatically processing massive volumes of data for you, and relaying to you the overall effective significance(s) of an entire situation in a very abstract form, which is easier for you to immediately grasp (so that you can actually employ that new knowledge) because you haven’t consciously followed all of the computations that went into that end result. It’s where you begin to blur the line between scientific approaches to thought and artistic approaches to thought, and I think it is usually a mark of great familiarity with and deep understanding of the subject at hand.


#5

^ What he said.


#6

Yeah, I figured when I first started the article that Carlsen was the product of computer chess experience. When you have software like Rybka that play stronger than pretty much any human while running on blazing multi-core systems, thinking time and good competition become non-issues and a player’s determination will allow him to rapidly advance. If the AI in fighters played like real top players and if you spent enough time practicing, players here would be a lot better. We do have much better online play nowadays, which is essential for those in areas with less competition, but you’re not always guaranteed a smooth match that mirrors offline’s fluidity, or that you’ll play against a decent opponent you can learn from.

There’s a huge difference that makes this analogy not directly apply to fighters though. In chess, the rules are very explicit. Everyone knows every facet of the rules. In fighters, you may know the motions but you don’t know the exact leniency of motions. You may even know general hitboxes but you don’t know the hitbox at every single moment or special exception cases. There are so many hidden minutiae in fighting games that if you don’t experiment (and get a bit lucky) or learn from others who’ve experimented, you could be playing years and not figure out special cases.

Case in point, even the basic concept of reversals was debated on for years during the early days of SF2 because there was no indicator. And because there were no rules, you couldn’t research any rules, and it wasn’t easy (or cheap) to test it out, players had to rely on personal experience. When they saw no meaty beating flash kicks in CE, good players added that guideline to their playing “algorithm.” With enough good guidelines and options, an algorithm becomes very strong and it’s easy to just follow what’s worked in the past. Consequently, you’re able to move in a looser manner and press more buttons with abandon (or because you know some special rules that benefit from doing so). And you start recognizing occurrences long earlier because you subconsciously remember so many times before where the opponent has gone down the same path and acted in the same manner.

But it’s actually not true that the best players aren’t technical. Sure, in the US, most OG players can’t tell you much about the mechanics (although they have the general gist in their algorithm already). But in Japan, top players generally know plenty of specific tricks that allow you to take the most advantage of the rules. It’s possible they know this because at a high level, a good algorithm does well against another good algorithm (although character choice matters more at this stage). You have to start reaching for marbles outside the box again in order to come up with something that will really give you the edge. And that’s where knowledge of more esoteric mechanics is necessary to make sure your algorithm is as streamlined as possible.

The other difference with fighters is the physical aspect. If someone has faster reaction times, they’re at a natural advantage. Plus, there are plenty of dexterous motions required. Pretty much everyone should be able to do them but if someone has hand-eye coordination problems, he’ll be a step back.

As for me, I’ve gained plenty of intuition in ST/HDR through thousands of hours of play. So most of the time, it’s easy to tell at a moment’s notice based on the opponent’s actions what his algorithm level is. One thing I’m not sure why but a lot of players at the same level follow similar patterns; this makes it easy to apply my own algorithm. The way to develop this algorithm is just to play more, play thoughtfully, and play competitively. That’s why players in Tokyo and Osaka are so blessed to have all-day arcades full of competition; a hyperbolic time chamber isn’t much exaggeration.


#7

Intuition imo is just a subconcious thought process that feels like a feeling because you didn’t actually “think” about it, or so you think.

It’s like the way we see…we absorb so much information but we don’t do it bit by bit. We just see a whole. Intuition is the brain’s seeing the whole picture. Usually comes with lots of experience because the more familiar you are with something, the more you can go on auto-pilot. The less familiar you are, the more you have to think, and this comes out as emotional/psychological/mental/social awkwardness.

To put it bluntly, experience is king. I bet this kid played a ton of matches versus super high level CPU chess opponents. At his age, he absorbed so much, and that’s why he’s dominating. I’m almost sure that’s why he’s so good. It even states in the article that he doesn’t have a physical chess board in his home.

To put it in SF terms, this kid basically played Daigo/Valle/Choi/Justin Wong etc. over and over and over. He played the best comp you can play, in his case a CPU.


#8

Just take Echo Black’s advice: get a hard on. :rofl:

Nawh, mad respect for playing The Last Blade 2, my favorite game of all time.

deadfrog is on point. It’s like music. It just happens and you’re not too sure how but it happened.


#9

unfortunately fighting games AI is not like chess in this regard, is that we are probably unable to win on a dexterity level as well, since they have milliseconds reactions (like SuperTurbo’s AI) upon receiving our controls. They are not susceptible to or able to perform “mind games” which higher level players used to play.

Since its harder to pit wits and strategy against them, more often than not, we are probably using some kind of loophole or pre-determined pattern in their programming when fighting them at “high levels”. Eg in SF IV, u can blow through the game with lariet on every level, with nearly every opponent. Less extreme examples, in the earlier Street fighter 2 series, if u crouch jab, the AI tend to jump at u, allowing u to do a AA to know them out of the air.

But supposed we* could* play a virtual Daigo/Valle/Choi/Justin Wong, that would be incredibly helpful to increasing our skills but until then, the only way is to play online matches i guess


#10

Btw excellent thread, I hope it doesn’t go to shit.


#11

I hope you didn’t think I was comparing CPU chess to CPU SF. I was just saying that Chess is a game where you can get arguably better comp by playing the CPU.

PS. Best way to improve in SF is to travel to outside tourneys, and play people that kick your ass. Nobody gets good online, cause online is shit.


#12

This is the most important bit I found in the article. One MASSIVE advantage that chess has over SF is that with chess you can easily set up and replay a very specific situation over and over again, fishing for the very best options. Take that and multiply it by every single situation, analyzed from a variety of different approaches and you have one very beastly player.

To take this one logical step further-- this kid basically played against Valle’s low forward footsie game from 9,000 angles before moving in to low short range options. That’s how deep chess gets on the holodeck.

Imagine if somebody came up with the Deep Blue of SF CPU AI ??? A CPU that would focus on zoning/baiting and always fishing for the absolute best option?

Also, a utility like a Program Pad, but taken one step further would be incredible. Imagine being able to manipulate both characters down to the pixel, then program and play through a whole slew of options. I realize you can do this with 2 PPads and lots of patience, but it would be excellent if this was streamlined for something like SF4.


#13

If they actually wanted to make the CPU GDLK in SF, then not Daigo, not anyone would have a chance. Reactions alone would kill most players. Just a simple algorithm that said for example (in 3S), if Ken blocks another Ken’s low forward, and if the Ken has meter, reversal Shippu. That’s just something super basic and obvious. I think developers should focus more on CPU AI and make it like similar to a top player. Then scrubs could level up at home even if they lived in some small town.


#14

No, on the contary, i meant that we can never get good in sf by playing cpu,
and i concur with ur statement abt tourneys


#15

Magnus Carlsen is a super-hero in Norway. He is currently ranked as #1 in the world with 2813,8. That’s completely insane, given his young age.

Regarding implementing chess software to fighting games, how about giving the AI a reaction time equal to a human’s (~210ms). That way, a person’s optical processing, muscle inertia etc will be negated by delaying the AI from processing information by the same amount of time.

Also, the best Chess software carries a database of almost every recorded game, so it can access f.ex. Carlsen’s playing history and learn from them. It’s plausible to think that the same principle could apply to fighting games (meaning the FG AI would learn from previous games played on that specific console)

Of course, learning completely from scratch probably wouldn’t get the AI anywhere. I was thinking something more in line with: “When human player Lariat, I have 5 coded options. Which option works most often against the human player?”


#16

In regards to the fact that you are “feeling” it: biologically, you are probably literally using more and other areas of your brain to work on a given problem. In addition to the part(s) of your head that are normally suited for such tasks, regions that are not specifically suited to that kind of task are now also pitching in, and even if they aren’t as efficient, they’re still unarguably adding more power. I would venture to guess that the phenomenon is somewhat related to synesthesia. It’s kind of like a reverse application of the idea that it’s easier to react quickly to seeing a huge area of flashed colour than a tiny spot of flashed colour.

That’s exactly the word I was thinking of too.

I think we’re all already at least somewhat familiar with this concept; this type of critical thinking is the norm in fighting games by necessity because these games just simply have wayyy too many variables for us to concretely and exactly take them all into consideration. After examining the entire range of actions that your and your opponent’s characters are capable of, and how those actions will interact with one another, and then looking at the life meters and super meters and other gauges and the timer, plus round points within that game, and potentially game points within that set/match as well, you’re also confronted with the truth that distances are measured in PIXELS (assume that they cannot be counted), and the game is played out in real-time: there is no turn-taking so anyone can theoretically do anything at any time, or at least at any 60th of a second. You’re examining all of this on the fly under conditions that are constantly and continously changing. Unless you’re just going to start arbitrarily omitting large chunks of intel from your decision-making (which is almost always in your disservice), sooner or later you just have to start making estimates and taking mental shortcuts if you’re hoping to distill everything that’s going on into a clear picture of what’s good to do and what’s bad to do… especially if you want to figure that out in a timely manner. I’m willing to bet it’s why our top players have such a hard time explaining some of the things that they do… especially something as hard to pin down as Ryu throwing fireballs at only the perfect times to keep someone out.

Compared to something like chess, for most people, the level of mastery one must first attain to engage this style of thought is probably comparatively lower, because this style of thought is just plain more effective for what we do than it is for what they do. To us, it’s just what we do, but to them, it’s probably an uncommon (and debatable) methodology. It boils down to best-suited strategies for managing and considering as much detail as possible in the context of a particular format. Either way it requires a pretty intimate knowledge of your game.

It seems like it’s basically a streamlining of the process wherein you’re effortlessly doing something like… let’s say… sifting through all the input we can get, eliminating quanta of information that are consistently negligible, selectively eliminating quanta of information that are conditionally negligible, then taking what’s left and ordering it by relative importance, and then doing all the weighing of risk/reward and so on and so forth to distill it all into one (or maybe a few) best choices to pick, which becomes the output of which you are actually conscious. Essentially you’ve learned to perfectly balance the increase of efficiency versus the sacrifice of accuracy, in order to make as many approximations as possible as you figure this stuff out without affecting the veracity of your outcome. To me, this level of advanced thought, or rather non-thought (how very zen hahaha), absolutely screams experience.

Sometimes it’s easier if we’re not aware of the entire mess, left out-of-the-loop so to speak, so that the very specifically conscious portion of our thoughts is free to deal with other things, whether they’re related to that task or not. Effectively, you’re making hard decisions into easy ones so that you can spend more brain time making other and/or harder decisions. It’s the same way you think about where you’re walking to instead of the individual steps that you take to walk. The better we get, the more and more complicated of these tasks we can trust to put on autopilot.


#17

How to become kickass. My mind is fucking blown, makes perfect sense.


#18

The SF4 AI is so disappointing when we have a much more complicated game like Virtua Fighter 5 using a much better AI, that really approaches the human element, almost. I love the AI in VF5, and I don’t know how they programmed such a feat, but I think if just a fraction of that kind of effort went into making a decent SF AI, it would be worth playing against.

Man, other “players” in VF’s quest mode don’t just have personalities but tendencies and sometimes patterns, in a realistic, human way. I fought this Vanessa that was a pretty standard DS Vanessa, but she was obsessed with throwing. She was good at mixing it up, and I’m no good at throw breaking, so she pretty much destroyed me, but to be fighting the computer opponent and go “crap, a throw is coming” and then see that throw, just excites me to no end. That kind of training is useful in the real world, against real opponents. A great thing about quest mode is the ability to go back and challenge a player like that again as well, so I could continue to analyze her behavior and attempt to adjust my own to beat her. It’s very cool stuff, and I’d like to see that in SF someday.


#19

Trying to better emulate top players through AI is such an absurd idea. How much time do you really expect these developers to spend on something that will go unappreciated by 99% of the consumers? Even if they did make something that sophisticated, how do you simulate different playstyles? How much do you really benefit from playing 1 kind of player?


#20

Trying to better emulate top players through AI is such an absurd idea. How much time do you really expect these developers to spend on something that will go unappreciated by 99% of the consumers? Even if they did make something that sophisticated, how do you simulate different playstyles? How much do you really benefit from playing 1 kind of player?