I was recently discussing match strategy with someone that sparked a provocative thought pertaining to the game as a whole and thought it would be nice to share. A player can make the right decision at every turn and still lose the match. To expand the context of that statement, it means that battling strategy is merely one aspect of the game. It is the one aspect players glorify and strive to perfect as it gives them meaning for playing. After all, if the player has no influence over the outcome of a game, it lacks its main competitive attraction. This philosophy alone leads to short sightedness of the game, as team match-up and luck play partial roles in the outcome of any match. An ignorance of the less desirable aspects of the game puts any player at a competitive disadvantage. The question this discussion seeks to answer is if you cannot always win through correct decision making, what significance does making the right move have in relationship to all other aspects of the game? Can a seemingly bad decision be a good decision in the right context? Do you as a player consider all aspects of the game before and during a match? What do you believe is the ideal balance between team construction, battling strategy, and managing probability factors?
OU: Strategy Precision of Play
- Thread starter Kevin Garrett
- Start date
Focus everything in team construction and battle strategy. You cannot manage probability factors. Managing probability means you have a second plan to everything, just in case, and there's no way you can do that in a team of six. Making bad decisions is also never right in any context, due to the nature of the game. It is not a bad decision to use EQ on a Flying type instead of your STAB Ice, cos there's a good chance your opponent will switch in his Steel type to take the attack, only to find themselves getting wrecked. For a player that knows what he is doing, every decision is right in theory, and it's only the result that matters. Yes, you can at times make all the right decisions and still lose because even though you predicted the switch, your Hydro Pump missed, or because your opponent's team seems to have been built only to counter yours. However, this will be a rare occurence. You will not always lose due to luck, and if you have made a good team, it is very unlikely something will be certain to break you, even if matchup seemingly doesn't favor you.
tl;dr You cannot consider probability and luck factors during a match. You can only consider the best move possible. If in the end luck turns against you, it happens. In the end though, if you really have made all the right decisions it's much more likely that you will win.
tl;dr You cannot consider probability and luck factors during a match. You can only consider the best move possible. If in the end luck turns against you, it happens. In the end though, if you really have made all the right decisions it's much more likely that you will win.
Let's put the line of thought from your post into a potential real game situation. You have a Pokemon active with approximately 30% of its health, access to a recovery move that heals 50% in one turn (Recover, Soft-Boiled, Moonlight, etc.), you are carrying Leftovers, you outspeed the opponent's active Pokemon. Stealth Rock is in play, so you understand that you will be not be able to freely switch your active Pokemon back into the game without sacrificing another Pokemon. Your opponent's Pokemon's strongest attack deals a damage range of 48-52%, allowing you to recover on average 6% each turn. For practicality sake, let's assume you need this Pokemon to win the game. Now, since you outspeed, you can just outright attack to get the KO and only recover 6% for the one turn. Alternatively, you can use 8 Recovers to outstall the PP of the move your opponent is using to do this 48-52%, which will give you the opportunity to heal 42% of the 70% you are missing. Thus, putting you in a seemingly more comfortable position to win the game.
Here's where the philosophy of you can't manage probability factors comes into play. If you attack, you don't put yourself at risk of a critical hit or a secondary move effect that could burn, freeze, poison, or paralyze your Pokemon. If you blindly say, "I need this Pokemon at full health because it puts me in the best position to win," without realizing you are giving your opponent 8 free turns to get either a critical hit or a secondary effect that will neutralize your attempt at healing, you will outright lose the match.
This is but one example that comes into play in almost every game a player plays.
Can you still say the player doesn't manage probability in a match?
Here's where the philosophy of you can't manage probability factors comes into play. If you attack, you don't put yourself at risk of a critical hit or a secondary move effect that could burn, freeze, poison, or paralyze your Pokemon. If you blindly say, "I need this Pokemon at full health because it puts me in the best position to win," without realizing you are giving your opponent 8 free turns to get either a critical hit or a secondary effect that will neutralize your attempt at healing, you will outright lose the match.
This is but one example that comes into play in almost every game a player plays.
Can you still say the player doesn't manage probability in a match?
If you /absolutely/ need that Pokemon at full health to win the game, then you have no choice but to stall out the PP. Otherwise, you lose regardless. The odds of a critical hit coming in the span cannot be more then 50% thus you /should/ win. Obviously doesn't always work out in practice since you can literally play hundreds of matches but odds are in your favor. Like any game that has luck involved such as Texas Hold 'Em, the better player will win more then the less skilled player but it just happens that the lesser can win due to the nature of the game. The best practice for any game is to figure out your best path to victory and work towards that goal. I literally have no idea how common it is now but the infamous last pokemon sweep directly falls into this category. Identify what your opponent has to stop said Pokemon, remove or weaken said Pokemon and strategically trade Pokemon with your opponent until you can get the cleaner out onto the field. You can still lose due to a timely critical hit, several turns of para, freeze, etc but you'll win more then you lose.
This is where I wanted the discussion to go, thanks Pride.
Just to move on from my previous post, the example I gave you lists your probability of successfully accomplishing your sequence of moves at 25.6%. The point I am making is that if you don't consider these numbers or know how to calculate them and go through your entire game plan pretending like your Pokemon are chess pieces with absolute control over the outcome, you are greatly mistaken. As sound as your plan for recovering your win condition may be, it leaves you with only a 25.6% chance to win. If a player doesn't manage their probability or luck factors, then the presumed best decision is likely the wrong decision.
To make this discussion more approachable to members here, I will say that I purposely picked a topic that has a gray area and requires deep thought and reflection about how all players should approach the game.
Edit: Fixed a rounding error, but now it's good.
Just to move on from my previous post, the example I gave you lists your probability of successfully accomplishing your sequence of moves at 25.6%. The point I am making is that if you don't consider these numbers or know how to calculate them and go through your entire game plan pretending like your Pokemon are chess pieces with absolute control over the outcome, you are greatly mistaken. As sound as your plan for recovering your win condition may be, it leaves you with only a 25.6% chance to win. If a player doesn't manage their probability or luck factors, then the presumed best decision is likely the wrong decision.
To make this discussion more approachable to members here, I will say that I purposely picked a topic that has a gray area and requires deep thought and reflection about how all players should approach the game.
Edit: Fixed a rounding error, but now it's good.
Yes, I can. If the Pokemon needs to be at a specific health for me to win, then attacking right off the bat will probably result in me losing anyway, so it isn't an option. Even if I do have a chance at winning with less than the ideal health, guaranteeing that I will win with PP stalling is still ahead. The chance of a critical hit exists, yes, but if it occurs then so be it; chances are that it won't, however, and there is no need to even take into account, it's an aspect of the game you cannot control. The secondary effect you mentioned doesn't have to do with probability, but I will answer to it for clarity's sake. Assuming your opponent does have this secondary effect, then we must also assume that you do not know the opponent's set. In other words, you're risking your must-have Pokemon against a huge potential threat without having scouted the threat first, which is not the best decision to make. Assuming you know the set, you act accordingly by either switching or keep up the fight. Also, if your opponent doesn't have a secondary effect to neutralize you, then he will most likely switch out himself and not stay in for a critical hit; the opposite decision will cost him the game in the majority of situations, and it's a bad decision to depend on mechanics you cannot manage.
Your example is very close to a matchup with Gothitelle's Literally Satan set (the main difference being you don't have the switching option against STag). Gothitelle stalls the opponent by trapping it and then using Rest and Calm Mind repeatedly, with many players expressing frustration over the fact that Gothitelle users extend matches by ridiculous numbers of turns in order to find themselves at the highest possible health when they finally destroy the trapped opponent. While critical hits do exist and can ruin this strategy (I myself have beaten a Gothitelle with a critical hit), in the vast majority of situations it won't happen and Gothitelle's enemy will lose, with the rest of the team following against a +6/+6 boost, hence the set name and its effectiveness.
You cannot control luck. You cannot manage something you can't control. Luck can ruin the most carefully calculated strategy, but statistically speaking it will do so in very few occassions. Therefore, there is no need for you to take probabilities into account.
Your example is very close to a matchup with Gothitelle's Literally Satan set (the main difference being you don't have the switching option against STag). Gothitelle stalls the opponent by trapping it and then using Rest and Calm Mind repeatedly, with many players expressing frustration over the fact that Gothitelle users extend matches by ridiculous numbers of turns in order to find themselves at the highest possible health when they finally destroy the trapped opponent. While critical hits do exist and can ruin this strategy (I myself have beaten a Gothitelle with a critical hit), in the vast majority of situations it won't happen and Gothitelle's enemy will lose, with the rest of the team following against a +6/+6 boost, hence the set name and its effectiveness.
You cannot control luck. You cannot manage something you can't control. Luck can ruin the most carefully calculated strategy, but statistically speaking it will do so in very few occassions. Therefore, there is no need for you to take probabilities into account.
You are admitting here that you do manage probability in your matches, which is a good thing. If you know you need a certain percentage of health to win, you aren't going to recover past that point because you are aware of the luck chances that are working against you. Though, in this example, they are slower so the only way to get to your desired health is to completely PP stall unless 56% is the health that can afford you to win the game.Chaos Jackal said:
First of all, the secondary effects matter. For the sake of the example, let's say it is Ice Beam. If you're frozen, assuming you don't have Heal Bell or the like and they aren't using Fire moves, that's an added 10% that goes into the equation against you. When the odds are in your favor, you naturally live with the so be it. However, what if you need to stall 16 pp, where the odds are no longer in your favor. That either means you need to plan an alternative strategy, or if it is your only remaining chance to win, there were likely previous incorrect decisions, poor team match-up, or luck leading to your situation.Chaos Jackal said:
Everyone knows you can't control luck. Manage does not mean the same thing as control. Managing luck means that you take it into account of your decisions and you move with the play that leads to the highest probability of you winning. If it helps you, think of it as you are aware of an entity that is constantly in play in the game and you utilize it to the best of your ability.Chaos Jackal said:
The point is that battling strategy has to require probability management. Otherwise, the strategically best play is not necessarily the most winning play. The examples that both you and I gave are somewhat shallow since it's not very practical to post an entire game situation to have a true simulation. For instance, if the plays that got the players in each example to their respective situations were wholly necessary. Nonetheless, they do work to show how players need to comprehend the things they can't control in a game if they want the best chance to win.
Chaos Jackal said:
If you had a choice between using Surf and Ice Beam as your last attack and you needed a crit to win the match, you would obviously choose Ice Beam, due to the Freeze rate, over Surf thus you're taking the move that would give you a more probable victory.
Just like your scenario, if you were facing against that +6/+6 Goth; you would place higher value any move that had an abnormal crit chance. Such as you have Stone Edge and Earthquake on the same Pokemon. Earthquake cannot miss and can still crit but the value of Stone Edge's 20% crit rate is the better despite that you could miss and deal no damage.
Kevin Garrett said:
Not quite. It's not a matter of whether luck works against me or not; it's just unnecessary. I need a specific amount of health to win. I got it. Now let's win.
Kevin Garrett said:
You will still lose if you don't stall him anyway, so what's the point? Either way, calculating the overall chance of not getting frozen can be quite misleading, because in the end each Ice Beam has a 10% chance of freezing the target, regardless of how many did or didn't. Probabilities and statistics are among the most unstable and uncertain aspects of maths, and depending on them isn't a good idea, especially in such cases. Say you can win without the ideal amount of health, which you would get by stalling Ice Beam to 0. Can you calculate the chances of you winning if you stall for less turns or outright shoot him in the face? I salute you if you do, because I personally can't read through all the potential moves and situations that will arise depending on my choice. And if you don't have something to compare your results with, what good will they do to you?
Kevin Garrett said:
That's the thing... you can't really utilize luck. It's an unstable element, and I personally do not trust it.
In the end, answers might depend on what each of us perceives as probability, what we consider managing of it, and just personal opinion. I do not believe you can manage probability, let alone try it. Would you let a Pokemon that does roughly 45% of damage against a Recover opponent in, hoping to land a critical hit?
Pride said:
As I said above, it's a matter of what you consider probability managing. You will lose anyway, so let's just go with that. You do not really manage probability; you know from the move's description that it can freeze, or that it is more possible to crit, and you're in a desperate situation.
I can see where you guys come from, and you are right from that point. But it's also a matter of what you perceive as probability managing. From your point of view, we're managing probability in such situations; from mine, we do not. Maybe I have a rather narrow view of probability, connecting it mostly to luck in Pokemon match and not to something else. Perhaps I see this from a wrong viewpoint.
I really like this conversation, but I need to go to sleep. University starts tomorrow. I'll be sure to follow up though.
That is the flaw of the example I gave for continued discussion, as it was absolute to the one play. There's no point in backtracking to broaden the example because it served its purpose in advancing the discussion.
The best approach for a competitive player is to weigh their options and the contingencies of those options against one another. How a player chooses to place value in those options is the gray area I alluded to in my previous post. It's a major distinguishing factor between players and it all boils down to the final question in my OP. Your original answer was that you do not focus on the things you cannot control. Do you make decisions considering the factors you can't control? Or if it is out of the reach of your decision making, do you not discriminate between your options because you consider it to be a lost cause?
I like most of the things Pride said to add to this discussion, but there is one statement he made that needs to be expanded on. He said that the most skilled player will win more times than lesser skilled players, and by extension win more times than luck will defeat them. However, success in Pokemon is sadly not based on the complete profile of games you have ever played. Tournaments are pressure cookers for all decisions because if you lose the amount of times the format specifies, you're done. A single game can decide your fate and reputation. The lines can get blurred in the course of a single match, leading to successive dealings with chance. There is no right way or wrong way to weigh the options. Much like a head coach deciding to go for a 4th down play or an onside kick in the NFL, calling an addition street in Texas Hold'em to hit your gut shot straight, and the like, the answer is, "it depends," on the specific situation in the match.
If you are a player who never takes chances, you will rarely be on the receiving end of good favor. If you are a player who constantly takes chances, you will have your fair share of big time wins, but equally much disappointment. There is a balance to strike in weighing probability into our decisions. Whether that balance is dependent on the number being above or below 50%, something you see in the way your opponent plays, or an instinct that you cannot explain, the balance needs to exist for any player to be successful.
The best approach for a competitive player is to weigh their options and the contingencies of those options against one another. How a player chooses to place value in those options is the gray area I alluded to in my previous post. It's a major distinguishing factor between players and it all boils down to the final question in my OP. Your original answer was that you do not focus on the things you cannot control. Do you make decisions considering the factors you can't control? Or if it is out of the reach of your decision making, do you not discriminate between your options because you consider it to be a lost cause?
I like most of the things Pride said to add to this discussion, but there is one statement he made that needs to be expanded on. He said that the most skilled player will win more times than lesser skilled players, and by extension win more times than luck will defeat them. However, success in Pokemon is sadly not based on the complete profile of games you have ever played. Tournaments are pressure cookers for all decisions because if you lose the amount of times the format specifies, you're done. A single game can decide your fate and reputation. The lines can get blurred in the course of a single match, leading to successive dealings with chance. There is no right way or wrong way to weigh the options. Much like a head coach deciding to go for a 4th down play or an onside kick in the NFL, calling an addition street in Texas Hold'em to hit your gut shot straight, and the like, the answer is, "it depends," on the specific situation in the match.
If you are a player who never takes chances, you will rarely be on the receiving end of good favor. If you are a player who constantly takes chances, you will have your fair share of big time wins, but equally much disappointment. There is a balance to strike in weighing probability into our decisions. Whether that balance is dependent on the number being above or below 50%, something you see in the way your opponent plays, or an instinct that you cannot explain, the balance needs to exist for any player to be successful.
Just for people wondering:Pride said:
if you have 8 turns to land a critical hit then:
Critical hits land 1/16 of the time
To do this will find the odds of hitting 0 critical hits and subtract it from 100%
15/16 times you will not get a crit
15/16 * 15/16 * 15/16 * 15/16 * 15/16 * 15/16 * 15/16 * 15/16 = 2562890625/4294967296=59.7%
So chances of getting a crit in 8 turns is 40.3%
So yes you should win!!
HOWEVER: let's say they're using T-bolt... 1/10 of the time it gets that "game breaking paralysis" (which makes sense since paralyze F's up your speed)
9/10 times you don't get the paralyze so:
(9^8)/(10^8) = 43.1% you don't get paralyzed
So you will be paralyzed 66.9% of the time O.O dang.
Mix that with chances of crit you have:
(9^8)/(10^8) * (15^8)/(16^8) = 25.6% to not get paralyzed or critted
So you WILL lose 74.4% of the time
Just thought I'd leave this post here for anyone wondering.
EDIT: yes I realize someone stated the probabilities sort of, I was just showing the math for anyone curious
There are actual absolute answers to some of the questions in this thread. The science of this exact topic is game theory. Here are some of the major conclusions (without any proof or detailed explanation. sorry, but to learn this much has taken months of research):
NOTE: This post is HUGE. Skip to the section with the list with bold text for the conclusion (tl;dr). Read through for some insight into how those conclusions were reached.
Firstly, there is, in any game state, an absolutely optimal strategy, which will result in the greatest overall record if used consistently. An easy (to follow, not to explain, as seen by the length of this post) example is Rock-Paper-Scissors.
If your opponent picks rock, the best strategy is always paper. If you pick paper, your opponent's best strategy is always to pick scissors. But wait! You don't know what your opponent's move will be. This is called a simultaneous game. You and your opponent make your decisions, with no knowledge of your opponent's decision before the "ply" (a turn in this case) is carried out (throwing your hand in the chosen shape).
So what's the optimal strategy for RPS? Well, suppose we assign the end of any game of RPS a value: 1 if you win, 0 if you tie, -1 if you lose. You would like, over time, to win as much as possible, and lose as much little as possible.
Draw or imagine a chart (I can't do that here, sadly) with each axis representing a player, with R, P, and S written along them. In the middle, write the values we chose earlier corresponding to each combination of moves. If you just chose Scissors over and over again, your opponent could easily determine that their best move is to choose Rock over and over again. Thus you get a net average score of -1. Obviously we can do better. The optimal strategy, as can be determined with a little thought, is to choose randomly between your three options with equal probability each (33.3%). This way, regardless of what move your opponent makes, you get a 1/3 chance to win, a 1/3 chance to tie, and a 1/3 chance to lose. Then we can multiple these probabilities with their respective "payoffs" (the values we chose last paragraph), add them together, and we get 0.
That's the best strategy: choose randomly, and over a very large period of time (look up the Law of Large Numbers in Statistics if you're not sure of this) you'll win precisely as often as you lose. You cannot do better than this. There is no way to guarantee a better average payoff than 0.
But there are actually Rock-Paper-Scissors World Championships (just like Pokemon)! How could you conceivably have a competitive games taken seriously if all they're doing is choosing randomly and hoping that luck is on their side? Well if you look back, you may notice (or you may have already noticed) that we made an assumption: we don't know what the opponent's decision will be. Humans are remarkably poor at choosing things at random. If you ask someone to recite numbers at random, you can predict, with exceptional accuracy, their next choice, based on just their past two. This is a flaw, and the best RPS players exploit this to their advantage. This depth of competition is not an inherent property of the game; it is an effect of the imperfections in the players. If everyone were perfect, the best strategy would, of course, be random choice, and that's not much of a game is it?
So Pokemon is kinda similar to RPS. The big difference between them is that Pokemon is a much more complex game (luck and all). You'll die long before you analyze every possible combination of decisions that leads from the beginning to the end of a battle. Even the most powerful computers in the world couldn't do it in any realistic amount of time. So now, you can't even prove an optimal strategy, even if you do assume both players are perfect. How could anyone even hope to play Pokemon strategically?!
Obviously, you can, but our problem becomes a lot more difficult. Ever heard of Deep Blue, the first computer to beat a Chess Grand Master? It faced a similar problem: a game with so many possible combinations that plotting out the full game was an impossibility, especially on regulation time limits. Deep Blue had to somehow evaluate the results of decisions without knowing anything about how the game might end. In fact, this is what good chess players really do (whether they realize it or not). The results of every decision get churned through the machinations of one's mind to produce an artificial estimate of whether that decision is any good.
Deep Blue did the same, albeit more precisely. It assigned every games state a value based on certain known factors -- the number of pieces each player had remaining (material), the overall number of moves available to each player (mobility), etc.-- and made a great big chart of each players decisions going back and forth some predetermined number of turns (since, again, you can't look through the whole game). At the end, all the conceivable game states resulting from that number of back-and-forth moves is assigned a numeric value, and an algorithm (the fundamental version of which is known as "Minimax", if you're interested in researching this) sorts through the chart to determine the theoretically optimal move -- the move that should put Deep Blue in the best position to proceed to win.
If you, say, wanted to write an AI which could play Pokemon competitively, you might go about it in a similar fashion to coding a chess computer (BIG grain of salt her, especially since Pokemon is a simultaneous game whereas Chess is not, but a better explanation would cost you the rest of your evening in all likelihood): make a chart of each players decisions over some number of turns, and evaluate each resulting game state based on how much of an advantage either player has with regards to winning. When you have a random factor, just average the values for the possible results based on the probability of each one (this prevents our AI from, say, choosing the win condition that has a .1% chance to win, 99.9% chance to lose, over the option to play the game out further for a better chance).
Just like Deep Blue works like a more mechanical version of a good Chess player, our Pokemon AI would end up working like a good Pokemon player. Realize it or not, they do the same process: consider each option and the corresponding options of your opponent, consider random effects, and weigh the potential payoffs. And so, finally, we can say that those things are the most important factors to being the best Pokemon player you can be:
That's all. I hope this clarified something for someone, as this took at least an hour to write. Congratulations if you read the whole thing!
NOTE: This post is HUGE. Skip to the section with the list with bold text for the conclusion (tl;dr). Read through for some insight into how those conclusions were reached.
Firstly, there is, in any game state, an absolutely optimal strategy, which will result in the greatest overall record if used consistently. An easy (to follow, not to explain, as seen by the length of this post) example is Rock-Paper-Scissors.
If your opponent picks rock, the best strategy is always paper. If you pick paper, your opponent's best strategy is always to pick scissors. But wait! You don't know what your opponent's move will be. This is called a simultaneous game. You and your opponent make your decisions, with no knowledge of your opponent's decision before the "ply" (a turn in this case) is carried out (throwing your hand in the chosen shape).
So what's the optimal strategy for RPS? Well, suppose we assign the end of any game of RPS a value: 1 if you win, 0 if you tie, -1 if you lose. You would like, over time, to win as much as possible, and lose as much little as possible.
Draw or imagine a chart (I can't do that here, sadly) with each axis representing a player, with R, P, and S written along them. In the middle, write the values we chose earlier corresponding to each combination of moves. If you just chose Scissors over and over again, your opponent could easily determine that their best move is to choose Rock over and over again. Thus you get a net average score of -1. Obviously we can do better. The optimal strategy, as can be determined with a little thought, is to choose randomly between your three options with equal probability each (33.3%). This way, regardless of what move your opponent makes, you get a 1/3 chance to win, a 1/3 chance to tie, and a 1/3 chance to lose. Then we can multiple these probabilities with their respective "payoffs" (the values we chose last paragraph), add them together, and we get 0.
That's the best strategy: choose randomly, and over a very large period of time (look up the Law of Large Numbers in Statistics if you're not sure of this) you'll win precisely as often as you lose. You cannot do better than this. There is no way to guarantee a better average payoff than 0.
But there are actually Rock-Paper-Scissors World Championships (just like Pokemon)! How could you conceivably have a competitive games taken seriously if all they're doing is choosing randomly and hoping that luck is on their side? Well if you look back, you may notice (or you may have already noticed) that we made an assumption: we don't know what the opponent's decision will be. Humans are remarkably poor at choosing things at random. If you ask someone to recite numbers at random, you can predict, with exceptional accuracy, their next choice, based on just their past two. This is a flaw, and the best RPS players exploit this to their advantage. This depth of competition is not an inherent property of the game; it is an effect of the imperfections in the players. If everyone were perfect, the best strategy would, of course, be random choice, and that's not much of a game is it?
So Pokemon is kinda similar to RPS. The big difference between them is that Pokemon is a much more complex game (luck and all). You'll die long before you analyze every possible combination of decisions that leads from the beginning to the end of a battle. Even the most powerful computers in the world couldn't do it in any realistic amount of time. So now, you can't even prove an optimal strategy, even if you do assume both players are perfect. How could anyone even hope to play Pokemon strategically?!
Obviously, you can, but our problem becomes a lot more difficult. Ever heard of Deep Blue, the first computer to beat a Chess Grand Master? It faced a similar problem: a game with so many possible combinations that plotting out the full game was an impossibility, especially on regulation time limits. Deep Blue had to somehow evaluate the results of decisions without knowing anything about how the game might end. In fact, this is what good chess players really do (whether they realize it or not). The results of every decision get churned through the machinations of one's mind to produce an artificial estimate of whether that decision is any good.
Deep Blue did the same, albeit more precisely. It assigned every games state a value based on certain known factors -- the number of pieces each player had remaining (material), the overall number of moves available to each player (mobility), etc.-- and made a great big chart of each players decisions going back and forth some predetermined number of turns (since, again, you can't look through the whole game). At the end, all the conceivable game states resulting from that number of back-and-forth moves is assigned a numeric value, and an algorithm (the fundamental version of which is known as "Minimax", if you're interested in researching this) sorts through the chart to determine the theoretically optimal move -- the move that should put Deep Blue in the best position to proceed to win.
If you, say, wanted to write an AI which could play Pokemon competitively, you might go about it in a similar fashion to coding a chess computer (BIG grain of salt her, especially since Pokemon is a simultaneous game whereas Chess is not, but a better explanation would cost you the rest of your evening in all likelihood): make a chart of each players decisions over some number of turns, and evaluate each resulting game state based on how much of an advantage either player has with regards to winning. When you have a random factor, just average the values for the possible results based on the probability of each one (this prevents our AI from, say, choosing the win condition that has a .1% chance to win, 99.9% chance to lose, over the option to play the game out further for a better chance).
Just like Deep Blue works like a more mechanical version of a good Chess player, our Pokemon AI would end up working like a good Pokemon player. Realize it or not, they do the same process: consider each option and the corresponding options of your opponent, consider random effects, and weigh the potential payoffs. And so, finally, we can say that those things are the most important factors to being the best Pokemon player you can be:
- Consider each option and the corresponding options of your opponent. One of the biggest flaws new players may make is failing to consider all of their options. Often times, even experienced players will overlook something: some move, switch, or some complicated play over several turns. This is also where metagaming (team building) comes into play; the best team is the one that supplies you with the most viable options at any point. Stronger, more versatile Pokemon can allow for more and better plays. Additionally, this is why double battles (VGC) are so different from single battles (Smogon OU): a tremendous increase in the number of options, putting pressure on players to train themselves to process the information even more quickly.
- Consider random effects. This is the "luck management" that has been debated a bit in this thread. Yes, this is a demonstrably real concept. It is not something that relies on talent or intuition, as for simpler cases you can always find a provably optimal solution, but it's still real. Like the rest of the decision making process, difficulty lies in evaluating all the possibilities quickly and precisely -- a skill that is developed over time. You should never forget to consider secondary effects, critical hits, paralyses, etc. when deciding which move is best. Going back to the weighted averages concept, this also influences team building in that consistent options are generally superior unless the risky option has a sufficiently large reward to make up for getting watered down by probability.
- Weigh the potential payoffs. This is perhaps the most absolutely critical, most overlooked aspects of competitive Pokemon battling, probably because of the work involved in even realizing it's there (see the rest of this post). Once again, most of the time, you will not have the capacity to look all the way towards the end of the game. Even the best players rarely look more than 2-3 turns ahead. You have to develop the skill of quickly assessing which of your options are viable, seeing the results of these solutions given all of your opponent's expected viable options, looking at the result, and seeing if it's favorable to you. Everyone has a different way of doing this, but to me the most universal concept is "momentum". It's a term that gets thrown around a lot, but generally "momentum" refers to either player's capacity (from a given position) to at some point take knockouts. Setting up a setup-sweeper provides a player with a colossal amount of momentum, but is typically costly to achieve. Every time you take a knockout, you actually lose momentum (!) because your opponent gets a free switch. Next time you choose a move, don't just think if that move is going to immediately get you a knockout. Think: am I going to be in a position to take knockouts if I do this? What about my opponent? How can I prevent them from gaining momentum?
That's all. I hope this clarified something for someone, as this took at least an hour to write. Congratulations if you read the whole thing!