Identifying your "edge" in Pokemon

[Firestorm]

Heh, I really do like that RMT point. The reason I haven't posted my team is because it works almost fully on surprise and through battling, I already know my glaring weaknesses (I'm too damn lazy to cover them, but I know I have them). I might post it just to see what people say once I quit Pokemon and finally move onto my crazy backlog of console games =( Still on the first character quest of Tales of Legendia and chapter 1 of Super Paper Mario... since April 21st, 2007.

Also, I don't know why Nintendo even bothered with IVs. There's no strategy involved like EVs and Natures. Nothing you have to think about. 31 is automatically the best (except of course for Hidden Powers, but those could easily have been done a different way). I mean, other than HP and Wish passing (where less speed is optimal) there's no reason to want a low IV.

 

[obi]

I'd just like to say that even if you aren't in a tournament, if you have a recognizable name, people DO steal your teams and / or build counter teams to anything you post.

 

[Dragontamer]

It is quite simple. To reward those who hunt for better IV values.

If all pokemon were the same, then that would be kinda boring IMO.

 

[Firestorm]

It is quite simple. To reward those who hunt for better IV values.

If all pokemon were the same, then that would be kinda boring IMO.

But in the end, they're all basically the same. Everyone tries for as close to 31 as possible in the stats that matter. Most Pokemon are used the exact same way as the other Pokemon of the species (you will not see a special sweeper Gyarados). I understand it gives them individuality or whatever you want to call it, but all it does is make you feel guilty about shoving 100 Pokemon in the box because they don't have the right IVs =(

 

[Dragontamer]

Yes,we all try for perfect 31 IVs. However, it is practically impossible to get more than a few perfect stats. So even then, pokemon breeders have to choose which stats get perfect 31 IVs.

While not physically constrained by the game like EVs (set at 510 max EVs period), IVs are practically constrained. No one has the time to hope for a perfect 31/31/31/31/31/31 pokemon

 

[chaos]

I'm going to speak on terms of Advance since I'm more skilled at it.

Synre gave his input and I'll give mine: I stay clear of RMTs. When I was first starting Advance, I never posted RMTs. When I first start D/P, I'll never post an RMT. When I want to beat someone, I don't scout for their team in an RMT topic or even read RMTs to see what kind of teams people use. At least in the context of NetBattle, it's completely redundant. If I want to see what people use, I'll... "battle." If I want to see if something works, I'll... "battle." Once you learn "Pokemon Theory" RMTs are useless, it's easy enough to identify your own weaknesses... which is why I advocate lots of reading material and ESPECIALLY practice as opposed to a post button.

This article didn't seem too helpful for me because anyone worth scouting wont post an RMT to scout and it doesn't seem to be about anything else except that.

 

[Firestorm]

Yes,we all try for perfect 31 IVs. However, it is practically impossible to get more than a few perfect stats. So even then, pokemon breeders have to choose which stats get perfect 31 IVs.

While not physically constrained by the game like EVs (set at 510 max EVs period), IVs are practically constrained. No one has the time to hope for a perfect 31/31/31/31/31/31 pokemon

I have a friend with 31 in about 4 stats and above 25 in the other two. I'm just saying that there's no thought put into "I need this amount of def to survive this, but this amount in atk to kill this" like there are in EVs. You always want the def/atk as high as possible. Of course, there's the fact that you'll always have a minimum def or whatever IV you want to get, but that's about it.

Maybe I'm just bitter at the 75 hours or so I spent breeding for what really are above average IVs. Nothing special. The amount of time needed to make a Pokemon with great IVs is ludicrous.

 

[McGrrr]

This article didn't seem too helpful for me because anyone worth scouting wont post an RMT to scout and it doesn't seem to be about anything else except that.

I agree, but I found myself in a dead end with regards to battle edge. I cheated a little by referencing my other thread too, and all I can really say is "battle lots and you will improve!!". However, I have updated the original post now, with something more substantial and hopefully a bit more useful for newer battlers.

 

[ekans dilos]

McGraw have you studied economics? I only took one year, but a lot of your terminology and way of thinking reminds me of it.

Anyway, you're right about competitor being more "even" than wi-fi, but it also takes some situations that would be extremely rare if at all possible in wi-fi (legendaries with 31 in the right stats and nature, hidden power, etc.) And yes some people will spend a year breeding for 31 in 3 or 4 stats, but really, how much of a difference does it make, as long as you can get 31 in speed where needed and 20-25 or so in other important stats; which isn't hard to do at all? (there's a thread about this atm)

 

[McGrrr]

McGraw have you studied economics? I only took one year, but a lot of your terminology and way of thinking reminds me of it.

Anyway, you're right about competitor being more "even" than wi-fi, but it also takes some situations that would be extremely rare if at all possible in wi-fi (legendaries with 31 in the right stats and nature, hidden power, etc.) And yes some people will spend a year breeding for 31 in 3 or 4 stats, but really, how much of a difference does it make, as long as you can get 31 in speed where needed and 20-25 or so in other important stats; which isn't hard to do at all? (there's a thread about this atm)

I am a second year economics undergraduate =] In the extreme example of an attack that averages a OHKO; one less point in the respective defense will often increase the OHKO chance by 6.25%. A lower HP IV is less significant.

 

[obi]

I'll be honest, when I post an RMT it's more of a "Man this team is cool" than an actual "Hey I need help!" I still take advice, of course, but I don't post new teams.

 

[ekans dilos]

Quote:
Originally Posted by Fat McGraw
I am a second year economics undergraduate =] In the extreme example of an attack that averages a OHKO; one less point in the respective defense will often increase the OHKO chance by 6.25%. A lower HP IV is less significant.

Heheh, I knew it. But of course, you're right that IVs come into play in extreme cases. That's basically what i said here:

Quote:
Originally Posted by Fat ekans dilos
The cases where a perfect IV is most important is when it's absolutely needed to perform. E.G., base 100 pokemon, who have to compete in speed with all other base 100s, of which there are MANY (Salamence, Staraptor, Raichu, Ninetales, Celebi/Jirachi/Shaymin/Mew/Manaphy, etc.) need to be at absolute max speed in most cases. Then you have cases in which it's vital to have max stats in order to 1HKO/2HKO (any special attacker with focus punch vs. blissey, for instance), or to survive more than 1 or 2 hits from a common counter. And there's the good 'ol 404 HP Sub (Seismic Toss seems to be going out of style though). Usually though, most of these can be fixed with EVs. The only one that can't is speed.

All these special cases aside, personally I'd settle for decent scores in the needed stats (25+) and average or better (15+) in the ones that aren't vital but still useful (hp/def/sp.def on a sweeper). Obviously only the used attacking stat needs to be good, the other doesn't matter (0+). A primarily defensive pokemon doesn't need to worry too much about it's offense, so if it has good defenses and say, 20 Atk/SpA, that should be fine.

To sum it up, in general the only stat that will actually make a life or death difference is speed.

So IMO to say that competitor is a more even playing field because it eliminates difference in IVs is correct, but a bit of an exaggeration, because these are only vital in a few cases and it generally is neither too hard nor too time consuming to breed a pokemon that can perform effectively by having the bear minimums. It may deal 1% less or take 1% more but in the end, as long as it goes first when it has to, and KOs/2HKOs/survives what it needs to, it won't matter in the big picture.

 

[Footnote]

Computers are in fact however, more unpredictable than humans. As a simple proof, try generating 100 random numbers without the use of die or coins. A standard computer algorithms will be able to do that and pass statistical tests of randomness. (Chi^2 test, etc. etc. Names of algorithms are LCG generators and that twister thingy... forgot its name) while humans tend to fail at these kinds of tests.

Just because a computer can generate random numbers more accurately than a human (which it can), doesn't mean it has an edge in unpredictability. In a given round of a Single Battle, you have, at most, 9 choices (4 moves and 5 Pokemon to switch to). Exceptions include Baton Pass and U-Turn, where you pick a move and a new Pokemon. It doesn't take much to see that most of your choices in a given turn are bad ones. Hence, a truly unweighted random algorithm for move selection is a bad idea.

Now, I know you weren't arguing that it was a good idea, but here's where I'm going: most of the time, you have 2 (maybe 3 or 4) valid options. For instance, if your opponent switches, Choice A is better. If your opponent stays in, Choice B is better. Maybe you need to decide between Pursuit and Sucker Punch or some such. If you want to introduce a factor of randomness so that your opponents can't predict you, it doesn't take long to flip a coin. At worst, you have to roll a die.

Although I haven't yet, I'm seriously considering including a coin flip when decisions like this come up. I'd be interested in hearing opinions on this strategy.

 

[nomargu46]

I'll be honest, when I post an RMT it's more of a "Man this team is cool" than an actual "Hey I need help!" I still take advice, of course, but I don't post new teams.

yeah, thats really the way (i assume) for pretty much all competitive battlers who actually know what they're doing. if i were to post a team it'd be more of a "hey check out my sweet gimmick" than a "i fully expect to be very successful with this team" post. still, i'd just like to point out that the tone of voice (tone of typing?) here seems to be a little bit on the bashing RMTs side, and if somebody's getting that impression I'm sure that isnt what anybody is meaning. or at least i dont think you guys are hatin' on the RMTs...

 

[Dragontamer]

Just because a computer can generate random numbers more accurately than a human (which it can), doesn't mean it has an edge in unpredictability. In a given round of a Single Battle, you have, at most, 9 choices (4 moves and 5 Pokemon to switch to). Exceptions include Baton Pass and U-Turn, where you pick a move and a new Pokemon. It doesn't take much to see that most of your choices in a given turn are bad ones. Hence, a truly unweighted random algorithm for move selection is a bad idea.

On the other hand, we were discussing how to find the best probabilities for the best randomness in the other thread. Which is more or less, how an AI would work.

One trivial game which has been solved for example is Rock-Paper-Scissors. It has been mathematically solved to show that 33.33~% of the time you should use Rock, 33.33% of the time paper, etc. etc. Now this is a simple and easy case. Nonetheless, with weighted rock-paper-scissors, it can also be solved.

And this applies to Pokemon because it is more or less a generalized rock-paper-scissors game. Instead of just 3 choices, you have upto 9 (4 attacks, 5 pokemon to switch to).

Now, I know you weren't arguing that it was a good idea, but here's where I'm going: most of the time, you have 2 (maybe 3 or 4) valid options. For instance, if your opponent switches, Choice A is better. If your opponent stays in, Choice B is better. Maybe you need to decide between Pursuit and Sucker Punch or some such. If you want to introduce a factor of randomness so that your opponents can't predict you, it doesn't take long to flip a coin. At worst, you have to roll a die.

Now Nash proved that in any game, there is a Nash Equilibrium. It can be a mixed strategy equilibrium however. And with that said, the weights can be arbitrary.

One method to doing the AI is to somehow solve for this Nash Equilibrium, the best probabilities of randomness assuming your opponent is also perfect. Of course, by the definition of Nash Equilibrium, it would be your optimal strategy even if the opponent changes his strategy.

Although I haven't yet, I'm seriously considering including a coin flip when decisions like this come up. I'd be interested in hearing opinions on this strategy.

Again, I'd say you shouldn't do that. The "best" probability is arbitrary. For all we know, the best case is choosing sucker punch 80% of the time, and choosing pursuit 20% of the time.

This may seem a bit like magic to you, but...

1. Nash did prove that one of these points exist. That there is a strategy that you gain nothing from deviating from it.
2. It -might- be possible to calculate these points.

So with that said, if a computer can calculate these points, and then make a perfectly random choice every time, with the perfect weights assigned to each decision, then we'd have an AI that would be very difficult to predict and very difficult to beat.

Course, we didn't reach that point in the discussion yet. So... yeah. I hope it is possible :-p

 

[Mr.E]

Actually, don't experienced gamblers and card counters have the edge over the house in Blackjack?

Another way to protect your edge is to simply not post teams at all. The most experienced players do not need the input of others; they can identify weaknesses themselves (or recognize them when they keep losing!). Granted, this is anticompetitive behavior, but it is also rational. Further, new players should be encouraged to find their own way beyond a point, otherwise there is no stimulation for innovation.

I have to disagree. You could not post any teams ever, but if somebody really wants to find out what you're using your team can be scouted fairly easily. Moves/pokemon/items are easy to notice and approximate IVs/EVs can be had from how hard your pokemon hit, how much damage they take, and what they do (or don't) outspeed. Unless you use a completely new team week in and week out people will figure out what you're using quick enough. It's a pain making a new team constantly, even if you only make 4-5 and switch between them, and if you're constantly switching teams you won't become very experienced or accustomed to using all those teams.

You don't even have to post your complete team and disclose every last detail about it, but talking to others is obviously helpful because somebody else could come up with a good idea that you didn't think up yourself. Maybe you didn't notice your Weavile weakness (*rimshot*), somebody pointed out a useful alternate item, or you could replace a pokemon for another one and get an even bigger edge that way. At the end of the day, once you're experienced playing your team and everybody knows what you're using anyway, you want the best team possible.

Just as an example, plus to toot my own horn, I can look at my GSC team. I never posted it for rating or anything like that, but I've used it for so long that any of the IRC regulars who have all played pokemon for years along with us [mcslaw] can tell you I use Encore 'Zam, my Zapdos has HP Water, and I use Light Screen in my Blissey's fourth moveslot, among other things. Do I give a damn if everybody knows this? Nope. My assloads of experience with this one team allows me to play better with it than I would with a new team that nobody's seen but I have no experience playing. By the time I become acquainted with the new team as well as my old team (YEARS of experience), everybody would know that team too and the cycle would just start again, so why bother making a new team at all?

The only time a RMT can really hurt you is if you're fighting a person for the first time and they scouted your team in advance just to make a counter-team. Even then, by the virtue of their team being a bunch of simple counters, your team is probably balanced better and you can still win with a little finesse. If you suspect it, you can always be a devil yourself and post a fake team that is liable to induce picking counter pokemon that are actually weak to your real team. :[ But I'm just being picky now. Or, if you have versatile pokemon like Dusknoir or Gengar that you can change movesets on a whim, you can change a move or item around and surprise the hell out of your opponent since they can't expect anything different than what they scouted.

Fact of the matter is, if I got those above average IVs, you could have gotten them too if you put forth a similar amount of effort.

This isn't necessarily true, but it's already been discussed enough by the rest of you.

Clearly those with more leisure time, and those who value their time less, have an advantage. Competitor removes this disparity.

Sounds like you're implying pokemon is a waste of time and those who play it do not value their time as much as others. :(

 

[Carl]

Well MrE, regardless of how easy it is to "scout" someone's team in a battle, a RMT is still more of a complete and accurate copy of what the battler in question is using without hardly any effort on the scouter's part. I'd much rather see everything in a nice compact format like a RMT because it's easier and people are much less adverse to doing something easy. This is why I can't remember the last time I actually posted a team looking for assistance. If you want to battle me 5 times to see my team, fine go for it, but I won't hand it to you on a silver platter.

I will agree with you however on the using a consistant team part. Using one team helps you almost memorize certain situations so you instantly know what to do when battling. I've used the same Advance team for the past 6 months or more and when I face another standard team I usually don't think at all because I know that when I see [pokemon] I will switch to [pokemon] or use [move] since this is by default my best option based on all the times I have encountered the situation before. Experience is key in just about anything if you want to see improvement.

As for the can of worms debate about IVs...

Breeding is essentially a level playing field yes. Everyone does have the same shot at getting a max IV pokemon. Where it can become uneven is the time factor. This makes it so only people with the most amount of time will have the strongest pokemon. That doesn't make them the best battlers, but it does give them an edge based only on pure effort. This is where I can indeed side with Dragontamer. Whether any of us wants to admit it or not, effort is a factor in a lot of competitions. In sports, effort is usually one of those intangibles the guys on ESPN love to talk about. Here in pokemon, effort is brought to life in the IV system, making up for the lack of physical activity on the behalf of the gamer. If you don't have the time or the willpower to put in this effort, I certainly won't fault you. It's too much for something that essentially provides nothing practical in return and usually doesn't make a whole lot of difference in battle compared to the other factors. Is breeding a skill? No. It is part of the game but I won't call anyone a great battler based on how much time they invest in breeding. Here's where a lot of people like NetBattle, it removes the weight of the skill-less component and focuses more on the skill of developing a team and using it well. People, myself included, will argue all day that this is the main factor in determining who's good at pokemon.

I guess my point is this. At the end of the day, nobody looks at how long Lebron spends in the weight room but they do look at what kind of numbers he puts up in the game. If he put in less hours in the gym, it could have an effect on how many times the Cavs win in a season, but I doubt it would be enough to lead the Cavs to a season record on par with the Celtics. They might win a few less games which might put them out of the playoffs. It doesn't make Lebron any less skilled at basketball though. The same goes for pokemon. Without those max IVs you might not win a pokemon tournament as often, but I'd be willing to bet if you were skilled enough in other aspects of the game you'd be in the finals still.

 

[Elminster12]

I post a lot on the GameFAQs Pokemon boards(too much, lol) And, I must admit, I am very hesitant to reveal my thought processes, novice though I am. My weakness is that I have somewhat of a McClellan complex, in that I am suspicious of advantages and thus tend not to press them as hard as I ought to, and am most dangerous when my resources are at their weakest. However, I do seem to have an intuitive sense for "pairing" for maximum defensive coverage...but I shall say no more!

Most often, I merely refer them to the appropriate analysis here at smogon(which they usually ignore for the guy suggesting they should make their Tangrowth a full-on sweeper with 252ATK/252SPE, unless one of the few other posters with a brain drops in). If I'm in a really good mood, I do some damage calcs to illustrate my point in raw mathematical terms. But rarely will I suggest anything but a "stock" moveset. I am loathe to do this because of what McGraw mentions...and also because I'd much rather them use the standard until they figure out everyone expects that because it's standard, and maybe then they start doing little variations and really getting into it deeply. I want them to innovate on their own, and not steal my movesets, in part because I don't want people to know my thinking until they've battled me a few times and (hopefully) lost a few times.

But, really, keep these coming McGraw. They're excellent food for thought...

 

[lilyhollow]

Great overall thread, but I must say that what interested me most was this:

battling skill is difficult to quantify and it may be impossible to improve beyond an individual's limitations.

to be honest I would have been happier had you expanded on this particular point, but a job well done anyway McGraw.

 

[Luck]

It is different for mods and real good battlers.I build team post to get rate nobody cares,I may be pretty good but nobody cares about my team.But say McGraw or Jumpman post team that kinda sticks in minds of some people,wow this their is pokemon named after them,If i beat them i am really good player.

battling skill is difficult to quantify and it may be impossible to improve beyond an individual's limitations.

I can't say what McGraw intended,but when i see that statement this pop in my head for example I am sure you seen people post the will never use Blissey, garchomp,etc the their unwillingness to be flexible in team building may cost them losses or I am not big number cruncher so times in battles,I keep pokemon when someone better at math would take them out.I will probably never be good at math so it it will be hard for me to improve in that area.

 

[Footnote]

On the other hand, we were discussing how to find the best probabilities for the best randomness in the other thread. Which is more or less, how an AI would work...

Ah, OK. I see now what you're saying. My apologies for misunderstanding your original point. This begs another question, however. It seems to me that if there is a deterministic algorithm that takes as input what the computer knows about your team (and its team) and returns a probability distribution for the AI to 'roll on', then it is possible for us as humans to determine what probability the machine uses to choose each option and from that derive a better strategy, unless all the nonzero probabilities in the distribution are equal to one another.

This may seem a bit like magic to you, but...

Yes, it does seem like magic. I guess my BA in Mathematics and Computer Science is for naught. ;) But, I wasn't an Economics major, so a lot of this stuff is new to me. It's really interesting, though, so please help me understand what I'm missing here.

 

[Dragontamer]

Ah, OK. I see now what you're saying. My apologies for misunderstanding your original point. This begs another question, however. It seems to me that if there is a deterministic algorithm that takes as input what the computer knows about your team (and its team) and returns a probability distribution for the AI to 'roll on', then it is possible for us as humans to determine what probability the machine uses to choose each option and from that derive a better strategy, unless all the nonzero probabilities in the distribution are equal to one another.

Only if the person works in conjunction with the machine. Chess players train on chess programs that analyze positions for them. Well, I do know of at least one program that does so anyway. (I'm not sure if they are precomputed or not, but regardless, the point remains)

I find it unlikely however that a human is going to use a computer however during the battle. Thats like when chess players use a chess program to calculate their moves for them :-/

Second, the algorithm exists yes. But is practically impossible to utilize. We need machines several million magnitudes more powerful before we can attempt a naive translation of the algorithm.

Yes, it does seem like magic. I guess my BA in Mathematics and Computer Science is for naught. ;) But, I wasn't an Economics major, so a lot of this stuff is new to me. It's really interesting, though, so please help me understand what I'm missing here.

Well then, I'm a computer engineer major to-be. I study this stuff because it is interesting, but it is certainly not my field :-p

Anyway, after watching "A Beautiful Mind", I researched what Nash actually did, and learned that the movie was basically all wrong. Lol. Anyway, I was just pointing out that it has been proven that Nash Equilibrium does exist, therefore, it is possible to search for the correct probabilities. Wikipedia has an excellent article on it. And "Artificial Intelligence, A Modern Approach" has a little tidbit on it as well. And really, this is where all that game theory fits in for me, not economics, but in AI.

I'm still reading it myself, so instead of sharing my ignorance, I'll point out that the algorithm is described on page 635. It is Von Neumann's technique called "maximin", not to be confused with min-max trees. (Von Neumann == maximin. Typical algorithm is minimax).

Course, pokemon is a fun game, so most likely this naive algorithm is intractable. With that said, strong heuristics come into play and volia. We've got a sliver of hope.

----------

Now, the Nash Equilibrium of the prisoner's dilemma is testify/betray. (check out wikipedia for an intro on prisoner's dilemma). This is because no matter what the opponent does, it is always better to testify. Wikipedia also has a good article on this so look it up.

 

[Elminster12]

It is different for mods and real good battlers.I build team post to get rate nobody cares,I may be pretty good but nobody cares about my team.But say McGraw or Jumpman post team that kinda sticks in minds of some people,wow this their is pokemon named after them,If i beat them i am really good player.

battling skill is difficult to quantify and it may be impossible to improve beyond an individual's limitations.

I can't say what McGraw intended,but when i see that statement this pop in my head for example I am sure you seen people post the will never use Blissey, garchomp,etc the their unwillingness to be flexible in team building may cost them losses or I am not big number cruncher so times in battles,I keep pokemon when someone better at math would take them out.I will probably never be good at math so it it will be hard for me to improve in that area.

I think I know what he's talking about. Pokemon is a game where you need to quickly and accurately assess a complex situation, and come to a decision, often quickly. This creates pressure, because, as a History major, I can tell you that things aren't usually spontaneous. Let's say you mispredict an EQ early on, and your Staraptor takes 45% instead of nothing. Ok, but what if you need Staraptor to switch into his Hera later on? Maybe Staraptor is dead, maybe it's doomed to be OHKO'd...but what if you had been right? To some extent, every little detail counts for something, and we all decide differently. Some of us err on the side of caution, some of us take a lot of risks, and some of us are just better at making good decisions than others in these situations. Some of us stay cool, and some of us panic under pressure, and some of that is stuff over which we have no control....

 

[davewaver]

McGraw while what you post is true, are you not just stating the obvious? What you are trying to teach is probably unteachable, each person has different levels of "skill" and "predicition." I am not saying I am good cause I dont play wifi because the actual games are worse than bad. Basically just "use your brain" is what i get from your post.

Also, over predicting is the same as predicting wrong. No offense, but anyone helped by this post probably aren't a real threat anyway (damn I sound stuck up).

 

[Mike B]

I just try to keep it simple by fitting as many gimmicks in as possible, changing pokemon and other moves along with a different mindset every battle... You can't plan around that ;p