1. Welcome to Smogon Forums! Please take a minute to read the rules.

# Switch rate

Discussion in 'Stark Mountain' started by Amazing Ampharos, Sep 25, 2007.

1. ### Amazing Ampharos

Joined:
Jul 11, 2005
Messages:
1,559
One thing I've noticed about a lot of battlers on shoddy battle just from talking to them is that they assume the best. That is, when planning their teams defensively, they ignore the prospect of added effects (such as Ice Beam freezing) or critical hits. In DP especially, many teams rely on "counters" that are merely 3HKO'd by their adversary. This means that one bit of bad luck sends their counter crashing down, and then they complain on the main chat about hax. This is where the concept of the switch rate comes in. Any given Pokemon can switch in n times to a 3HKO move before the odds of that added effect or crit exceed the odds of those things not happening (thinking in terms of team building so using cumulative odds). The switch rate is the number switch which switches the overall advantage of the switch in from the defender to the attacker. If the defender has to switch into that attack fewer times than the switch rate, the defender has the advantage. If the defender has to switch in equal or greater times than the switch rate, the attacker has the advantage.

This is modeled by move, not by individual switch in. I'm ignoring Sand Veil, Serene Grace, Compoundeyes, Super Luck, and other factors that would change the properties of the moves, though you could certainly do the calculations for those cases. The main presumptions are:

1. The Pokemon being hit is hit while switching in.
2. The Pokemon being hit will lose if either the crit or the added effect happen, even if hit from full health.

General case: .5 >= (move accuracy)(1- ((1-crit rate)(1-added effect rate))^switch rate)

Moves:

100% accurate no effects (Earthquake, Hidden Power, Aura Sphere, etc.)

Example:

Leftovers Garchomp Earthquake vs Milotic

Switch rate:

11

In general, relying on these 3HKOs for defenses are very safe. Moves like Draco Meteor don't even have enough PP to reach the limit, and only a really hyper-defensive strategy is likely to exceed this.

100% accurate 10% effect (Ice Beam, Flamethrower, Thunderbolt, elemental punches, Psychic, etc.)

Example:

Specs Porygon-Z Ice Beam vs Dusknoir

Switch Rate:

5

This is probably a lower number than you expected given the high result from Earthquake. Yes, that freeze rate on Ice Beam has a huge effect after all. I should note that at 4 the odds are nearly 50/50 very slightly in the defender's favor (about .6%). These switches are usually assumed to be safe but really should only be relied on a few times. The odd example used here was the first I could find despite how common these moves are.

100% accurate 20% effect (Crunch, Shadow Ball, etc.)

Example:

Choice Band Tyranitar Crunch against Swampert/Hippowdon

Switch rate:

3

You don't want to be switching into these moves more often than you have to. Not even a great tank like Hippowdon can really be counted to consistently come in on Tyranitar because of the very low switch rate.

85% accurate no added effect (Megahorn)

Example:

Heracross vs tanks neutral to bug

Switch rate:

14

I only included this silly case because of how prominent Heracross is. You can feel pretty safe even if your best counter is 3HKO'd by Megahorn, assuming you have a way to heal back the damage.

85% accurate 10% added effect (Fire Blast, Meteor Mash)

Examples:

Specs Moltres/Heatran Fire Blast vs Blissey
Metagross Meteor Mash vs Hippowdon (among others)

Switch rate:

6

I feel so happy to be able to include a case involving Blissey, and yes the burn does ruin her in that switch-in (the little bit of extra damage takes away her safety margin from the repeated Fire Blasts entirely). Metagross also has a ton of these with Meteor Mash; I just used a simple to calculate example. These aren't all that safe, but they are at least more reliable than the Ice Beam style move cases due to the reduced accuracy.

80% accurate high crit rate moves (Stone Edge, Cross Chop)

Example:

Leftovers Tyranitar Stone Edge vs Skarmory

Switch rate:

8

If only high crit rate moves were 25% like in RSE, this would be 4. Anyway, despite the fact that people tend to fear critical hits a lot more than usual from these moves, they really aren't a terrible much more dangerous than standard moves with perfect accuracy (note that this suggests that you switch into every single PP of Stone Edge). The added effect moves are a bigger threat overall.

90% accurate 70% effect (Charge Beam)

Common cases:

Switch rate:

1

There's a lesson here. Don't switch Gyarados into electric attacks.

I hope people were able to get the mathematical point I was trying to make. When planning your team to switch into threats, don't assume they won't get crits or freezes or whatever. Plan from the start that you're on a probabilistic timer. Can you eliminate the threat without having to switch in as many times as the switch rate? If your answer is no, you don't have the threat covered. If yes, you should be able to handle the threat over half the time.
2. ### AeroblacktylThe pizza doesn't scream in the oven! LOL!

Joined:
Feb 19, 2005
Messages:
8,336
So by your logic, if someone challs me with freeze clause off, I'll automatically lose cause there's no way I can prevent that....
3. ### AgostonF

Joined:
Aug 22, 2007
Messages:
368
This is sort of.. over the top in my opinion. You can't always expect crits or freezes.
4. ### glassbutterfly

Joined:
Jun 26, 2007
Messages:
733
i think what he's saying is that there is no absolute counter when luck is factored in and how counter's switch-ins are limited.
5. ### AeroblacktylThe pizza doesn't scream in the oven! LOL!

Joined:
Feb 19, 2005
Messages:
8,336
Well then it's impossible to ever counter anything by that logic. If the same theory is applied, Blissey wil NEVER wall Regice cause it risks getting frozen?
6. ### n3ther

Joined:
Jun 11, 2007
Messages:
247
Your counter-argument isn't very good since Blissey has natural cure and you probably have another pokemon that resists ice on your team as well.

Also Leona you shouldn't put words into other peoples mouths. Ampharos never asserted that there is no absolute counter. He is telling people what they should already know: sometimes a counter pokemon will lose due to chance. Maybe that lucky freeze or that lucky crit.
7. ### Tracker

Joined:
Sep 21, 2007
Messages:
520
Great theory, are you an economist?
8. ### ultimifier

Joined:
May 13, 2006
Messages:
1,095
being a probability freak, and thinking of this when making my teams, i think this is actually useful, for pokemon other than blissey, or switch-ins who would OHKO back 100% of the time (like heracross or something)

good job though
9. ### husk

Joined:
Jan 9, 2005
Messages:
2,083
For future reference: Draco Meteor has 90% accuracy, so it has a "switch rate" of 12 which is even more in favor of the defending player.
10. ### TSPhoenix

Joined:
Jun 21, 2007
Messages:
384
On classic case of this is stuff missing with Ice Fang, you'd think it'd be a pretty reliable move with 95% acc and two added effects, but I've had situations where fangs missed two or three times in a row.
11. ### Thunderpup

Joined:
Aug 16, 2007
Messages:
1,530
What this means is this - You don't want to waste a turn switching when the number is lower.

Charge beam is 1 because chances are it just makes them stronger.
Earthquake is high because you can safely switch more often than charge beam. After a special attack increase, your wall might be fucked up now.
12. ### Re\$Ullus

Joined:
Jul 29, 2007
Messages:
37
This is confusing and unnecessary...
Trying to apply logic to randomness is ummm a paradox.
13. ### AeroblacktylThe pizza doesn't scream in the oven! LOL!

Joined:
Feb 19, 2005
Messages:
8,336
That's besides the point whether or not Blissey has natural cure. I could have a team of 6 Heatrans and by tihs logic I am at risk of not being able to beat a mono Regice team because it could freeze every signle time should I forget to click Freeze clause.
14. ### Amazing Ampharos

Joined:
Jul 11, 2005
Messages:
1,559
Oh, that slip up on Draco Meteor's accuracy was pretty sloppy. The general idea is the same though.

Yeah, the point isn't to expect to always get unlucky. The point is to figure out how many times you can switch in on stuff before it would be lucky on your part to not have been critted or frozen/burned/whatever. Blissey can switch in on Regice's Ice Beam and reasonably expect to not be frozen, but if you try switching in on Regice 5 times, odds are that one of those times you will get frozen (though Natural Cure and the fact that Regice is nowhere close to 3HKOing Blissey make it pretty inconsequential). Cases like that in which the crit or the freeze don't really matter are outside of the scope of this; the assumption was that the crit or the freeze meant doom for the target, as is the case in CBtar Crunching on Hippowdon.

About Regice vs Heatran, this argument would pretty much say that it's dumb for the Heatrans to just keep switching instead of attacking and actually killing the Regice. I'm sure you agree with that.
15. ### AeroblacktylThe pizza doesn't scream in the oven! LOL!

Joined:
Feb 19, 2005
Messages:
8,336
Then what was exactly the point of this, it's pretty much common sense from as far as Tobybro in RBY trying to Rest against a Starmie that the more chance you give someone to attack, the more likely they're going to get lucky.
16. ### Dragontamer

Joined:
Jun 9, 2007
Messages:
2,068
Wow. I was thinking the exact damn thing. Except I was going to post it later when I figured out the details. Looks like you figured the details before I did. Good Job.

EDIT: Although, you did it in a slightly different manner. You're analysis is entirely involved in defenses. My (to-be) analysis revolved around offensive ones. (IE: Fire Blast is effectively better than Flamethrower on the average... 6 shots)

Either way, I highly recommend to everyone here to think about what was said seriously. Thinking in long-term averages will make you win in day-to-day battles as well as up your win overall.

Also, food for thought, all of these are just the Geometric Distribution. And formulas predict the 95% confidence rate that say... you won't get frozen, or that you won't get hit within x number of hits. I forgot the precise formula, but it is out there.

AA, what do you think about doing an analysis based on say, 80% confidence rate or so?
17. ### TSPhoenix

Joined:
Jun 21, 2007
Messages:
384
Yes and no. I mean if you can switch into Fire Blast and take 20% or stay in take 80 then kill them sometimes you are better off staying in if the burn chance on the 20% damage taker would be more detrimental than losing 80% on whatever you have in.

Its a minimix and maximin thing, where you take a greater minimum loss to prevent a higher maximum loss. Criticals not accounted for.
18. ### david stoneFast-moving, smart, sexy and alarming.

Joined:
Aug 3, 2005
Messages:
5,152
First off, given enough switches and a known probability of an effect, you may not be able to tell what happens on any given turn, but you can be fairly confident in what will happen over several turns. That's what this is. It doesn't care about any individual turn (and really, one turn is mostly irrelevant, it's the sum of the turns that's important). AA stated exactly what the purpose is.

If you estimate that, on average, you have to switch in something on a Pokemon 5 times before you can finish it off, and the calculated switch rate is 4, then you will, in fact, usually lose, assuming you fit the criteria of "win if nothing happens, lose if something happens". He's saying that you can't just say "OK, if everything goes my way, I'm safe", when the odds are against everything going your way. No, MoP, this doesn't mean you have to assume everything will go against you, and I'm fairly confident you realize that your all-or-nothing approach to arguing about luck here is erroneous. It's actually possible that some side-effects will occur, and others will not, and in fact, this is the most likely outcome. If you have an 80% chance to lose with the strategy you are using, then I think you'd agree that it's not exactly the best strategy.

Joined:
Aug 26, 2007
Messages:
348
Er in one place up there you said Draco Meteor is 100% percent accurate its not its 90% accurate
20. ### Dragontamer

Joined:
Jun 9, 2007
Messages:
2,068
This is also really applicable to defenses and walls, as AA stated above, most of the time you assume you are safe with a 3-hit KO. (switch in, roost/recover, come back later)

However, that is an assumption, and after a few switch ins, the chances of you failing increase dramatically. This offers an interesting perspective on the issue of walls, as a Wall is "6% chance he wins (critical), 94% chance nothing happens". In the case of ice beam, it is more like 84% chance nothing happens, 16% chance I lose (critical or freeze), etc. etc.

EDIT: Damn, I need sleep. I'm just parroting what was already said...
21. ### Pneuma

Joined:
Jan 9, 2007
Messages:
1,370
Do keep in mind PP of attacks as part of the switch rate.

Draco Meteor against Dusknoir with maxed SpDef could never occur more than 4 times a match (unless the salamence is @Leppa lol).
22. ### Kumar

Joined:
Dec 19, 2004
Messages:
3,107
Luck IS an integral part of pokemon, but this is taking it too far imo.
23. ### Re\$Ullus

Joined:
Jul 29, 2007
Messages:
37
You make Pokmon sound like taxes or something. I thought it was common sense that randomness exists in this game. Yea u might get burned or it might miss *gasp*, but you shouldn't change how you play based on the possibility of a haxxx..... that's all I'm trying to say.
24. ### Dragontamer

Joined:
Jun 9, 2007
Messages:
2,068
However, the "probability of a hax" is on the order of ~60% chance of happening. If you...

A) Actually have a battle that long

or

B) Play enough battles where you get in a "switch in" situation (say, a tournament with 5 or so rounds)

Then some of these switch rates (like Swampert vs T-Tar) not only have a chance of happening, but virtually become a statistical certainty.
25. ### Sovereign

Joined:
Nov 4, 2005
Messages:
774
Honestly, I don't think a single formula exists for switching rates. Imo, it depends more on whether or not the opponent is a noob. Some obvious switch predictions can be backfired if the opponent decides to be stupid (thus accidentally getting the edge).