Azure Heights Pokémon Laboratory
Main Menu
What's New?
Attacks
Pokédex
Compendium
Weird Science
Forum
Chat
Contributors
Links
Founders

All Attacks
A-D
S

E-K
T-W

L-R


Machines
HMs

TMs

Types
BUG
FGT
GHO
ICE
PSY

DRG
FIR
GRA
NOR
RCK

ELE
FLY
GRO
POI
WAT

Others
Confusion
Critical Hit
Evade/Accuracy
Multi-Turn
One-hit KOs
Paralysis
Sleep
Stat Modifiers


Confusion
Evade/Accuracy
One-hit KO
Attack Type Base PP Pow Acc Avg Effects TM HM
Double Team NOR NDA 15 - 100 - raise Evade 32 -
Minimize NOR NDA 20 - 100 - raise Evade - -
Flash NOR NDA 20 - 69.5 - lower victim Accuracy/ light dark caves - 05
Kinesis PSY NDA 15 - 79.7 - lower victim Accuracy - -
Sand-Attack NOR NDA 15 - 99.6 - lower victim Accuracy - -
Smokescreen NOR NDA 20 - 99.6 - lower victim Accuracy - -

Analysis & Advice
General

The effects of Evade and Accuracy modifiers are cumulative for a maximum of six uses; further uses have no effect.

Whether using an Evade-increaser or an Accuracy-decreaser, by far the greatest benefit comes from the first two uses of the move. This provides one basis for deciding how many times to use the move before switching to offense. Here are some other considerations:

  • How many hits can your Pokémon take from the current enemy? Suppose you decide to use Minimize a few times before mounting an attack. Even if you win the initiative, don't forget that your opponent could get lucky, and hit you on every single turn. (As usual, it's a big help to go first: that way even the first of your enemy's attacks has a chance of missing.) Depending on how many HP you lose, it may be wise to turn to offense ASAP.


  • Can you recover the damage you may suffer while using the Evade/Accuracy modifier? These moves are particularly effective when combined with Recover, Softboiled, or Rest. Armed with one of these moves, your Pokémon can use the Evade/Accuracy modifier until it is in danger of fainting. Then, use the HP-restoring move to get back to full strength (and if desired, use the Evade/Accuracy modifier a few more times, until the maximum of 6 has been reached.) This combination by no means guarantees victory; nevertheless it is so deadly that it behooves all good trainers to have at least one Pokémon on their team that knows either Swift or Haze.


  • There are some attacks which ignore modifications to Evade and Accuracy. For example, Accuracy-decreasing moves will not prevent a Pokémon from successfully using moves on itself (eg, Recover). Evade-increasers won't help a Pokémon to avoid Swift, Haze, Bide, or Transform.
Evade-increasers vs. Accuracy-decreasers

There is no clear advantage between Evade-increasers and Accuracy-decreasers. On the one hand, enemy Pokémon who have had their Accuracy decreased are able to switch away (unless their trainer is out of Pokémon), negating the penalty you have imposed on them. However, this is not such a bad thing: you get a free hit on whichever Pokémon replaces them. Also, suppose that you decrease the Accuracy of an enemy who nevertheless defeats your current Pokémon. The next Pokémon you use will still enjoy the advantage of fighting an inaccurate enemy.

The advantages of Evade-increasers are the converse of those described above. Since you're affecting the statistics of your own Pokémon, it doesn't matter if the enemy switches away or faints. This is especially suitable for the combination tactic described above. However, if your Pokémon is defeated, the next one you bring in will have no advantage.

Another consideration is that your Pokémon cannot miss when increasing its own Evade, but it can miss when trying to lower the Accuracy of an opponent, particularly if it is employing Evade/Accuracy modifiers of its own and/or your Pokémon is using Flash or Kinesis.

There may be occasions when you can exploit the fact that 1 Evade-increaser combined with 1 Accuracy-decreaser has about the same effect as 3 uses of either type of move alone. Perhaps you can't decide how to use that last attack slot for your Horsea or Koffing, for example; there would be some value to keeping Smokescreen, but also using a Double Team TM. However, it's a dangerous gamble to devote an entire attack slot to a hit-probability advantage of less than 10%.

Double Team vs. Minimize

Double Team and Minimize have the same effect. You should therefore choose between them according to preference, or possibly PP (Minimize has 20; Double Team has only 15). However, since it is hard to imagine a Pokémon using more than 15 PP of an Evade-altering move in a single battle, PP is probably not a major consideration.

While you can use both attacks interchangeably, this is not a way around the six-use maximum. Only the first six uses of both attacks combined will have any effect. Since the attacks are functionally identical, the order in which they are used would make no difference.

Flash vs. Kinesis

All Accuracy-decreasing attacks have the same effect in battle, if they hit successfully.

Flash has more PP than Kinesis, but the latter's higher accuracy is most likely more important for your Kadabra than 5 more PP. For other Pokémon (and for Kadabras/Alakazams not in the Yellow version of the game), Flash is not totally useless, but its inaccuracy is definitely something to consider. If you're not specifically planning on lowering the Accuracy of your opponents, you might be better off with Double Team.

While at least one of your Pokémon should learn Flash for the useful out-of-battle effect, for combat purposes, Flash is definitely an attack you'll want to think hard about before deciding it's right for your Pokémon.

Sand-Attack vs. Smokescreen

Sand-Attack and Smokescreen are the most reliable Accuracy-lowering attacks. Smokescreen has more PP than Sand-Attack, but it's a moot point for comparison purposes, since no Pokémon learns both of these moves, and there is no TM for either of them. Both are fine attacks.

Research & Data

The following table (1) summarizes the effects of successive uses of the Evade-increasing moves (Double Team and Minimize):

Uses: Evade:
1 34.1%
2 50.0%
3 55.8%
4 59.6%
5 67.1%
6 71.6%

It is doubtful that these values are precisely what the game is programmed to produce. For example, it's implausible that the 5th use of an Evade-increasing move provides greater benefit than the 4th use. Despite the large number of trials recorded, sampling error may be as much as 2 or 3% in some cases. (back)

The following table (2) summarizes the effects of successive uses of the Accuracy-decreasing moves (Flash, Sand-Attack, and Smokescreen):

Uses: Accuracy:
1 67.4%
2 53.3%
3 42.7%
4 *40.4%
5 *32.9%
6 *28.4%

Again, these values are approximate only. Entries marked with an asterisk are not based on actual Accuracy data, but were inferred from the Evade results. We believe that the same mathematical function produces the hit frequencies for both Evade- and Accuracy-altering moves. Discrepancies between the Evade table and the Accuracy table for uses 1, 2 and 3 are probably just sampling error; however, this is only an hypothesis. This page will be updated as more data is collected.

The probability that an attack will hit is described by the formula
P = M x A x (1-E), where

P is the probability that the move will hit,
M is the accuracy of the move being used,
A is the Accuracy of the attacker, and
E is the Evade of the defender. (3)

In the absence of attacks which modify these statistics, the normal value for Evade is 0%; the normal value for Accuracy is 100%. Relative real Speed, relative base Speed, and relative Level have no bearing on the Evade or Accuracy percentages of the battling Pokémon. (4)

To use the formula, it is easiest to convert the percentages to decimal values. For example, suppose that Pidgeot uses Sand-Attack on Raticate during one turn, and then uses Double Team the following turn. Sand-Attack reduces Raticate's Accuracy to about 67.4%, while Double Team increases Pidgeot's Evade to about 34.1%. When Raticate uses Quick Attack, the chance of success will be Quick Attack's accuracy (0.996) times Raticate's Accuracy (0.674) times the complement of Pidgeot's Evade (1 - 0.341 = 0.659), which is approximately 0.442, or 44.2%.

Other attacks have less chance to hit. Suppose that Raticate had used Take Down, which has an accuracy of 84.4%. The chance of success would be Take Down's accuracy (0.844) times Raticate's Accuracy (0.674) times the complement of Pidgeot's Evade (1 - 0.341 = 0.659), which is about 0.375, or 37.5%

There are several attacks that do not use this formula; the best-known of these is Swift. Anecdotal evidence supports the 99.6% accuracy value for this move reported by Necrosaro. That is, Swift may miss on very rare occasions, but its accuracy is not a function of the attacker's Accuracy or the Defender's Evade. Hence, the probability that Swift will hit is simply P = M. The fact that Swift (and some other attacks) can even hit Pokémon who have used Dig or Fly suggests that the effect of these moves is to change the attacker's Accuracy to 0, or the defender's Evade to 1, or both. In the full formula, either of these changes render the defender "unhittable", but in the restricted formula (ie, P = M) they are of no advantage.

Other moves that appear to ignore Accuracy and Evade include those that a Pokémon uses on itself (eg, Recover), Haze, Bide, and Transform.

Ongoing Research

What other moves use the restricted P = M formula?

What are the Accuracy percentages for the 4th, 5th, and 6th uses of moves that decrease this statistic?

Footnotes

All data were collected using the no$ emulator, running at 6 or 10 times normal speed. Most enemy Pokémon were L2 and L3 Pidgeys from the area just south of Viridian City; for some experiments L2 and L3 Dittos were used. There were three other subjects: a L4 Weedle, a L28 Beedrill, and a L20 Slowpoke. Gameshark codes were employed to teach Pokémon attacks that they would not ordinarily learn.

1. Evade Modifier Trials
Weedle vs. Pidgey
Double Team once, 764 trials. Gust hit 63.6% of the time.
Minimize once, 510. 65.3%
Beedrill vs. Pidgey
Minimize once, 510. 66.3%
Double Team once, 510. 69.0%
Slowpoke vs. Pidgey
Minimize once, 510. 66.5%
Minimize once, plus Double Team once, 765. 48.9%
Double Team x2, 765. 51.1%
Double Team x3, 765. 44.2%
Double Team x4, 765. 40.4%
Double Team x5, 765. 32.9%
Double Team x6, 765. 28.4%

Beedrill was much faster than Weedle in both real and (obviously) base Speed, and also much faster than Pidgey. The fact that the accuracy for Gust INCREASED by 7.4% as the Speed of the target (Beedrill vs. Weedle) increased suggests to me that relative Speed does not provide a noticeable advantage in this regard. Note also that the Double Team/Minimize discrepancies are looking more and more like sampling error. (back)
2. Accuracy Modifier Trials
Beedrill vs. Pidgey
Sand-Attack once, 510 trials. Gust hit 68.2% of the time.

Slowpoke vs. Pidgey
Flash once, 510. 67.5%
Smokescreen once, 510. 66.5%
Sand-Attack x2, 765. 53.3%
Sand-Attack x3, 510. 42.7%

This seems to be following a pattern sufficiently close to that for Double Team that the fourth through sixth applications of Sand-Attack were not tested. (back)
3. Evade and Accuracy are separate values.
The fact that Evade and Accuracy are separate values that are multiplied together to produce P was shown by the following test:
Slowpoke vs. Pidgey
Minimize once, plus Sand-Attack (against the Pidgey) once, 765 trials. Gust hit 44.2% of the time.

This result was predicted perfectly by the formula. (0.996 x 0.659 x 0.674 = 0.442)

Another test was done against Dittos using Egg Bomb, to test how P would be affected when the accuracy of the attack was significantly less than 1.0.
Slowpoke vs. Ditto
Sand-Attack once, 510 trials. 47.6%

This result was quite close to the formula's prediction. (Egg Bomb has an accuracy of 74.6%. 0.746 x 0.667 x 1.0 = 0.498.) (back)
4. Base values for Evade and Accuracy. Speed has no effect.
That Evade is normally 0% and Accuracy normally 100% is suggested by the fact that unless modifiers have been used, the accuracy values for attacks seem correct. Attacks that are 99.6% accurate, for example, miss only very rarely.

Evidence that relative Speed (real or base) and relative level are not factors in Evade/Accuracy:
Weedle vs. Pidgey
Minimize once, 510 trials. Gust hit 65.3% of the time.
Beedrill vs. Pidgey
Minimize once, 510. 66.3%
Slowpoke vs. Pidgey
Minimize once, 510. 66.5%

Another experiment pitted a L100 Electrode against a L2 Ditto that had copied a low-level Pokémon that knew only Egg Bomb. The accuracy of Egg Bomb against the Electrode over 250 trials was 78.4%: somewhat higher than the 74.6% figure reported for that move by Necrosaro.

Another 250 trials were run, under the condition that Electrode used Agility 3 times before Egg Bomb's accuracy was tabulated. The frequency of hits was actually slightly higher (79.2%); presumably this was sampling error. (back)
Sampling Error

Inferential statistics is all about deducing characteristics of a population by examining one or more subsets of the population, called samples. For example, consider the population of all humans. If you knew the height of every human on earth, it would be fairly straightforward to compute the average human height (call this value H). Since it's preposterous to try to learn the height of everyone on earth, you might try to find the average height of just 100 people (call this value S). This would give you a much better sense of H than you had before, but it's not very likely that S will be precisely equal to H. The difference between them is called sampling error.

Getting back to Evade- and Accuracy-modifiers, consider the population of all attempts to hit a Pokémon that has used Double Team once. This is an infinite population, and the proportion of hits vs. misses will be exactly equal to whatever is written in the game code. When we compute the proportion of a sample, even one with 765 trials, it is doubtful that the value will be precisely the same as for the population. Hence, these data are probably not equal to what is written in the game code. (back)

Please ignore this subliminal message. Please ignore this subliminal message. Please ignore this subliminal message. DRINK COKE Please ignore this subliminal message. Please ignore this subliminal message. Please ignore this subliminal message.