Sleep Talk strategy and statistics

obi

formerly david stone
is a Site Content Manager Alumnusis a Programmer Alumnusis a Senior Staff Member Alumnusis a Smogon Discord Contributor Alumnusis a Researcher Alumnusis a Top Contributor Alumnusis a Battle Simulator Moderator Alumnus
When a Pokemon is hit by a Sleep move, it gets 1-4 "fast asleep" messages and prior to waking up. This means that you have a 0% chance to wake up on the first turn of sleep, a 1/4 chance to wake up on the second turn of sleep, a 1/3 chance to wake up on the third turn of sleep, a 1/2 chance to wake up on the fourth turn of sleep, and a 100% chance to wake up on the fifth turn. I will consider the strategy for when to use Sleep Talk in a few cases. This analysis does not consider issues of PP, as that becomes too complex and mostly irrelevant.

The first case is when a Pokemon is put to sleep with the following move set: Rest, Sleep Talk, superior move, inferior move.

Superior move is the move the Pokemon would select if it were awake, essentially. Inferior move is the move that is better than no move, but worse than the superior move.

The first turn it's obviously best to Sleep Talk as you have a 0% chance to wake up, so you only have a 1/3 chance to do nothing (as opposed to a 100% chance if you select any move but Sleep Talk). If you are faster, this is the first after you are put to sleep. If you are slower, this is the turn they are putting you to sleep.

The second turn, you have a 1/4 chance to wake up. Selecting Sleep Talk therefore has a 50% chance to fail to do anything useful (wake up or select Rest). You have a 1/4 chance to select your superior move and a 1/4 chance to select your inferior move. Compare this to selecting the superior move outright, which has a 1/4 chance to use your superior move and a 3/4 chance to do nothing. Sleep Talk here is clearly superior, as it replaces a chunk (25% of the total) of "do nothing" with "use inferior move", which is still an improvement.

The third turn, you have a 1/3 chance to wake up. Using Sleep Talk therefore has a 5/9 (55.6%) chance to do nothing. You have a 2/9 chance to select your best move and a 2/9 chance to select the inferior move. Compare this with selecting the superior move, which has a 1/3 chance to use your best move and a 2/3 chance to do nothing. Here, you will use your superior move an extra 1/9 of the time, but you will also fail to do anything an extra 1/9 of the time. If your superior move is less than twice as good as your inferior move, use Sleep Talk. If your superior move is more than twice as good as your inferior move, use the superior move. For instance, if the two moves are Return and Brick Break against Tyranitar, Brick Break is such a superior choice that it is better to use Brick Break than Sleep Talk.

The fourth turn, there is a 1/2 chance to wake up. Using Sleep Talk therefore has a 2/3 chance to do nothing, a 1/6 chance to select the superior move, and a 1/6 chance to select the inferior move. If you select the superior move outright, you have a 1/2 chance to select the most useful move (compared with Sleep Talk's 1/3 chance to select any useful move at all), making selecting the superior move a clear winner.

You clearly should use a move other than Sleep Talk on the turn you'll be waking up, the fifth turn.



The other somewhat common scenario is a Pokemon with 3 useful moves and Sleep Talk (so no Rest). This helps out Sleep Talk a bit. Let's assume the move set is Sleep Talk, superior move, average move, inferior move.

The first turn you use Sleep Talk for the same reason as in the last example.

The second turn, you have a 1/4 chance to wake up, a 1/4 chance to select the superior move, 1/4 chance to select the average move, and 1/4 chance to select the inferior move. Using the superior move outright is a 1/4 chance to select the superior move. This is another clear win for Sleep Talk.

The third turn, you have a 1/3 chance to wake up and a 2/9 chance to select each move. Using the best move gives a 1/3 chance to use the superior move. As before, you have an extra 1/9 chance to use your best move, but this time, you have a 1/3 extra chance to do nothing. Here, you have twice the chance to at least do a move by using Sleep Talk, but the move you use will be slightly less powerful. This is again a battle-specific judgment, but it is a little better for Sleep Talk than the previous example.

The fourth turn, you have a 1/2 chance to wake up and a 1/6 chance to select each move. Using the best move gives you a 1/2 chance to use the superior move. By comparison, Sleep Talk gives you a 1/2 chance to select any move at all, so using the superior move is better.

The fifth turn, you should clearly use the superior move because you're certainly waking up.

In both of these cases, Sleep Talk is the dominant strategy for turns 1 and 2, while the superior move is the dominant strategy for turns 4 and 5. The third turn of sleep has the best strategy depend on battle conditions.


There is another interesting case, however. What happens when the Pokemon using Sleep Talk is something like Rest, Sleep Talk, Thunderbolt, Will-O-Wisp Rotom-A, and the opponent is Heatran? Then, Thunderbolt is the good move, Rest is the useless move, and Will-O-Wisp is the bad move, actually worse than doing nothing.

The first turn, Sleep Talk gives a 1/3 chance to make things worse than doing nothing, a 1/3 chance to do nothing, and a 1/3 chance to make things better. If the good move is a greater good than the bad move is bad, then Sleep Talk is the winner. If the bad move is worse than the good move is good, then anything but Sleep Talk is equally good.

The second turn, you have a 1/4 chance to wake up. Selecting Sleep Talk therefore has a 50% chance to fail to do anything useful (wake up or select Rest). You have a 1/4 chance to select your good move and a 1/4 chance to select your bad move. Compare this to selecting the good move outright, which has a 1/4 chance to use your good move and a 3/4 chance to do nothing. Therefore, selecting your good move dominates selecting Sleep Talk, as they have the same probability of good payout, but the good move replaces the bad payout of Sleep Talk with a neutral payout.

The third turn, you have a 1/3 chance to wake up. Using Sleep Talk therefore has a 5/9 (55.6%) chance to do nothing. You have a 2/9 chance to select your good move and a 2/9 chance to select the bad move. Compare this with selecting the superior move, which has a 1/3 chance to use your best move and a 2/3 chance to do nothing. Selecting the good move is the best option. You have better odds of the good move being used, and zero odds of the bad move being used.

The fourth turn, there is a 1/2 chance to wake up. Using Sleep Talk therefore has a 2/3 chance to do nothing, a 1/6 chance to select the good move, and a 1/6 chance to select the bad move. If you select the superior move outright, you have a 1/2 chance to select the good move, making selecting the superior move a clear winner.

You clearly should use the good move Sleep Talk on the turn you'll be waking up, the fifth turn.

If one of your moves is better than doing nothing, and one of your moves is worse than doing nothing, you should select the good move in every turn but the first turn. On the first turn, you select Sleep Talk if the good move is a greater good, and you select the good move if the bad move is a greater bad. If the good payout is just as good as the bad payout is bad, Sleep Talk is as good as selecting Rest on any turn.

Another option is that you have Sleep Talk, superior move, inferior move, bad move.

If you select Sleep Talk turn 1, you have 2/3 chance of positive payout and 1/3 chance of negative payout. Which move is the smart choice (the superior move or Sleep Talk) depends on the relative values. If the benefit from the superior move + the benefit from the inferior move is greater than the loss from the bad move, use Sleep Talk. If the reverse is true, use the superior move.

The second turn, you have a 1/4 chance to wake up, a 1/4 chance to select the superior move, 1/4 chance to select the inferior move, and 1/4 chance to select the bad move. Using the superior move outright is a 1/4 chance to select the superior move. If the bad move is worse than the inferior move is good, then selecting the superior move is the winning choice. If the reverse is true, Sleep Talk wins.

The third turn, you have a 1/3 chance to wake up and a 2/9 chance to select each move. Using the best move gives a 1/3 chance to use the superior move. This is yet another battle-specific judgment call.

The fourth turn, you have a 1/2 chance to wake up and a 1/6 chance to select each move. Using the best move gives you a 1/2 chance to use the superior move. By comparison, Sleep Talk gives you a 1/3 chance to select any good move at all, with the added loss of a 1/6 chance of being worse than doing nothing, so using the superior move is better.

The fifth turn, you should clearly use the superior move because you're certainly waking up.

If you have two moves that are better than nothing and one move that is worse than doing nothing, turns 1, 2, and 3 are all judgment calls based on battle conditions, while turns 4 and 5 should always be dealt with by using the superior move.

The final cases are those in which it is (almost) never wise to Sleep Talk. For instance, if you have Sleep Talk, a good move, and two bad moves, this is worse in every situation for Sleep Talk compared with two good moves and a bad move. In this situation, the only time you might Sleep Talk is on the very first turn, but the good move has to be as good as both bad moves combined are bad. After that, it is always better to select the superior move.

To determine when to use Sleep Talk while Resting, that is the same as doing turn 1 twice (0% chance to wake up) followed by turn 5 (100% chance to wake up).
 
I've actually considered this in the past, after seeing a Jap Milotic abuse Rest, Sleep Talk, Surf and Hypnosis. First of all, this strategy is greatly feasible, if used with a bulky Pokemon, this strategy can be devastating.
 
I've actually considered this in the past, after seeing a Jap Milotic abuse Rest, Sleep Talk, Surf and Hypnosis. First of all, this strategy is greatly feasible, if used with a bulky Pokemon, this strategy can be devastating.
It's possible that a Pok'emon could be asleep on the opposing team already, and this set could potentially activate the sleep clause and net you a loss, however.

As for the original post: Very insightful. That's why I love this website. I never looked at RestTalking in that light, and while I haven't got around to employing a RestTalk strategy yet (I breed all my pokes, don't play that Shoddy), I will look at it with this post in mind when I choose to.
 
I wonder if you would consider factoring in the chance the opponent has to miss the sleep inducing move. Probably irrelevant but its irritating when the situation is for the opponent to use Sleep move and miss as you use sleep talk and then enter into guessing games.
 

obi

formerly david stone
is a Site Content Manager Alumnusis a Programmer Alumnusis a Senior Staff Member Alumnusis a Smogon Discord Contributor Alumnusis a Researcher Alumnusis a Top Contributor Alumnusis a Battle Simulator Moderator Alumnus
This really isn't a "rest talk" strategy so much as "Sleep Talk when they put you to sleep.". Sleep Talk when you've used Rest is so much simpler because you know exactly when it's going to end. Sleep Talk under Rest is basically "Do I have one more that's good and another that isn't too bad?", if yes, use Sleep Talk. If not, don't.
 
Obi, it might be helpful if you put up a couple of charts with this information so that people that are more visual learners can grasp the concepts. Good use of basic game theory to demonstrate the best choices in a game of pokemon, though. A+ thread.
 
Early Bird means there is a max of two turns possible for sleep (5, cut in half, rounds down)? What are the probabilities for waking first turn? Is it cut in half, remain the same?
 
Early Bird means there is a max of two turns possible for sleep (5, cut in half, rounds down)? What are the probabilities for waking first turn? Is it cut in half, remain the same?
For Early Bird, you have a 25% chance to wake up immediately, 50% to wake after 1 turn, and 25% to wake after 2 turns. So, if you've seen no "Pokemon is fast asleep" messages, you have a 25% chance to wake up next turn. If you've seen 1 "Pokemon is fast asleep" message, you have a 66.7% chance to wake up next turn. If you've seen 2 "Pokemon is fast asleep" messages, you have a 100% chance to wake up next turn.

This pretty much means an Early Bird Pokemon should only select sleep talk if it hasn't seen any fast asleep messages yet, or used Rest the previous turn.
 
Good read indeed.

Just as a side note, Roar/Whirlwind adds yet another take on this, since Sleep Talk improves it (against slower Pokemon) by giving it a priority of 0. This obviously wouldn't fit such a simple, concrete formula as all the situations above since there are several factors that determine whether or not it would be more beneficial to increase Roar's priority or not. It would essentially depend on the degree to which you are "okay" with letting the opponent act or how valuable it is to act first. But yea, just another situation to consider.
 
Yes, but Facade does not get boosted power.

Obi, it might be helpful if you put up a couple of charts with this information so that people that are more visual learners can grasp the concepts.
Completely agreeing with this. For someone like me, it's difficult to read through this whole thing. If you could make a chart with turn x and move y with percentage z at the intersection, it would be much easier to understand.
 
Good read. I used to just use Sleep Talk until turn 4, but after reading this I'm definitely changing my thought process.
 

Lockeness

(e^(i╥))+1=0
I don't often use the sleep talk strategy ,but this thread was interesting. Your use of game theory and applied mathematics to battling strategy is very thought out. An excellent read.
 
Great read!!!

I actually use Sleep Talk (and no Rest) on a Venusaur of mine in UU (other moves being Leaf Storm, HP Ice, Sludge Bomb). It started out as simply a psuedo-sleep status absorber, but I never really thought through the percentages of the chances Sleep Talk could work throughout the sleeping turns.

Thanks for furthering my knowledge about this stratagy.
 
I don't often use the sleep talk strategy ,but this thread was interesting. Your use of game theory and applied mathematics to battling strategy is very thought out. An excellent read.
Truly was an excellent read. I try to use game theory wherever i can when battling.

Early Bird means there is a max of two turns possible for sleep
Glad someone mentionned that.
 

obi

formerly david stone
is a Site Content Manager Alumnusis a Programmer Alumnusis a Senior Staff Member Alumnusis a Smogon Discord Contributor Alumnusis a Researcher Alumnusis a Top Contributor Alumnusis a Battle Simulator Moderator Alumnus
With Early Bird in play, the odds of each length of sleep are as follows:

1-3

25% chance to sleep 0 turns
50% chance to sleep 1 turn
25% chance to sleep 2 turns.

However, the odds of waking up on a particular turn are as follows:

25% chance to wake up on the first turn
66.7% (2/3) chance to wake up on the second turn
100% chance to wake up on the third turn.

The reason the probabilities change is that if you get to the second turn, we already know that the waking up on the first turn event has not happened, so we need to remove it from our probabilities entirely. Now the only options are to wake up after 1 or 2 fast asleep messages, instead of 0, 1, or 2.

These probabilities mean that the very first turn you have been put to sleep (0 fast asleep messages yet), the best move is the same as in the second turn of any non-Early Bird Pokemon.

The second turn, your odds of waking up are better than in any of the scenarios I mentioned earlier, and as the odds of waking up increase, the payout of using the superior move increase as well. In every case I analyzed, the final turn had the superior move being the clear winner, so in this case, that will be an even stronger case against using Sleep Talk.

The third turn you are definitely waking up, so you should certainly not use Sleep Talk.
 
While I can greatly appreciate Obi's effort and vast intelligence, I don't really get the point of this thread.

This isn't some new breakthrough its just common sense. Simple math..
 
While I can greatly appreciate Obi's effort and vast intelligence, I don't really get the point of this thread.

This isn't some new breakthrough its just common sense. Simple math..


You'd be surprised at how little common sense people have these days....


great thread anyway obi
 
Good stuff. Most are too lazy to do the math, unfortunately, and I admittedly include myself in that category....and I'm actually good at math. Again, much appreciated, and definetely will be implemented when running Sleep Talk on my pokes.
 

Users Who Are Viewing This Thread (Users: 1, Guests: 0)

Top