At what level does the Sleep spell stop being effective?

Question

Answering this homebrew question reminded me of the peculiar nature of the sleep spell -- on level one, it is an absolute beast: common opponents like kobolds or goblins have around half a dozen hp, so the 5d8 hp sleep effect on average removes three or even four of them, with a single action and no save. And even removing a single opponent with no save is pretty impressive.

However, the spell quickly becomes less useful as you gain levels and your opponents have a lot more hit points. It still can retain some usefulness, for example for finishing of mobs or monsters that have been softened up with a fireball or other attacks by your party, but eventually you stop preparing it.

My question here is -- at what level is the actual turning point reached, at which the spell, upcast to the highest level available, can be expected to lose usefulness to put to sleep (a) multiple opponents and (b) a single one?

Clearly, this will vary a lot based on the individual opponents. You can encounter goblins when you are level ten, or an ogre when you are level one, and it may retain some use for those weaker opponents you run into for a longer time. For this question, please assume opponents in the form of monsters whose CR matches the average party level.

Answer 1

Simulations of the use of Sleep in combat suggest that the spell's usefulness declines steadily and is consistently low starting around level 8

I think the other answers address the question clearly, but I thought it would be useful to add some context from a simulated-play perspective. My motivation comes from the observation that comparing Sleep against individual monsters is somewhat problematic as it doesn't represent the fact that encounters typically include multiple monsters, some of which might be very easy to neutralize with Sleep. Here, I'm answering the question, "If a full spell-caster, in a campaign of 5 PCs, specialized in casting Sleep in combat, how effective would their Sleep spell be at neutralizing enemies?"

To get at this, I put together a small dataset of simulated encounters across levels 1—20. At each level, I used the ChaosGen D&D 5e encounter generator to generate 10 random encounters for a player of 5 PCs at the appropriate level of which 4 encounters were ranked "easy", 3 were ranked "medium", 2 were ranked "hard", and 1 was ranked "deadly". I ran 2,000 simulated "days" of combat encounters at each level; in each day, the party would have 3 combat encounters, randomly drawn from the 10 generated encounters of the party's level. The spell-caster was assumed to cast Sleep using their three highest level spell-slots each day, one per encounter in order. All enemies were assumed to be in the range of the Sleep spell and were assumed to have 85% of their hit points when targeted. For each day, the total hit points of all enemies faced was calculated along with the total hit points of all enemies that were successfully neutralized using Sleep. The graph below shows the percentage of total enemy hit points that were successfully neutralized per day in terms of the party's level (blue), as well as the percentage of total enemy hit points that would be successfully neutralized if the spell-caster were able to maximize all of their Sleep rolls (red). The plot-points in the graph show the median value across simulations, and 50% of all simulations result in values between the error bars.

Clearly there are a lot of assumptions used in these simulations, but the point is to get a general sense of the average effectiveness of Sleep in actual play as a party levels up. Overall, these results support the observations in the other answers--i.e., that Sleep is primarily useful in Tier 1; though its effectiveness remains above or around 20% (measured in terms of total neutralized enemy hit-points per day) until level 8.

The level-to-level noise in the simulations likely comes from some combination of the encounter generator's algorithm and its database of monsters. At certain levels, especially above L15, the encounter generator only produced a fairly small set of different encounters, and any accidental bias toward fewer monsters with more hit points or toward more monsters with fewer hit points can have a big effect on the simulation. (This isn't the generator's fault really; I had to manually enter all the monster hit-points so to save time I limited it to one land type and only used entries with links to the monsters' stat cards).

Answer 2

Sleep is only worth it up to CR 1

Here is a table of the expected hp from sleep and for monsters on average using those in the Monster Manual, or using the guidelines given in the DMG for creating monsters of a given challenge rating, including the number of monsters you could put to sleep based on them:

CR	Sleep hp	MM hp	DMG hp	Sleep/MM	Sleep/DMG
0	22.5	3.5	3.5	6.4	6.4
1/8	22.5	8.2	21.0	2.7	1.1
1/4	22.5	14.8	42.5	1.5	0.5
1/2	22.5	19.7	60.0	1.1	0.4
1	22.5	28.2	78.0	0.8	0.3

By the numbers

As can be seen, already by CR 1, the average expected monster hit points outpace what sleep can put to sleep. You will not catch up at higher levels with upcasting -- you get to add 9 points of sleep for every spell level or two player levels, while the monster hit points increase on average by 13 per challenge rating in the MM (26 per spell level), or by more than 15 by DMG (30 per spell level).

As the question is defined, even on the first level sleep would not be worth it. That of course is nonsense, as you encounter a lot of smaller groups of monsters that are of CR 1/8 or 1/4 at that level. So, it does not make sense to look at this from a player level perspective, it is better to look at this from a monster hp and numbers perspective.

The issue is not primarily that you only face enemies with higher hit point totals at higher levels, the issue is that, as encounter XP values increase, you face larger numbers of enemies with the same hit point totals as before. And while removing four kobolds from a group of six is a powerful move, removing four kobolds from a group of sixty is a much weaker use of your single action for the round.

Play experience

In my own experience, you continue to encounter CR 1/8 to CR 1/2 monsters well into tier two play, and sleep may still useful as a cheap way to deal with them. However, at these levels, such monsters tend to come in larger groups, so you want to be able to remove all of them in the area, not just a few. Keeping a preparation slot for sleep has increasing opportunity costs in this context.

After entering tier two, you also can fireball and kill those monsters outright. A fireball can be expected to remove all monsters with less than 14 hp, no matter whether they save or not. I found myself no longer preparing sleep, even when dealing with low-CR monster mobs, somewhere around level 7 or so.

Answer 3

Sleep is worthwhile largely at levels 1-3.

This is due to the area and range. Burning Hands does 3d6, averaging to 10 or so fire damage. If the enemies are clustered together (which is very common, especially if they have already acted and moved up to attack) it can easily hit 3 or 4 of them, for around 40 or so damage (some will save). Critically, if they survive this attack, they will still be injured by it and be easier for your allies to defeat.

Sleep does on average 22, but it's not damage - it's sleeping enemies. Any excess that isn't enough to knock another foe asleep will be lost, and going from 'least hp first' means that's fairly likely to happen. Instead of knocking out Bob the orc sergeant, it knocks out one of the goblins under Bob's command, and then doesn't have enough hp left to affect Bob.

At which point, the affected enemy is now asleep. They can be woken up by their friends, likewise by any attack that fails to knock them out. Not only can sleep simply roll poorly and fail to affect any enemies, it can also be negated by turn order, situation, or a low damage roll from an ally.

This is in situations where it's relevant at all, such as goblins, kobolds, and other CR 1/4, 1/8, and (more rarely) 1/2 enemies.

However, there is a saving grace. Sleep has a range of 90' and affects a radius of 20'. There is no 1st level spell that has a similarly long reach and aoe. Burning Hands, Frost Fingers, Thunderwave, all originate from the caster and don't even go as far. Entangle is the only real contender in terms of range and aoe, and it's likewise a bit underwhelming in actual effect.

If you want to do things to enemies that are far away from you, in an aoe, and they have low hp, Sleep will do very reasonably compared to your other actual options, which are like Magic Missile or Firebolt. Given that one of the major threats to casters at this level is shortbows, this is actually kinda relevant.

Even at level 3, when you get access to second level spells, well, Shatter and Snowball Storm are pretty cool but the enemy might be spread out beyond their 10' or 5' aoes. When the enemy is few in hp but large in number and spread out with shortbows, Sleep becomes again a good choice.

But as you get more second level spell slots, and especially when 3rd level spells become available, sleep immediately loses any utility even in the small niche where it was 'ok' rather than 'terrible'.

Answer 4

It doesn't stop being effective

While it stops being as useful later on due to hit point inflation, as others have noted, it is a uniquely high damage spell which offers no save. This is especially useful against legendary enemies, which often have legendary saves or against enemies with good saves.

Consider a typical enemy encounter at level 20. An Adult Red Dragon.

You could use a typical damage spell on them, meteor swarm and with their legendary resistance and fire immunity you'll do an average of 10d6 damage, or 35.5 damage. You could also use sleep upcast to level 9 for 21d8, which does an average of 94.5 knock out power. An adult Red Dragon has 256 hit points on average, so after pushing them well below half health you can disable them in a single spell.

This also allows you to bypass any defensive measures that stop direct damage spells, like magical clouds.

If you can find some way to maximize it, perhaps with a DM with a flexible interpretation of maximize damage, you can reliably one-hit knock-out the monster with the spell, which is more than finger of death can do.

This is more useful if your DM gives you a good idea of hit points

If your DM likes to homebrew monsters with inflated for your CR hit points or not give any information on their hit points it's much more of a gamble using sleep. If you can make a good guess, sleep can be very effective up to level 20.

Likewise, if they conceal heavily what a creature is good at saving at or alter it, it's less useful.

Answer 5

Death isn't the only, or necessarily most useful, way to remove enemies from the fight

If keeping enemies alive for questioning, rehabilitating or even ransoming is useful in your games then the Sleep spell never loses its usefulness — it just means you need to wear your enemies down more before you can use it.

My question here is -- at what level is the actual turning point reached, at which the spell, upcast to the highest level available, can be expected to lose usefulness to put to sleep (a) multiple opponents and (b) a single one?

The premise of your question seems to come more from a video games min maxing than a role play perspective. Sleep can down an enemy (or enemies) without killing them, letting them survive to be questioned later. Spells don't normally have the ability to deal non-lethal damage:

Sometimes an attacker wants to incapacitate a foe, rather than deal a killing blow. When an attacker reduces a creature to 0 hit points with a melee attack, the attacker can knock the creature out. The attacker can make this choice the instant the damage is dealt. The creature falls unconscious and is stable.

The usefulness of this utility depends on your game. I often play in/DM games where there is an element of uncovering mystery where questioning people is essential to uncovering information. In one game (level 7) we were investigating some robberies — after uncovering the next likely target we waited and ambushed the burglars. When the fight turned the captain fled faster than any of us could run (all those rogue dashes), we wanted to question them and the best ranged option we had to knock an opponent out at range was sleep. The story progressed smoother and, since the captain was the son of a noble family, we avoided some negative consequences the DM had planned for us if we had chosen to kill him.