Sword of Spirit
Legend
First some background.
What does the game assume the players know about DM dice rolls?
Claim
The 5e baseline assumes that the DM rolls the dice in front of the players, but doesn't tell them the total. Ie, you see that the attack roll against you is a 17, but you don't know what the final total (with modifiers) is.
Evidence
"Cutting Words: ...You can choose to use this feature after the creature makes its roll, but before the DM determines whether the attack roll or ability check succeeds or fails..." PHB p.54
"Combat Inspiration: ..after seeing the roll, but before knowing whether it hits or misses..." PHB p. 55
There is no counter-evidence that I am aware of, although there are other abilities that work differently. For instance, Bend Luck (PHB 103) functions similarly, but lacks some of the language, leaving its functioning somewhat ambiguous and requiring DM adjudication.
The DMG mentions rolling in open or rolling behind the screen as valid options for the DM. It does, however, fail to provide any advice for how to handle abilities such as those I referenced above if the DM does keep the dice hidden. Likely that is an oversight. It would stand to reason that, at a minimum, Combat Inspiration requires the DM to share to d20 result with the player.
What information about the attack does shield let you have before casting?
Claim
Shield says you know, prior to casting, whether or not the attack hits. It doesn't specify that you know by how much it hits.
Evidence
"Shield:...Casting Time: 1 reaction, which you take when you are hit by an attack..." PHB 275
There is no mention of dice rolls or attack results other than "hit".
As an aside, a related feature, Defensive Duelist works similarly. In this case it includes "potentially causing the attack to miss you." What bearing, if any, the phrasing in Defensive Duelist should have on interpreting shield is unclear. It might imply that there is uncertainty about whether the use of Defensive Duelist would cause the attack to miss (my assumption), but it could (less likely, IMO) also just be leaving the door open for other abilities that might turn a miss back into a hit and wanting to phrase Defensive Duelist in such a way as not to imply that its effect can't be later negated by another feature. In any event, I only mention Defensive Duelist for its similarity to shield, it is a different effect and i direct evidence for how shield should work.
The Parry monster feature functions similarly to Defensive Duelist. The only reason it might be relevant to a discussion is that it is usually the DM who is having the monster use it, and it is generally assumed that the DM knows all the information about the attack (ie, d20 roll, attack total, and whether or not the attack hits). I suppose the DM could just ask players if they hit a specific AC when they make an attack, "do you hit AC 14?" and therefore know the d20 roll (because it's on the table) but not the PC's total attack roll result (unless you have their bonus written down or memorized), but I doubt that is a majority technique.
With that background, I'd like to discuss, "How is shield supposed to work?" and "Which is most balanced?" A further question will be brought up later.
1) How is shield supposed to work?
It think my claim about what the 5e baseline assumes players should know (d20 rolls, but not total roll results) is solid. If that is the case, shield appears to grant the additional information of whether that attack hits or not. Since the player knows two out of the three pieces of information, they have more information to use in deciding whether to take the risk or not than someone would with Cutting Words or Combat Inspiration, which grant only one out of the three pieces of information.
But maybe the books suck in explaining themselves. Some people prefer to say that shield not only allows you to know whether or not you hit, but the exact number. Part of this might be because many people use the "Does a 16 hit your AC?" horribly non-optimal DM attack procedure, so they assume the game assumes that. I've made a counter-claim, with evidence (which anyone is free to challenge with counter-evidence). Anyone unwilling to abandon that procedure (just write down the PC's AC on a scrap of paper so you know whether or not a monster hits and don't have to ask every time--this isn't 3e, AC almost never changes!) can't use the 5e rules assumptions, and so must use an alternative such as telling the player the total attack roll result, or not telling them anything beyond "you're hit!" Either such option shifts the power of shield away from how the rules assumptions present it.
Then you have another reason one might not be able to use the standard assumptions: hidden dice rolls. For some people, that's a game play style. For others (such as my group) it's a logistical requirement: we play online and the DM isn't going to enter the stats in for every monster that's fought, or try to roll a bunch of d20s in an online tabletop when the game will move faster rolling physical dice. Let's not make any assumptions of dice fudging here. That's another play style issue that isn't relevant (for instance, we use hidden rolls (on the DM side) out of necessity, but don't fudge rolls).
So there are about three ways one could run shield. The "baseline assumption" method I claim the game assumes; an alternative of "full disclosure" so the player will always block the initial attack when they cast the spell, or a "hit only" method where the player has no direct knowledge of the roll at all.
Another point that could alter how we might choose to run it would be creating PC-Monster parity. How do we run it in such a way that the monsters aren't getting an unfair advantage in using it over the PCs?
2) Which is most balanced?
Considerations
1) Shield is a reaction to cast. The typical caster of shield is unlikely to have many other uses for their reaction. Opportunity attacks, and a total of 4 other spells on the same spell list (3 for sorcerer) are the most likely alternatives.
2) Shield has a "rider" in addition to its usage against the first attack: it raises your AC by 5 until the start of your next turn.
3) Shield also automatically negates magic missiles.
4) Shield is a 1st level spell that retains its effectiveness at all levels of play, without needing to be cast in a higher level slot.
To me, that looks like a really good spell.
I think the weakest way of running it, the "hit only" way, would still be a viable 1st-level spell. It might be on the weaker side, but I would probably still prepare it if my DM ran it that way. For about the cheapest spellcasting resource and opportunity cost available (other than making it a cantrip, the only way to make it cheaper would be to remove either the V or S component) I can raise my AC by 5 points per round and negate magic missiles, all without having to give up my action or bonus action, and not needing to cast it until I know it has a chance of immediately helping me. Compare that to buff spells you have to cast in advance and use an action or bonus action on and which might never even become relevant (if you aren't attacked for instance), and it seems worth it to me.
The "baseline assumption" version is my preferred balance point. I get everything I just mentioned, plus, I have additional information to allow me to make a more educated guess as to whether it's going to immediately help me. I think it is one of the best 1st level spells, arguably the best, and definitely a must-have.
With the "full disclosure" version it is a must have spell. They might as well just make it a wizard class feature, because anyone who doesn't take it doesn't understand the game. Now you are guaranteed to always (as long as you have the 1st level spell slot to spend) be able to block an attack that hits you by 5 or less, plus the rider abilities. It is bordering on overpowered in my opinion.
There are additional considerations that I haven't addressed much. One of them is how frequently you'll need to defend versus multiple attacks in a round. If that rarely comes up, then the spell isn't as effective. If it comes up all the time (such as for an Eldritch Knight or Bladesinger), then that is an extremely powerful benefit. Or if you run into a lot of enemies throwing magic missiles at you, then the spell's usefulness also skyrockets. I'd say as a baseline (and that's what we're going for, not what works best under uncommon campaign assumptions) we might assume multiple attacks are neither rare nor extremely common (though at high level, if a wizard is targeted by an attack, the likelihood they are going to be targeted by multiple attacks is probably higher), and that being targeted by magic missile is pretty rare for PCs.
A final question to address, which we might want to wait until later for, is:
3) How would one approximate their chosen method with hidden rolls and monsters?
This is a tough one that requires some sort of home brew technique. I'm going to need something like this in my online game where DM rolls are hidden, and any DM with NPC shield-casters who cares about a rough PC-NPC parity will have to come up with a way of deciding how to approximate the same decision-making process the PC makes based on the limited knowledge granted to them, unless you go with the "full disclosure" technique, which avoids this issue.
What does the game assume the players know about DM dice rolls?
Claim
The 5e baseline assumes that the DM rolls the dice in front of the players, but doesn't tell them the total. Ie, you see that the attack roll against you is a 17, but you don't know what the final total (with modifiers) is.
Evidence
"Cutting Words: ...You can choose to use this feature after the creature makes its roll, but before the DM determines whether the attack roll or ability check succeeds or fails..." PHB p.54
"Combat Inspiration: ..after seeing the roll, but before knowing whether it hits or misses..." PHB p. 55
There is no counter-evidence that I am aware of, although there are other abilities that work differently. For instance, Bend Luck (PHB 103) functions similarly, but lacks some of the language, leaving its functioning somewhat ambiguous and requiring DM adjudication.
The DMG mentions rolling in open or rolling behind the screen as valid options for the DM. It does, however, fail to provide any advice for how to handle abilities such as those I referenced above if the DM does keep the dice hidden. Likely that is an oversight. It would stand to reason that, at a minimum, Combat Inspiration requires the DM to share to d20 result with the player.
What information about the attack does shield let you have before casting?
Claim
Shield says you know, prior to casting, whether or not the attack hits. It doesn't specify that you know by how much it hits.
Evidence
"Shield:...Casting Time: 1 reaction, which you take when you are hit by an attack..." PHB 275
There is no mention of dice rolls or attack results other than "hit".
As an aside, a related feature, Defensive Duelist works similarly. In this case it includes "potentially causing the attack to miss you." What bearing, if any, the phrasing in Defensive Duelist should have on interpreting shield is unclear. It might imply that there is uncertainty about whether the use of Defensive Duelist would cause the attack to miss (my assumption), but it could (less likely, IMO) also just be leaving the door open for other abilities that might turn a miss back into a hit and wanting to phrase Defensive Duelist in such a way as not to imply that its effect can't be later negated by another feature. In any event, I only mention Defensive Duelist for its similarity to shield, it is a different effect and i direct evidence for how shield should work.
The Parry monster feature functions similarly to Defensive Duelist. The only reason it might be relevant to a discussion is that it is usually the DM who is having the monster use it, and it is generally assumed that the DM knows all the information about the attack (ie, d20 roll, attack total, and whether or not the attack hits). I suppose the DM could just ask players if they hit a specific AC when they make an attack, "do you hit AC 14?" and therefore know the d20 roll (because it's on the table) but not the PC's total attack roll result (unless you have their bonus written down or memorized), but I doubt that is a majority technique.
With that background, I'd like to discuss, "How is shield supposed to work?" and "Which is most balanced?" A further question will be brought up later.
1) How is shield supposed to work?
It think my claim about what the 5e baseline assumes players should know (d20 rolls, but not total roll results) is solid. If that is the case, shield appears to grant the additional information of whether that attack hits or not. Since the player knows two out of the three pieces of information, they have more information to use in deciding whether to take the risk or not than someone would with Cutting Words or Combat Inspiration, which grant only one out of the three pieces of information.
But maybe the books suck in explaining themselves. Some people prefer to say that shield not only allows you to know whether or not you hit, but the exact number. Part of this might be because many people use the "Does a 16 hit your AC?" horribly non-optimal DM attack procedure, so they assume the game assumes that. I've made a counter-claim, with evidence (which anyone is free to challenge with counter-evidence). Anyone unwilling to abandon that procedure (just write down the PC's AC on a scrap of paper so you know whether or not a monster hits and don't have to ask every time--this isn't 3e, AC almost never changes!) can't use the 5e rules assumptions, and so must use an alternative such as telling the player the total attack roll result, or not telling them anything beyond "you're hit!" Either such option shifts the power of shield away from how the rules assumptions present it.
Then you have another reason one might not be able to use the standard assumptions: hidden dice rolls. For some people, that's a game play style. For others (such as my group) it's a logistical requirement: we play online and the DM isn't going to enter the stats in for every monster that's fought, or try to roll a bunch of d20s in an online tabletop when the game will move faster rolling physical dice. Let's not make any assumptions of dice fudging here. That's another play style issue that isn't relevant (for instance, we use hidden rolls (on the DM side) out of necessity, but don't fudge rolls).
So there are about three ways one could run shield. The "baseline assumption" method I claim the game assumes; an alternative of "full disclosure" so the player will always block the initial attack when they cast the spell, or a "hit only" method where the player has no direct knowledge of the roll at all.
Another point that could alter how we might choose to run it would be creating PC-Monster parity. How do we run it in such a way that the monsters aren't getting an unfair advantage in using it over the PCs?
2) Which is most balanced?
Considerations
1) Shield is a reaction to cast. The typical caster of shield is unlikely to have many other uses for their reaction. Opportunity attacks, and a total of 4 other spells on the same spell list (3 for sorcerer) are the most likely alternatives.
2) Shield has a "rider" in addition to its usage against the first attack: it raises your AC by 5 until the start of your next turn.
3) Shield also automatically negates magic missiles.
4) Shield is a 1st level spell that retains its effectiveness at all levels of play, without needing to be cast in a higher level slot.
To me, that looks like a really good spell.
I think the weakest way of running it, the "hit only" way, would still be a viable 1st-level spell. It might be on the weaker side, but I would probably still prepare it if my DM ran it that way. For about the cheapest spellcasting resource and opportunity cost available (other than making it a cantrip, the only way to make it cheaper would be to remove either the V or S component) I can raise my AC by 5 points per round and negate magic missiles, all without having to give up my action or bonus action, and not needing to cast it until I know it has a chance of immediately helping me. Compare that to buff spells you have to cast in advance and use an action or bonus action on and which might never even become relevant (if you aren't attacked for instance), and it seems worth it to me.
The "baseline assumption" version is my preferred balance point. I get everything I just mentioned, plus, I have additional information to allow me to make a more educated guess as to whether it's going to immediately help me. I think it is one of the best 1st level spells, arguably the best, and definitely a must-have.
With the "full disclosure" version it is a must have spell. They might as well just make it a wizard class feature, because anyone who doesn't take it doesn't understand the game. Now you are guaranteed to always (as long as you have the 1st level spell slot to spend) be able to block an attack that hits you by 5 or less, plus the rider abilities. It is bordering on overpowered in my opinion.
There are additional considerations that I haven't addressed much. One of them is how frequently you'll need to defend versus multiple attacks in a round. If that rarely comes up, then the spell isn't as effective. If it comes up all the time (such as for an Eldritch Knight or Bladesinger), then that is an extremely powerful benefit. Or if you run into a lot of enemies throwing magic missiles at you, then the spell's usefulness also skyrockets. I'd say as a baseline (and that's what we're going for, not what works best under uncommon campaign assumptions) we might assume multiple attacks are neither rare nor extremely common (though at high level, if a wizard is targeted by an attack, the likelihood they are going to be targeted by multiple attacks is probably higher), and that being targeted by magic missile is pretty rare for PCs.
A final question to address, which we might want to wait until later for, is:
3) How would one approximate their chosen method with hidden rolls and monsters?
This is a tough one that requires some sort of home brew technique. I'm going to need something like this in my online game where DM rolls are hidden, and any DM with NPC shield-casters who cares about a rough PC-NPC parity will have to come up with a way of deciding how to approximate the same decision-making process the PC makes based on the limited knowledge granted to them, unless you go with the "full disclosure" technique, which avoids this issue.
