5.4 Update: Raid Cooldowns Overview

Contents
Healer Throughput CDs
Non-Healer Throughput CDs
Damage Reduction CDs
Methodology

Back in 5.3 I made a blog post about raid cooldowns and their comparative strengths, which you can find hereI went into a lot of detail about the different types of raid CDs and when they are most useful, wrote a freakin’ dissertation on Spirit Link Totem that still applies, and compared the survivability each of the damage-reduction CDs would bring to a raid taking constant heavy damage.

However, the numbers in that post are now outdated – because they were generated using Throne of Thunder gearing, compared to Throne of Thunder mechanics, and because a few cooldowns and class mechanics have been changed. Here I hope to provide a more relevant set of numbers and graphs, and I’ll try to keep the amount of words I throw at you to a minimum. If you think I’ve missed some nuance, check the previous Raid CD post, because I probably covered it there :)

Healer Throughput Cooldowns

Below is a table showing the maximum raw throughput of each healing class’s cooldowns. Remember, these numbers are generated with absolutely no overhealing taken into consideration; the amount of overhealing dealt by each cooldown is dependent not only on raid composition but also on timing, strategy, and the amount of damage you’re taking at the time. You’re rather unlikely to see this amount of effective healing from these cooldowns, but when used at appropriate times the relative strengths of these cooldowns ought to be the same.

If you care at all about the gear profiles I used to generate these numbers, see the Methodology section at the end of this post.

Healers’ Throughput CD Comparisons
Cooldown Raw Healing Comment
Tranquility (10-player) 3.5M +12% w/NV, +15% w/ToL, +28.8% w/both
Tranquility (25-player) 8.4M +12% w/NV, +15% w/ToL, +28.8% w/both
Revival (10-player) 1.8M Also removes Disease, Magic, Poison debuffs
Revival (25-player) 4.6M Also removes Disease, Magic, Poison debuffs
Light’s Hammer 2.9M
Spirit Shell: PoH 3.4M Assumes Archangel & BT
Spirit Shell: PoH, 4p t16 5.1M Assumes Archangel & BT
Divine Hymn (10-player) 3.9M Worst-case scenario of 50% of ticks self-buffed
Divine Hymn (25-player) 9.5M Worst-case scenario of 48% of ticks self-buffed
Lightspring 2.8M Glyphed for 2 extra “charges”
Healing Tide Totem (10-player) 4.2M Average raid health of 40%
Healing Tide Totem (25-player) 10.1M Average raid health of 40%
Ancestral Guidance 8.2M Average raid health of 40%: UE-HR, 5x CH(4)-RT
Ascendance 5.7M Average raid health of 40%: pre-cast UE-HR, 5x CH(4)-RT, HR, CH(4)-RT

There’s not a lot to talk about here that hasn’t already been said at length, but I won’t let that stop me! I have a few things I’d like to say about those shocking Shaman numbers. First of all, the “40%” number. Keep in mind I’m suggesting this as an average raid HP only during the time that you’re using the cooldown. The basic assumption here is that you use raid CDs to respond to periods of low raid health, not when the raid is on high health, and that most CDs are used during periods of pulsing AoE damage so that raid health is still fluctuating while your CD is active. Since this information is aimed at people who are pushing progression, I don’t think it’s an unfair number to choose, but of course you can use this information to reverse-engineer numbers for your own raid if your raid’s health is not dropping that low.

This time, I have included only the healing that would be directly attributed as “Ancestral Guidance” or “Restorative Mists” for those cooldowns, not the total healing you’d generate while the cooldown is active. Counting that in pushes your healing-done-while-AG-is-active up to around 13M and your healing-done-while-Ascendance-is-active up to 11.4M.

Specifically regarding AG (although true for Ascendance as well): In patch 5.3 when I wrote my previous post I was asked about the relative strength of AG with Chain Heal compared to AG with Healing Surge. Back then, my answer was that the AG/HS rotation did only about 100k less healing than the optimal AG/CH(4)-RT rotation, so Shaman in 10-player formats, who couldn’t get every bounce out of Chain Heal, would definitely gain throughput by switching to an AG/HS rotation.

The answer has changed in 5.4, though, with Chain Heal’s buffs meaning that a CH that hits only two targets, so long as the first target was Riptided, will deal more healing than a single HS will. So here’s the run-down:

  • AG with HS spam: 5.9M healing
  • AG with CH(2)-RT spam: 5.8M healing
  • AG with CH(3)-RT spam: 6.9M healing

What this indicates is that if even one of your Chain Heals bounces to 3 targets, even if the rest only bounced to 2 targets, you’re better off using Chain Heal than Healing Surge even in 10-player formats. I can’t think of many fights where this is actually impossible; but if it is, on say Heroic 10 Iron Juggernaut for example, you’re probably better off glyphing the Glyph of Chaining and alternating CH and HSx2 spam than just going with HS alone. You’d be even more better off convincing your raid to stay clumped up enough that you could get unglyphed Chain Heals to bounce at least 3 times, though :)

Non-Healer Throughput Cooldowns

Here’s a look at the raw healing delivered by hybrid tanks/DPS using their healing cooldown abilities:

Non-Healers’ Throughput CD Comparisons
Spec Cooldown Raw Healing Comment
Balance Tranq (10) 3.7M HotW
Tranq (25) 8.9M HotW
Feral Tranq (10) 3.9M HotW, no weapon swap
Tranq (25) 9.4M HotW, no weapon swap
Guardian Tranq (10) 2.8M HotW, no weapon swap
Tranq (25) 6.6M HotW, no weapon swap
Shadow Symbiosis Tranq 493k 10-player raid
962k 25-player raid
Vamp. Embrace (10) 2.6M Unglyphed
3.5M Glyphed
Vamp. Embrace (25) 6.5M Unglyphed
8.6M Glyphed
Ele Healing Tide Totem 1.2M 10-player raid
2.8M 25-player raid
Ancestral Guidance 3.2M Just damage-to-healing conversion
4.7M Pre-cast HR
Enh Healing Tide Totem 1.2M 10-player raid
2.9M 25-player raid
Ancestral Guidance 3.7M Just damage-to-healing conversion
5.0M Pre-cast HR
Warrior Rallying Cry (10) 1.4M 2 tanks @1Mk HP, 8 others @650k HP
Rallying Cry (25) 3.4M 2 tanks @1M HP, 23 others @650k HP

There’s not a lot that’s changed here. Shamans now have Healing Tide Totem baseline, so you can expect your raid’s Shamans to have both HTT and Ancestral Guidance. Note, however, that the DPS Shamans’ version of HTT is pretty lackluster, and as expected Ancestral Guidance well outperforms it. While it is still a healing gain to pre-cast a Healing Rain before activating Ancestral Guidance, there are fewer cases in SoO where this is possible for an Elemental Shaman without giving up some DPS, and since Glyph of Healing Storm no longer affects an Enhancement Shaman’s Healing Rain, it doesn’t really make a huge difference.

I don’t believe many Druids take Heart of the Wild these days, so you could certainly reduce those numbers a fair bit if you wanted to discount that Talent choice. For Balance Druids, HotW provides a 50% boost, so divide the numbers above by 1.5; for Feral/Guardian Druids, HotW is a 100% boost, so halve the numbers above. I would do this myself, but I don’t want to make a chart so long I could lay it across the ocean and walk to LA :P

Damage Reduction Cooldowns

These cooldowns mitigate damage based upon the amount of damage being dealt to the raid, and therefore you have to know what abilities you are mitigating in order to determine how much damage will be prevented. If you’ll recall, I’m using three mechanics from Heroic Siege of Orgrimmar encounters: Iron Juggernaut’s Seismic Activity, a heavy magical damage DoT; Spoils of Pandaria’s Mogu Elders, a one-off, strong magical attack; and Thok’s Deafening Screech, a frequently-occurring physical damage attack.

To display the relative power of each cooldown vs. each mechanic, I’ve done up a little bar chart. Since some of these mechanics are more powerful in 25-player than in 10-player, I’ve also separated the two raid formats into their own graphs:

Damage reduction cooldowns in 10-player raids

Damage reduction cooldowns in 10-player raids

While I won’t spend too much time going into detail about this, I do want to note that although Devotion Aura does not mitigate any of the damage dealt by Thok’s Deafening Screech, remember that it is still by far the most important raid cooldown your raid can use. In this particular case it acts as a healing throughput CD, allowing healers to continue their usual chain of spellcasts without fear of interruption.

Damage reduction cooldowns in 25-player raids

Damage reduction cooldowns in 25-player raids

Similarly, Spirit Link is stronger than it appears in the Thok situation as well, because the health redistribution makes it less likely that the players who are struck by Shock Blast or other random spike damage will die. This is handy when you are trying to keep two separate groups so that you can go as low in health as possible without pushing the boss into Frenzy for Blood! early, or for when you are stacked up trying to push into that phase, and need to heal less to fall beneath the 50% health threshold, but don’t want to die.

Oh – and I left Zen Meditation off because it’s better used as a personal CD than a raid one, especially with the vast number of mechanics that it doesn’t really do much for.

Methodology

This time around I was a little lazy. Rather than attempting to build BiS lists on Ask Mr. Robot and either trust their stat weights or tweak them myself, I simply used the gear that the healers and DPS in my guild were wearing as of last week. This resulted in a set of gear profiles around the 566-570 ilvl mark.

I am going to trust that my guildies know what they are doing with their own stat weights, and have optimised their gear far better than I could with the AMR tool. However, I’ll note that their choices may be biased toward 25-player raiding rather than 10-player, so healers who do 10-player raiding may find these stat allocations a little unsatisfactory. Feel free to leave feedback! :)

I’ve also included all relevant raid buffs, including Intellect flasks and food.

Healer Stats in 566-570 iLvl gearing
Restoration-Druid-Icon mistweaver monk icon holy-paladin-icon discipline-priest-icon discipline-priest-icon restoration-shaman-icon
Resto Druid Mistweaver Monk Holy Paladin Disc Priest Holy Priest Resto Shaman
Intellect 28371 27176 30636 29448 32395 30809
Spellpower 42004 42552 45183 42341 46942 52627
Mastery 39.80% 19.13% 51.27% 53.33% 42.25% 80.02%
Haste 37.91% 39.19% 39.05% 7.66% 17.36% 30.10%
Crit 17.70% 44.92% 24.25% 40.97% 19.50% 31.01%

The disparity in Intellect/Spellpower values comes around because some classes are valuing secondary stats way more highly than primary (particularly Mistweavers and Disc Priests, but also Resto Druids just reaching the high Haste breakpoint). You’ll note I’m using a pretty Mastery-heavy build for the Shaman stats; that’s mostly because I was most recently progressing on H Thok when I grabbed this information. Mastery is really nice there, but I also tend to value it more highly than most other Heroic-progressing Shaman since I place more significance on clutch healing/cooldowns over maintenance healing.

For the DPS spec cooldowns that rely on damage done to generate healing, I used the SimC DPS value on the WoW-Progress page for each character. Again, yes, it’s quick-and-dirty, but I think it’s close enough for this sort of work.

Non-Healer Stats in iLvl 566-570 Gear
balance-druid-icon feral-druid-icon guardian-druid-icon shadow-priest-icon elemental-shaman-icon enhancement-shaman-icon
Balance Druid Feral Druid Guardian Druid Shadow Priest Ele Shaman Enh Shaman
Spellpower 43429 31094 22543 41869 44676 34915
Haste 22.34% 26.64% 22.89% 43.18% 33.57% 43.59%
Crit 47.03% 51.64% 33.39% 29.37% 24.32% 36.28%
SimC DPS N/A N/A N/A 267,011 299,571 339,281

Finally, I chose the following boss abilities to use as mechanics to be countered by damage reduction cooldowns:

Pulsing Physical Damage: Thok’s Heroic Deafening Screech. I chose a 2.4-second interval between Screeches as that’s the interval of the final 15 or so Screeches my raid experienced on Heroic Thok; each screech dealing 580k Physical damage per player in 10-player formats, and 630k Physical damage per player in 25-player formats. Cooldowns were evaluated by how many Screeches they could fit within their uptime if timed well. Shock Blast and other prisoner-related debuffs and abilities were not included.

Pulsing Magical Damage: Iron Juggernaut’s Heroic Seismic Activity. This pulses every second for 60 seconds, dealing approximately 78,200 Nature damage per player in 10-player formats, and 86,000 Nature damage per player in 25-player formats. Since it is possible to stack early in this phase or between the second and third Shock Pulses, I have assumed the entire raid fits within an Anti-Magic Zone, Power Word: Barrier, Smoke Bomb, or Spirit Link Totem.

Large Single Attack: Spoils of Pandaria’s Heroic Mogu Elders (the guys in the Massive Crates) deal a large magical-damage AoE every 12 seconds that strikes every player in the quadrant. It is 20% larger per Stone Statue active, and since sometimes you can’t have every Stone Statue dead in time, I calculated its damage with 2 Stone Statues alive. This deals 486k damage per player in 10-player formats and 535k damage per player in 25-player formats. I calculated it assuming there were 5 players in the quadrant for 10-player raids, and 13 players in the quadrant for 25-player raids. Note that one cooldown, Demoralizing Banner, lasts long enough to cover two successive AoE attacks; this was factored in to its calculation.

About these ads

About Dedralie

Stuff about me!
This entry was posted in Discipline, Druid, Guides, Holy, Monk, Paladin, Philosophy, Priests, Raids, Shaman, Theorycrafting. Bookmark the permalink.

19 Responses to 5.4 Update: Raid Cooldowns Overview

  1. sorryman105 says:

    That was very informative, thank you so much for doing this. This question may not be vey relevant but I don’t use twitter and couldn’t think of abetter place to ask. I’m in a ten man guild and I’m using the 15k haste break point on my shaman (we are going to be progressing on Malkorok). What did you happen to use?

    • Dedralie says:

      I can’t quite reach the 15k breakpoint myself (I’m 400 shy with a Heroic Prismatic Prison of Pride!), but I’m hoping to get there and try it out. Most of my progression this tier has been with 7613 Haste + Ancestral Swiftness, putting me at 30% Haste when raid-buffed. Just this week I went up to 10118 Haste (which is 30% Haste when raid-buffed *without* Ancestral Swiftness) and have been playing around with Elemental Mastery, but I’m not sure I like it.

      On Sat, Oct 26, 2013 at 5:09 PM, healiocentric

      • sorryman105 says:

        I detest EM…but it might be because I heal with two shield classes and I end up over-healing every time.

      • hoofdee says:

        have you spent any time playing with the totem breakpoints? with some experimenting, i’ve found 9400 is about what i need for an extra reliable totem tick (supposed to be 8812 or so). i found this number by swapping haste gear in and out while standing by myself and counting ticks (i was in ele spec with AS at the time).

        it seems to be effective; however, i’ve never tried crawling through my logs after a raid to verify that i get the extra tick every time.

        regarding your CD summary, how do you account for the lost RS healing when choosing AG? even if we disregard the RS healing done outside the burst window, you should at least figure out how much you lose inside the window (both from raw healing, and glyph reduction). also, what fights do you choose RS vs. AG?

  2. sorryman105 says:

    Drat, If I could get my armory to update I could show you my gear setup. Do you have any advice for Malkorok? I’m healing with a Disc and Holy Paladin (although the disc priest may try Holy Healing for that fight). I kind of think I over-invested for it. My intellect is at 26k with 566 ilvl.

    • Dedralie says:

      I have some tips in my Malkorok guide ( http://healiocentric.wordpress.com/2013/09/24/soo-9-malkorok/ ) about things you can do for it, but ultimately without knowing what your exact problem is it’s hard to give solid advice :-)

      If it is just that you look low on the meters, the “Ancient Miasma and You” section goes into great detail about this, but basically as a Shaman you are at a disadvantage on the meters. This is not because you are doing poorly; it is because the other healers are cheatier than you. Direct healing shows up on meters as “effective” even if the target’s shield is 100% full. Meanwhile, smart heals ignore entirely every player with a green Barrier debuff, even though their shield may be as low as 85% capacity. Since Shamans have the greatest proportion of smart healing (HR, CH, HST, AA, HST), we look weakest on this fight. So long as you are casting heals on players with red or yellow shield debuffs, you are doing your job and shouldn’t fret.

      If you really want to boost your numbers just so you look more competitive on meters, you can take AG and pop it right at the start along with Ascendance. AG will be back up for the first Blood Rage phase and you can use Ascendance again for the second Blood Rage.

      I can’t go into much more detail here without logs (and, well, because I am not at my computer right now) but I hope this info and the linked guide help! :-)

  3. Tiberria says:

    I think that using 40% HP as the metric to evaluate the strength of raid cooldowns is inflating the output of the Shaman cooldowns significantly. H Thok (during progression – 3 weeks later we all outgear it so not so much) is the only fight this tier where I have seen average health pools/output levels of HTT go to anywhere near that level.

    Average overheal of HTT across the first 11 bosses is about 60%. That indicates that the average HP level is way higher than 40%. You should probably use no less than about 70% as the baseline for these calculations. Otherwise, the mastery gain is just being massively inflated.

    • Dedralie says:

      I don’t pop HTT when the raid is at 70% health. Do you? Perhaps your raid should drop a healer. :-P

      Seriously though I think people (probably people who run that RSS mod since I got this 70% statistic from multiple people) are confused about what I mean when I say “average raid health at 40%”. I do not mean for the whole fight; I mean *for the time that HTT (or other throughput CDs) is being used*. I certainly find the raid at that low health at the end of H Norushen, H Galakras, after the first Shock Pulse on H Galakras, during a Massive Mantid on H Spoils esp if I or the other healer have Set to Blow, on H Thok as you note (and which incidentally is one of the mechanics I am using to judge CD strength), and so far, during Overcharged Magnetic Crush on H Siegecrafter, though admittedly we are just learning this fight so that may change as we get more comfortable with the encounter.

      I have routinely seen 9M+ HTTs even on farm content (with 20-30% overheal) and am getting about 60-75% of my AG/Asc values, and my guild overheals everything with 2 disc priests and 2 holy pallies. Perhaps I am a seriously anomalous case, although you’d think if I was getting _so much more_ out of my CDs than every other Shaman I’d rank a lot higher ;-)

      You can of course easily adjust these numbers to reflect whatever Mastery rating and average raid HP you like, it’s very simple. But since this is meant to be a tool for people to use during Heroic progression I do not think it unfair to assume CDs are being used when players are at low health across the board. But this is why I wrote my assumptions out to begin with – the transparency of the method means you know exactly how to adjust the numbers for your own raid, and if it is too daunting, well, I have a spreadsheet on my What’s New in MoP / Shaman Changes / Extra for Experts page where you can plug in your own stats and see these cooldowns on 100% hp, then adjust for your raid’s average health there.

  4. Tiberria says:

    No, I rarely see our raid’s health drop below 70%, even during “cooldown situations” like P2 of Galakras, Swelling Pride on Sha, Warsong on Nazgrim, the later phases of the fight on Norushen, at all on Spoils, magnet phase on Siegecrafter, at all on H Paragons, etc, etc. If I waited to only use it when the raid dropped to 40% HP, I wouldn’t be using it at all on 75% of pulls. Also, even if you use it with the raid at 40% HP, you can’t assume that the raid stays at 40% HP for the 10 second duration. It’s a lot of healing, and unless you regularly see 0% overheal HTT drops (which I didn’t even see on Thok progression), it’s a pretty safe bet that the average HP during the duration of HTT is nowhere near 40%.

    I just think that assuming 40% HP levels, 80% mastery (when the most commonly accepted gearing strategy is to minimize mastery) and 0% overheal is painting a best case and unrealistic picture of the relative strength of HTT, and to a much larger extent Ascendance and AG. I also think assuming 0% overheal on Tranq and Divine Hymn makes both of them look way stronger than they realistically are. Both spells by their very nature are even more prone to a lot of overheal than HTT is.

    • Dedralie says:

      The difficulty in all of this is that I cannot predict how much overheal each CD is going to do on average. I mean, I could perhaps if I went through every parse on WoL and took an average, but there’s serious issues with this – players who have ranked are probably those whose overall overheal is low, or for whom the situation arose where their cooldowns were more powerful or effective than usual, or who were solo-healing, or etc. etc. And should I correlate for talent choices (PI vs ToF for Disc Priests, NV vs HotW for Druids, etc.), raid composition, strategy differences?

      So either I can bias the results by assuming an overheal percentage that may or may not represent what anyone would see in a raid, or I can bias the results by reporting “this is the maximum throughput you’ll get” and everyone can look at their own stats and overheal percentage and work it out for themselves. I choose to do the latter, because reporting maximum values and making it clear that they are maximums is less manipulative of the data than making the adjustments myself. In the linked ’5.3 Raid Cooldowns’ post I covered a lot of the drawbacks of all of these raid cooldowns, including the fact that DH and Tranquility are prone to a lot of overheal, but in the interest of brevity I just threw links in to the previous post rather than rehash or copy-paste old ground.

      Basically, I guess I write with the explicit assumption that people can look at the data I’ve provided, and as you have done, identify places where the assumptions I had to make in order to be able to generate _any_ data misalign with their own experiences, and then adjust my data to better align with their own experiences. Nobody expects to see a fully effective Tranq or DH outside of the first 8 seconds of Malkorok, and I trust that the people who read my site are intelligent enough to realise that these raw throughput numbers aren’t what you’re going to necessarily see in a raid. But this is true for any theorycrafting – we calculate the maxima, you guys apply that information intelligently. That said, I’ll edit the table to make it clear it’s RAW healing, and maybe throw a few more words at it, so that people who aren’t used to being given raw data will at least realise that’s what’s happening :)

      (For reference, on my last farm night, excluding the Thok re-kill because it’s not part of this log, I did:

      * An average of 9.4M raw healing with each HTT, at ~40% overheal;
      * An average of 5.9M raw healing with each Ascendance, at ~70% overheal;
      * An average of 6.7M raw healing with each Ancestral Guidance, at ~40% overheal.

      I’m pretty happy with the model, and especially for those HTT numbers I know, for example, that some of them were very high overheal/very low effective heal, like during H Protectors when I drop HTT just because it seems like I should drop HTT on every fight, while some of them were very low overheal/very high effective healing, like the HTT I drop right after I come back from the Test of Reliance on H Norushen.)

  5. Bombadil says:

    Great work! I appreciate the time it took for you to compile this information and set it, so that the community can digest it and grow.

    Shaman are wonderful. They always have been.

  6. Nari says:

    Awesome as always, still in the middle of reading.
    if i may say so, i do feel, that for a true comparison 1 variable is still missing in these calculations.
    and that is: Time
    there is a huge difference in casually instant popping a revival, or using spirit shell and standing still spammig PoHs for ~10 sec.
    maybe somehow… use a short fixed same time window with a spell priority cast list, or average these numbers to 1 GCD or… i don’t know.
    and….
    I love you :)

  7. Pingback: MoP In Review, Part 1: Healer Representation | healiocentric

  8. Rainbo says:

    This is very belated, but I love this post, and I can’t believe the amount of work you put into it!!

    I’m especially interested by your evaluation of Spirit Shell. I thought that calling SS a “raid CD” was exaggerating, but I see now that it’s not exaggerating at all. It’s ridiculous. I noticed that in your chart, you made no mention of Inner Focus, and you made no mention of 2p T16 (which is 10% crit during Archangel). Are those factors included?

    • Dedralie says:

      Hi Rainbo,

      Sorry it took me so long to get to this! You’ve probably forgotten all about it by now, haha.

      Inner Focus is not included, but 2pt16 is included in the 4pt16 row. I haven’t included Inner Focus because I wasn’t sure if Priests now use it during Spirit Shell or if they use it to generate DA outside of Spirit Shell, so I just left it out. You could add another ~400k for Inner Focus.

      • Rainbo says:

        Thank you for the reply! I was worried you’d forgotten about this post XD

        From what I understand, it’s rare for priests to be using 4p and much more common to be using 2p, so I’m actually more interested in 2p-only numbers. But that’s not particularly important. I continue to admire just about every post you make :D

        I actually used your posts on Spoils and Blackfuse to craft my tanking strategy for those fights!! Your guides are that thorough!

  9. emmet says:

    I love the math you have going, but I have to disagree with the stats you’re giving discipline priests. Haste has absolutely no value for us, and 20% crit in 570 is virtually unheard of. At that point you should be rocking at least 35%.

    • Dedralie says:

      I think you are looking at holy priest stats by accident. My disc priest stats (which I took from my guild’s disc priest’s gear) include over 40% crit and only 8%-ish haste (to allow 5 casts of PoH under SS with BT and the 4pc, since we raid 25s). :-)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s