There seems to be a severely incorrect widespread assumption of how spell haste affects the global cooldown. I have read in numerous threads from numerous sources that spell haste is useless for flash of light because it will not shorten the global cooldown, yet all evidence leads to the contrary.

In my just-for-fun spell haste set, I have 166 spell haste rating, which calculates out to 166/15.76=10.53% spell haste. 1.5s*(1-.1053)=1.34s flash of light cast time. My in game cast bar shows a 1.3 cast speed, which is consistent with this decrease. In practice during raid situations, I have not once encountered a global cooldown related error chain casting flash of lights.

One experiment I have spent an extensive amount of time with is one minute flash of light drills. In my normal raid gear, with an average latency (somewhere around 150ms), I consistently cast 32 flash of lights per one minute. In a perfect world, the number would be 40, but obviously latency and user interface speed are an issue. This puts my flash of light at an average real cast time of 60s/32=1.875s.

In my spell haste gear, My one minute benchmark is 35 flash of lights. 60s/35=1.714s. 1.875s-1.714s=.161s. This is accurate as the theoretical increase from earlier was 1.5s-1.34s=.16s. The significance of this is that if the global cooldown was causing the global cooldown to limit the effectiveness of spell haste, there would be no significant increase in my one minute healing benchmark.

A solution that accepts the non-shortening of the global cooldown, yet still provides meaningful results from spell haste usage is that spell haste only combats the effects of ping, and if I could get my actual cast time less than 1.5+some small amount of interface latency, that I would hit a global cooldown wall. The spell haste gear does not yet exist to test this though.

Regardless, spell haste does provide a meaningful affect to flash of light, regardless of common conception. This being said, the sacrifices you must make to regen to obtain this effect often do not outweigh the gain, and at this point I rarely use my spell haste gear outside of trash healing.

It is worth mentioning that unless you have an incredibly low latency, the benefit you gain from stop casting will 'probably' out weight any spell haste itemization at the moment.
Did you try your experiments with stop casting? Or perhaps with Blood Lust and stop casting? Under these situations it is much more obvious that the GCD is not effected by haste.

I think you need to look over your "evidence" once again, it is a well proven fact that spell haste does NOT alter the GCD. Ever tried FoL spam with the [Scarab of the Infinite Cycle] proccing and a stopcasting macro? Keep in mind not using a stopcasting macro generally puts .1-.3 seconds onto your cast.

Fortunately, next patch there will be no need for stopcasting macros, as a blue post mentioned that queries to cast will now be sent to the server, not the client. I'll try to link the post for you guys.

Tested it for fun and did 36 casts in 1 min with 0 spell haste using a stopcasting macro.

I just did some extensive testing on ptr running consistenly around 150ms. With the new auto-stopcast, I'm averaging 35 fol/min without any haste and 38 fol/min with 10.5% haste. I'm never visibly hitting the gcd. I tried it with my scarab equiped as well. I did not hit any gcd restriction to my casting, though 38fol/min was still the average considering trinket uptime per minute will not give you a full extra cast regardless.

I found a shaman to heroism me, and though I wasn't able to run too many tests because of heroism cooldown, the few I did averaged 41fol/min. This result would not be possible with a 1.5 global cooldown, though it does seem that heroism does not affect it. A strange thing I noticed was that my casts were still over 1 second each on my cast bar, while they should be roughly .9 seconds with heroism up. Perhaps it's possible heroism does not stack with haste rating related haste buffs and subsequently non rating haste does not affect the global cooldown?

Regardless, the conception of flash of light not being affected by spell haste is incorrect, global cooldown decrease or not. I have never seen evidence or documentation pointing to global cooldown not being affected by spell haste. I do often read people stating that as a fact or assumption though. If someone wants to link something with some empirical evidence feel free.

For your empirical evidence, as stated in this thread:

However, if this testing is on the PTR, than it is certainly possible the mechanic has been changed in 2.3, which would be a very interesting fact. I will personally look into it, and if you could post more of your raw data that would be helpful as well. If blizzard did change this mechanic, then I'll be a very happy paladin

Also, there is another issue with your data. You said without any haste you were getting 35 FoL's per minute... in an idea situation you should be getting 40 (or 39 to be fair if you have a realistic ping that's fairly low). The 35 suggests to me either CPU latency or a very large connection latency. Working it out, you are averaging...

35(x) = 60
x = 1.714s FoL cast times, which indicates an approximate latency of 200 ms.

Also, even hasted you aren't hitting the proper 1.5s FoLs, you are actually at...

38(x) = 60
x = 1.579s FoL cast times, which means YOU ARE NOT HITTING THE GCD.

For the 41 casts with Heroism/Bloodlust, PLEASE PLEASE PLEASE be sure you aren't giving yourself an extra second. If you are sure, then you may be right.

Edit:

Also, please keep in mind the PTRs are very unstable and often don't have some operational models due to poor implementation or changing code.

Also AFAIK on the PTR right now spells that are 1.5s cast or less are not working as intented with the new "replacement"/"fix" to /stopcasting and will be changed to work properly.

The only way you'll get more FoL/min with haste is if you weren't using /stopcasting, but if you want to maximize your HP/s you need to use /stopcasting anyway in which case haste again doesn't help (for FoL). And as you can see you're not getting 44 FoL/min which you should be getting since 40 FoL/min is what you should get with no haste using proper /stopcasting with 0 errors (which is actually not possible with most connections due to ping changes, but if you cast for say 62-63 seconds with very low ping variance getting 41 FoLs is quite reasonable, with or without haste).

I think we beat the horse so hard his spirit died already... spell haste doesn't speed up spells limited by the global cooldown.

As for crit VS +heal, let's make it simpler - with your crit setup, if you don't crit, the MT dies, with my setup either I have the ability to keep him alive *every time* or I don't, but with high enough +heal should be able to keep him up every time, or else illidan would've still been alive due to people occasionally not critting their heals when a crit was needed to save the MT, which in a long fight would be very likely to happen at least once.
Since the first days of WoW, crits had been very misleading to just about anyone I've been talking to... Sure you see big numbers once in a while but when you look at the actual DPS/HPS gain from increasing your crit VS increasing your +heal, you see crit just isn't worth it.

Regarding the Crit vs Heal argument, I have to agree with Galz.

My biggest issue personally with stacking crit is that even if you have 50% crit, its still not RELIABLE. As a healer, you DO have to worry about overhealing. While crits do boost HPS, I primarily view them as a means to regain mana on the long term, because +healing scales better and is reliable.

You dont have to worry about overhealing on crits.

Anyone have any numbers for the BoL changes, so we can try to model them and get that information up?

crit vs. healing hmm?
Whats up with "if you dont crit MT dies"? Thats just bullshit :P
Just go with healing to 2100/2200 raid buffed, that should be enough for anything.

Well Fanvast, you've asked that question at an interesting time. The past page or so of posts have been debating the feasibility of a HL based build after 2.3 hits, due to the downrank penalty now added to BoL, as well as the nerf to Tier 5's set bonuses. Considering you have mostly Arena gear which has high amounts of crit, I would still say that your best bet is to use Holy Light primarily, but you are going to have to stop using lower ranks that were commonly used before (such as 4 & 5) and start using higher ranks (say 7 at the lowest) and cancel a lot.

At least, that's my view.

With low rank HLs no longer being efficient, FoL should be the most efficient, in which case I really don't see why you would not use FoL in a situation where it provides enough HP/S and only swap to HL when more HP/s is needed. Remember someone with high crit has a -% reduction to both HL and FoL mana cost so even if there is a HL rank that's just as efficient and provides just as much HP/s as FoL, your chance to crit will really not affect this.

HL getting 6% crit from talents doesn't make it better to stack crit on it. The only thing that improves the value of crit is casting more spells and/or more expensive ones. Using a cheap max rank FoL or a cheap low rank HL benefits the same from extra crit assuming you picked an equivalent rank of HL.

Of course if FoL's HP/s isn't sufficient for the fight, you're naturally wanting to cast more, higher ranks, and thus crit becomes better, but it's not the sheer fact of spamming holy light that makes crit good, it's the fact you spam high rank holy lights that makes it good. Crit is just (roughly) no better for low rank HL than it is for FoL.

Expanding on what Galz said, because your gear has high amounts of crit, if FoL isn't sufficient (as you said it generally isn't in your post) then your best bet is to move to higher ranks of Holy Light to get the most out of your gear.

I don't see how gear has to do with it. When you need more HP/s, you use max rank holy light no matter what your gear is or your raid will die... And if you don't need high HP/s no point wasting mana on high rank holy light so you switch to something more efficient with less hp/s.

You might want to add in [Tome of Diabolic Remedy] as a viable trinket selection for 2.3, so everyone knows about it. It's definitely one of the nicer choices available, in my opinion.

According to standard rules of probability, if your %chance to crit is <50%, your odds of crit'ing are better the fewer spells you cast. In fact, if you cast more spells, your ratio of crits to non-crits will likely be worse.

In other words, unless I am mistaken here, you're better off casting fewer and larger heals and AVOID the consistency of your crit% until you exceed 50% chance to crit.

The extra 6% seems like gravy, as the math already favors HL in this regard at crit percentages lower than 50.

Again, unless I am mistaken?

I've already added it to my gear spreadsheet, I just need to figure out about how much passive +healing the on Use part of it is worth overall.

edit: Going off of the math from the Healing trinket thread, I'm going to use the formula D/C*H. Thus giving the ToDR a passive +66 healing to go along with the 18mp5.

Not to be offensive, but I don't see how rules of probability have anything to do with your post. If you crit 40% of your heals, 60% of 40% of the mana you spend will be returned. If you crit 50% of your heals 60% of 50% of your mana will be returned. All on average of course (but when you cast many heals in a real fight it tends to be close to the average rather than far from it, and anyway if you cast just a few heal your (weighted) chance to be under the average is the same as being above it).


TBH having more crit does increase the value of additional crit value, but at least since the patch it's really not that much:

20% crit (before talents) paladin with 1800 healing and 90 mp5 spamming super mana
pots in a 10 minute fight from the spreadsheet using all his mana on holy light max rank (spreadsheet not updated for 2.3 afaik for lower ranks) is 469,171 healing done. Add 1% crit and you have 5580 more healing done, or 1.19% more healing done. Will need about 72 +heal to get the same benefit.

Bump that pally to 40% crit before talents, and he'll do 599,072 healing. Give him 1% more crit and it'll add 7727 healing, or 1.29% more healing. Will need about 78 +heal to get the same benefit.

So as you can see while crit does give you more when you have more crit, the difference
between the gain of a 20% pally and a 40% pally is rather small. Therefore the difference in the effect on the 6% crit on holy light VS flash of light is even going to have a smaller effect on the value of crit.

Now let's change that pally to do FoL spamming:

With 20% crit he would heal for 703,071 and gain 8212 healing from 1% crit, or 1.17% increase. Would need 40~41 +healing to get the same benefit.

With 40% he would heal for 892,861, and 1% crit will add him 11,199 healing, or 1.25% more. Would need ~43-44 +healing to get the same benefit.

As you can see the gains from crit for a HL paladin are marginally smaller than for a FoL paladin, and the gains from extra crit increase only marginally when you have a LOT more crit as a starting point.
The only reasons for a FoL paladin to *not* stack crit is that he just gains that much more from additional +healing compared to a HL paladin, and generally spends less mana to heal the same amounts of HPs so gains less from crit. That doesn't mean high crit should automatically point at HL, as you can see both at 20% and 40% crit FoL is significantly more healing over the course of the 10 minute fight, which also leads me to saying that the only reason to cast HL over FoL is if you need burst HP/S *now* (aka your target might die if you try to flash it up) rather than raw healing over the course of the fight.

No offense taken. Essentially, this deals with how to handle unfavorable odds.

Assumption: Crit% chance less than 50%.

In this scenario, the odds are stacked against you getting a crit on a given heal. As such, the only way to "beat" the odds, or to get an unconventionally high percentage, is to reduce the number of casts. To capitalize on the fact that you're casting less, you want to gamble more mana on the few casts you do, in fact, make. HL clearly favors this strategy.

That's how probability relates. The more casts you make, the closer to the average you'll get--the fewer casts you make, the closer to the extremes you'll get. The danger is, of course, that you will fall to the opposite extreme more often than to the favorable one. In other words, taken from a probability table:

HL:
"You get unlucky and get few crits." is a non-negligible possibility.
"You get lucky and get many crits." is a non-negligible possibility.

FoL spamming:
"You get unlucky and get few crits." is a non-negligible possibility.
"You get lucky and get many crits." is a negligible possibility.

The only realistically conceivable way to beat the house, therefore, is to cast the biggest heal you have as few times as possible.

The question becomes, is the possible gain from crit luck worth the loss of efficiency of FoL spamming (which is admitted)?

1. In any given fight the amount of heals you'll be casting is large enough for the variance in %crit to be rather small, and while sometimes noticeable, on average it'll deviate equally both in your favor and against you, regardless of if you cast 1 spell or 1 million.
Casting a heal every 4 seconds in a 5 minute fight will result in 75 heals. Say you have 40% crit chance, you will crit on average a total of 30 of those heals. The variance of this total will be 70*.4*.6=16.8, which means you will likely not see your # of crits go up or down by more than 16.8^0.5=4.1, and extremely extrelemly unlikely to have it go up or down by more than ~12. This is with taking rather small number of heals, in most fights you will heal a lot mroe than that wether you want to or not, both due to length and hp/s needed, making the relative difference in # of crits even smaller. When you cast enough spells, the variance of your crit % goes down by a factor of N (number of casts) and the difference you'd expect seeing thus goes down by a factor of N^0.5.

When you cast less spells total, you have a higher chance of getting 0 crits, as well as a higher chance to get everything crit (with any crit chance)! On average though you get the same benefit from crit as if you would've casted 1 million heals. Therefore casting more heals or less heals doesn't statistically favor any kind of crit ratio. The only thing that makes crit stronger when you cast a lot is the fact you're spending more mana and thus getting more mana returned, especially if it's using holy lights that spend more mana per second as well as spend more mana per healing done.

Just a little HP/s VS HP/mana comparison I've ran using the spreasheets (1 for HP/s and another for amount healed over a fight with your mana) posted around these forums:
We all know FoL is more efficient and if it provides enough HP/s to keep people from dying it should be used, but then again you just have to holy light sometimes, significantly dropping your efficiency in order to boost HP/s.
I went and did something similar to what people do when comparing +heal to regen when downranking - checked how much +heal you'll need to make FoL as much HP/s as holy light, and checked how much mp5 you'll need to be able to have as much HP healed with holy light as you would normall heal with FoL.

Base assumptions: took 1800 heal, 20% spell (not holy) crit, 90 mp5 and 500 int gear, no shadow priest, 10 minute fight with mana pot on cooldown, full buffs/flask/food/oil but no party buffs. You can run it with other setups but my guess is your results will not differ greatly except possible if you add shadow priest as it's the only thing you could possibly change that would significantly tweak the results (as it buffs only your mana pool and nothing else, while gear upgrades even if big will most likely buff all of your stats and have little effect on the results).

Results:
FoL needs 860 +healing to match holy light in HP/s.
Holy light needs 216 mp5 to match FoL in healing done in that 10 minute fight.

Divide the 860 +healing by 216 mp5 and you get almost 4 +heal per mp5. I would say that means that for each 4 +heal you get, you're basically getting 1 mp5 through not needing to holy light as often. Granted 4 heal will rarely make the difference HP/S-wise, but so will 1 mp5 on your mana pool... At the end, on average, 4 +healing gives you "hidden" 1 mp5 by allowing you to FoL more often, on top of the overall boost to your mana efficiency. Of course if you count that "hidden" value of +healing, you need to ignore its "burst" effect as the 1 mp5 is gained because you're still using the same hp/s.

Basically this gives another way to look at HP/s vs HP/mana. I wonder how useful would it be to look at it this way, as it still doesn't take into account the "with higher +heal I have higher max possible burst when needed" factor, but it definitely bumps +healing up when it comes to raw efficiency style the spreadsheet (which can be found at [Paladin] Values of Int/SpellCrit/Mp5/+Heal in 2.1.0 )

If I didn't make any huge errors and if my approximations aren't totally completely worlds off, +healing is by far the most effective stat to stack for paladins in terms of item budget, with even the 18 healing gem overcoming the 9 heal 2 mp5 gem in terms of efficiency.

What could make my point irrelevent: If you're already lacking HP/s when spamming holy light, you don't really gain mp5 as you need the extra HP/s, however that still means you're going to be stacking +healing... If you're already spamming 100% FoL then additional +heal will not make you cast HL less which makes the gain (beyond what's shown on the spreadsheet, as in the extra 1 mp5 per 4 +healing) non-existant.