Good Afternoon! Morning…? Evening…?
Good Day! I am here to discuss and dispel a very common misconception about a certain strategy that existed prior to Mercy’s rework:
“Hide and rez was common”.
As some of you might already know, I have already covered the effectiveness (or lack thereof) of the aforementioned strategy in overwhelming depth in other locations on the Forums. Likewise, I have also highlighted the underlying cause of said tactic in those same locations. Considering that I have beaten those topics to death, I see no reason to further discuss those in this specific thread unless prompted to do so in the replies. Instead, I will be discussing the results of these two factors on in-game applications as indicated by the statistics.
Shall we begin?
First, let’s lay out some hypotheses that we can test. These hypotheses are derived from the assertion that hiding prior to the execution of Resurrect was common. The first hypothesis is rather obvious, as it is quite literally the aforementioned assertion.
Hypothesis #1: Hiding to use Resurrect was a common strategy before Mercy’s rework.
The other hypothesis that we can extract from this assertion is not directly stated. However, considering the argumentative context in which the assertion is most often stated or implied (including the Developer Update), along with the implications of of “was” in the assertion, it is reasonable to conclude that the second hypothesis is included in said assertion.
Hypothesis #2: Hiding to use Resurrect is no longer a common strategy after the rework.
With these two hypotheses in mind, we can progress to what symptoms these hypotheses would produce if we were to assume that they are both true.
I’m about to launch into a paragraph in which virtually everything I say is just me stating the obvious. However, there is reason in my doing so.
“Hide and rez” was a tactic in which the Mercy player, after charging their ultimate ability, would disengage from the two teams, leave the combat area, and go off somewhere to hide, ensuring their personal safety. During this period of hiding, they were not contributing to the fight, as that would reveal their position and thus contradict the notion of “hiding” in the first place. Mercy’s most common and most effective method of contribution to the team was her healing stream, which was most needed during teamfights. As such, when a Mercy player hid to use Resurrect, they were not contributing to the fight, and therefore, they were not healing.
I’ve heard a variety of claims on how long this period of hiding would typically last. Some say 15 seconds. Others say 20. I’ve even heard a few ridiculous claims of 40 seconds [Citation]. Regardless, it is a significant amount of time to not be participating, and therefore not healing, especially considering that teamfights are where Mercy would be busiest when it came to healing.
So let’s operate under the assumption that both of our hypotheses are true. What do the above two paragraphs mean?
Anyone who has seen some of my other threads knows that a post of mine on this scale is never complete without some numbers… so there’s a hint.
If our two hypotheses are true, it would mean that Mercy’s average healing prior to the rework would be significantly lower than Mercy’s average healing after the rework. If a Mercy player is hiding, they are not healing, and therefore their drop in the bucket of global healing averages is smaller than which it would be if they were not hiding. If our first hypothesis is correct, then enough players would be abusing this strategy to cause a deflation in Mercy’s global healing averages. If our second hypothesis is correct, then Mercy’s healing averages are/were no longer deflated after her rework.
Inversely, if our first hypothesis is incorrect, then Mercy’s average healing dealt before the rework was not deflated, as the vast minority of players who used “hide and rez” would have a negligible effect on the healing averages. If our second hypothesis is incorrect, then Mercy’s healing rate would still be deflated post-rework.
If only we had a way to test these hypotheses… Oh wait… We do.
I just so happen to have a bunch of data that I recorded off of Overbuff on different dates for Forum purposes saved somewhere. Some of it is from before the rework, some is from after the rework. Those data points and their respective forum posts for which I used them are shown below. Feel free to check the citations to verify that those data points are actually from the time periods I claim them to be from. Considering that these data points were almost always used to make a point through adversity, a quick fact-check from the people I was debating with at the time would have disproved any incorrect data points if they were far off the mark. Some data points were not posted on the same day they were recorded because… writing massive posts takes time.
Note that all post-rework data shown below was also recorded before Mercy’s healing nerf went live.
- Mercy’s average healing/game as of June 28th, 2017 (pre-rework): 11905. [Citation]
- Mercy’s average healing/game as of August 12th, 2017 (pre-rework): 11912. [Citation]
- Mercy’s average healing/game as of February 18th, 2018 (post-rework): 11926. [Citation]
- Mercy’s average healing/game as of March 21st, 2018 (post-rework): 11926. [Citation]
- Mercy’s average healing/game as of May 31st, 2018 (post-rework): 11603. [Citation]
- Mercy’s average healing/game as of June 30th, 2018 (post-rework): 11752. [Citation]
- Mercy’s average healing/game as of July 22nd, 2018 (post-rework): 12132. [Citation]
- Mercy’s average healing/game as of August 7th, 2018 (post-rework): 12085. [Citation]
Writing those out so it’s easier to look at in context…
So… Where is the big increase in healing dealt that we expected? If “hide and rez” was really a common strategy, then surely we would see some drastic improvement in Mercy’s average healing rate, right?
But… we saw a peak increase of about 200 health/game…
And a trough decrease of about 300 health/game.
Something doesn’t add up.
Numbers don’t lie or fall victim to misconception. A person can.
Our set of numbers can’t coexist with both of our hypotheses. At least one of our hypotheses must be incorrect.
Of course, this isn’t the first place I have discussed this contradiction… I’ve discussed this on Twitter before, and came across an interesting counterpoint. Someone replied to me, saying that there was a problem with my conclusion based upon the data. Just before Mercy’s rework hit, changes were made to Control maps in Competitive Play. These changes turned Control victories from the first team to win 3 rounds to the first team to win 2 rounds. This would shorten the duration of Control matches, and ultimately decrease Mercy’s average healing/game.
I considered their argument. In other words, I sat down and did some math.
The changes to control maps meant that the longest Control match was 3/5ths as long as it used to be (maximum of 5 rounds reduced to 3), meaning it had a reduction in duration of 2/5ths. Likewise, the shortest Control match was reduced in length from 3 rounds to 2 rounds, a reduction in length by 1/3rd.
For four of the six post-rework data points, there were 18 maps total and 4 Control maps. I decided to be generous to arguments of the person I was debating with and include the first two post-rework data points, when there were 17 maps total and 4 Control maps. I’d just pretend that the additional non-Control map was irrelevant. Thus, the probability that a player plays on a Control map is 4/17ths.
I decided to be even more generous and assume that Mercy was picked as often on Control maps as on non-Control maps, contrary to the fact that her actual popularity was lower on Control maps than on non-Control maps.
With these numbers in mind, we can calculate the expected reduction in Mercy’s healing averages.
The high end of healing reduction we can expect from these changes: (4/17)(2/5) ≈ 9.41%.
The low end of healing reduction we can expect from these changes: (4/17)(1/3) ≈ 7.84%.
Which, if we multiply these by Mercy’s pre-rework healing averages, we would expect a drop in average healing between 933 HP and 1121 HP from the Control point changes… a drop that is mysteriously absent from Mercy’s healing numbers post-rework.
There must be a reason that the numerical drop is absent. That reason must be the use of a certain ability… An ability that influenced Mercy’s healing averages, an ability that, quite notably, was and still is hated by many…
Did I have you fooled for a second?
Anyway, we can say without a doubt that Valkyrie increases the healing Mercy outputs. You can’t grant a hero an ability that increases their healing potential fivefold for a period of time and seriously expect that hero’s healing averages to stay where they are.
But that raises a question: Just how much healing does Valkyrie deal? When I replied to the argument about Control points on Twitter, I used hypotheticals that were hilariously low compared to Valkyrie’s maximum numerical potential (IE: Valkyrie only increased Mercy’s effective healing rate by 1.5 times for its duration) to disprove the argument.
But I’m not one for half-measures, and hypotheticals can be… inaccurate. Upon finding a way to work the math without using said hypotheticals, I decided I would do so. Hypothetical models become meaningless if contradicted by reality.
How am I going to do this, you might ask? I need a way to isolate the additional healing dealt in Valkyrie from the healing dealt by Mercy’s base kit, including the healing that would have been dealt without Valkyrie.
Thankfully, the Patch notes have given me exactly that; the last time Mercy was changed, Valkyrie’s healing output was increased by 20%, and it’s charge requirement was reduced by 16.67% (contrary to what the patch notes would have you believe). Now all I needed were some monthly data points to work with from before the Valkyrie buffs. I have a single weekly data point from the week after the healing nerf recorded, but in regards to monthly data points… I have no such data recorded.
But someone else does.
I dug through the Forums and found this post, in which someone reported that Mercy’s average healing for the past month according to Overbuff (the same source as all of my data) was approximately 10700.
I can provide the second data point. Today (January 19th as of the time I am typing this), Mercy’s average healing dealt over the past month is 11128.
Because Valkyrie was the only thing that changed between the two data points, we can safely assume that the Valkyrie changes lead to the increase in healing averages. In other words…
ΔAverage healing = ΔHealing added by Valkyrie in this patch
We also know that the amount of healing added by Valkyrie in this patch is a function (specifically a multiplier) of the amount of healing that Valkyrie already added to the average before the patch:
ΔHealing added by Valkyrie in this patch = (Healing output increase)(Healing added to average by Valkyrie before the patch)
ΔAverage healing = (Healing output increase)(Healing added to average by Valkyrie before the patch)
- The change in average healing is 11128-10700 = 428 health.
- Valkyrie’s healing output was increased from 50 to 60, a 20% increase.
- Valkyrie’s charge requirement was reduced from 1950 to 1625, meaning that Valkyrie charges 20% faster (new charge time is 83.33% of the old charge time, 1/.8333 = 1.2, or a 20% increase).
- Combining the 20% increase in ultimate charge rate with the 20% increase in healing rate gives us a 44% increase in healing dealt by Valkyrie (1.2*1.2 = 1.44, or a 44% increase).
428 HP = (0.44)(Healing added to average by Valkyrie before the patch)
Which simplifies to:
Healing added to average by Valkyrie before the patch = 972.73 HP.
That’s (approximately) how much healing Valkyrie was adding every game when Mercy had 50 health/second of base healing, 50 health/second of Valkyrie healing, and a charge requirement of 1950. I also know that after the most recent patch to mercy, the healing increase provided by Valkyrie went up by 44%. Therefore, I can predict that the amount of healing dealt by Valkyrie after the patch is 1400.73 HP, which we’ll just round up to 1401 because that makes the number prettier.
So what does this have to do with the healing done by Valkyrie before Mercy’s healing nerf? Well, Valkyrie healed at the same rate back then as it does now. It also charges at virtually the same rate, due to the nerf to Mercy’s base healing. Before the healing nerf, Valkyrie had a minimum charge time of 36.8 seconds through healing. Today, Valkyrie has a minimum charge time of 36.1 seconds through healing. Therefore, we can generalize the additional healing provided by Valkyrie today to the additional healing provided by Valkyrie before Mercy’s healing nerfs.
That number happens to more than counterbalance the amount of healing we expected to lose from the Control map changes. With the Control point changes and the introduction of Valkyrie, we would already expect a net increase in average healing dealt, if we expect any change at all.
So what does this mean in regards to “hide and rez”?
The equation of pre-rework healing averages vs post-rework healing averages already has difficulty working out without “hide and rez” in the picture. Without “hide and rez” in the picture, the numbers make enough sense that we can say that the missing 200-400 health from the post-rework numbers is just due to variation. If we assumed our hypotheses to be true and tried to factor in “hide and rez” to the equation, the numbers would stop making sense because reality would differ severely from our numerical assumptions.
To be blunt about it, there simply is no room for “hide and rez” in the equation. Our hypotheses cannot coexist with the numbers we have. Our assumptions are mutually exclusive to that which we know to be true. Therefore, our assumptions are incorrect.
Considering the negligible difference between the healing averages, Mercy’s pre-rework numbers were not deflated. The reality of the situation is that the employment of “hide and rez” as a tactic was so exceedingly rare that the vast minority of situations in which it was employed had no impact on Mercy’s healing average. If it were common, as the uninformed individual would like to claim, we would have seen a major increase in healing dealt between the rework and the healing nerf.
We saw no such increase. At least one of our hypotheses is incorrect, and just one of them being incorrect invalidates “hide and rez” as a reason to rework Mercy from 1.x to 2.x. In other words, either the rework failed, or it was pointless in the first place.
But that’s not something we don’t already know.