Combat Logs Scores

What exactly do some of the scores mean on our site? Well, I’m here to shed some light on them.


Comparison to Successful Kills

Even though every attempt shows a percentile, you’re only being compared to other player’s stats on successful kills. Why is that? Because we only want to compare you to what actually… works! This applies to all comparisons, like personal DPS metrics, team metrics, and even boss mechanics on the infographics.

Percentile by Item Level

This shows how well you performed compared to other people of the exact same item level AND item level for your spec. Previously, there were ‘buckets’ but that lead to people dismissing the number if, say, they were at the bottom of the bucket. I know I did… And since we want this number to be meaningful, we now give you the exact percentile for each item level.

This shows up in 2 spots: the data view and the infographic view (both shown below).

Our SV Hunter, Suffokate, was better than 95% of all other iLevel 673 SV Hunters. He was better than 88% of all other players, regardless of iLevel or spec.

iLevel_DataView iLevel_Infographic

Percentile for your Team

There are 2 interesting numbers that really stand out for me. The team metric for “Is that good damage?” compared to “But in it good for our gear?” Let’s look at these first, then I will explain why I like them so much.

Is that good damage?
This number shows how well we do compared to teams killing the boss. It isn’t taking gear or anything else into account. It’s a straight up comparison of our DPS (whether a wipe or kill) to teams who’ve killed the boss. This is a good DPS benchmark to see if our DPS is good enough to kill the boss ourselves.

It might seem like our team isn’t doing well, ranking only at 42%. But keep in mind, that means we are doing better than 42% of teams who have killed the boss. So if we are wiping, DPS isn’t the main problem, it’s likely some other mechanic.


But, is it good for our gear?
This metric shows how well your team is performing for their gear level. It’s taking everyone’s individual item level & spec performance (like 95% for suffokate in the example below), and averaging them together. Our team is doing very average (50% is the median). Suffokate is in the 95th percentile, so that means we have a few people much lower than 50%, lowering the average.Scores_GoodiLevel

Comparing these 2 percentiles:
These 2 numbers tell me a story. If the first number is higher than the second number, it means you are outgearing a fight (your success is being helped by higher levels of gear). If the first number is lower than the second number, it means you are undergeared but your skill is helping you kill bosses.

In my example, we currently fall into the ‘undergeared’ category. Our players are performing at 54% for their item level and spec, BUT we only rank at 42% overall. That means we are less geared than a lot of teams who have killed  this boss.

What does 50% really mean?

You’re smack in the middle, that’s what it means. It’s the median (not the average). I’ve run into a lot of people who see 50% and think it means they are bad. It doesn’t mean you’re bad, in fact, quite the opposite. It means you’re doing the same as most people killing the boss.

50% is a great benchmark for what most people are doing to kill a boss.


On our site, we color-code the infographic circles, as shown in the picture above.

At 50%, that means you’re doing what most people are doing who kill the boss. This is true for all percentages on the site, from DPS to boss mechanics. If most people take 10% of their damage from rocks falling from the ceiling, and so do you… it doesn’t mean you’re bad, it means you’re doing what most people are doing when they kill a boss.

Sure, you’ll get people who argue that you should be perfect. We all want to improve, but you don’t need perfection to kill a boss. Mythic has less wiggle room obviously, but each set of statistics is based on the difficulty you’re doing. So if 50% of mythic players are taking 10% of their damage from rocks falling from the sky when they kill a boss… yep, it means you don’t need to execute that perfectly to get a kill 🙂 Mr. Robot isn’t here to judge you, here’s here to put things into context and help you.

Other scores

Sometimes you will see a score out of 10, instead of a percentile. It’s the same thing, really. I particularly like the scores in the overview section, when I first load an infographic. On the right, you can see quickly how well your team did on each mechanic (they are sorted from best to worst). This is corresponds to the colored borders around the navigation at the top.

For example, we scored an 8.3 on Shatter (notice the mostly filled in green border on it at the top too). But we didn’t do so hot on avoiding damage… scoring a 1.6 with a barely filled in, orange border at the top.


 Still new to Infographics? Learn more here.


  1. OMG! i’m on the blog! w00t! 🙂

    Thanks for the write up! it really helps explain some of the figures.

  2. As a data analyst, I’ve been pretty impressed by these infographics, myself. I feel you folks are really onto something here with the smooth way you’ve presented this information.

    I don’t know your sample sets, but are you confident that reducing the ilevel buckets to an effective bucket size of 1 ilevel in the comparison is pretty fair, based on the data you have?

    (Aside: I mean, I presume you’ve narrowed the bucket width to 1, either by flooring significant digits after the decimal or rounding them, rather than making a comparison at some nth decimal).

    Are you seeing a pretty smooth correlation at each “discrete” ilevel in terms of performance? Or are there a lot of cases where a person may be at, say the 50th percentile at their ilevel, but compared to 1 or 2 ilevels higher, they might be at the 60th percentile — just due to not having enough samples (or better/worse players at different discrete ilevels)?

    Basically — I’m just asking if any kind of skew is minimal, based on the amount of data you have, from narrowing your ilevel bucket range.

    • I’m sure there’ll be some sort of skew; Compare someone with 4 tier pieces + high ilvl weapon to someone with no tier pieces, and a low weapon. Or someone that’s gotten unlucky and hasn’t got many gear pieces with ‘good stats’ on it.

    • So far it looks good. There are a ton of data points, and that’s one of the reasons we waited to reduce the sample to 1 iLevel – to make sure the sample size was big enough 🙂

      And thanks for the compliments, we really appreciate it, especially from a data analyst!

  3. I disregard all of these silly arbitrary scores.

  4. Hi,

    It’s Haaggis here from @failcapped.

    I’d just like to mention that I don’t think healers should be made to pre-pot like dps are on your info-graphics.

    It’s pretty much fight dependent but i know quite a lot of the fights start of with little or no healing to be done.

    Just a thought.

    • We don’t ding healers or tanks on not prepotting. The scores only assume combat pots should be used.

    • I would argue that both tanks and healers should. Tanks because it’s both a survival and dps improvement. Healers because why not? They’re dirt cheap and will help them do a little extra healing/damage/whatever they do for the first 20ish seconds. Heck, I shield-tronic shield “pre-pot” as well lol (it counts as a health pot so can do both)

      Maybe that’s too tryhard for s penalty on the infographics page, but no reason not to do it.

  5. Cesar Julian Papagiani

    March 4, 2015 at 10:29 am

    what a big load of crap!
    never trust mr. robot.

  6. Hmm… that seems really awesome and in many points better, that warcraftlogs looks like. Is it possible – just like in warcraftlogs – to add a replay button (If yes, i won’t need warcraft logs anymore) and can Guildies look for the Guild?

    I’m our raid-lead and i upload our logs on warcraft logs pretty often, and my guild likes to look at the calender structure to see how they perfomed. Is something like that implemented as well?

  7. umm butterfly effect anyone? So the Carried teams will rule out this way. This essentially slams all toons specs and classes into the majority or bosskills but not by individual stats If the tanks and heals carried the group with a few dps you still have shitty dps stats being recorded and those in turn go into the pool of oh yep this is best… but its not best… its a majority got a boss kill with this good luck with your group being even close to the majority. This now make mrrobot lose usefulness as a stand alone raider optimization

  8. Артём Брынцев

    March 15, 2015 at 9:29 am

    Товарищи разработчики, а русский язык когда на сайте будет?

  9. Hi there,

    First i would like to thank you for your tool, its really great.

    I would like to point out something concerning the “But, is it good for our gear?” percentage. You told us that : “This shows how well you performed compared to other people of the exact same item level AND item level for your spec” but is it related to the boss ?

    I think it is but as you didn’t mention it…

    Thank’s and sorry for my english mistakes !

  10. Hi there,

    First i would like to thank you for your tool, its really great.

    I would like to point out something concerning the “But, is it good for our gear?” percentage. You told us that : “This shows how well you performed compared to other people of the exact same item level AND item level for your spec” but is it related to the boss ?

    I think it is but as you didn’t mention it…

    Thank’s and sorry for my english mistakes !

  11. Not sure if this has been covered or not but Damage Reductions should really be included for heals, Especially Devotion Aura and Hand of Sacrifice.

  12. Jacobo Martinez Crespin

    September 27, 2015 at 4:00 am

    I’ve a question. i’ve been working with datasets of wow logs, but i would like to know, what is the most representative percentile? we can think that 50% because it represents a normal player but the majority of class/specs not to be rise up in the same way, for example, the increase of mage frost between (50-95)% percentil there aren’t the same that for example lock destruction, so…if we take 50% as reference maybe we were taking an imprecise value. I think that the best percntile is 95% because in this we can see almost the best performance on each class/spec where there are a community of players in this percentile. Why i dont take 100% percentile? because as 10% percentile has an anomalous data (for example, people that die in combat) 100% use to be a anomalous value that only 1 player has achieved.

    Could you give me your opinion about that?

    • 50% would be the most representative. One of the reasons they differ so much between 50 and 95%: that shows which classes are hard to master, or easier to master, for example.

      If you want to show what the masters of each class can do, then 95% is the right number to look at.

© 2019 Mr. Robot's Blog

Use the tools on AskMrRobot.comUp ↑