Sunday, July 06, 2014

DPS Feedback Idea - Historical Meter

As you know, I am of the opinion that the reason of lot of DPS players play badly is not because they don't care, or are innately bad. Instead it is because they lack the required feedback necessary to improve.

Currently, the best tool for feedback is DPS meters. But while DPS meters work, they are a very blunt instrument. They don't account for differences in gear, or fights, or even tell what number you should be aiming for.

I think a better DPS meter could be made, but it would probably require the game developers to implement.

Currently, DPS meters compare you to the other players in the current fight. It would be better if the DPS meter compared you to the overall historical performance of people with your item level.

Let's start by recording everyone's performance on individual boss fights. Note the boss, DPS done, and the item level of the player. Once you aggregate all the records, you can tell for a given boss and item level, what the top DPS was, or what the median DPS was.

That gives you a target number. If the top DPS on Boss A at i500 is 10k, you can tell the player after Boss A: "You did 6k damage. The top DPS was 10k." That is concrete feedback. The player can't blame her gear or the fight mechanics.

Of course, using the top DPS mark is probably bad, because it would be a very lucky parse and probably individuals doing something excessive to hit that mark. A better target number would be something like one standard deviation above the median. Or possibly target the range between the median and one standard deviation.

The advantage of using this mechanism, which looks back at the history of all the people doing the fight is that it nullifies variables and fight mechanics. Because the amount of data collected is large, a few lucky parses or exceptional players do not skew the results. It provides a viable target number that people know for a fact is within the capabilities of the class and gear.

As well, this doesn't necessarily involve the entire raid. You aren't being compared to other people you know, but to the entirety of the WoW community.

If feedback is vague, you can always make excuses as to why you don't need to improve. For the DPS to improve, they first need unequivocal proof that improvement is necessary. This Historical DPS Meter would provide that feedback.

13 comments:

  1. Except that number will be skewed on some fights due to meter padding and thus someone doing the fight correctly will be told they're doing poorly.

    For example, people attacking Horridon with damage taken debuff while adds need to be killed. Multidotting all three Paragons. Ignoring MCs/adds to tunnel Garrosh. Etc.

    ReplyDelete
  2. That sort of thing would only work for PvE boss fights. In PvP it would be even less useful because of the ebb and flow of the game itself.

    Still, in a limited situation (boss fights) it would help on things such as rotation.

    ReplyDelete
  3. If you are comparing it to some average (or median) value, doesn't that mean that half the players will naturally fail to perform? It also wouldn't be able to account for the number of times you have fought the boss. Someone on their 15th kill should be doing more than someone on their first.

    ReplyDelete
  4. Adding to the problems above: it must also take into account the raid composition. If the rest of your raid has higher level the fight is shorter, so the CD uptime is higher. This is why when you look at rankings to see how you perform you must look for people with roughly the same gear, a fight duration which is approximately the same and you must compare damage per target to make sure you're comparing with someone who has the same assignment as you. Overall it's a lot of work for something which is probably not worth it. When I look at my raiding (and myself in particular), the main problem is not really dps but survival.

    ReplyDelete
  5. Even a system-level historical meter is just as blunt, for much the reasons Balkoth said.

    Very few fights are just standing there and wailing on the boss. Very few are even moving around and wailing on the boss. Most of the fights involve engaging mechanics in ways that just don't get shown up in a pure "damage done" tally. What does a historical DPS meter tell you about how well you handled Immerseus or Galakris, for example?

    ReplyDelete
  6. @Balkoth, I think you are underestimating the Law of Large Numbers. Meter padding, unusual situations etc are outliers, and will get swamped by sheer numbers.

    @Aelos, Don't forget that this is sorted by item level. On your 15th try, you will have a much higher item level than your first try, and so your performance will get sorted into a different bucket.

    @Helistar, again, Law of Large Numbers. Yes, unusual raid comps will produce higher numbers. But unusual raid comps (in the other direction) can produce lower than expected numbers. That's why you target the median to first standard deviation as your goal.

    @RJ, the idea is fight-specific. You would only be comparing Immerseus versus Immerseus performance.

    Finally, I think you guys are focusing on the wrong audience. The target audience for this are the players who don't know that they are playing badly. The ones who don't hit cap, who don't use DoTs, who don't use the right rotation.

    To improve, they first need to know that they can improve. This historical meter gives them a target number, something they can work towards achieving.

    ReplyDelete
  7. I think perhaps instead of a historical DPS meter, you instead have a comparison to the average required DPS for a boss.

    To elaborate, I assume that when Blizzard designs a boss, its health and its enrage timer, they determine some level of average expected DPS that is required to down the boss in time. A comparison to this number might be more effective for unaware underperformers.

    ReplyDelete
  8. Law of Large Numbers. Yes, unusual raid comps will produce higher numbers.

    But this goes against what you want to obtain. What you say is that you will get very accurate average numbers, so the comparison will be useful only if your raid composition is the same as the average. If you are in one of the tails it's automatically useless. And it's the same for assignment: you need to compare your dps to people who are doing the same assignment, otherwise it's useless as a tool to tell you how well you're doing.

    What you are proposing is no different that what can already be done with raidbots. And it performs much worse than simply having someone with the same class/spec in your raid and comparing to him.....

    ReplyDelete
  9. I think Nyo is on a better track here. For my raiders, I'd often run some calculations to say, "Hey, this is the minimum raid DPS we need to ht enrage, so this is what each DPSer needs to be bringing. Get on it!" And that was generally sufficient information to get people to improve.

    If Blizzard provided a minimum guidepost, basically doing the calculation for me rather than me having to do it myself, it gives a minimum performance rating and ergo, an easy benchmark to compare yourself to. Still falls apart to a certain extent for fights where someone has a job that's unorthodox.

    ReplyDelete
  10. Rohan: My argument doesn't matter if it's comparing Immerseus to Immerseus; the point actually matters to exactly that!

    With how many gimmicks are in each fight, how is someone supposed to know what "Historical DPS for Immerseus: 1000 DPS" actually means? Does it mean the average is being great on the boss, and not understanding the blobs? Does it mean you're a great AoEer of the blobs?

    How do you determine what the historical average is for fights like Galakris, where you have different distinct roles for players. You can't just split it between Tower or Adds, and merging the two groups produces a number that's meaningless to either. Blackfuse is a similar way, because on higher difficulty levels you need players dedicated to dealing with the fight gimmicks.

    And even if you could solve all this, how does any of that tell a player how well they actually followed their role. A player could be flailing, and yet still meet the "target" because of a variety of reasons, ranging from overgearing to ignoring fight mechanics.


    What you're asking for is not something you should be looking for in a raid system. If you want a tool that tells players they're doing their rotation or gearing poorly at a system level, you should be looking at the Proving Grounds, and suggesting that it be made class and spec specific.

    ReplyDelete
  11. "@Balkoth, I think you are underestimating the Law of Large Numbers. Meter padding, unusual situations etc are outliers, and will get swamped by sheer numbers."

    Did you ever actually do Horridon on LFR? Like 75% of the raid used to ignore the adds and tunnel the boss...which means you have a very accurate and very large dataset of people tunneling boss (who takes extra damage) and ignoring adds.

    Or, as mentioned, how about someone doing belts on Blackfuse? Or helping the tank kill shredders (who take 80% less damage)?

    ReplyDelete
  12. Many of these complaints could be addressed with a few other stats, such as "Average damage taken".

    Beside that, I suspect that once someone gets the idea that they can improve in one way, they will find other ways. Getting that initial idea is critical.

    ReplyDelete
  13. "Many of these complaints could be addressed with a few other stats, such as "Average damage taken"."

    Wouldn't address any of the ones I raised.

    ReplyDelete