Wednesday, May 4, 2011

The Enemy is not Optimization

I think people mix two things up when it comes to fear of inefficiency and optimization:
The process of optimization and having an optimized character.

Let me say very clearly that the process of optimization is not only unavoidable, but actually desired in MMORPGs. The whole point of offering customization of characters on a power-level, is to encourage players to optimize; to engage in interesting, meaningful decisions. And I love it.

The problem with WoW and Rift is not this. The problem is that the process was removed; removed by crowd-sourcing. While I leveled, I tinkered around with my builds in Rift and I had a few nice ones. But when I was almost max level I couldn't resist my curiosity and had a look at RiftJunkies and the official forums. I was bombarded with highly optimized builds that, I have to admit, were sometimes significantly better than the ones I came up with. And then: game over.

I remember in classic WoW and sometimes TBC when a new item dropped I would do some calculations on my own about how much of an update it would be. I had done my own little theory-crafting; without tools for statistical analysis and without differential evolution techniques for numerical optimization. These rules of thumb worked nicely.

Then, when I returned to WotLK, this game was over. Instead of optimizing myself I started RAWR and let it run an (almost brute force) simulation of my gear. I found out that I wasn't so far off with my rules of thumb. But, of course, RAWR was better.

To make things worse, there were the raids that required me to specc in a different way, because they didn't understand the speccs, but just copied them. That probably was even a logical way to handle things for them. But it also destroyed my fun at optimizing myself even more.

So let me say this very clearly: The enemy is not optimization. The enemy is crowd sourcing. And the more tools you give the crowd (like export of combat logs), the worse the problem will become. Hey, even as a part of ElitistJerks this wasn't fun anymore! You would just do the statistical analysis and numerical optimization. As difficult as these may sound: once you have figured it out, it's boring.

This kind of post cannot go without this Extra Credits video.


  1. Well written but the Internet is here to stay. You will never again have choices that can be solved by mathematics.

  2. How much of this is inevitable and how much is due to the design of the game? To me the main problem is that MMOs can be solved using a spreadsheet. That isn't true of many other games:

    -Chess has been analyzed over decades using powerful computers, and no one's found a unique optimal set of moves. Many different opening moves are playable.
    -Poker has been studied intensively, and I don't think anyone's discovered a unique optimal betting strategy.
    -Magic the Gathering is regularly analyzed by the hive mind, and still a fairly wide variety of decks are playable.

    Part of this may be the difference between playing against the AI as opposed to other players, but part of it might be the game design as well.

    So the big question: Is there any way that combat systems could be designed so that they can't be solved by simple spreadsheet analysis?

  3. Tolthir, Kring, while the internet will not go away, you can make things more difficult. If MtG had a database of all matches ever played in every detail, people would use this database for optimization.
    This would probably yield astonishing results, but the normal player would now have to use these results. And if he wanted to engage in optimization (what MtG is actually all about!), he would now have to use statistical analysis.

    It is possible to make a calculation-problem too difficult too solve. Something Extra Credits don't mention (understand?), either. Ask any scientist: Most mathematical problems are unsolveable!

    The export of the combat log for statistical analysis is an extremely powerful tool. And it is possible to take it away!

  4. The problem is that MMOs try to be all games to all people.

    Back in the days of Adventure Games in the 90s you had to be really into them to solve the increasingly sophisticated puzzle. Any puzzle in another genre would have been so simplistic by comparison that a diehard adventure gamer would have mocked it.

    But that's ok because the people who liked that had their own niche.

    The firestorm on the Rift boards arose from an irate post by a player who wanted to be included in the optimal playstyle but did not herself want to have to optimise. She wanted to join a premade that needed a healer as a Justicar and they told her they'd only accept her with a healing soul.

    I actually quite liked the gearscore/link achievement pugs in Wrath. Why? Because before then you didn't get pug raids. Now you have a raid with a minimum standard but that's better than no raid. Could be hard to attain that standard but once you got it done you were set.

    Optimisation is a playstyle, and one which, in a game that aims to please a broad range of players needs to be partnered by an alternative. I'd like to see more open world raids, like Rift's raid rifts, where raid leaders, especially on pvp servers are motivated to grab every warm body if they want to beat the content.

  5. Disagree. Optimisation absolutely is the enemy -- it just creates a tedious pre-stage to actually playing and enjoying the game.

    Frex, look at Portal. No character optimisation in there. Or weapon optimisation.

  6. There are some differences with M:TG and WoW.

    M:TG is PvP, Raiding is beating a static script.

    M:TG depends a lot on "luck", randomness has a much bigger impact then in raiding. It's a different kind of optimization, you try to minimize the negative impact of "bad luck". In WoW you try to optimize for the best performance and if "bad luck" happens you call for a fast wipe. M:TG decks are designed to recover, WoW raids aren't as you can easily try again.

    In M:TG you do have a rock, scissor, paper system and if your opponent plays scissor against your paper you can't just log to an alt or use dual-spec. (Besides using a limited sideboard.) That puts a limit on your optimization as you need to create e.g. a "rocky paper" deck.

    There is no "single best" solution against an unknown opponent. But there is one against a known static script.

    And there are known "killer decks" and most people play the same few decks at tournaments. I'm not sure if that helps mine or your argument. :)


    About the combat log removal. How can you optimize what you can't measure? Science can't exist without measuring. Removing the combat log would remove the ability to optimize. Maybe you might think that you optimize but you would just be guessing.

  7. Kring, estimated guesses is what science is all about. If we had a database of everything, e.g. astronomy would be quite boring (but very successful at explaining the universe).

  8. Even M:tG is quite susceptible to optimisation.

    There's an optimal amount of land based on the casting costs of spells in your deck.

    I've played tournaments against people who died with a handful of shivan dragons and sengir vampires they couldn't cast (and that's without me even destroying their lands, they just didn't get the cards).

    I've played against people who draw land after land while my Black Knight kills them 2 life at a time.

    And there are a vast array of cards that are simply not tournament functional (Thallids).

  9. Spinks, to consider optimization the enemy is a legitimate opinion, I think. But that means to disable all character customization that has an effect on character power.

    I mentioned it as number(5) in my last post. Point is that for a lot of people, me included, optimizing their character is a lot of fun. But we want to do it ourselves. To (be forced to) copy/paste a specc is no fun at all.

  10. Nils: That's fair and I think I'll write a longer blog post on this subject to expand further. In particular I'm sympathetic to people's wishes to keep developing/ progressing their character but I think mechanical optimisation is the wrong tool for MMOs to use to do it.

  11. Great discussion. Just a couple thoughts:

    @nils--I'm not sure hiding the data solves the problem, because the hive mind will still figure things out by experimenting. What I was thinking about was making the game either deeper or more varied so that the game wouldn't be easily solvable even with all of the data (like chess, poker, mtg, etc).

    @Kring--Perhaps part of the solution is to make raids less scripted?

  12. Tolthir, I agree that hiding the data does not solve the problem, but it limits it. MMORPG-history is proof.

    Otherwise, see last post for the six methods to fight the inefficiency phoebia.

    This post is about the fact that for many people the optimization itself is not the problem, but rather the fact that other people do it for you.

  13. Its the same problem we have in real life. Too much information. Give people too much information about something and they'll do dumb things like convince themselves Vaccinations are dangerous.

    hell walking across the road is dangerous, eating something small enough to choke on is dangerous....(LMAO) Just not as dangerous as sitting home starving to death.

    The people that cause the most problem in wow are the ones that would suggest in real life we blend all our food because it's more efficient than chewing it. And safer too.

    Games are more fun when you can play them without being told how to play. that's what crowd sourcing took away.

  14. I don't think the best approach would be to make crowd sourcing the enemy. Hiding data would be frustrating, people want to know how they are improving.

    I think the answer really relies on implementation and variability. It's one thing if getting stronger is dependent on one formula, one stat increase. It's another thing if there are multiple systems that contribute in different ways to getting stronger, each way having it's own pros and cons. I'll be making a post on it shortly.

    If anything I think crowd sourcing can be a fantastic way to realize potential if the system is complex enough. There just can't such a precise "best" way to approach things, preference has to be a major factor.

  15. Theoretically it's possible to number crunch your way through all of the possible combinations of souls within Rift to calculate optimum builds but I've seen little evidence of this other than players posting builds based on a samples of log parsing and best guesses; surely there is still scope to improve rather than giving in and taking the first 'optimised' build you see on the interweb?

  16. I have to agree that optimization is not the enemy. Crowd-sourcing, nice.

    "Here set these points here and do this in this order."

    I think that was one of the reasons Blizzard moved most of the class mechanics towards a priority system. They might not be able to control 'optimal' builds based on the community, but they could force players to decide and make choices.

    Priority systems require (some) evaluation of what's available, resources and adjustment as opposed to 'rotations' which were very easy to follow, perform and automate (macros) in some cases.

    I think Blizzard went a little too far in trimming down the talent trees, now there is little actual choice and the 'best' options are a fair bit more obvious. They might almost be better to drop talent trees and just have subclasses. :(

  17. I don't think the best approach would be to make crowd sourcing the enemy. Hiding data would be frustrating, people want to know how they are improving.

    The problem is that data automatically generates crowdsourcing in this day and age of the internet. If you give someone data, they crunch it then they post it somewhere then people feel obligated to use it.

    I agree in spirit with you but human nature will never allow "partial" use of the data. If it's there it'll be used to its maximum efficiency and then beyond for things it was never intended for.

    A good real world example is car insurance companies using your credit report to set your rates.

    I've said it before the Armory was the absolute worst thing the Wow devs ever came up with. The good of it is totally overwhelmed by the predictable human desire to find an easy path. And the fact that everyone finds the easy path means the devs have to make things stupidly hard to be challenging. It's created a feedback loop that makes it keep getting worse.

  18. If the internet is not going away, then there will always be someone to tell you how to win. In single player games you can just go look up a walkthrough, or even just watch someone else play the whole game on youtube. If you actually want to play a game you usually wouldn't do either of these things, but they are there and people do use them.

    Suppose you were playing a horror FPS like Dead Space. You are progressing through the game nicely and having fun, but at some point you are in an area and you can't figure out what you are supposed to do next to continue the game. How long do you try to figure it out yourself before you just go look it up? After a few hours (much less for most people, I would think) of retreading the same area over and over you'll probably decide the puzzle you are on is an impediment to your good time rather than a part of your good time. So you look it up, hit yourself in the head for not seeing the switch, get past that puzzle, and keep enjoying the game.

    MMOs are not really that different. People start by playing the game and having fun without optimization, but eventually hit a situation where they feel like progressing on their own devices would just be a pain. That's when it's time for crowd sourced information.

    The thing is, there is no one in the world that beats WoW hardmode raid bosses without significant optimization. Everyone must rely on the crowd sourced information if they want to do the hardest things.

    I really think the solution is to just not make things so hard, and accept the fact that some people will leave the game because they personally find it so easy. Tuning the challenges to the top 0.1% doesn't seem like its worth the consequences.