This is a very very early implementation of the research screen.
We have a lot of work to do on research as we’ve been experimenting with lots of different ideas for the past few months.
The goal was to have something very different from Galactic Civilizations but also something that is easy for people to add their own techs, distinguish different factions with their own techs, and allow for infinite researching.
In this UI, the player has researched the ability to research 3 things at once (no penalty). This option gives us a lot more flexibility in terms of letting people make interesting choices on what kind of civilization they want to develop (we’ll explain more on this as we get closer).
More to come.
Quitters aren't a problem if we're getting rewarded for winning like WarCraft 3 on Battle.Net where you unlock avatars the more victories you accumulate. So if that system was in place I would be really happy to face a quitter as my record would get better and better and I would be one victory closer to my new lobby avatar
And I can't seriously believe the people who are debating about scientists keeping notes to save some of the research and stuff like that....I've only played Age of Wonders out of the AoW/MoM/MoO trio but if you lose a city which gave you 20 research then what you are currently researching will take longer unless you devote more magic income to research....
Scoutdog, reply #89
I don't like exclusive techs. I would rather be able to research contradictary techs, but you will have to choose the effects you want for you empire. I prefer that there would be some in game reason why something doesn't work well, such as the effects of 2 or more techs not mixing very well.
Perhaps you might have civics (like in civ 4) where you have to choose how you govern your civilization in multiple categories. For instance, for Elemental we could have a civic category called slavery (or maybe even better name, labour) where you could choose "light slavery", "heavy slavery", "no slaves", or even "free the slaves" policy. You would have to pick one (unless contraditions are allowed, but that would be another topic), and you could always research more later.
That would work, but it takes away the "tough choice" quality of the research that I was attempting to create. I rather like the idea of making an irreversible decision that could impact your empire forever. Sort of like the morality in GC2.
I didn't Galciv 2 ethics very much. It favoured evil too much. I also tolerated by the "tough choice" largely because I didn't feel I was missing out on whole lot near the end of the game. I considered the bonuses to ship defense you got as the "good guys" more valuable than high priced, eventually made obsolete by end game weapons you could get as evil (even if they were very useful until made obsolete).
Clearly, we are trying to push this game in two opposite directions...
I'm also not the biggest fan of strict mutual exclusivity - whether it be in research or some other aspect of the game. I don't hate it, but I think a 'soft' form of exclusiveness works better: it allows people to mold their strategies more to their own personal preferences while still preventing players from getting the best of all words.
To give an example, let's say there are two techs, "laser accuracy" and "laser power." With regular mutual exclusiveness, researching one prevents you from researching the other - you have a binary choice between accuracy and power. With a soft form of mutual exclusiveness, there are several ways of going about it. One method is to make researching the second of the two take significantly longer or cost much more than it would (or a combination of the two). Another method is cause each tech to have reduced effectiveness if you research both (or, when actually using these techs, choose whether to have a very accurate but weak laser, a very strong but wildly inaccurate laser, or an all-around mediocre laser) - how this would be be done depends on the nature of the tech and the game.
In the vast majority of circumstances I think some form of soft mutual exclusivity is more fun than straight up "if you researched this then you can't under any circumstances, no matter how situations have changed, and no matter how much you are willing to put into it, research that." Sometimes that is the way to go, though. In my example of laser research, I'd say no. If there's something that fundamentally changes some part of an aspect of your nation, maybe. Things like "free the slaves" vs. "enslave everyone!" don't seem fundamental to me if they're including as research. If position regarding slavery is determined by faction choice, on the other hand, that's an entirely different matter. But to me, if I can research it, then I can choose whether or not to actually use it, regardless of whatever other technologies or policies I have at my disposal. However, I'm all for not being able to use certain techs/policies in conjunction with others - or getting severe penalties to their effectiveness if you do.
A combination of both methods where appropriate seems good to me.
Well, best to keep all of our options open, for the modders if nothing else. I realize that the GC2 ethics system is screwed, but the idea of snap choices that can effect the outcome of the entire game has always been appealing to me.
The problem with "tough choices" is that they're usually not very tough. One tech will be better then the other inevitably, and people will choose that one.
That's the problem with blocking things. If I can only have one or the other, I'll ALWAYS take the one that is better in a majority of cases. You just wind up with tech that never gets reserached at all because doing so locks you out of some other, better tech.
I'm going to have to agree here. The only tough choices I find when playing these games are usually what race/magic school to choose. All other choices are mostly predetermined from that point since certain things work better with the strategy/playstyle I've already chosen.
A huge amount of effort would need to be put into balancing those choices. A similar dilemma was seen in the quest rewards in AOW series where you would choose between two rewards. One of them was almost always the clear winner and you didn't even need to think about choosing.
Stardock has quite the advantage here though. Since they already analyze matches with GC they can do a similar thing with choices. Let's say they program in a random event where you choose to lose X (+/- a random #) of your mana or lose (Y +/- a random #) of gold. The program then records and submits which option players pick most and what random ratios or variables were found to make the decision the most balanced. Then the game either modifies the choice in the next patch or on the fly. Trying to balance things like "Gain X trolls or Y swordsmen" are just really tough to balance when you have to consider things like special abilities, cool factor, and other things that don't fit into an equation very well.
Stardock does not analyze matches in GC. I know this because I play GC on my computer, and I don't send them my games and they don't reach into my computer and take them. Sure, they might analyze them internally, and when people send in bug reports with a save game, or here and there during the beta process...
But even more to the point, that's still not a great way to go about doing things... For one, they would need a very large sample size of (very) similar choices to be able to come to any worthwhile conclusions, and their results would be biased towards people who play more often. And even more importantly, this method would not work as intended regarding things like mutually exclusive research or magic - they might find a perfect 50/50 split, but that could mean several things: it could mean people choose each half the time (ideal), but it could just as easily mean half of the players always choose one and half always choose the other (not so ideal).
Yes, they could track all of that information and try to work from there, but it'd be an enormous effort and ultimately I don't think it would help them do a better job than if they just use their gut, and include decisions that will give them pause.
The AI in the GC series uses the tactics of the most successful players. It really doesn't matter if you personally aren't giving them information since tons of other people who submit to the metaverse are. Elemental's devs already revealed they'll have AI's that learn from you as you play them. Either way, this is irrelevant as there's nothing stopping them from doing so in Elemental. Even the most privacy concerened player could still excuse them for using information collection if it was only collected during alpha/beta. They could always give users an opt out button, but would you really care enough to opt out of anonymously improving the game AI and balance though? This isn't a big evil Microsoft.
You go with your gut and design a choice or rules for choices and then you use the statistics to help refine it. It doesn't take a lot of effort to do. You can get a nice picture of a player's empire and playstyle by sending a miniscule fifty vaiables to a server when they make a choice (though some of them could even be helpful without almost any related data). If that doesn't fit your network or players then you just have the game automatically submit completed games or autosaves to a server. There's a lot of ways to go about this with various levels of intrusiveness. A lot of times, the question you are trying to ask might only involve mainly involve a couple variables.
Sample sizes of 1000 are fine and not hard to come by when the collection is automated. Having that data point you to possible problems is a trivial task as even a glance looking at percentages can be quite insightful. In my experience even the most minimal data analysis has always paid off compared to how much time I spent implementing the collection.
The problem of getting a situation where the same half of players choose one option and the other half always choose the other is solved by unique id's as is done with most scientific studies. You'd realize the two populations of players and look at other variables. Of course, looking at broad and sweeping choices like exclusive technologies are harder to look at than the near immediate player rewards/punishments I suggested. However, because it is harder to balance, having a more objective way to see the results of choices is even more important. It's better to analyze info they collect from the beta testers' actual games than having the devs try and simulate it themselves (wasting time, including their own biases, etc.).
At the very least, this type of statistical analysis will point out common trends and game imbalances. If done properly, it can also provide possible soltutions and then evaluate how effective they were.
In my experience, I've found that statistics often lie and if you do a cursory analysis, or if you don't pay enough attention to context (or even worse - a combination of the two) you are likely to miss big issues and cause more problems than you'll solve.
If Stardock decides to do any sort of statistical analysis on matches, then they have to fully commit to a rigorous and comprehensive methodology. If they don't, they won't gain anything - the really obvious things that they'd catch from a cursory analysis they'd find anyway through feedback; and the more subtle things would fly under the radar. And even worse, a cursory analysis could easily point towards imaginary problems. And yeah, maybe using really thorough statistical analysis of player matches will help things along, but I personally don't see it having much of an advantage over using their instinct, their own preferences, and player feedback - not enough of an advantage to make setting up the data-gathering and analysis processes time- or cost-effective, though. This is just my opinion - I'm not saying your idea is wrong or bad - just that I think it isn't worthwhile.
Also, no one ever said that Elemental's AIs will learn from you as you play them. Dozens of people on the forum including employees of Stardock have said how cool that would be (and maybe they're even going to attempt it), but they never at any point said that the Elemental AIs will learn from you. The closest thing to that is that Brad said he'll periodically look at the favored and most successful player strategies and update AIs to mimic and counter them. Those are two very different things.
Alright, so we've got a reasonable disagreement here on the merits of data collection & analysis that didn't degrade into a steaming pile of junk. That's a pretty amazing feat on the internet.
"but what I hope to do is to build a cloud of multiplayer AI players that I can constantly be updating and enhancing in near-real time based on the data I get back from them on how they were beat, what exploits were used against them"
Sounds pretty indistinguishable from the bot learning from the player's perspective, which is really all I'm concerned about. All AI are just a collection of tricks anyways so the actual methods are pretty irrelevant so long as the player thinks it's effective and doesn't "fall" for the same thing every time.
Either way it looks like those multiplayer bots are getting scrapped anyways. I totally missed this line:
"2. I’m killing off the bots concept. People hate them. I thought they were cool but they’re too much work only to be hated. So there won’t be artificial players."
I am glad to see how the devs are eager to use game data to improve things though and I hope they still use it for the normal AI.
Curious..... from my skimming of the forum, people's "hatred" sounded suspiciously like praise.....
Only hatred I remember was demigod players being mad at getting stuck with them as teammates and opponents when they were looking for humans. No bots checkbox = problem solved.
We are so off topic.
I see what you mean, but I still think that the subtle difference between that and AIs learning from you is extremely significant. What Brad said just amounts to lots of frequent mini-patches to the AI - this is wide and sweeping and affects everybody the same. AIs that actually learn from each game you play is very different and would result in AIs on each person's computer behaving differently. They'd pick up on preferences, strategies, etc that you do and that work against you - you specifically. Ultimately this learned behavior could then be retrieved from all consenting players by Stardock and integrated into better 'base' AIs and the learning process could begin all over again.
Another difference is that AIs that learn from their matches against you is a huge stepping stone on the way towards AIs that can learn much better within the course of the game, and thus understand changing situations and adapt to them much better than current static AIs could ever possibly do.
Sorry for off topic... But at least we're having an intelligent discussion, so I don't feel too bad
Well since essence does not recharge, how to use that will be your tough choice.
Sorry, I had to hassle you after the ribbing you gave GW in the FAQ thread lulz.
I understand the difference between things actually learning and the illusion of learning with mini patches. I never had any misconceptions about that I just didn't think anyone would actually read so literally into that one sentence of my post. Since that wasn't actually a main point of my post I just summed up the whole frequent-AI-patching-based-on-player-data as "AI Learning".
I would prefer that no actual major AI learning is even attempted. Adding a system where the AI learns would just add too long to the development cycle. You have to worry about it coming to the wrong conclusions and also learning bad habits from players. Considering how you opposed relying on data analysis, which would be limited and all the AI would have to rely on, this probably wouldn't go so well. Even trying to have the AI learn post match is pretty unreasonable considering how many factors can differ when using such a small sample size with so many random variables.
I think having a system where different AI play styles are chosen based on conditions instead of them trying to learn mid match is far superior. On normal or below you could have the AI just randomly choose a playstyle and on hard or above the AI can change playstyles depending on what it sees the player doing. Having it recognize some obvious player warmongering and responding by changing from say, the expansionist to defensive playstyle is relatively simple compared to trying to make it learn.
Data based mini patches would then improve all of these playstyles over time. The AI would not permenantly learn from a individual player though, it would simply change to a different predefined tactic midmatch. GC2 Seems to follow a fairly similar system and it's definitely one of the best AI's I've seen in any game.
I'm with you 100%. I've pretty much been arguing for what you said in your first paragraph for months now, in various threads here and there. And any patches to AI (or any other aspect of the game) are always welcome, of course.
Depends on how the levelling works. If it takes the same amount of EXP to level up each time, it's just a matter of "working to refill essence by killing bad guys (or however you get exp). If the amount increases with levels, then it would be possible to "run out" of levels (and thus essence) when the requirements became so huge that you could never realistically reach them.
I'm of this opinion as well. In short, a robust dynamic AI system will be key for the longevity of the franchise. That coupled with personality based activity will make the game environment very rich and challenging.
Yep, he was talking about how much Demigod players hated it. The forum reception here was more positive.
temporarely back from vacation....
Leaving again tomorrow...
Nice update thanks
Usually researching 3 items at once will only split the research efforts thin, - researching one at a time would yield the three techs in the same amount of time, but the player benefits from the effects of the technologies earlier.
Would 'research without penalty' mean the player has effectively tripled his research output, or is research based on random success rolls, so researching three techs at once increases the chance on a success?
Otherwise multiple research is a red herring, seems useful but distracts from the only correct option 'always research one tech a time'.
Opinions?
Well, tech research had better not be based entirely on RNGs.... the random element will completely nerf any balance in the game. That's not saying a little randomness isn't fun, but once it becomes the main aspect of teching.....
as I said before, since the ability to research 3 things at once is an UNLOCKABLE ability (which implies a definite benifit), it's possible that when you research 3 things at once you literally add all your tech points to each tech, increasing your output threefold. (There might be a small penalty, though: triple research output sounds like a BIG boost.)
There are many great features available to you once you register, including:
Sign in or Create Account
