A brilliant new balancing idea

[RCIX]

One of the biggest problems with strategy games is ensuring all units are fair and useful. This problem has plagued many a game (including this one at times ), and can often make or break the success of a game. To that end, i propose an "Overseer" object within the game. This overseer periodically samples game stats (like KTL ratio versus other types of units, quantity built, basically any stat that shows an aspect of the unit's performance), then uses these numbers to determine how effective the ship is performing. To be clear, by "object", i mean game class that you instantiate behind the scenes. It would run roughly every hour i imagine.

This number can be directly acted on (if controlling imbalances in the middle of a game is vital), or sent to a server for the devs to analyze and act upon. I suppose the feature would be most effective in doing both, as it could correct the game balance for the duration of the game then notify you guys of a possible game imbalance. The advantage of this is that you can use a more standard performance quantifier set and let the "overseer" determine how to best make units fit that role.

So, whaddya think? Is it feasible? useful? will it help?

[rubikscube]

first, not big enough of a player base, even with all players that's a must install, one person can change the results by a few percent if not controlled. otherwise great idea.

Is it for people who volunteer or a must install?

if so, does the beta tester only play "official" versions, as in newest beta, everything turned on no cheats with decent difficulty?

though i would love to hear what x has to say.

[rubikscube]

though i admit it's much better than making a poll

Also it will be interesting to understand what it means, building things with low health a lot isn't abnormal, it just means you're doing it wrong kinda. And depends how much you micro so even you collect the data, idk if you can understand if a unit should be buffed or not simply based on their kd ratio. but i will say it would be nice to introduce new ways to use a ship that's hardly used as in a buff.
And since people have their preferences, it might be strongly to their preferences. 

[RCIX]

I thought it would probably be included in all AI war installations. So for example, you pick... Tanks for example. You find they are absolutely great, they can rampage through anything and do a great job on their own. As a result,
 * The relative build count for tanks over other fleetships would be up
 * The KTL for tanks would be abnormally high for thing it's supposed to be weak against
The OVerseeer would see that and say "hey tanks are too good, lets make bombers (a counter to them iirc) more effective against them, and reduce the strength of tanks against bombers ". The exact numbers would depend on how out of whack tanks are, but that's basically what it would do.

The idea is that the Overseer includes the swarminess or powerfulness factor in determining the balanced-ness of fleetships, so that it knows that it would take thousands of swarmships for them to be OP. You'd also have to find a way to collect several more kinds of stats about all ships, that would be useful to the process of determining whether to buff or nerf a unit.

The only problem i can see (right now) is for things like munitions boosting. Imagine if this was in effect when light starships were OP: entire fleets would be shown as being overpowered and receive nerfs! A solution might be to consider all damage dealt as a result of multipliers to be considered to be dealt by munitions boosting ships.

[rubikscube]

what about nutjobs like me that like to mess with the games? i suppose cheats off and everything else on, perhaps that only that version can be counted as newbies tend not to mess with "decent difficulty"


As i said, if you micro enough, a lot of stuff is just plain strong. plus the bonus of boosting as you said and shields and fleet composition. I say even anything is detected wrong, it will be a too small of a fix ( plus 2 damage ) and anything that's horribly overpowered would be said here mostly. But for sure the data collected will be interesting , i wouldn't know how to read it but x probably will know.  Perhaps a controlled environment is your bonus ship plus the ai fleet composition. (is the ai fleet composition top secret  ) anyways, use that composition to fight a decent enemy fleet and see the results.
still, WAY too many variables but data collecting i agree.

[RCIX]

It would be all in the tweaking of sensitivity of the Overseer; if you could get it just right it would catch things like the Light Starship (which IIRC survived for quite a while before people reported it), without reacting to micro-ing of ships. Though perhaps you might want it to react on that level, to keep you working and changing around your strategy.

[rubikscube]

back to the techy stuff, even though you are not the developer, will this slow down the performance?
How would we send this information? Does it as you say in real time checks for it every few hours and groups it up and organizes for x?

What if we don't have internet at the moment:D (you always have to theorize that)

What if we don't play normally and do some bad strategic moves and loses whole fleet against something.

What if use exploits (unless it's an exploit finding software)

What if people call it spyware (i'm just being suggestful)

Well, from your replies i can tell your answer to What if we don't play normally and do some bad strategic moves and loses whole fleet against something. is that the software is very sensitive, it does not concentrate on pure kd ratio but ratio against other ships.

what if some ships were MADE to destroy the other

[RCIX]

 * Depends on how heavy the algorithm is and how often it's run.
 * The game would automatically try and open a connection to a server at Arcen games, then would send (i imagine) a compressed form of the data (which would be decompressed on the other end.
 * In that case, it could simply not send that batch of data.
 * If you were making bad strategic moves, it might be nice to have a bit of a bosst to get back on your fleet, though something would have to be done to prevent this from becoming an exploit (perhaps the Overseer would ignore relatively brief spikes of "underpoweredness", or the sample rate would be such that you'd have to ram ships into suicide missions constantly for hours to get a benefit)
 * that's part of the point of the Overseer, it detects that players are abusing overpowered ships and nerfs them to concentrate If you're talking about cheats, then the Overseer should probably be turned off when cheats are on.
 * There is that, and there should probably be a "turn off game balance data collection" option in the interface.
 * The overseer would concentrate on analyzing the relative numbers; if a given ship is performing too well against what counters it or what it counters; i imagine that a general KTL ratio isn't that useful.

[superking]

i did think at once point that it would be good to have the game remember the stats from every game, which could then be autosubmitted

by having a program that compiled all of the submissions, a statistical database could be constructed which could be used to divine how well each unit is performing

[x4000]

Well, that sort of algorithmic approach sounds good on paper, but it's not something I'm a fan of. The reason is context. You'll notice that when someone claims that unit x is overpowered, I want to know why they think so. Often it might be because the opposing ship mix is just really poor against that unit, or the player is using some clever tactic like the old munitions boosting snaking or kiting, or whatever else. With just raw percentages, I'd have no idea what to fix.

I already run something like what you arE decribing in the form of the strong/weak data simulation, and it strips out as much context as possible. So when I want to use a contextless raw set of performance data, that's what I use. But that only goes so far, which is why I rely on intuition and user reports of specific explotative behaviors, often waiting for multiple confirmations. The system works VERY well, I think, but we are constantly adding new content and changing various mechanics, which means we are then also constantly rebalancing; that's just a fact of an ever growing game, it doesn't mean it isn't working.

[Buttons840]

I've proposed this in other forums for other RTS games RCIX.  I think you must be my evil twin.  It's kind of funny considering our initial battle over making control nodes free, if you remember that?  We had different views on that, but otherwise we have much in common I believe.  I play AI War, SupCom2, and started looking into Dwarf Fortress recently.  I also enjoy programming as you do.  But I digress...

Anyways.  I agree with X, and have never considered using such a system for AI War specifically.  I do think it could work for multiplayer competitive RTS games where lots of data could be gathered.  You could then adjust the scores automatically every week or so, giving rarely used units a buff, etc.  It would be kind of cool to look at these changes as say "wow, assult bots are really up in attack power this week, I better give them a try," etc.

[Fleet]

I too enjoy programming, and have a special place for Dwarf Fortress in my heart. Pretty specific group of people Arcen collects here on the forums

[x4000]

Indeed.

[RCIX]

Makes sense!

[superking]

dwarf fortress will turn into a religion, where the faithful shalt one day be rewarded with a 3D graphics engine