 Poll
|
| Question: |
Do you feel like the cost values of ships must be re-calculated?
| Definitely! |
| Most, yes. |
| Some, yes. |
| Too few. |
| No way! |
|
|
|
Author
|
Topic: Cost balancing project (Read 23760 times)
|
Captain_Smith
Guest
|
The more I think about this, it would be possible to make such an analysis for the AI alone (it plays the same for everybody, can't say that for any human player, even though it is trash at playing), assuming something was coded into a copy of UQM. You'd have to make sure you use proper statistical analysis on the results, given the full data generated. For example, you would have to track victory counts and crew, but recognize that beating a ship at 1/4 or 1/2 strength is different than beating it at full strength. This is especially true, since there are a lot of matchups that are impossible for a ship against a full strength adversary, but very possible against a weakened one.
Basically for a statistical module, the game would need to track winners & losers, as well as losses by both in terms of percentage of total crew lost. Yes the numbers for the Syreen would skew things a bit, since you'd have to base it off of 12 crew instead of 42. But there's a lot of thought to be had still on a proper way to go with this.
Of course, all bets are lost if the AI is changed in any way to "improve" it.
|
|
|
|
|
Logged
|
|
|
|
Valaggar
Guest
|
Ahoy, Captain, the AI is better at some ships (the quick-reflexed-based ships) and HORRIBLE at others (those for which you need to use your Brain). It's by no means a judge of the ships.
As to starting crew, you only need to test for both ships at full strength, since the ships play equally well, no matter how damaged they are.
Of course, the AI may be improved - Elvish and Novus rambled a bit about using a learning algorithm, but... it's only a dream.
|
|
|
|
|
Logged
|
|
|
|
|
|
Squisherxxx
Zebranky food
 Offline
Posts: 44
|
Wow, I've been of the boards for a while (Everything going wrong at work these last couple weeks). Firstly, I would to thank Shiver for his complements. Secondly I would like to make several comments in regards to peoples posts:
As far as the need to ballence goes, it is true that it is not nessessary. However, I do believe that doing so does provide something positive for online supermelee.
People have mentioned that the costs of the ships are dependant upon the skill levels of the players. While this is true, I do believe that the costs should reflect a player who is considered "good". I came to this conclusion based on the concept of "optimal strategy". Two poorer players seeing an imballence between two ships can easily be flipped due to a change of strategy, while two veteran players have analyzed the play enough to have the strategies evolved to a point where it is unlikely that a radical strategy will upset the ballence any longer. Although new strategies can be determined, they generally dont swing from dominating of one side to the other. To give an example, Valaggar give the Kohr-Ah -1.5 vs Chenjesu and +2 vs Earthling, when these numbers are actually reflective of poorer players, and ballence based on this will be virtually unplayable for good players.
As for the testing of all combinations; I dont think that this is nessessary. I believe that if I drafted up the +/-'s of all the ship combos, and If Angus did the same, we would only have to play out the matches which had differences of opinion.
As to how the algorithm is currently being done, I believe it is flawed. I believe that the ship +/- levels should be based upon the COST of the ship, and that the results of the algorithm would produce ADJUSTMENTS to the current costs. I explain further: Valaggar currently lists for the Chmrr the following values:
Umgah +2
Spathi +2
Shofixti +1
Now, it is clear that none of these ships can beat the Chmrr, but the spathi has no chance to damage, while the 4 Umgahs (7 food each) have a decent shot at taking out the chmrr. My proposal is this: Make the +/-s based on the current food level. A shofixti is a great choice vs a Chmrr, and it should be reflected as such. Based on current cost values, a Chmrr should be -2 vs a Shofixti. Based on the way Valaggar is currently doing it, the Shofixti has 0 as the highest value it can acchieve, as it can not win without dying also (generally speaking). Also, many people would agree that the current costs are close to what is appropriate, and a full overhaul is not nessessary; a simple balance will do.
Additionally, there are other factors which should be accounted for: Ship Cost vs Ship vulnerability. Consider a ship which simply pummells every ship, but has no chance of even damaging one certain ship. By your system, this would be a high cost ship, which would destroy 1 enemy ship, then lose the next fight as the opponent will pick that ship to which it is weak. Ships with low cost are inheirently stronger (relative to its cost) because even if they have a glaring weakness to a high cost ship, they can still be cost effective so long as they make your opponent choose that high cost ship, which will subsequently die quickly to your next choice.
Anyone who has played tiberian will quickly recognize this. So long as a fleet of low cost ships can collectivly destroy any high cost ship, they will always at the advantage because the high cost ship, despite being good against most ships, will still destroy 1 ship and lose to the next ship, something wich a high-cost ship fleet cannot afford to do. Thus, it is a ships disadvantages which should be the largest impacting factor to ship cost. Again, this can be easily seen by considering the effectivness of a ship which is strong vs all but weak vs one compared to a ship which is stong vs half, but weak against none. Clearly the latter is the better ship.
|
|
|
|
|
Logged
|
|
|
|
Captain_Smith
Guest
|
Valaggar:
You are right about the AI not being a judge of the ships, but at the same time, it's something that is a constant if you wanted to try to figure out the relative strength between ships. Of course, you get a skew in results given the poorer players versus the good players, but it's much less than if you consider two players. Not to mention, the game mechanics are different between local and online games anyway (specifically movement timing, which is the most important factor in the game), and that plays a factor into things.
Squisherxxx:
There definitely would be a difference of opinion in most of these. If I could understand the spreadsheet as it was done, I would probably draft much different numbers than Valaggar did, simply because my skills would be different, I'd be better at some ships than others, and a whole host of other things. As far as your ship observations go:
1) A shofixti is indeed a great thing for a Chmmr, even if it dies, it wipes out a number of the zapsats (3 if you're really good), making the Chmmr vulnerable to much more. Also a shofixti is much capable of taking out a few ships without using the glory device, so any observation based on the glory device alone is flawed.
2) I take it you did not see the video that was posted a few months ago regarding what a Spathi can do to a Chmmr. http://uqm.stack.nl/wiki/SuperMelee_demonstration_videos#Defeating_a_Chmmr_Avatar_with_a_Spathi_Eluder So this actually reveals that there's another factor in there, preceived notions of impossibility. One thing I've found personally is that my game got much better when I gave up the notion that any matchup is "impossible so I shouldn't even try it".
|
|
|
|
|
Logged
|
|
|
|
|
|
Elvish Pillager
Enlightened
Offline
Posts: 625
|
Well, his numbers for Ur-Quan...
Ur-Quan:
Ur-Quan 0
Chmmr 0
Kohr-Ah 0
Chenjesu 0
Orz -2
Yehat +1
Utwig -1.5
Mycon +1
Pkunk -0.5
Mmrnmhrm -1
Melnorme +1
Spathi -2
Druuge +1
Slylandro +1.5
Supox -2
Arilou +1
Androsynth +1.5
Syreen +1.5
Earthling +1
VUX +0.5
Ilwrath +2
Zoq-Fot-Pik +2
Shofixti -1
Umgah +2
Thraddash -2
It's better than he says against, at least, Chenjesu, Orz, Utwig, Pkunk, Spathi, Supox, Shofixti, and Thraddash.
(Especially Thraddash. What the hell is up with that number?!)
|
|
|
|
|
Logged
|
My team of four Androsynth and three Chmmr is the most unfair team ever! My mod |
|
|
Cedric6014
Enlightened
Offline
Gender: 
Posts: 701
|
Ok, I think i understand. The method may work but so far the data is very bunk. We will need 1 experieinced player to go through Valagger's template, and for it to be subsequently peer reviewed by 2 or 3 others.
|
|
|
|
|
Logged
|
|
|
|
Valaggar
Guest
|
AAAAAAAAAAAAAAAAAAAAAAARGH!!! THAT WAS IT! I KNEW! I KNEW!
I wrote -2 for Ur-Quan vs Thraddash, then I tested again and I discovered that I was wrong... but I forgot to change the value!
 ...
Plus, I suck at Ur-Quan, not because I lack skill with it, but because I like shooting like mad, even if I don't have any chance of hitting.
Ur-Quan will be reviewed immediately.
Ok, I think i understand. The method may work but so far the data is very bunk. We will need 1 experieinced player to go through Valaggar's template, and for it to be subsequently peer reviewed by 2 or 3 others. Yes. Do it. Please!...
As for the suggestions related to taking into account HOW MANY ships are needed to destroy the <current> ship... yes, in fact I did kinda like that - I converted:
+2 to 8
+1.5 to 4
+1 to 2
+0.5 to 1.(3)
0 to 1
-0.5 to 0.75
-1 to 0.5
-1.5 to 0.25
-2 to 0
(the numbers on the right represent how many <current enemy> ships must be lost to defeat the <current> ship)
More precision is unattainable since the match-ups depend on skill and luck too.
As to the SuperMelee demonstration video with the Eluder defeating an Avatar - it's against Mr. Awesome!
|
|
|
|
« Last Edit: April 12, 2007, 10:16:22 am by Valaggar »
|
Logged
|
|
|
|
|
|
Squisherxxx
Zebranky food
Offline
Posts: 44
|
Actually, any match vs AI has absolutly no credibility. Show me a human player losing that matchup then we'll talk.
|
|
|
|
|
Logged
|
|
|
|
Tiberian
*Smell* controller
Offline
Gender:
Posts: 335
|
Playing single battles and calculating the outcomes is a very wrong way to make adjustments to ship values.
The only proper way to re-think the values would be a real-time discussion with lots of people who know everything there is to know about melee. That is never going to happen.
Trying to fix something that is not broken will often lead to breaking it.
|
|
|
|
|
Logged
|
|
|
|
Captain_Smith
Guest
|
Now, it is clear that none of these ships can beat the Chmrr, but the spathi has no chance to damage, while the 4 Umgahs (7 food each) have a decent shot at taking out the chmrr.
Actually, any match vs AI has absolutly no credibility. Show me a human player losing that matchup then we'll talk.
Typical arrogance. It's unfortunate how this board has gotten in the last few months. I'm thankful I haven't gone into #uqm-arena as much as I might have had.
|
|
|
|
« Last Edit: April 12, 2007, 05:56:09 pm by Captain Smith »
|
Logged
|
|
|
|
|
|
Valaggar
Guest
|
Rather than flaming the boards, better test the match-up on #uqm-arena.
Playing single battles and calculating the outcomes is a very wrong way to make adjustments to ship values.
The only proper way to re-think the values would be a real-time discussion with lots of people who know everything there is to know about melee. That is never going to happen.
Trying to fix something that is not broken will often lead to breaking it. No, a real-time discussion is not necessary. We can simply have the melee experts (you and the others on #uqm-arena) make a few tests and judge the match-ups.
Also, cost balancing would be a good thing for tournaments.
|
|
|
|
|
Logged
|
|
|
|
|
