Difference between revisions of "Talk:Proposals/Obsolete Proposals/Campaign"

From UFO:AI
Jump to navigation Jump to search
 
(No difference)

Latest revision as of 11:33, 13 August 2013

Hacked off the first round of brainstorm due to refinement and inclusion in main article. --Wanderer 05:36, 13 May 2007 (CEST)

Zoning the Geoscape

Instead of actually zoning the geoscape into regions, I had a brainstorm I'm not sure of it's feasibility. If you put 'Location Markers' on the map, say, center of Antarctica, Eastern Europe, Western Europe, North/South America, so on, and then when a mission occurs use whichever marker is closest as an indicator. Associate each map or random build with a point of origin, and when a mission occurs near that PoO, it can read through the list and find the correct map.

The nice part about this is you could be as specific or generic as you want. Surround New York City, London, Paris, whatnot with more generic 'Woodlands', 'Farmland' markers. Use a Marker type once or many times, to help control more specific areas. --Wanderer 18:46, 11 May 2007 (CEST)

--BTAxis 14:10, 12 May 2007 (CEST):

It's probably easier to do it that way from an implementation point of view, but I think it might be hard to define specific areas that way because it's hard to tell exactly when one marker takes priority over another marker. For example, take the Sahara desert. Missions inside of it should be desert-themed, but missions outside of it should not. Using markers, it's going to be a very hard job to get the desired behavior because the Sahara desert isn't shaped very conveniently.

Choosing the Interest Events

The event with the highest number that is still smaller than CategoryInterest 
will be the event that is selected, with a small random modifier applied to 
prevent all multiple instances of the category type to be duplicate. 
For example, CategoryInterest + (RAND() - 0.5) * SQRT(CurrentInterest). 

This is definately different then what I remember discussing as being the selector. At early game this modifier isn't enough to even move the CategoryInterest (CurrentInterest of 50, say, so we're looking at ~7 * .0-.5, an average difference of around 1.5-2.), and in late game still not really much to move it (CI 1000 = 100 * .0-.5, average .25, 25 points). There is almost no variability in this functionality for the selector. Especially when you consider according to the older selectivity method, you'd need, at 1000 CategoryInterest, a 100 point gap for a mission for it to have 10% selectivity.

I know you'd like a more effective way of selecting off the mission list to weight it at the top, but (CategoryInterest * RAND() ) still seems like the most effective way to go. If you really want to weight it, how about a function like:

( ( CategoryInterest * .5) * RAND() ) + ( CategoryInterest * .5). This keeps you in the upper half of the category mission list while providing more volatility in the mission selection. --Wanderer 17:37, 14 May 2007 (CEST)

--BTAxis 19:39, 14 May 2007 (CEST): Changed it. Thanks.
EDIT: How about this one: CategoryInterest * SQRT( (RAND() * .7) + .3)? You're basically limiting the event range to 30%-100% of CategoryInterest, and the root skews it to the top. A cube root would skew it even more.
You'd actually be more around 54%-100%, which wouldn't be so bad either. (Square Root of 0.3 is .547722). It would definately weight towards the top though, very very heavily. I haven't done the complete math but if I'm guessing right about 30% of your missions will be in the top 3%, and 60% of your missions in the top 10%. That's a LOT of weight. Is that your intent? --Wanderer 20:06, 14 May 2007 (CEST)
I just realized the first point myself, but you beat me to it. The skew can be adjusted by weaking the power: ((RAND() * .7) + .3) ^ (0.7) would result in a weaker skew than a SQRT, and would allow for lower level events as well. The "hard" skew by the multiplication and addition within the power can, of course, also be adjusted or discarded if that is the better choice. with these values, a RAND() of 0.5 gives us ((0.5 * 0.7) + 0.3) ^ (0.7) = 0.74, so 50% of the missions would be in the top 25%. a RAND() of 0.8 gives us ((0.8 * 0.7) + 0.3) ^ (0.7) = 0.9, so about 20% of the missions appear in the top 10%. How's that? --BTAxis 20:18, 14 May 2007 (CEST)
This would work pretty well, actually, and would be tweakable in the long run if necessary to a more effective rate if needed. It would max off the bottom at around 43%, giving more variety. Now, if we really want a top weighted value with a lower bound of 30%, we need to work the calculation to simply adjust the 70% part.
So, if we start off with a base of (CurrentInterest * .3) + weighting, and then play with the weighting part to create a bell curve. So, we work off a base value of (CurrentInterest * .7) and then give it a multiplier of weighting as above. So we could use something like the calculation of ( (CI * 0.3) + ( ( CI * 0.7) * ( RAND() ^ 0.8) ).
So with a RAND() of 1, we end up at 100%, obviously. At RAND() 0.8, we'd get .83 of the difference. Working off 100 CategoryInterest, This would be .83 of 70 + 30, for 58.6, for a total of 88.6. Pretty close to 20% of missions in top 10%. At Rand() 0.5, we get a weighting of .57, for about 40+30, 70. This means 50% still in top 30%. However, as we drop, the curve steepens faster. At 0.25 Rand(), We get a .32 weight, for an additional 23, equating out to 53. 75% of missions in top 47%. At 0 RAND(), we get a flat 30, meaning 100% in top 70%.
I'd like to see the bottom 50% of RAND() have more volatility, but at least this can give us a good variable base to start from. You could almost increase the Rand()'s exponential modifier to 0.85 or so and really steepen the lower end of the curve without significantly affecting the upper portion.
One thing to keep in mind with this method is that the widening spread as you go higher in the mission files needs to be not quite as large, or your volatility here won't be enough to keep 'repeating missions' out of the queue where the law of averages is not necessarily always going to apply.
--Wanderer 21:31, 14 May 2007 (CEST)
I'm not entirely sure what you mean by that last part. But I like the rest of it, and I think an exponent of 0.85 would probably give us better results in this scheme. --BTAxis 21:49, 14 May 2007 (CEST)
Good, I like the 0.85 too. That eases that. What I was mentioning in the end was regarding the 'selection value' climb in the CategoryMission files. In the original method, I was expecting the 'most recently available' mission to have about a 15% selectivity rate, somewhere between 10-20%, anyway. So, at 10 Interest, it'd only be 2 selections wide. At 100 Interest, however, it'd be 15-20 points wide. With this weighting method, to make sure we can clear the lower bound of the uppermost mission so at least in the top 25% of selection we'd have two that would be chosen from, we'll have to tighten this range, and perhaps mix up the missions more. My comment about the Law of Averages is that we can only count on it with a very large sample. In a short run, it's very easy to hit the same mission in the top 25% two or three times in a row, especially when we include weighted event category selectivity and we hit a couple of recons/harvests in mid-late game. --Wanderer 23:05, 14 May 2007 (CEST)
It's going to be pretty impossible to come up with a unique event for every single entry in the event table anyway, so as long as the nulls get hit at a satisfactory rate, I'm not too worried about the chance that the same event might be selected a couple of times in a row. The important thing to me is that they won't necessarily get selected twice. So providing the event tables are sane, I foresee no major problems. --BTAxis 23:45, 14 May 2007 (CEST)

Base attacks

When a base attack is spawned, the UFO or UFO group should fly to a base that is randomly picked, the selection being modified by the bases' weights. A base's weight is a function of how well-known that base is to the aliens (the base will make itself more known the more it attacks nearby UFOs) and the degree of alien infiltration and/or XVI infection of the area the base is located in. --BTAxis 16:10, 27 July 2007 (CEST)


should a base be 100% known if it has taken out an enemy ship with missiles--Rioting pacifist 00:48, 8 July 2008 (CEST)


2012 Adjustment Considerations

The following is the beginning of an attempt to prevent too many UFOs from spawning and getting the balance a bit better.

1. Our current setup seems to increase the number of UFOs during the campaign according to a specific plot equation. Y-axis controls the likelihood of a mission spawning (roughly) and X-axis controls alien interest (consider it the progression of time).

The main problem with this is that there is no way to properly balance the number of UFOs in a given period of time if we rely only on this equation. No matter what we decide is the right UFOs-per-month balance, the game will always progress from not enough to too many. The equation used is important for the early game, where we want to ease the player in with fewer missions, and the late game, where we want to overwhelm the player. But the majority of the campaign should take place in some kind of sweet spot where there are not too many missions to overwhelm the player. The game should become more difficult during this period not because of an increase in UFOs, but because of an increase in the weaponry/number of aliens that each UFO carries.

This is just a preliminary comment. I am still looking into other aspects of the system to find good numbers, but I wanted to get comments from the devs about this structural change. Would we agree that the mid-game needs to occur with about the same number of UFOs per month? --H-hour 23:33, 6 January 2012 (CET)

I haven't played campaign for 2 years, so I don't feel too competent as far as balancing goes. But as all formulas seem to rely on interest level, we should start with a complete list of all the things that interest level has an effect on imho. Secondly, we need a list of everything that can not be achieved with a simple formula like "0.5 ufos per day. In other words: let's collect the requirements before talking about the implementation.--Duke 04:26, 7 January 2012 (CET)

I feel that a simple formula for number of UFOs could work in the mid-game. But here is a list of things that we need that complicates it:

  • UFOs should increase slowly at the beginning before they reach a standard rate.
  • UFOs should increase dramatically at the "end" (a certain alien interest level) to overwhelm a player that has not progressed quickly enough.
  • Random UFOs should come in bursts, not just a single chance each day. Could probably be solved by something like (3 possible UFOs, each has a 10% chance of being an actual UFO). So we could have bursts of 3 UFOs in a day, but it would be uncommon. (This is basically the current implementation, except that many more UFOs can be spawned in a day.)
  • Selecting the UFO mission type. There's already a complex system involved that chooses what type of mission to spawn when a UFO is actually spawned, based on specific alien interest levels in certain mission types. See the section below.
  • Selecting the UFO mission difficulty. This is the most important. Higher interest levels should trigger larger alien parties and bigger ships to make it more difficult.
  • anything else I'm missing?

We do face one problem with using the saves data: there doesn't appear to be any historical data on increase of alien interest levels. So I can pull out how many UFOs are being spotted throughout the course of the game, but I can't match that to specific alien interest levels --H-hour 10:55, 7 January 2012 (CET)

Graphs on campaign.png

Alien Interest

Overall alien interest is set to increase once every 18-22 hours depending on game difficulty. On a very hard game save near the end of the current campaign (226 days) the total interest was only 319. Our current max alien interest is set to 1000. I think there is a problem here. It should be set lower, or research progress needs to happen more slowly. This is complicated by the fact that an alien base ends the game at the moment, so the campaign might progress farther at some point. But this number can be adjusted then.

If the save data is correct about the alien interest, then equipment_missions.ufo needs to be adjusted as well because many of the advanced equipment sets only occur at interest levels above 400.

Also, if we reduce max alien interest, we will want to reduce the number 50 in line 268 of cp_missions.c. At 400 interest it will create only 7-9 aliens:

	numAliens = max(4, 4 + (int) ccs.overallInterest / 50 + (rand() % 3) - 1);

Individual Mission Interest

Numbers accurate as of 2012-01-07

 /* successful base attack */
 	INT_ChangeIndividualInterest(0.3f, INTERESTCATEGORY_RECON);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_HARVEST);
 	INT_ChangeIndividualInterest(-0.5f, INTERESTCATEGORY_TERROR_ATTACK);
 	INT_ChangeIndividualInterest(-0.5f, INTERESTCATEGORY_INTERCEPT); 
 /* unsuccessful base attack */
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BUILDING);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_BASE_ATTACK);
 /* alien base constructed */
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_TERROR_ATTACK); /* only if it is a subversion mission */
 	INT_ChangeIndividualInterest(0.4f, INTERESTCATEGORY_XVI);
 	INT_ChangeIndividualInterest(0.4f, INTERESTCATEGORY_SUPPLY); 
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_HARVEST);
 /* alien base mission failed */
 	INT_ChangeIndividualInterest(0.5f, INTERESTCATEGORY_BUILDING);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BASE_ATTACK);
 /* alien base destroyed */
 	INT_ChangeIndividualInterest(+0.1f, INTERESTCATEGORY_BUILDING);
 	INT_ChangeIndividualInterest(+0.3f, INTERESTCATEGORY_INTERCEPT);
 /* successful harvest mission */
  	INT_ChangeIndividualInterest(-0.3f, INTERESTCATEGORY_HARVEST);
 	INT_ChangeIndividualInterest(0.2f, INTERESTCATEGORY_RECON);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BUILDING);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_XVI); /* only if XVI researched */
 /* unsuccessful harvest mission */
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_INTERCEPT);
 	INT_ChangeIndividualInterest(0.03f, INTERESTCATEGORY_BASE_ATTACK);
 	INT_ChangeIndividualInterest(0.03f, INTERESTCATEGORY_TERROR_ATTACK);
 /* successful intercept mission */
 	INT_ChangeIndividualInterest(0.3f, INTERESTCATEGORY_RECON);
 	INT_ChangeIndividualInterest(-0.3f, INTERESTCATEGORY_INTERCEPT);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_HARVEST);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_XVI); /* only if XVI researched */
 /* unsuccessful intercept mission */
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_INTERCEPT);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BUILDING);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BASE_ATTACK);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_TERROR_ATTACK);
 /* successful recon mission */
  	INT_ChangeIndividualInterest(-0.2f, INTERESTCATEGORY_RECON);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_HARVEST);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_SUPPLY); /* if base exists */
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_XVI); /* only if XVI researched */
 /* unsuccessful recon mission */
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_RECON);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_INTERCEPT);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_TERROR_ATTACK);
 /* successful base supply mission */
 	INT_ChangeIndividualInterest(-0.2f, INTERESTCATEGORY_SUPPLY);
 /* unsuccessful base supply mission */
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_SUPPLY);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_INTERCEPT);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BASE_ATTACK);
 /* successful terror mission */
 	INT_ChangeIndividualInterest(-0.2f, INTERESTCATEGORY_BASE_ATTACK);
 	INT_ChangeIndividualInterest(0.03f, INTERESTCATEGORY_HARVEST);
 /* unsuccessful terror mission */
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_TERROR_ATTACK);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_INTERCEPT);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BUILDING);
 	INT_ChangeIndividualInterest(0.02f, INTERESTCATEGORY_BASE_ATTACK);
 /* successful XVI mission */
 	INT_ChangeIndividualInterest(-0.3f, INTERESTCATEGORY_XVI);
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_HARVEST);
 	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_BUILDING);
 /* unsuccessful XVI mission */
 	INT_ChangeIndividualInterest(0.1f, INTERESTCATEGORY_INTERCEPT);
  	INT_ChangeIndividualInterest(0.05f, INTERESTCATEGORY_TERROR_ATTACK);
 	INT_ChangeIndividualInterest(0.01f, INTERESTCATEGORY_BASE_ATTACK);
 /* once every day (increases interest in supplying base */
 	INT_ChangeIndividualInterest(AB_GetAlienBaseNumber() * 0.1f, INTERESTCATEGORY_SUPPLY);


Pre-2.4 Conclusions

I've had a look through the code for generating UFOs, and increasing/decreasing nation happiness, and here are my suggestions. These suggestions are designed to address the worst problems (too many UFOs, nation unhappiness early in game) with a minimal solution to reduce the risk of seriously upsetting game balance for the 2.4 release.

Recommendations

1. Increase the cycle for mission spawning from 4 days to 8.

cp_campaign.h
#define DELAY_BETWEEN_MISSION_SPAWNING 		8
  • This will reduce the number of UFOs spawning by half.

2. Reduce the maximum possible missions per cycle from 40 to 25.

cp_campaign.h
#define MAXIMUM_MISSIONS_PER_CYCLE 			25
  • This will further reduce the number of UFOs spawning, and it will also prevent the numbers of UFOs spawning on average per cycle from exceeding 8, which is a hard limit in the game anyway. The peak (8), will only be reached around 1000 alien interest, far ahead of our current campaign (which seems to be peaking around 400).

Reasoning

The above two changes will drastically reduce the number of UFOs. Here's how I think they will effect other aspects of the game:

Nation Happiness:

Currently, the only things negatively effecting nation happiness are missions the player does not attack or completely fails (also UFO sales, see below). As a result, the high number of UFOs was making it very difficult for players who do not expand quickly enough. UFOs were showing up in unseen areas causing nation happiness to drop early in the game (<-- my hypothesis). Reducing the number of UFOs should slow down nations getting angry early in the game.

By reducing the number of UFOs, the player will have fewer to sell to nations. This should not effect things too much. The current code suggests that selling to one nation just angers others, so selling UFOs actually works (in aggregate) a negative effect on nation happiness:

cp_campaign.h
#define HAPPINESS_UFO_SALE_GAIN				0.02
#define HAPPINESS_UFO_SALE_LOSS				0.005

cp_uforecovery_callbacks.c
UR_DialogStartSell_f
	/* update nation happiness */
	for (i = 0; i < ccs.numNations; i++) {
		nation_t *nat = NAT_GetNationByIDX(i);
		float ufoHappiness;

		assert(nat);
		if (nat == nation)
			/* nation is happy because it got the UFO */
			ufoHappiness = HAPPINESS_UFO_SALE_GAIN;
		else
			/* nation is unhappy because it wanted the UFO */
			ufoHappiness = HAPPINESS_UFO_SALE_LOSS;

		NAT_SetHappiness(ccs.curCampaign->minhappiness, nat, nat->stats[0].happiness + ufoHappiness);
	}

If I understand this correctly, selling a UFO increases one nation's happiness 0.02 and decreases the other seven nation's happiness by 0.005, leading to an aggregate happiness effect of -0.015. At first I thought this was a mistake, but it might actually work well, because the player can channel the effect to those nations it is unable to effectively defend. Those it can defend will be happy enough because of the positive results of missions. This aggregate negative effect from something that seems like it should give a positive effect may actually cleverly balance the game mechanic.

There is a separate problem with mission performance/nation happiness, but that needs a separate solution.

Economy/Research/Production/Disassembly:

Players are likely to have less money because they have fewer UFOs to sell. As a result, the research/production will probably be slowed down a bit in relation to game time. But because the player will face fewer UFOs over that period of time, it will feel like the game is progressing more quickly. This means that the player may have less research completed by the time certain alien interest levels are achieved, but my guess is that this won't put the player in a very bad position. All indications from player's comments suggests the game gets pretty easy in the mid-to-late game, where this effect will be strongest.

Employees:

The player will face fewer UFOs between intervals where it receives additional soldiers/scientists/workers. This could mean that the player has more available employees than they need. But since this risks making one aspect of the game easier rather than more difficult, I think it is okay to leave as-is for 2.4 and adjust for 2.5 based on feedback.