Talk:Proposals/Obsolete Proposals/Stats, Experience and Mission Count in 2.5-dev

From UFO:AI
Jump to navigation Jump to search

Most of the data for this is available at the following URL:

http://docs.google.com/spreadsheet/ccc?key=0AoMT-GwBufRGdDMwanlOeVBPYmVIMXZPSGN6MldORUE#gid=3

Calculating Accuracy

Accuracy and weapon skills, don't improve effective accuracy enough. The below chart shows the accuracy (as displayed in % chance to hit when aiming) of various weapons with various stats (45/40 = 45 weapon stat / 40 accuracy).

  • 20/20 = average rookie
  • 45/40 = low vet
  • 60/50 = high vet
  • 80/60 = superstar (unexpected, definitely not more than 1 or 2 per game)

Accuracy-with-weapons.png

The gain between an average rookie and a high vet is just not enough, I think. Especially not between an average rookie and a low vet. I'd like to see a high vet achieve the accuracy gains of a superstar, maybe even a little more. But its not easy to figure this out. The chart is based on % chance to hit, but this is probably not even related to the calculations done to actually fire the weapon, which are more complicated.

The following description of the math from DarkRain is useful:

accuracy = 1 - (accuracy_ability / 100 + weapon_skill / 100) / 2
(which equals to: take the average of both, divide it by 100 and then rest that from one)
first for parabolic throws (grenades and such):
first, the angle (pitch, yaw, roll) for a perfect throw to the intended target is calculated, then   it is modified this way: 
pitch += <random number  between -1 and 1> * 2 * (spread1 * (0.5 + 1 * accuracy))
yaw += <another random number between -1 and 1> * 2 * (spread2 * (0.5 + 1 * accuracy))

now for straight shots (bullets, beams, knives, anything travelling in a straight line):
again the pitch, yaw, roll angles for a perfect shot is calculated, and then modified:
first get two Gaussian distributed random numbers with median at 0 and standard deviation of 1,   let them be gauss1 and gauss2, 
then calculate a secondary injury multiplier (this one was here before wounds were implemented and  I haven't touched it):
if HP / maxHP < 0.5 injurymultiplier2 = 1 + 0.2 * ((1 / (HP / maxHP + 0.5)) - 1) * 100 / mind
else injurymultiplier2 = 1
then:
pitch += gauss1 * (spread1 * ((0.5 + 1 * accuracy) * injurymultiplier2))
yaw += gauss2 * (spread2 * ((0.5 + 1 * accuracy) * injurymultiplier2))
if the shooter is crouched, each of the above two is additionally multiplied by the crouch factor  from the fireDef

Thanks to DarkRain's spreadsheet, I think I've got an idea of how to adjust the accuracy calculation to dramatically increase the improvement of veteran soldiers. Currently, the accuracy calculation is:

accuracy = 1 - (accuracy_ability / 100 + weapon_skill / 100) / 2

If we modify this to:

accuracy = 1 - ((accuracy_ability - 10) / 50 + (weapon_skill - 10) / 50) / 2

Then we end up with regular vets (50 skill, 40 accuracy) getting a significant bonus to accuracy, equivalent to the bonus we'd get with a super vet (80 skill, 60 accuracy) using the existing formula.

See the spreadsheet called "Scratch Sheet" http://docs.google.com/spreadsheet/ccc?key=0Ajol7mzNZeF-dGpaOVpRdXFQQy1hck9vU1Zxb29MbEE#gid=4.

You can change the values at the top and see the old equation accuracy versus the new equation accuracy. This is a pretty dramatic change, and SuperVets (those who manage to get 80 skill and 60 accuracy) will really be extraordinary shots. I expect this to be pretty rare, though, and I think players will really notice soldier improvement (not just assume it based on stat improvement) Any thoughts? --H-hour 18:10, 6 August 2013 (CEST)

I third this proposal! I say we give it a try. --DarkRain 05:15, 7 August 2013 (CEST)

Speed Growth Mechanisms

I'm not really a fan of how speed is improved. It encourages tedious maxing out movement wherever possible, even using the last 4 or 6 TU of each soldier. Maybe that's fine and not a big deal, but I wonder if we should consider some alternatives. Here are two: --H-hour 00:02, 19 March 2013 (SAST)

  • Large Movements: Give a soldier bonus points for moving with more than 75% of his TU in a turn. We could also give a lot of bonus for using 75% of his TU and getting a kill in the same turn. This would not eliminate the tedious maxing, but would at least eliminate all the using up of any tiny excess TU. This may be more difficult to track and code, though, and would be less reliable for balancing stat gain.
    • This is too complicate and impossible to explain to the player. I like the cumulative aproach much more. --ShipIt 07:53, 19 March 2013 (CET)
  • Cumulative Growth: Like HP, we could just make this one a function of all the total XP accumulated elsewhere.

Stat Maximum

  • I am in favour of this. The gain is too slow atm. But I still think there should be a limit to what stats a human physically can reach. This limit should depend on the starting value instead being an absolute number, to reflect the physical conditions of the soldier. --ShipIt 08:49, 19 March 2013 (SAST)
    • With a maximum number set it should be possible to tweak the values in a way to make it possible to get to the max within ~30 missions. --ShipIt 10:24, 19 March 2013 (SAST)
      • Stats will max at 100, but I don't think this is what you want. I'm not entirely sure I agree with you on this. If soldiers hit an absolute max cap regularly, this could create too much of a disincentive to keep the soldier at home, as further experience would be wasted on them. It would also probably render meaningless the bragging rights you get with your top soldier (remember when you and I traded screenshots of our top soldiers?) if the player ends up with several soldiers all topped out around the same numbers. HOWEVER, I would be in favor of tapering the experience curve more, so that after a certain amount of stat gain the soldier must get a lot more exp for further development. This would still provide that incentive to spread the exp around but would be a softer limit, allowing top soldiers to still develop some (albeit slowly) and players to play how they want (note that some skills, like strength, will really not be able to get to 100 even if they go in ~60 missions). If you look at the first graph below, the equation I've proposed is the orange line. I like how this equation tapers early -- the first 2,500 exp gives twice as much of a stat boost as the next 2,500 exp. But it really straightens out around 5,000. Maybe DarkRain knows some math that would taper the post-5,000 more so that it flattens out a bit? Or maybe we just use some code that calculates the bonus for the first 5,000 exp differently from the second 5,000 exp? What do you think about this? --H-hour 14:22, 19 March 2013 (SAST)
        • Note: with the system I describe the soft max will be different based on initial stats. --H-hour 14:22, 19 March 2013 (SAST)
        • The most direct way of altering the curve is changing the exponent: moving it closer to 1 will make the soldier improve less quickly at the start compared to later stages, but he will reach higher values more easily. On the other hand, moving it closer to 0 will make the soldier improve more rapidly at the start, but almost not at all once he becomes a veteran. If a good growth rate can be achieved changing this, fine, if not then we can think of a new method for calculating the stats growth. --DarkRain 19:07, 19 March 2013 (SAST)

Exp-to-stat-curves.png

        • Here are a couple of curve options. They don't seem to flatten out quite as much as I would like and I wonder with the other equations if the stat gain is too quick at the start. But I'm not sure... opinions? --H-hour 23:44, 19 March 2013 (SAST)
        • I just updated the image with another curve that uses a flatter equation for all exp over 5000. It doesn't look like such a dramatic difference, to be honest. THoughts? --H-hour 00:04, 20 March 2013 (SAST)
          • Is that the right image? I don't seem to see any difference at all --DarkRain 19:45, 23 March 2013 (SAST)
            • The caching in the wiki is severe. Try clicking on the image, then CTRL-R refreshing the image page. It should update with the new image. --H-hour 21:03, 23 March 2013 (SAST)
    • On a related note: should we look into adjusting the max experience per mission? I noticed that some of the averages in your games were way below what the max per mission would allow, will this provide an incentive for players to grind to max experience? On the other hand, the stats with the larger difference between the average and the max are also the hardest to grind efficiently -weapon skills- due to limited ammo and targets, so maybe this isn't such a concern after all? --DarkRain 19:07, 19 March 2013 (SAST)
      • I thought I had responded to this but I guess it was just in my head. I think this is a really difficult thing to know in advance and plan properly for. This is because I don't know if stat gain for any particular stat is based on a few missions with high values or based on receiving low values every mission. In the first case, increasing the max gain could lead to a large increase in overall exp received during a campaign. In the latter case, increasing the max gain wouldn't change much. I think it's worth separating which skills might be which. The following is just guesses: --H-hour 16:54, 3 April 2013 (SAST)
        • More exp comes from a smaller subset of the missions where the soldier performs very well:
          • accuracy I think it would make sense for this skill to max at 2/3 the max of the weapon skills, so that it couldn't grow as fast as a weapon skill but wouldn't grow too slowly.
          • mind This stat is still a bit up in the air until psionics is invented. I suggest leaving it as it is for now.
          • weapon skills If we set all weapon skills to 150-per-kill, that means the 4 kills will hit the 600 max (actually 680, but would need 5 kills for that I guess). I don't know about others, but it was pretty rare for any individual soldier to get more than 4 kills in a mission, so I'm guessing we're not hitting the max here. My feeling is that the max could be kept at 600 for this, simply because it's really not that easy to get more than 4 kills in a mission, so the incentive for grinding has a strong counter-balance.
          • hp (maybe) Health might be getting a big bonus from some missions if a soldier reaches their weapon skill max. The max on this seems huge, and HP is already growing pretty well compared with other abilities. I'd need to check a graph that shows exp-per-mission for it, but maybe this max can be reduced.
        • Exp is fairly consistent from mission to mission:
          • strength Strength seems fairly good as it is, and would just need to be adjusted to fit the new formula as the proposal page specifies.
          • speed I think I need to generate a new graph for this one, to show exp per mission, to know if we're vastly under the max-per-mission or hitting it. The max will obviously need to be adjusted if we follow the current proposal to raise its rate by 3x.

Ability calculations

In /game/q_shared.h I found the following:

#define GET_HP( ab )        (std::min((80 + (ab) * 90/MAX_SKILL), 255))
#define GET_ENCUMBRANCE_PENALTY(weight, max)	(1.0f - ((weight) > (max) * WEIGHT_HEAVY ? WEIGHT_HEAVY_PENALTY : (weight) > (max) * WEIGHT_LIGHT ? WEIGHT_NORMAL_PENALTY : 0.0f))
/** @todo Skill-influence needs some balancing. */
#define GET_ACC( ab, sk )   ((1 - (((float)(ab) - 10) / (MAX_SKILL / 2) + ((float)(sk) - 10) / (MAX_SKILL / 2)) / 2))
#define GET_MORALE( ab )        (std::min((100 + (ab) * 150/MAX_SKILL), 255))
#define GET_TU( ab, md )		(MIN_TU * (md) + (ab) * 20 / MAX_SKILL)

This suggests that the EXP->points equation is different than for the skills. I'm noting this here to investigate precisely where and how these numbers are used, to see if refinement is needed after the changes (with, ie, morale->panic/rage, etc). --H-hour 18:59, 8 August 2013 (CEST)

GET_HP calculates HP from strength and this macro is used only in one place (G_Damage(), g_combat.cpp) to calculate the max HP an unit will have after being healed in combat, TBH I'm not really sure what this is supposed to achieve, it seems that this number will be higher than the soldier's max HP most of the time anyway, I would like to remove this macro. --DarkRain 01:08, 9 August 2013 (CEST)

Useful Graphs

Exp-to-stat-equations.png

Skill-exp-per-mission.png

Skill-stat-per-mission.png

Ability-stat-per-mission.png