Ground Combat Balance Discussion Build V21 Experimental 4!

[Kalthon]

That does indeed explain alot, any chance you can check the range modifiers for Sebillians too, they seem to be proficient snipers but have never hit me within 5 tile range. Inverse range modifiers or something?

[Chris]

You're actually right, the aliens aren't receiving the short range hit bonus (it's not inverted though). Ugh. No wonder they're so passive - they've got no incentive to get up close to your troops.

Good spot there!

EDIT - actually, it may just be that the short range hit bonus isn't included in the debug output for the accuracy calculation. We'll look into it anyway.

[Betuor]

I probably just missed this, but when did guards get put on light scouts? I like it, but it does make it a whole lot more lethal. Good incentive to bring a rocket.

[StellarRat]

You're actually right, the aliens aren't receiving the short range hit bonus (it's not inverted though). Ugh. No wonder they're so passive - they've got no incentive to get up close to your troops.

Good spot there!

EDIT - actually, it may just be that the short range hit bonus isn't included in the debug output for the accuracy calculation. We'll look into it anyway.

You going to get a quick balance update out to fix the short range bonus thing, the .5 difficulty modifier, and the LMG accidently getting nerfed? Or can you tell us how to fix the XML ourselves?

[Chris]

The aliens not getting the short range bonus thing wouldn't be an XML update, it'd be a code issue. However, it's a false alarm anyway. Turns out I was reading the debug output wrong.

The other two aren't worth a new post by themselves and I don't want two versions of the balance patch going around one thread and confusing people.

Betuor - it was a while ago now, just before V20 Stable I think.

[Betuor]

Betuor - it was a while ago now, just before V20 Stable I think.

It's been a couple more builds than I thought since I last played it. So much for me choosing a higher difficulty. Nice addition to make the game a little bit more challenging in the beginning now. I like it.

[Max_Caine]

Max, I still don't understand the fundamental premise of your argument. You're arguing that because a shot can kill a Xenonaut outright in most cases, it's more beneficial for them to take two shots at say 45% each than it is for them to take a single shot at 95%?

Mathematically that's simply not true, though. The optimal course of action there is to take the single more accurate shot as it's a more efficient way to spend their TU to cause maximum casualties. If the two shots were at 50% then they would be better, but that's not the case at the moment.

*sigh*

I guess that unless I gather adequate footage to show what the AI is doing, you're not going to be convinced. Okay, I'll try one last time, this time using the numbers from the game.

Let's take early game. The Caesan guard, standing, armed with a plasma rifle hits (unmodified):

A single burst shot: 42%

A snap shot: 56%

A normal shot: 70%

An aimed shot: 84%

The AP cost for each aim type is very close to the next, from 20 AP for a snap to 28 for an aimed. A Caesan guard has 50 AP, so they should select the most appropriate shot type, i.e. the most accurate. Aimed cost of 28 AP for, that leaves 22 AP (7 tiles) for movement which is plenty. No brainer, right?

Yes, and no. Chris, you look at the individual probability for each shot and say "well, this is 42% and that is 84% - no question that the shooter should always select 84%". What I have been trying to get at is you're not looking at the cumulative probability that one of the shots fired within the AP available will hit the target. Going back to that online calculator, there's (rounding up) a 20% chance that out of three trials (shots) by a burst rifle that no shots will hit. However, there is a 42% chance that exactly one shot will hit, and a 38% chance that more than one shot will hit. Cumulatively, there's an 80% chance that one or more shots will hit, which for a burst shot is very close to what an aimed shot can do.

Similarly, for two snap shots (which equal the AP cost of a burst shot) there's a 49% chance exactly one shot will hit, and a 31% chance one or more shots will hit, for a cumulative total of 70%. While aimed shots beat snap shots, those two shots are cumulatively, the equivalent of a single, normal shot.

I believe the AI hedges its bets (which is a smart move). But I'll never prove this to you without evidence, so I'm going to go off and get some.

[StellarRat]

Similarly, for two snap shots (which equal the AP cost of a burst shot) there's a 49% chance exactly one shot will hit, and a 31% chance one or more shots will hit, for a cumulative total of 70%.

I don't know about that calculator, but my math is showing that with two 56% chance to hit shots the odds that AT LEAST one round will hit are 81%. There is a 31% chance that two rounds will hit. That means the average damage inflicted per turn is much higher with two snapshots vs. one aimed shot. That's why I've been saying that the aimed shots are not done correctly. The math doesn't support it's use in most cases. However, without a full analysis at all ranges it's not possible to comment on whether or not there is some range/cover combo where aimed shots are better, there should be though. Edited February 13, 2014 by StellarRat

[Max_Caine]

StellarRat, I've deleted my post because I've realised I need to gather actual proof that what I say is happening is happening. There's no argument other than actual hard evidence that will work. Soz.

[StellarRat]

StellarRat, I've deleted my post because I've realised I need to gather actual proof that what I say is happening is happening. There's no argument other than actual hard evidence that will work. Soz.

OK, but I think you are on to something.

[Skitso]

That does indeed explain alot, any chance you can check the range modifiers for Sebillians too, they seem to be proficient snipers but have never hit me within 5 tile range. Inverse range modifiers or something?

I've seen similar close range behaviour from the sebs (second post): http://www.goldhawkinteractive.com/forums/showthread.php/8997-V21-E4-Xenonaut-died-with-health-left

There has to be something there!

[Chris]

No, I don't need proof that it's happening in game - it's not relevant at the moment. I just need a maths lesson so I can understand the problem you're bringing up.

Basically, you need to prove to me that it's advantageous for the AI to burst fire with a plasma rifle rather than take an aimed shot. The numbers are:

<Set1 ap="25" accuracy="85"

<Set2 ap="30" accuracy="100"

<Set3 ap="35" accuracy="120"

<BurstFire ap="40" accuracy="40"

I'll look at the post again tomorrow, but as far as I'm aware three shots at 40 accuracy using burst fire gives you 120 cumulative accuracy for 40 AP spent, whereas you can get one shot at 120 cumulative accuracy for 35 AP with an aimed shot. Therefore the aimed shot is better. You need to prove my maths wrong before I understand the problem, if there is a problem.

EDIT - having read your mathematical proof above (using admin magic), what you're saying seems to agree with what I'm saying. An aimed shot is a more efficient use of AP than burst fire is, and two snap shots have a cumulative 80% chance of hitting with at least one shot - making them the same as burst fire.

However, burst fire is 10 AP cheaper than two 2 snap shots, so why would the aliens ever use two snap shots? And an aimed shot is 5 AP cheaper than burst fire, but is 4% more effective according to your figures.

Therefore, what I'm asking you is why you think it would be beneficial for the AI to not be using aimed shots. We can discuss what's actually happening in the game once I understand the assumptions you are making. Everything I've seen so far suggests that the AI should be taking aimed shots in preference to any other action.

[Tegron]

Is it normal that in my first mission under normal difficulty the aliens are killing and hitting my guys without missing at all, even when they are suppressed?

[KateMicucci]

That's normal if you get too close to them. Are your guys behind cover? They don't last long in the open until you get some armor. Scout with a shield, car, or somebody ugly and keep everybody else out of alien's visual range.

[StellarRat]

No, I don't need proof that it's happening in game - it's not relevant at the moment. I just need a maths lesson so I can understand the problem you're bringing up.

Basically, you need to prove to me that it's advantageous for the AI to burst fire with a plasma rifle rather than take an aimed shot. The numbers are:

<Set1 ap="25" accuracy="85"

<Set2 ap="30" accuracy="100"

<Set3 ap="35" accuracy="120"

<BurstFire ap="40" accuracy="40"

I'll look at the post again tomorrow, but as far as I'm aware three shots at 40 accuracy using burst fire gives you 120 cumulative accuracy for 40 AP spent, whereas you can get one shot at 120 cumulative accuracy for 35 AP with an aimed shot. Therefore the aimed shot is better. You need to prove my maths wrong before I understand the problem, if there is a problem.

EDIT - having read your mathematical proof above, what you're saying seems to agree with what I'm saying. An aimed shot is a more efficient use of AP than burst fire is, and two snap shots have a cumulative 80% chance of hitting with at least one shot - making them the same as burst fire.

However, burst fire is 10 AP cheaper than two 2 snap shots, so why would the aliens ever use that? And an aimed shot is 5 AP cheaper than burst fire, but is 4% more effective according to your figures.

Therefore, what I'm asking you is why you think it would be beneficial for the AI to not be using aimed shots. We can discuss what's actually happening in the game once I understand the assumptions you are making. Everything I've seen so far suggests that the AI should be taking aimed shots.

You can't add the 40% together to come up with accuracy. You have to multiply the chances to miss out and subtract that from 1 then multiply by 100. So, .6 x .6 x .6 = .216 x 100 = 21.6% chance to miss with burst fire or an 78.4% chance to hit with at least one round. You also have a 16% chance to hit with two rounds and a 6% chance to hit with all three in the same burst. Also, accuracy above 100% doesn't help you as a hit is a hit. The average damage over time for in this case is going to be very close. If you assume the average HP inflicted from a plasma rifle is 50 HP (just made that up, BTW). In 100 separate firing "events" with aimed fire you'll do 5000 points of damage 100x50=5000. For burst fire it's tricky. You've got on 2850 HPs from single hits (79 - 22) x 50 = 2850, 1600 points from double hits (16x100), and finally 900 HP in triple hits (6 x 150) for a total of 5350 in damage. Someone please correct me if I'm wrong. For snapshots vs. aimed the difference is way more pronounced. Snapshots are far better.

Also, and this is important, a Xenonaut can often survive a single hit and get under cover before being killed then you can use first aid and get them right back into the fighting, but double and triple hits will probably kill a troop in Wolf instantly so the survivability factor also needs to be taken into account. That's a BIG deal, IMO. Also, I'm not factoring in suppression, destruction of cover, etc...

I'm not really sure the "efficient" use of AP is the proper way to look at this in any case. As, in mind at least, it's more a question of single firing events. AP efficiency doesn't matter much if you can't fire more than once per round. There aren't too many cases where either the aliens or the Xenonauts get more than one shot of any kind off before they either kill the target, they are killed or one of them goes back under full cover. I'm much more fearful of weapons that can kill with a single firing then ones that take multiple hits to kill even if the damage over time is equal.

Edited February 13, 2014 by StellarRat

[Max_Caine]

Tegron, while I'm busy modelling risk in the background, I'll take a moment to answer that question. The answer is yes, aliens can. The damage a plasma weapon does is not affected by suppression. Accuracy is not affected by suppression. What is affected by suppression is AP, which is halved. A Caesan guard at 25AP can fire a normal shot at 24AP for a 70% chance to hit, or a snap shot at 20AP for a 56% chance to-hit. A Sebillian guard at 30AP can fire an aimed shot at 28AP for a 72% chance to hit, a normal at 24AP for a 60%, or a snap at 48%. Meaning that aliens have (in general) a 1 in 2 chance of hitting a solider.

[Tegron]

To be fair the guys were not fully behind cover. I was behind a wooden farm like fence. But I'll try again and see.

Another thing that I don't understand is When my fighters early on intercept a scout, if its automated its a win 100%, but if I try to do it manually its 100% fail.... Maybe I'm doing it wrong, are there youtube/video tutorials how to manually fight properly? If there is a link would be great!

[KateMicucci]

I think that I am with Chris in not understanding exactly what you're saying the problem is

Aliens only use burst/snap shots because it is the most efficient, and you would prefer they also use normal and aimed? Aliens are using fire modes that aren't TU efficient? And also what is the solution to the problem?

[StellarRat]

I think that I am with Chris in not understanding exactly what you're saying the problem is

Aliens only use burst/snap shots because it is the most efficient, and you would prefer they also use normal and aimed? Aliens are using fire modes that aren't TU efficient? And also what is the solution to the problem?

My understanding is that there doesn't appear to ever be a good reason to use anything other than burst and snapshots with the current AP and accuracy numbers. That applies to some of the Xenonaut weapons too. The solution is to fiddle with the numbers a little more. IMO, the accuracy of aimed fire needs to go up or the accuracy of snap and burst fire needs to go down. Like I said without an analysis of damage over time vs. AP at the full spread of ranges it's really hard to say at this point. This is perfect problem for some spreadsheet and graph analysis.

[Waladil]

I don't know what Max_Caine's original post was but Chris, probability doesn't work like that. I don't know how mathematically smart the AI is, but you CANNOT simply add the accuracy values and call them cumulatively equal.

Simple proof for this concept:

Take a simple six-sided die. Then roll it six times. Count the number of times you rolled a 3. Do this 5-10 times. You'll notice that you usually get about one 3 per six rolls. However, you do not ALWAYS get exactly one 3. Sometimes you get zero. Sometimes you get two. It's even possible you might get six 3s. Now, take that same die, and set it down on the table so that it shows a 3. How many 3s did you get? One.

So six iterations of a 1 in 6 chance is different than one iteration of a 1 in 1 chance.

I'll provide some better XenoNumbers now.

A sniper, who is aiming at a Caesan guard in cover, has 90 TUs and doesn't need to turn or move before shooting.* She has the following shot choices:

Two aimed shots for 42 TUs each (46% chance to hit per)

Three snap shots for 30 TUs each (31% chance to hit per)

In addition, there are a few sub-par options such as two normal shots. We don't need to deal with those.

Since a sniper rifle has 100 accuracy for snap shots, and 150 accuracy for aimed shots, they have exactly the same cumulative accuracy. The two aimed shots are marginally more TU efficient, costing only 84 rather than 90. Therefore, AI calculating merely cumulative accuracy and TU costs will choose the two aimed shots.

Percentage analyses:

2 iterations of 46%;

P(0 hits): 29.16% P(1 hit): 49.68% P(2 hits): 21.15%

Average # of hits: 0.9198 (0.92)

3 iterations of 31%;

P(0 hits): 32.85% P(1 hit): 44.27% P(2 hits): 19.89% P(3 hits): 2.97%

Average # of hits: 0.9296 (0.93)

Discussion:

The two possibilities resulted in extremely close average hit numbers. Certainly closer than I expected. However, there is still a distinct difference between them. In addition, the shot choice that is likely to produce the greatest absolute damage is to fire three snap shots, which is the TU-inefficient choice.

While the two choices resulted in similar average results, they are different in ways that are important to a discerning player (or AI). Taking the two aimed shots will hit more reliably; if the target is weak and you're more worried about DEALING damage rather than MAXIMIZING it, then the two aimed shots are a superior choice (being about 3.5% less likely to result in no damage at all). Contrariwise, if the situation is bleak and the player needs a miracle, then taking three snap shots has a slim chance of hitting all three times, perhaps just providing that very miracle.

Also of XenoNote is the suppression system which encourages making more shots rather than less, even if that results in less accurate fire. So in summary, anyone with the goals of doing as much damage as possible while suppressing as much as possible (read: everyone including the AI) would choose to fire three snap shots.

*This is a 100% real situation EXCEPT that I increased the sniper's TU count to 90 in order to allow for that arrangement of shots. Her %chance-to-hit values are completely unchanged from what the game calculated.

[Max_Caine]

Waladil, I wish I hadn't deleted that post now. Le sigh. Anyway, I had found that the cumulative accuracy of a burst shot, 2 snap shots or an aimed shot as fired by a Caesan guard to be very close. I had come to 84% for the aimed shot, 80% for the burst hitting with at least one shot and 79% for snap hitting with at least one shot.

The AI has a distinct AP advantage if it fires a single aimed shot - 22 AP over 10 AP, which translates to a greater freedom of movement (7 tiles over 2), or the possibility of another shot, a snap shot. Furthermore, at that stage in the game the chance of any plasma shot killing a squaddie is very high - the base damage is 65 with a mitigation of 20. 65 is more than most starting troopers. The variance introduced by the random element can result in a minimum of 32pts to a maximum of 97. With an mean HP of 54 (based upon starting 20 games and looking at the hire screens 20 times), damage caused must drop by a minimum of 19% to not kill. So I, who wants the AI to use normal/aimed shots, am looking at the situation and asking why does the AI not use normal/aimed shot, because I see it consistently using snap/burst (depending on weapon).

-------------------------------

On a totally different and totally about the update subject

Chris, Jackal armour isn't much good now. My people keep dying in it, and thanks to this debate, I've just worked out why. It's that mitigation of 20. It means Jackal armour only absorbs 15pts of damage, and that's not nearly enough to even mean "almost dead" except when the RNG smile on me and put the random damage modifier in my favour. My current strategy in the game is to tech first to Foxtrots then to Wolf ASAP, because Wolf is the first protective suit of armour I can wear.

Edited February 13, 2014 by Max_Caine

[StellarRat]

Waladil, I wish I hadn't deleted that post now. Le sigh. Anyway, I had found that the cumulative accuracy of a burst shot, 2 snap shots or an aimed shot as fired by a Caesan guard to be very close. I had come to 84% for the aimed shot, 80% for the burst hitting with at least one shot and 79% for snap hitting with at least one shot.

The AI has a distinct AP advantage if it fires a single aimed shot - 22 AP over 10 AP, which translates to a greater freedom of movement (7 tiles over 2), or the possibility of another shot, a snap shot. Furthermore, at that stage in the game the chance of any plasma shot killing a squaddie is very high - the base damage is 65 with a mitigation of 20. 65 is more than most starting troopers. The variance introduced by the random element can result in a minimum of 32pts to a maximum of 97. With an mean HP of 54 (based upon starting 20 games and looking at the hire screens 20 times), damage caused must drop by a minimum of 19% to not kill. So I, who wants the AI to use normal/aimed shots, am looking at the situation and asking why does the AI not use normal/aimed shot, because I see it consistently using snap/burst (depending on weapon).

Well, Max, I have serious doubts the AI calculates "chance to kill" vs. "how to maximize damage" to do that the AI would have to know how much HP the target had to start with which would be "cheating" as that is not known by observation. The only thing it could know is what kind of armor the target is wearing and how much damage it's already done to the target (which would probably also be cheating.) Also, given that there is variance in weapon damage assuming X number of hits is going to kill your target probably isn't a very wise course of action. You might get a low "roll". Finally, there is really good chance you aren't going to get another chance to engage the same target, so going for max. damage is really the better choice.

My question is whether or not aimed fire is ever a better choice for max. damage. IMO, at some range it should be otherwise it's really pointless option to even have in the game. I would hope that at longer ranges it pays to use aimed fire. IMO, snap and burst should be much less accurate than they are now and useful only at close range or for suppression.

Also, I believe there shouldn't be much AP difference between snap and burst fire. A three shot burst is basically one trigger pull on an autofire capable weapon. The only reason you choose a single shot is to conserve ammo. Something the Xenonauts normally wouldn't be concerned with.

Edited February 14, 2014 by StellarRat

[KateMicucci]

Chris, Jackal armour isn't much good now.

I see a huge increase in my soldier's survivability with Jackal. Most of my troops are able to survive taking a hit or two with it. So I disagree that skipping to Wolf is a good idea, at least if you care about keeping your promoted troops alive in the early game.

[StellarRat]

On a totally different and totally about the update subject

Chris, Jackal armour isn't much good now. My people keep dying in it, and thanks to this debate, I've just worked out why. It's that mitigation of 20. It means Jackal armour only absorbs 15pts of damage, and that's not nearly enough to even mean "almost dead" except when the RNG smile on me and put the random damage modifier in my favour. My current strategy in the game is to tech first to Foxtrots then to Wolf ASAP, because Wolf is the first protective suit of armour I can wear.

I agree with Max here. The situation with Jackal has gotten much worse because the AI gets plasma rifles much sooner in the game now. It's nearly useless against those. So, there is a VERY small window where Jackal is useful now and that makes it not worth the workshop time and cost to even bother with it in the first place.

[Tegron]

I think that I am with Chris in not understanding exactly what you're saying the problem is

Aliens only use burst/snap shots because it is the most efficient, and you would prefer they also use normal and aimed? Aliens are using fire modes that aren't TU efficient? And also what is the solution to the problem?

No I was saying that no matter what the aliens did it was always a "hit" or "Kill" where my guys were "missing" the aliens.