Combat mechanics

[deducter]

The random combat results are irrelevant for MP, since luck always average out in the end. I've never won or lost a MP game because of luck, and I have played probably 150ish games.

[Rudankort]

There is something wrong the way the generator creates numbers. I call it grouping when it happens.
It's like there is one first hidden roll that has an equal chance to land between 1-100, but then most other rolls form a cloud around this virtual number.
If I would visualize the random number generator results it would look like a long barreled shotgun pattern. A few pellets will fall astray at the far edges but most would form a cloud around the aiming point. And this "grouping" happens far to often and is then the source of strange combat results.

Let say you attack a target with your ROF 140%, strength 10 flametank and have a 10% chance for a kill, so if dice rolls between 90-100 it results in a kill. If that first hidden dice roll land, let say on 95 most of the 14 dice rolls land between 85-100, leading to out of bounds results. You can reverse the scenario to a 90% kill chance, if this first hidden dice roll goes between 1-10, most of the other rolls will be in the lower digits. Leading to an rather surprisingly ineffective attack.
If that would happen here and then, than this would be just bad luck/good luck. Shit happens right? However when I see this grouping pattern on a regularly basis, than this is a hint of an underlying problem with the generator.
What I would like to see is this grouping to be gone. This "dice rolls fall into the same range" get fixed, and each dice having an equal chance to land between 1-100. That is all what I could wish for.
I don't have a problem with spread in battle results. I think this spice up the game and it did it in all 5-Star series games I ever played. However within reason and without loaded dices.

It may seem that the problem lies indeed in the "grouping". But how to prove this?

I've just made a simple test. First, I resolved a combat 1000 times using existing random number generator. Then, I resolved the same combat 1000 times using a huge array of "true" random numbers obtained from http://www.random.org/. At least in theory, random numbers generated by this web site have a perfect uniform distribution, you cannot expect a pseudo-random generator used in games to perform better. A quote from this web site:

RANDOM.ORG offers true random numbers to anyone on the Internet. The randomness comes from atmospheric noise, which for many purposes is better than the pseudo-random number algorithms typically used in computer programs.

In both tests I calculated error as:

abs(AttackerResult-AttackerPrediction) + abs(DefenderResult-DefenderPrediction)

This means that any point of difference between prediction and actual result was counted. In my test I used Poland scenario and attacked polish cavalry with german infantry. This is a rather bloody combat with predicted result 4-4. So, if actual combat resolution were 4-4, the error would be 0. If combat resolution gave 3-4, 5-4, 4-3 or 4-5, error would be 1. Etc.

Here are my results:

Code: Select all

[0]	60    49
[1]	186   190
[2]	249   269
[3]	238   227
[4]	165   159
[5]	68    62
[6]	22    28
[7]	8     14
[8]	4     2

First column is error, second column is how many times existing random generator created this error, and third column is the same for "true" random generator. As we can see, their performance was pretty similar. I don't see any advantage of the "true" number generator in this test. So, the question is, are you guys really sure that using a true random generator in this game would change the situation (and your impression of it) all that much? I seriously doubt this. In fact, most likely, you would not see the difference.

[Longasc]

"Things cannot be random if I have extremely bad luck twice in a row! This should not be possible!"
At the same time people tend to forget the positive results rather quickly.

I really don't think the random number generator is broken, it's rather player perception. And well, you can't fix players with a patch.

What I would like is a checkbox for "seed" or no "seed" in single player games, but as I would only need this to replay some moves for demo purposes and otherwise not at all it is only a minor wish. A similar checkbox could also exist for "chess" mode.

But there is a problem. Not everyone is a forumite and I am NOT sure if adding these options for the public would not rather make the game worse for them.

Part of the fun is the random element, taking a chance and being prepared for the worst. For instance sometimes a single, lone and unsupported Infantry can win against enemy Infantry but it could also lose. Risk it or wait for reinforcements? If the results would be fixed like in chess mode, the situation would be very "tame" and slightly less exciting. But it would also be more predictable of course. An example that you can't have your cake and eat it.

[Mordan]

Funny results give an advantage to the defender. a bouncing unit out of reach. good random result never benefits the attacker. Why? Because the units you provided for the attack just lose a turn doing nothing. and when the unit bounces out of reach, the turn is lost for all attacking units.

i just don't like how randomness plays out. in practice it gives you situation that are wholly unrealistic. like a 9 Rus Tank attacks a 5 Brumbar with no ammo. Guess what? No casualties on both sides. The bloody unit does not even bounces!! While the prediction was like 0-3. You can't advance on the square adjacent to the city. Ennemy keeps the town.

Every turn there is a whacky result which makes the game . You can't really take risk on the manoeuvre side of units since as soon as a real bad random result occurs everything goes to the toilet, while good result don't really help. it slows the gameplay.

I NEVER complained in PG II multiplayer games. and luck was a factor but not so outlandishly as here in PC.

@Longasc yea we forget them but i sure feel bad for the other player when i see it happening. i don't have any sense of glory or accomplishment. just outlandish luck. because tell me in the examples I gave you the rationale for an engine to produce such results on the battlefield consistently?

if there was LUCK tokens to give to units to prevent such events, it could be more interesting which would allow some pushes to actually have a chance to succeed. Right now I don't even want to play since the success of your tactical moves is based on 1)overwhelming force 2)luck

[Rudankort]

The point is, the way randomness works in Panzer Corps stems from its combat mechanics, not quality of random generator. And this combat mechanics is by now established. Players are used to it. Good players know how to use it to their advantage. We cannot possibly change all this and suddenly make PzC a completely different game. So sorry guys, but we are not going to change status quo in the patches.

However, I've made a mental note that existing randomness is not to everyone's liking, and in future games I'll see if this can be improved.

[huertgenwald]

However, I've made a mental note that existing randomness is not to everyone's liking, and in future games I'll see if this can be improved.

[naughtybalrog]

Thanks Rudankort for thinking of ways to imporve this area

[Longasc]

The consistent outlandish luck of the AI and other players ...

You don't fix this by "fixing" the random number generator. You want to prevent things that in your opinion should not happen. Make things more "predictable".
That's rather something I wish for the bonus SE unit assignments which are totally random.


But has anyone an idea how to make random less random and more predictable so that "outlandish" things don't happen? -> the "chess" mode is the only solution IMO.

[soldier]

Funny results give an advantage to the defender. a bouncing unit out of reach. good random result never benefits the attacker.

I'm not sure how random results could only benefit the defender ?
A retreating unit might allow you to advance and take an important hex on the field. Even if you can't always destroy your enemy you should at least reduce his his reinforcements. If your enemy units keep bouncing away then its likely your not leaving enough forces to destroy them. You just need a reserve unit to follow up the attack and mop up the leftovers . Check the angle you attack from, Its not always beneficial to just push the enemy away from you.
The AI units often retreat or bounce when i attack them with powerful units or combined arms assaults but i'm rarely forced to leave the weak units to restrengthen and haunt me next turn. Thats a waste of time and effort

[deducter]

If the confusion is how retreat/surrender mechanics work, the answer is simple: a unit will retreat/surrender if all its remaining steps are suppressed and its entrenchment at the time of attack is 1 or 0 (after the attack, entrenchment = 1 will go to entrenchment = 0). The unit will try to retreat in the opposite direction of the attacker, although it can retreat "forward" if there's no other choice.

As for units retreating and reinforcing, one thing to keep in mind is that killing units is not necessarily that important in MP. Often neutralizing a unit is just as good. For instance, knocking an artillery down from 10 to 3 is enough to effectively take the artillery out of the fight. It at the minimum must spend a turn to reinforce, more if you can park a unit or two next to it, then it has to spend 2 turns: 1 to withdraw, 1 to reinforce. That's effectively as good as destroying it. Don't get fixated on individual units or combat results.

[Mordan]

Funny results give an advantage to the defender. a bouncing unit out of reach. good random result never benefits the attacker.

I'm not sure how random results could only benefit the defender ?
A retreating unit might allow you to advance and take an important hex on the field. Even if you can't always destroy your enemy you should at least reduce his his reinforcements. If your enemy units keep bouncing away then its likely your not leaving enough forces to destroy them. You just need a reserve unit to follow up the attack and mop up the leftovers . Check the angle you attack from, Its not always beneficial to just push the enemy away from you.
The AI units often retreat or bounce when i attack them with powerful units or combined arms assaults but i'm rarely forced to leave the weak units to restrengthen and haunt me next turn. Thats a waste of time and effort

I'm talking about extreme results. I'm doing quite fine thx for the usual retreating. The game engine works fine in this aspect. By the way, I don't play single player much. Mostly MP games.

Just tell me how the following isn't absurd which happened to me in MP Dniepr: T34 attacks infantry on Russian side of river. Infantry retreats one square to Dniepr River. 3 Strength left. Fine. I planned for that. Not wanting to take risks i launch Artillery attack. Infantry down to 1. 1 suppressed. NOW TELL ME I attack with a 10 strength infantry, ableit form a river square, the 1 infantry just bounces away into German territory. This is one of the most extreme case that happened to me in 40 MP games. There should be a threshold where OVERRUN is activated, because this is insane amount of luck. How can 10 times strength infantry can miss a single sitting duck on river banks that has been pounded by artillery? There are too many cases like those IMO in the current engine. Like 88mm AA miss or get 1s or 2s. Then when I'm losing and without a single plane anymore, this AA get a 8 shot!!! Never happened to me before ever. Why can't I have a nice 0-1-2-3-4 distribution? Which leads me to the question are the dice rolls loaded to help the losing side?

I really think LUCK tokens should be implemented as a game feature when an engine produces such results. The idea of LUCK token is : each player has 2 or more luck tokens which they can give to units they want. Those luck token protect those units from bad luck but does not give them good luck.

[deducter]

Singling out 1 case of extreme bad luck in over 40 games is exactly that: 1 case of extreme bad luck. Each game has hundreds of combat results, and in 40 games there were probably over 10,000 combat results. Even if 100 of those results are very bad luck, that's still 1% of the total.

I've lost plenty of MP matches. Every single one of those was because the opposing player was better than me. I have never won or lost because I got lucky.

[soldier]

Like 88mm AA miss or get 1s or 2s. Then when I'm losing and without a single plane anymore, this AA get a 8 shot!!!

lol, you gotta love it when your units start putting in superhuman efforts but the battles already over... Its a bit late guys

[LeChevalier]

I'm talking about extreme results. I'm doing quite fine thx for the usual retreating. The game engine works fine in this aspect. By the way, I don't play single player much. Mostly MP games.

Just tell me how the following isn't absurd which happened to me in MP Dniepr: T34 attacks infantry on Russian side of river. Infantry retreats one square to Dniepr River. 3 Strength left. Fine. I planned for that. Not wanting to take risks i launch Artillery attack. Infantry down to 1. 1 suppressed. NOW TELL ME I attack with a 10 strength infantry, ableit form a river square, the 1 infantry just bounces away into German territory. This is one of the most extreme case that happened to me in 40 MP games. There should be a threshold where OVERRUN is activated, because this is insane amount of luck. How can 10 times strength infantry can miss a single sitting duck on river banks that has been pounded by artillery? There are too many cases like those IMO in the current engine. Like 88mm AA miss or get 1s or 2s. Then when I'm losing and without a single plane anymore, this AA get a 8 shot!!! Never happened to me before ever. Why can't I have a nice 0-1-2-3-4 distribution? Which leads me to the question are the dice rolls loaded to help the losing side?

I really think LUCK tokens should be implemented as a game feature when an engine produces such results. The idea of LUCK token is : each player has 2 or more luck tokens which they can give to units they want. Those luck token protect those units from bad luck but does not give them good luck.

Use the prediction (after selecting your attacking CTRL+click on the target) or press L key after the engagement. You'll see much info there.

I also read about your post in another thread:
Retreating is not directly determined by randomness, if you knew when an attacked unit retreats or not. The dice rolls determined how many strength points of target unit get killed and/or suppressed. Take a look at Tips & Tricks sticky thread's topics about retreating.

[Chris10]

like most people i like randomness but too much randomness kills the joy of randomness.

in other words you like randomness as long as it works as you expect it to work in every situation ...
unpredictable results are part of warfare...there is this ol saying: "No battle plan survives the first attack..."

[huertgenwald]

...there is this ol saying: "No battle plan survives the first attack...

I'll stick to : "I love it when a plan comes together" (Col. Hannibal Smith)

[Dragoon.]

There is something wrong the way the generator creates numbers. I call it grouping when it happens.
It's like there is one first hidden roll that has an equal chance to land between 1-100, but then most other rolls form a cloud around this virtual number.
If I would visualize the random number generator results it would look like a long barreled shotgun pattern. A few pellets will fall astray at the far edges but most would form a cloud around the aiming point. And this "grouping" happens far to often and is then the source of strange combat results.

Let say you attack a target with your ROF 140%, strength 10 flametank and have a 10% chance for a kill, so if dice rolls between 90-100 it results in a kill. If that first hidden dice roll land, let say on 95 most of the 14 dice rolls land between 85-100, leading to out of bounds results. You can reverse the scenario to a 90% kill chance, if this first hidden dice roll goes between 1-10, most of the other rolls will be in the lower digits. Leading to an rather surprisingly ineffective attack.
If that would happen here and then, than this would be just bad luck/good luck. Shit happens right? However when I see this grouping pattern on a regularly basis, than this is a hint of an underlying problem with the generator.
What I would like to see is this grouping to be gone. This "dice rolls fall into the same range" get fixed, and each dice having an equal chance to land between 1-100. That is all what I could wish for.
I don't have a problem with spread in battle results. I think this spice up the game and it did it in all 5-Star series games I ever played. However within reason and without loaded dices.

It may seem that the problem lies indeed in the "grouping". But how to prove this?

I've just made a simple test. First, I resolved a combat 1000 times using existing random number generator. Then, I resolved the same combat 1000 times using a huge array of "true" random numbers obtained from -snipped- URL . At least in theory, random numbers generated by this web site have a perfect uniform distribution, you cannot expect a pseudo-random generator used in games to perform better. A quote from this web site:

-snipped- URL offers true random numbers to anyone on the Internet. The randomness comes from atmospheric noise, which for many purposes is better than the pseudo-random number algorithms typically used in computer programs.

In both tests I calculated error as:

abs(AttackerResult-AttackerPrediction) + abs(DefenderResult-DefenderPrediction)

This means that any point of difference between prediction and actual result was counted. In my test I used Poland scenario and attacked polish cavalry with german infantry. This is a rather bloody combat with predicted result 4-4. So, if actual combat resolution were 4-4, the error would be 0. If combat resolution gave 3-4, 5-4, 4-3 or 4-5, error would be 1. Etc.

Here are my results:

Code: Select all

[0]	60    49
[1]	186   190
[2]	249   269
[3]	238   227
[4]	165   159
[5]	68    62
[6]	22    28
[7]	8     14
[8]	4     2

First column is error, second column is how many times existing random generator created this error, and third column is the same for "true" random generator. As we can see, their performance was pretty similar. I don't see any advantage of the "true" number generator in this test. So, the question is, are you guys really sure that using a true random generator in this game would change the situation (and your impression of it) all that much? I seriously doubt this. In fact, most likely, you would not see the difference.

I thought I set to email notification, I almost forgot about this.
Anyway I started to notice this first when the game came out. I thought it was just my imagination. Like you're tricked into thinking certain events always seems to happen on full moon, when in fact it's just the light in the sky making things more memorable.
To rule out this is my imagination because some attack goes haywire I started to take screenshots. For this purpose the Windows 7 own Snipping Tool served great as you can take screenshots of just the result window.
Here some screenshots from my last session 3 weeks ago in AK.

EDIT: I'm not allowed to post URLS. But I send you a PM on the Matrix forums with the link to my Imgur Photo Album.

But there is something you make me remember that I have not considered previously. While I have taken screenshots to document this "grouping"several times in a single scenario, I forgot to count the total number of battles fought. You're right there is a difference is this happens 15 times in 100 battles or 500 battles.

Well, the weekend is ahead a new DLC is out. Let's talk again after I finished this new DLC I'm eager to play.