What is your alternative to an RNG???
RNG - Civ's Detractor?
- Thread starter CivCube
- Start date
It's been my experience with statsitical analyzation for my Combat Calculator that the Civ III RNG does give properly distributed results (e.g. bell curve) over many trials. I did over 11,000 trials in my Retreat study and the RNG seems fine. I have also did 15,000+ tests on unit combat in Civ III to check it against my calculator. The Civ III RNG is fine.
Also, how would you produce 'bell-shaped curves' without an RNG... It is integral to the process.
Also, how would you produce 'bell-shaped curves' without an RNG... It is integral to the process.
It does compare the stats now though... The formula for the attacker winning the round is A/(A+D+DBonuses). Are you saying all battles should be resolved by this formula without regard to hp? I don't understand what you are trying to say.
Bomber Escort: atm, the RNG gives NO bellshaped distribution of the number it comes up with to compare to the win/loss number. Since it comes up with random numbers the probabilities fit in the long run - but for 1 single fight the chance of getting an absurdly high or low number is the same as getting an average number. This means, that win probs depend on the A and D stats and the boni.
What I am saying is that average numbers should be more common, too, which would simply make extreme results for single HP fights rarer than they are.
I think Bamspeedy did quite some nive work on probabilities a while bakc (last fall?).... I'll try to find it and then explain to you what I mean.
Fact is: as it is now, the RNG makes rare results come too often - on a small map three screwy figths in a row can break you!
What I am saying is that average numbers should be more common, too, which would simply make extreme results for single HP fights rarer than they are.
I think Bamspeedy did quite some nive work on probabilities a while bakc (last fall?).... I'll try to find it and then explain to you what I mean.
Fact is: as it is now, the RNG makes rare results come too often - on a small map three screwy figths in a row can break you!
The way I understand it works is:
Attacker: 2
Defender: 10
Defender Bonus: 10%
Attackers Chance of Winning is (2/(10+1)) = 0.181818
If the RNG generates a number less than this, the attacker wins the round if it doesn't it loses. The last number the RNG generates has no bearing on the next number. In the case above the RNG has a 8,182 chance out of 10,000 of getting a number that is not in the attackers favor. Your right, bell shaped distributions cannot apply to single cases, since the word 'distribution' implies more than one case. I understand your statement, but I do not see how you would make certain random numbers more common than others... they would no longer be 'random numbers'
I believe the problem is truly this:
People focus too much on the A/D values and not enough on the hp. For example:
Which unit is stonger in a battle, a...
6.4.1 1hp unit or 3.2.1 2hp unit
At a glance it may appear to be the 6.4.1 unit, but it isn't...
6.4.1 1hp attacking 3.2.1 2hp (No Bonus) = 6.4.1 win is 56.25%
3.2.1 2hp attacking 6.4.1 1hp (No Bonus) = 3.2.1 win is 67.34%
The experience matters more than is gathered at a glance... A better example is:
4.6.1 3hp Regular Rifleman v. 4.1.1 5hp Elite Longbowman
Longbowman attacks Rifleman on Grassland winning 53% of the time.
This also stresses the importance of Bombardment and why us modders get so upset that the AI doesn't use it effectively. It is a huge advantage to knock the opponent down even one hp.
My point is that the A/D value can sometimes be deceptive if taken alone and not in the context of hp, terrain etc... To say a spearman beat a tank is not enough information... to say an Elite Spearman beat my unfortified 1hp Tank on Grassland is... BTW this has a approximate 42% chance of this happening and if it does happen you deserve it. If I was playing a game with you a this situation happened and I could attack you with a spearman I would do it, maybe even twice if the first attempt failed... Occasionally I have even seen the AI attack redlined units with spearmen and win...
EDIT: A general rule of thumb that I use for quick calculation of strength is this:
If my attacking unit has twice the hp of the defender I halve the defenders defense... example from above:
4.6.1 3hp Regular Rifleman v. 4.1.1 5hp Elite Longbowman has a 53% chance of winning (from my calculator).
A/D Calculation only: 4/(4+6+.6) = 37.73% (this is also the chance of the longbowman winning each round)
A/D Calc modded as above: 4/(4+3+.3) = 54.79% (half the D, 10% of that for the bonus)
Notice this is very close to the actual number which is 53%... Just thought I share this statistical observation I've made after 20,000+ trials...
Also, if my attacker has half the hp of the defender, I halve the attackers Attack value... This gives an excellent reference for quick calcs in MP games.
Attacker: 2
Defender: 10
Defender Bonus: 10%
Attackers Chance of Winning is (2/(10+1)) = 0.181818
If the RNG generates a number less than this, the attacker wins the round if it doesn't it loses. The last number the RNG generates has no bearing on the next number. In the case above the RNG has a 8,182 chance out of 10,000 of getting a number that is not in the attackers favor. Your right, bell shaped distributions cannot apply to single cases, since the word 'distribution' implies more than one case. I understand your statement, but I do not see how you would make certain random numbers more common than others... they would no longer be 'random numbers'
I believe the problem is truly this:
People focus too much on the A/D values and not enough on the hp. For example:
Which unit is stonger in a battle, a...
6.4.1 1hp unit or 3.2.1 2hp unit
At a glance it may appear to be the 6.4.1 unit, but it isn't...
6.4.1 1hp attacking 3.2.1 2hp (No Bonus) = 6.4.1 win is 56.25%
3.2.1 2hp attacking 6.4.1 1hp (No Bonus) = 3.2.1 win is 67.34%
The experience matters more than is gathered at a glance... A better example is:
4.6.1 3hp Regular Rifleman v. 4.1.1 5hp Elite Longbowman
Longbowman attacks Rifleman on Grassland winning 53% of the time.
This also stresses the importance of Bombardment and why us modders get so upset that the AI doesn't use it effectively. It is a huge advantage to knock the opponent down even one hp.
My point is that the A/D value can sometimes be deceptive if taken alone and not in the context of hp, terrain etc... To say a spearman beat a tank is not enough information... to say an Elite Spearman beat my unfortified 1hp Tank on Grassland is... BTW this has a approximate 42% chance of this happening and if it does happen you deserve it. If I was playing a game with you a this situation happened and I could attack you with a spearman I would do it, maybe even twice if the first attempt failed... Occasionally I have even seen the AI attack redlined units with spearmen and win...
EDIT: A general rule of thumb that I use for quick calculation of strength is this:
If my attacking unit has twice the hp of the defender I halve the defenders defense... example from above:
4.6.1 3hp Regular Rifleman v. 4.1.1 5hp Elite Longbowman has a 53% chance of winning (from my calculator).
A/D Calculation only: 4/(4+6+.6) = 37.73% (this is also the chance of the longbowman winning each round)
A/D Calc modded as above: 4/(4+3+.3) = 54.79% (half the D, 10% of that for the bonus)
Notice this is very close to the actual number which is 53%... Just thought I share this statistical observation I've made after 20,000+ trials...
Also, if my attacker has half the hp of the defender, I halve the attackers Attack value... This gives an excellent reference for quick calcs in MP games.
BE: actually, it is done in numbers between 0 and 1023.....
now, if the RNG works as it should you get the numbers 1, 55, 999, 375 and 0 with roughly the same proability.
Imagine is that was not the case - it is not hard to write a number generator that draws numbers at random from a base that has a bell distribution of numbers....
now, if the RNG works as it should you get the numbers 1, 55, 999, 375 and 0 with roughly the same proability.
Imagine is that was not the case - it is not hard to write a number generator that draws numbers at random from a base that has a bell distribution of numbers....
In my youth, I enjoyed playing chess. I was quite good at it and enjoyed the non-random element of the game. The best player won. No excuses.
After several years, I stopped improving and stopped enjoying the game. I could beat players who were worse than me, draw with players who were about the same and lost to players who were better than me. The fun went.
I took up backgammon and bridge. Both have a random element. Both have high skill levels. Good players win 90% of the time - but not all. There is always a chance. I've beaten international players in bridge tournaments when I got lucky - and have lost to players who arent as good too sometimes. You know what - I enjoy the unpredictability of it all much more.
A Civ that was too mechanistic, that held no surpises, that doesnt give the underdog a chance would not keep my interest long. I hate games that I've 'worked out'. Some games you can win every time by following a certain strategy. Shortly after that I stop playing them.
I've played Civ now for a long time - through 1 2 and 3. I like randomness. I like the RNG. Sure in the long run it all evens out. Every study showing 100+ tests shows it to be fair. But this time and in this battle, I may win - or lose. And long may it remain so.
After several years, I stopped improving and stopped enjoying the game. I could beat players who were worse than me, draw with players who were about the same and lost to players who were better than me. The fun went.
I took up backgammon and bridge. Both have a random element. Both have high skill levels. Good players win 90% of the time - but not all. There is always a chance. I've beaten international players in bridge tournaments when I got lucky - and have lost to players who arent as good too sometimes. You know what - I enjoy the unpredictability of it all much more.
A Civ that was too mechanistic, that held no surpises, that doesnt give the underdog a chance would not keep my interest long. I hate games that I've 'worked out'. Some games you can win every time by following a certain strategy. Shortly after that I stop playing them.
I've played Civ now for a long time - through 1 2 and 3. I like randomness. I like the RNG. Sure in the long run it all evens out. Every study showing 100+ tests shows it to be fair. But this time and in this battle, I may win - or lose. And long may it remain so.
col, I fully agree with you on the need for randomness - but you talk about 90%!
it is the evening out after 100 battles that gets on my nerves: if you get stiffed for 20 battles in a row you have already lost, and that happenes regularly in PTW. The same, btw, applies the other way round: if you win 20 battles by pure luck - what have you planned for?
Superevie
Latina Goddess™
Actually there are gong to be natural diasters in conquests. I'm not really looking foward to it though, unless it happens to another civ than mine.
Also, some of the terrian changes with global warming. It isn't a happy event, but it is a random one notheless. 
I like some of the randomness, but some of it is ridiculous. I hate when I lose an entire stack of horsemen to two stupid spearmen. It makes me so mad!
These two statements cannot be reconciled with each other...
I don't believe this is possible and still retain the 'randomness'
The RNG is fine and is working properly. Human nature tends to notice patterns where none exist. To test this, and this would take some time. Record the outcome of every single battle for three games. A/D values, terrain, etc. and post the results... You'd see that the results are similar to what a combat calculator would predict.
well, well, seems you are trying very hard not to understand:
the RNG makes the rare results come too often for the game we play
get it!
the RNG is NOT SUITABLE FOR PTW!!!!!
Moderator Action:
--Padma Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
sorry for my outburst, but I get real tired of people always tring to find error or mistakes in my posts about the RNG instead of trying to get twhat I mean. Been ther, discussed this with 100,000 people. and tired of doing it again.
Moderator Action: Apology accepted.
--Padma
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
Moderator Action: Apology accepted.
--Padma Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
It's in your mind... the RNG is fine.
I understand perfectly what you are trying to say and you are in error. Please read my previous posts. The RNG is fine and this has been demonstrated several times by several different people. Its all mathematics... the RNG is not flawed, a basic understanding of mathematical statistics is required to see it though, personal experience and memory are unreliable over many trials. It's not a matter of opinion but fact.
anarres
anarchist revolutionary
I think carlos means he still wants the numbers to be random (i.e. unpredictable from one to the next), but he doesn't want the underlying distribution they come from to be uniform.
BomberEscort, when you consider the RNG in civ you are thinking of random numbers that are chosen from a uniform distribution - that is, where every number possible has exactly the same chance of occouring as every other number.
carlos, when you talk about the RNG you would like to see in civ you envisage an underlying normal distribution - the numbers are still random (i.e. not predictable from one to the other), but you get more of the 'mid-range' numbers.
Here is a normal distribution. the x axis is the number to be generated, the y axis is the probability of generating that number.
note that for the current RNG you would have a horzontal line through the graph (all numbers have same prob of being picked).
Now why you two end up shouting and arguing like kids is beyond me, but meh, no skin of my nose....
P.S. BomberEscort - carlosMM is no idiot and he does understand about probability.
I know that a series of wins that are against the odds can swing the game and that is what you hate cMM but in my opinion that is part of what makes the game interesting. Civ must have a chance for the underdog to win.
Of course it is a matter of subjectivity as to whether unlikely results happen too often or not. It may only need one battle at a key time to swing against the odds for the game to swing. Real battles may be like that too. Its better to be lucky than skilful
Strategically we need to plan for the possible outcomes of a battle. Sure we can attack at 60% odds in our favour or wait for 90% odds in our favour. Either carries gains and risks. I personally - and I suspect you too - tend to attack with the odds slightly in my favour. I think you have to acknowledge that this carries a big risk.
I think the RNG is fine. The odds of winning a victory ARE in line with what you would expect when it is trialled 100 times. Its just too bad if you happen to be unlucky. I want the rare events to occur with a probablity roughly what the A/D numbers suggest they should.
Civ contains that element of chance and I for one think it is a much better game for it.
Of course it is a matter of subjectivity as to whether unlikely results happen too often or not. It may only need one battle at a key time to swing against the odds for the game to swing. Real battles may be like that too. Its better to be lucky than skilful
Strategically we need to plan for the possible outcomes of a battle. Sure we can attack at 60% odds in our favour or wait for 90% odds in our favour. Either carries gains and risks. I personally - and I suspect you too - tend to attack with the odds slightly in my favour. I think you have to acknowledge that this carries a big risk.
I think the RNG is fine. The odds of winning a victory ARE in line with what you would expect when it is trialled 100 times. Its just too bad if you happen to be unlucky. I want the rare events to occur with a probablity roughly what the A/D numbers suggest they should.
Civ contains that element of chance and I for one think it is a much better game for it.
