Patch comments

[chequita guevara]

I'm using patch 1.13b right now, but since this wasn't changed in 1.15b it should still be valid.

We have this new "four-die-roll" system for figuring combat. I'm not sure it's working. In my most recent game, three of my walled cities with fortified infantry were rolled over by cavalry. With between two and three times the defense v their offense, I should have seen a pile of cavalry corpses before they took any of those cities, let alone three in a row.

What happened?

[swat-spas2]

I believe this feature was removed in patch 1.11 (which wasn't released). Should be included in the readme for the patch file somewhere.

[Sarxis]

Yes, it was removed. I don't remember seeing it reintroduced in the latest versions. Was it?

[swat-spas2]

Not that I can tell. I don't really see why it should be to be honest. I'm generally happy with the combat model as it stands now. The 4 die roll as discussed (frequently) introduced alot of imbalance, mostly as far as intra-(not inter)era combat went.
The example provided is mostly just bad luck. Occasionally you will get a streak like that even randomly. I'm ok with that. You will get an opposite 'streak' now and then that benefits you as well.

[Jaybe]

chegitz guevara, if you would send me a PM when you change your avatar to something "less disturbing" I would appreciate it.

Meanwhile, I will have to ignore you, despite your worthy posts.

[ducki]

There was such an outcry from the community against this, with plenty of statistical analysis showing that it would adversely affect same-Age units to an astonishing degree, that it was abandoned.

It's also important to note, if you go back and read the original language, this "fix" was intended to remove the "perception of streakiness" in the RNG.

The RNG is about as good and random as you can get from a computer RNG. The human mind is designed to detect "unfortunate" patterns (as well as anomolies)- you don't complain when you get a string of victories with horsemen vs. fortified spearmen in cities, but if the AI were to win 3 in a row or more in the same fashion, we'd be decrying the RNG.

It's all about perception. There's nothing wrong with the RNG as repeated tests have shown, and the proposed fix had nothing to do with fixing the RNG - it was all about "fixing" the player's perception of the RNG.

[DrSpike]

Yeah it never happened, thank goodness. Now they have time to come up with a suitable way to attain their objectives.

[ducki]

I don't think they need to attain that particular objective.
The "problem" was that players perceived the RNG to be streaky. The best way to address that is to educate the players in regards to Pseudo-Random Number Generators, human psychology in relation to patterns and anomolies, and statistics.
So you occasionally get a super-spearman that defies logic, so what? Random numbers are just that. If there's a .01% chance that the oldest unit in the game can defeat the newest, guess what that means?
It means that it's possible for a spearman to defeat a tank on any given combat round.
Not bloody likely, but possible, and when it happens, you take notice because it is anomolous - that's human nature. I don't think the combat system needs changing, really. I'd much rather see programmer times spent making the AI a bit more... intelligent, less predictable, less manipulatable.

[Jaybe]

I'd much rather see programmer times spent making the AI a bit more... intelligent, less predictable, less manipulatable.

YES! Anything that the AI is to do should be based on Probabilities (e.g., 75-95%), not on absolutes.

[MattPilot]

since we are on the subject of RNG's, can someone expain this to me please?

How do random numbers factor in with the Attack & defense value?

If everything depends on the RNG, then why have attack/defense values?


As i understand it, a fortified spearman in a walled-town (100% bonus total), would have a defense of 4 points (2+100%). Assuming an attacker with a strength of 2, how would you calculate that?

I suppose you would get a % value (what is that equation?), and how does the Random number go with that %?

Thanks for any help!

[MrWhereItsAt]

I am not sure how exactly it is done, and I am sure someone will turn up with a better understanding than I, but for the moment here is a way in which this COULD work.

In the example you describe, you are ignoring terrain benefits as well, but let us continue to do so for simplcity's sake.

The defender has a 4 in 6 chance of winning each round of combat until one unit is dead. Thus, one way of bringing in the random number is that a random number between 1 and 6 is drawn for each round - if any number 1-4 are chosen , the hp is lost by the attacker, whereas if 5 or 6 are chosen the defender loses an hp. This would repeat until one or the other unit is lost.

There are further complications in combat such as promotion/generation of an MGL and retreat odds that would require additional random numbers to be drawn. Furthermore, things are made more complex again by the observation that luck tends to continue - the random numbers are not totally random, and so if you have been experiencing bad luck the trend is for that to continue for an unspecified number of random number rolls.

[Catt]

Originally posted by MattPilot
since we are on the subject of RNG's, can someone expain this to me please?

How do random numbers factor in with the Attack & defense value?

If everything depends on the RNG, then why have attack/defense values?


As i understand it, a fortified spearman in a walled-town (100% bonus total), would have a defense of 4 points (2+100%). Assuming an attacker with a strength of 2, how would you calculate that?

I suppose you would get a % value (what is that equation?), and how does the Random number go with that %?

Thanks for any help!

A & D values form the basis of the combat system. Using your example, a fortified spearman in a walled town (assume grassland) would have a defense of 3.7 (50% bonus for walls; 25% bonus for fortified; 10% bonus for grassland -- 85% bonus to base defense of 2 means 3.7 effective defense). Assume an attacker with its "2" A value (let's say an archer to avoid discussing retreat chances of the horseman).

The basic combat calculation for an attacker's chances in each round of combat (each HP) is A / (A+D) where "A" and "D" are the effective values for the participants reflecting any bonuses available to their base values. Obviously, a defender's chances would be D / (A+D).

In our hypothetical, the basic value would be 2 / (2 + 3.7) or 2 / 5.7 -- in percentage terms, the attacker would have a 35.08% chance of success in each round of combat.

In order to determine whether the attacker wins or loses an HP, a random number is generated by a random number generator (RNG). Common computer RNGs work on a scale from 0 to 1023, meaning a possible random number from among 1024 possibilities. The RNG result would be compared to the probability of the the event that "calls" on the RNG -- in our case, it is one round of combat between our 2 attacker and our 3.7 defender. Our attacker wins 35.08% of the time. 35.08% of the possible 1024 outputs of the RNG is 359. Our attacker will win an HP if the RNG returns a value among the highest 359 possible outcomes of the total of 1024 (could also be lowest -- main point is that it is 359 of 1024, or 35%). If the RNG value returned is 665 or higher, the attacker wins; if 664 or lower, the attacker loses. Each HP requires a new random result from the RNG to determine whether the "odds" were met or not for that specific event.

In our example, assume that a vet attacker lost the battle, but managed to knock 2 HPs of the defender -- all told 6 HPs were lost, meaning that 6 "calls" were made to the RNG. The RNG output and practical in-game output might look something like: 879 (A win); 650 (D win); 1102 (A win); 600 (D win); 124 (D win); and 273 (D win).

It might be helpful to think of it as: (1) A vs. D sets the odds; (2) the RNG is a dice roll to determine whether the good guys (you!) beat the odds.

Catt

EDIT: Walk away while composing a post, come back to finish and submit, and you run the risk of a 20-minute differential cross-post! MrWhereItsAt summed it up nicely before me. END EDIT

[chequita guevara]

So, let's see, 10 defense, walled, fortified, in a desert,=17.5 defense. attacker strength 6.
6/23.5 or roughly 1 in four. So I should have killed 20+ cavalry to my six infantry, instead of the three that did die. I got hosed.

I was curious that he didn't use his forces to counter-attack my invasion. At least my main invasion is succeeding while the AI's only taking my far flank.

[vmxa1]

You are correct about the def bonus 175. If you put 30 calvs against 6 infantry and run 10 trials it comes out 21.8 survive.

So you kill 8 on avg.

Edit: I ran that with the calvs attacking the inf. You have to factor in the retreat, it is large. I used all vets on both side. It would actually be more likely that the calvs would have a number of elites and the inf none. In that case you kill less. If the inf attack, then the bonus do not exist.

[swat-spas2]

Actually shouldn't it be 18.5D? 10% terrain, 50% walls, 25% fortified = 85%... each cav should have a good shot (1/3ish) of running away, but only about 7% of winning against fresh vet infantry. I've not often seen 30 AI cav attack, but if it were to happen, it should be ugly for both sides. Bad luck on your part.