Is reloading battles an exploit?

[psweetman1590]

I beg your pardon, but I don't quite get what you're saying... yes, larger sample sizes give a greater conformity to the predicted rates, but what has that to do with what I said?

I'll take a stab and assume you meant that a 4% chance to win ought to mean that 4% of all battles happen that way. It ought to be so, but it is not necessarily what is going to happen. If you flip a coin 20 times, it might come up heads each time - not what's supposed to happen, but it CAN, and that's what is being debated here.

If a 24/25 situation occurs in the game, what happens is that the computer will pick a random number between 1 and 25 inclusive. Any number other than 25 will result in a win for the modern armor we're talking about, and 25 indicates a loss. It is possible that four 25's will happen in a row, no? This chance is, actually, purely random, because there is no way to predict WHEN it will happen. We can tell how often it will happen, to an extent, however.

 

[timerover51]

Psweetman, I understand what you are saying. I design miniature and board war games where we throw actual dice, and miniatures or counters represent the actual units. In one case, I had one of my game tester's take four shots are pointblank range at a target, chances of missing were 10%. He MISSED with all four shots, 1 in 10,000 occurence. Only time that happened. You roll dice, sometimes you get weird results. However, when I consistently loose hitpoints from my Modern Armor, which is boosted to a 36/24/3 unit, to an unboosted spearman, I start wondering about the RNG. Loosing Modern Armor to a spearman in a city, I have no problem with that, considering what I have down to increase the defense of cities, but when it consistently happens in the open, on grassland or plain tiles, I do start wondering. Sit back, break out the dice, start rolling them based on the combat ratio, and do not get the same consistent results of lost hitpoints that the game RNG gives me. Have I determined what odds that it does take to replicate the game RNG, no, I have not, as I have other demands on my time. Hopefully, this fall I can do some analysis, and figure it out. But something odd is going on,

 

[anaxagoras]

The problem as I see it is that what people THINK they are comparing is often different from what they are ACTUALLY comparing. We can talk about Modern Armor versus warriors all day, but how many of us have really put that to the test? How many times are your MA really facing warriors? But cognitively, we often feel the same way when the MA loses to, say, a cavalry, so we lump the experiences together in a general category we think of as "losses that weren't supposed to happen". And, frequently, we forget the defensive bonuses. Perhaps that cavalry we were supposed to beat was actually on a hill, so our 24-attack monster unit actually had roughly a 17% chance to lose each round, not 4%. But we were still "supposed" to win, so the loss is an injustice.

Also, I often cognitively label the loss of a single hitpoint of an extremely dominant unit as an injustice. But the (unmodded) MA versus a warrior, even on level ground, has nearly a 1-in-20 chance of winning any given round, and multiplied by four rounds for a veteran defender means I have something like an 18% chance of losing at least one hit point. That's one combat in five, so the odds aren't really all that long. But to me it seems like they should be.

I often feel that my elite units are having worse luck than my vets and even regulars. But every time I've tried to put it to the test, I have found that I am mistaken.

Personally, I think the RNG is just a mechanical, unbiased RNG, and that my brain, which is carefully designed to look for patterns, is imposing an unjustified order on something that is really chaotic.

A powerful thing, evolution.

 

[timerover51]

Actually, Anaxagoras, I am running into quite a few of my modified Modern Armor verses spearman/warriors/medieval infantry types, as it is part of my reworking of the various combat ratings. The combats that trouble me are taking place on grassland terrain, with whatever defensive bonus that gives. I am taking damage to the Modern Armor is virtually every combat, and if I had not boosted the hitpoints of the armor by 4 and use a 2/4/6/8 system of combat experience, I would have lost probably 3 or 4 units of the Modern Armor. I have lost one of my modified tanks to a spearman, and one to a longbowman, again on grassland, with the longbowman being the attacker in the one case. Not sure of any of the saves I have would help, as I do not have "Preserve Random Seed" in use.

Intuitively, this strikes me as odd. I understand that the combat system does not allow for certain units to be unable to inflict damage on other units, such as galleys on battleships. That is quite easy to do in miniature and board games, which I have designed and play quite a bit. This clearly is a factor in my disatisfaction with the combat system. One can also argue that modern tanks are not as effective as WW2 era tanks, say the Sherman, at combatting large number of low-tech combatants, but again that cannot be reflected in the current combat system. Short of cracking the source code and taking apart the RNG, I cannot say for certain that there are biases built into it, but my perception is that there appears to be. I understand that this has been argued extensively in other posts, with no conclusive detemination being made. I will continue playing the game, but I am not happy with the combat system.

 

[anaxagoras]

Again, remember that MA v. warrior is quite a bit different from MA v. spearman or MDI, and your experience system makes it MORE likely that you will lose at least one hitpoint in a combat, not less (even though it also makes it more likely you will win the overall combat).

If you are facing a vet defender, you now have six combat rounds in which you could lose a point. Your modded MA against a vet spear on open ground has about a 94% chance of winning each round, but with six combat rounds guaranteed, you now run slightly more than a one-in-three risk of losing at least one point. That is almost exactly the same as an unmodded unit, which has only a 91+% chance to win each round, but only has to go four rounds to win. In both cases, your chances of losing at least one round is very close to 36%.

(The open terrain unmodded defense bonus, by the way, is 10%.)

I'll bet if you keep a log of combats, you will find that, given a sufficiently large sample size, you'll get very close to the expected results. If you do keep a log, remember to only log the combats that meet your specified criteria - better terrain or different defense factors have different expectations and should go in a different log.

 

[timerover51]

I think that in at least a couple of the cases, the Aztec unit doing the damage was a regular, but not positive on that. I can see your point with respect to the increased hitpoints for both sides giving more time for damage to occur. The combat system is a pure "step" system, loosing one hit point at a time, and not allowing for either instantaneous kills or multiple hit point loss in a combat round. I can understand the reasoning for that, from my own studies of actual combat damage results where you have the range of a near direct hit resulting in no damage to the proverbial "golden BB", where total destruction occurs as a result of a hit that should have inflicted minimal damage at most. The step system of hit point loss or the "cookie cutter" approach, where every unit inside of a certain radius is considered damaged while all units outside of that radius are considered untouched is the standard used by the US military, and in theory, represents the average result. However, the step approach works best with similar units at similar tech levels. The results of Omdurman in 1898 or the First Day on the Somme in 1916 are not reproducable using the Civilization RNG and combat system, but are the classic cases of what happens when a major difference in technology of equipement or tactics occurs. One side is massacred at very little loss to the other side. When teaching weapons effects and damage classes, I try to stess that while the step loss or "cookie cutter" approach is useful for trying to predict a large number of outcomes, it cannot predict any individual outcome, and even in predicting large number of outcomes, it is only as good as the information used in its preparation. Certain variables cannot be easily modeled, and if those are present, any prediction is a waste of effort, and can lead to a huge gap between actual and anticipated results, with corresponding false expectations of the amount of effort required.

Basically, it boils down to there are things you can do better with computers, i.e. being able to game on a solitaire basis without guaranteed success, and there are things that you can do using people, miniatures, and dice, which is more accurately duplicate actual combat results. I prefer achieving the latter, but with Civilizations, I will have to accept the former.

 

[anaxagoras]

That's a great summary, as usual, timerover. The Civ system certainly isn't perfect. But I don't think it is biased in favor of lesser units, either. It just doesn't match our expectations very well, given the unit labels and the system design.

 

[Tribute]

You could make a game of reloading. Play Always War and allow yourself the ability to reload.

The catch? You can't ever lose a single unit or city. Otherwise, you consider the game a loss. You must use preserve random seed, obviously. So..., the only way to make the RNG change is to reload the save and make attacks in a different order. If you find yourself trapped in a situation you can't get out of, you lose! (No reloading to anything besides the prior save.)