Getting hit WAY MORE than the percentage the game suggests

[botski]

I am on my second playthrough of Book 1, to which I give Basilik a hearty round of applause for giving me such a wonderfully intense, difficult, satisfying gaming experience!!, and I am wondering about why I am getting hit so often as a fighter. My first game was as a mage and I never got physically hit (or else I would die) so I generally avoided it.

I am fightig goblins at about level 3. Early in the game.

the last three goblin encounters went as such:

Goblin #1 - Chance of hitting me was 33%
He hit me 8 out of 13 times.

Goblin #2 - Chance of hitting me was 57%.
He hit me 11 out of 11 times!!!!!

Goblin #3 - Chance of hitting me was 32%
He hit me 7 out of 7 times.

What the heck. Is that percentage buggy? Are there other percentages going on behind the scenes? if so, I would think they should be displayed to us so we can make informed decisions.

Cool! thanks for any tips.

[Kreador Freeaxe]

Hey, Botski,

This has been discussed a bunch of times, but longer term examination of happenings shows that the percentages really are correct. Consider it like flipping a coin and getting heads 11 times in a row. It can and does happen all the time. However, your perception of being hit is more memorable than your perception of NOT being hit, so there may well be fights where your opponent never lands a hit on you, but you don't remember it as much.

Here in Vegas it's like the people who talk about how much they won this night or that night, but somehow they're always broke before their next paycheck arrives.

[botski]

Thanks for you reply, K.F.,

I will look for the past discussions, but I am not imagining this. The opponent is hitting me at a rate 2-3 times that what is posted (of course, if the percentage listed is 77% then my statement is false, more like 100/77th of the time ). For each time he hits me 7/7 times on a 33% chance, I should be seeing an encounter of him hitting me 0/7 or even 1/7 - but that doesnt happen.

[TheBuzzSaw]

This goes way beyond Eschalon games. I have noticed this issue in a wide variety of RPG and strategy games. Battle for Wesnoth comes to mind: it will claim I have 80% chance to hit, but I miss all three attacks; it will claim the opponent has 20% chance to hit, but I get hit all three times. In various RPGs, the case is the same. The percentages are way off from the actual results. How on earth are the percentages calculated???

[Randomizer]

Percentages are the chance of that particular hit happening. You can work out the percentage chance of a string of hit and/or misses occurring and see that streaks are not probable but entirely possible.

This is why for proper probability calculations you have to do a Monte Carlo simulation where you repeated play the same combat sequence over and over to see how many times you get a streak and how many are closer to what you expect.

[TheBuzzSaw]

We know how random sequences work. We're pointing out that the overall distribution does not hold true. I was not complaining about 80% happening to miss 3 times; I was trying to point out that despite ALWAYS having 80%, I ALWAYS miss all 3 times. That is the issue. The point is that these occurrences SHOULD be rare, but they are not. I run into the problem highlighted in the OP all the time in RPGs.

[CrazyBernie]

It just means that your computer sucks at rolling dice.

Pseudorandom number generators aside, I've seen it from both sides... I've been on the recieving end of a beating from an enemy with a 20% chance to hit me, and I've single-handedly destroyed a swarm of enemies when I should have had problems defeating one.

The problem is, we only tend to notice it when we're getting the beat-down. If you tracked your hits and misses throughout the whole game, I'm willing to bet that the numbers would prove to match up with the percentages. BW has even addressed this in the past by creating an app that simulated Book I's "dice rolling engine." Sometimes it would appear to swing lopsidedly to one side, but it would *always* average out.

So I'd argue that the overall distribution *does* hold true, even if it isn't on a "per game" basis. Statistically speaking, there'd be far more people complaining about it if it were the case... 0_o

[ReligionIsDumb]

I would just like to add that I actually do Monte-Carlo simulations for my job. RNGs are very good these days at producing the desired uniform distribution over a large number of subsequent rolls. However, depending on how the game acquires the seed, you can get systematic behavior.

Say, for example, that the default game seed for the RNG is 1234. Say that this random seeds makes the first 10 rolls suck really bad due to pure chance, but as you play longer, this evens out. Well, what if you quit and load the game again. Do we go back to the 1234 seed? Two common methods to get random seeds are as follows. 1. Everytime the RNG generates a random number, the seed changes. Save this seed on exit and you continue on from where you left off and you don't start with the default seed. But if a crash occurs or read/write error, you can end up stuck with the same seed again when you load and get crappy rolls for your first 10 rolls again. 2. Some games grab a seed using some sort of number in the computer (system time for example). But if the game is running in an environment where it doesn't have access to the number it is looking for. It can get stuck using the default seed on startup every time producing systematic behavior since, typically, after 10 crappy attacks you have to load the game...and your seed reverts to the default or seed that was saved with your game save.

To the OP, the developers of the game would need to tell you how they do the RNG, but if you have done a mathematical analysis and are getting a skewed distribution consistently, I would ask if the computing environment is unusual (wine or some other OS emulator or a strong security setting that may prevent a game from having read access to system variables such as the time) or if you notice the bad attacks immediately after a game load or immediately on startup. Seed acquisition is the true bane of RNGs. Rarely does one find a problem in any of the standard RNG algorithms. They are very good.

Just food for thought.

RID

[SpottedShroom]

Book I and II both seem do a decent enough job of seeding their RNGs, since you can save before e.g. opening a chest and reload if you don't like the outcome. Book II also gives you the option of including your RNG state with your saved game for additional challenge.

But I'm pretty sure the people who think the RNG cheats are just getting a run of bad luck, and that neither the algorithm or the seeding process are at fault.

[SigilFey]

This is something that is part of the nature of all RPG's (and a good many other non-RPG games out there) that rely on random number generators for anything. Truly, until some genius manages to write ten million lines of code for almost infinite number sequences, this will always be a problem.

The problem lies with how a computer generates "random" numbers. No program can generate a truly random result--it must follow rules. So, if I want a computer to generate a random number between 1 and 10, I must specify which number values it is limited to. I must also write code to specifically instruct it not to choose the same number over and over again. All of this code, already, opens up a thousand possibilities for error in programming or lines of code that override or exclude other lines. Think about how hard it would be to create code that told a computer:

"Choose the number 3. But not always. Choose a different number once in a while. But not a number that's not on the list. But don't ALWAYS choose another number. Choose 3 sometimes. Just not too many times in a row. As for the number 4..."

This is why most RNG's are not actually "rolling dice", they're creating random lists ahead of time then assigning the next, pre-calculated number to the next software call for a random number. In the case of RPG's with stats that increase your chances to strike an opponent or cast a spell, there are even more pre-calculations involved that can result in even more funkiness. It's actually next to impossible to get a good system going. On the one hand, excluding lower values completely can result in such regular and high numbers that the challenge gets sapped out of the game. On the other hand, too much calculation can result in such random highs and lows that it seems as if the characters in the game never really improve.

In short, I, too, have had some moments of utter frustration, where a little red lizard was able to take my level 5 character down to half health, or my 84% chance-to-hit result in nine consecutive misses (despite two successful dodges). On the whole, though, Eschalon's system is a LOT more balanced that games like...sayyy...The Temple of Elemental Evil.

[CrazyBernie]

I must also write code to specifically instruct it not to choose the same number over and over again.

Huh? Where's the randomness in telling an RNG not to select the same number? If I can roll a six on a six-sided die five times in a row, shouldn't my RNG be able to as well? Seems like you're suggesting that you make an RNG that's already pseudo-random even less random. A true random system doesn't mean getting a different result every time.

[SigilFey]

Huh? Where's the randomness in telling an RNG not to select the same number? If I can roll a six on a six-sided die five times in a row, shouldn't my RNG be able to as well? Seems like you're suggesting that you make an RNG that's already pseudo-random even less random. A true random system doesn't mean getting a different result every time.

Yeah, I know it sounds odd, but that's exactly what RNG programs do. It's not like it tells the computer to NEVER to select the same number twice in a row, but with a poorly coded RNG, there are certain circumstances in which a RNG can associate a single numerical value with a certain call by the software. So coding is done and redone (bug-fix style) to ensure that differing values are assigned to all calls.

I'll bring up ToEE again as an example. One of the biggest problems with the game upon release (and for about 2 years afterward...) was that characters created with Strength values over 18 started receiving more critical misses during combat than a Charlie Chaplin impersonator could actively TRY to rehearse and perform. The problem was the random number generator was attempting to avoid excessive rolls of critical hits, and automatically defaulting to a set value of "1" whenever an attack roll fell within a certain value with an associated Strength bonus over a certain value...phew! The RNG needed to be recoded to instruct the program to select numbers other than "1" more often. It was a fluke. A bug.

Yep, it would be possible to roll four critical misses in a row on a 20 sided die in real life. But in real life, you would have a human being as a DM that could adjust the flow of the game and create a chance for recovery and success, making the game fun! Imagine the frustration of a player who specifically created a strong fighter/tank class character, only to be outrageously and regularly penalized by critical miss after critical miss, making a Halfling Mage/Thief a better toe-to-toe melee fighter in practice than a Half-Orc Barbarian with a Strength of 22. Not fun. No fun at all. Hence, they fixed it. Eventually...

...sort of...

[SpottedShroom]

Yeah, I know it sounds odd, but that's exactly what RNG programs do. It's not like it tells the computer to NEVER to select the same number twice in a row, but with a poorly coded RNG, there are certain circumstances in which a RNG can associate a single numerical value with a certain call by the software. So coding is done and redone (bug-fix style) to ensure that differing values are assigned to all calls.

That's not how a pseudo-random number generator works. If you are interested in the subject, Wikipedia has a reasonable summary including pseudo-code implementations of several popular PRNGs. See http://en.wikipedia.org/wiki/Pseudo-ran ... _generator

I'll bring up ToEE again as an example. One of the biggest problems with the game upon release (and for about 2 years afterward...) was that characters created with Strength values over 18 started receiving more critical misses during combat than a Charlie Chaplin impersonator could actively TRY to rehearse and perform. The problem was the random number generator was attempting to avoid excessive rolls of critical hits, and automatically defaulting to a set value of "1" whenever an attack roll fell within a certain value with an associated Strength bonus over a certain value...phew! The RNG needed to be recoded to instruct the program to select numbers other than "1" more often. It was a fluke. A bug.

What you describe isn't a bug in the PRNG, but a bug in how it was used. It wasn't that the numbers weren't really random, just that the program wasn't asking for the right ranges or interpreting them correctly.

[SigilFey]

@Shroom,

You are correct on both counts, and so am I. The Wiki is a good read and makes pretty clear sense on how an RNG program works. Better than I could, obviously. But it's only explaining the point I was trying to make more clearly. Primarily, even the Wiki refers to the program as a "Pseudo-" (Imitation of a-) Random Number Generator. Pseudo, becasue the best a computer can do is "simulate" randomness. It cannot actually be random. In the end, all commands, no matter how complex to begin with, come down to a series of switches the computer's processor must throw, and each switch has only has two options--"1" or "0". "On" or "Off".

And you are exactly correct in your second response--all RNG's work exactly as described. In fact, all computer programs do exactly what they were programmed to do. It is we imperfect human beings who accidentally give them the wrong instructions that create bugs, errors, or wierdness in programs. This is again what I was trying to point out.

That point is that Eschalon comes far closer to balanced random number generation than a lot of more complex systems used in the past, but when any RNG is utilized in a program, it is virtually impossible to create a system that avoids all wierdness. Here, especially, all it takes is a quick reload, and things usually balance back out.

[SpottedShroom]

That point is that Eschalon comes far closer to balanced random number generation than a lot of more complex systems used in the past, but when any RNG is utilized in a program, it is virtually impossible to create a system that avoids all wierdness. Here, especially, all it takes is a quick reload, and things usually balance back out.

I think you're still not understanding how this works. Reloading the game will have no effect on the quality of the random numbers coming out of the generator.