Understanding the Game engine - HELP WITH ITEMS NEEDED

[sees_all1]

EDIT:

 

I need help finding items that provide very specific amounts of bonuses, particularly f2p armour/weapons that provide low Attack bonus, OR low Strength bonus, and items that provide low Defense Bonus, esp. in the category of the attacking one.

 

example: Bronze b-ax provides 9 strength bonus, nothing else or

 

Bronze Pick provides 2 slash defense, nothing else. etc.

 

 

 

If you can think of any items that fit these categories, please post with their stats, and if its a weapon, what kind of attacks it has (i.e. trains defense with its slash bonus of +2)

 

 

 

 

 

Alright, after a hijacking a rant and having a lengthy discussion/argument with Compfreak, I've decided to try and conduct several experiments to find the underlying methods used in Runescape, particularly the ones determining the probability of a hit.

 

 

 

Critical Assumptions:

 

-Your own defense, magic, or range has no outcome on the probability of you hitting on your opponent with melee.

 

-All numbers generated are random, or pseudo-random.

 

-The formula for determining a hit does not change from trial to trial.

 

-The dice rolled to determine hits are integers, from 0 to the stat itself

 

 

 

[hide=The first test - probability of 1 d and 1 a, no bonus, P to hit]To start, I'd like to know the following scenario:

 

What is the probability of hitting on a character with 1 defense, with 1 attack, 1 strength, and no bonuses?

 

 

 

Next, I'd like to explain the setups I will be using.

 

I will create 2 or more characters, go to the Duel Arena, and repeatedly duel. For each hit, I will record its value. One character will have auto retaliate off, with its style as accurate, while the other will be hitting, with its style as defensive. That way, the only experience gained is for a stat that doesn't effect hits.

 

 

 

The results:

 

Sub Test 1: 31 miss, 12 hits

 

Sub Test 2: 15 miss, 11 hits

 

Sub Test 3: 16 miss, 11 hits

 

Sub Test 4: 34 miss, 12 hits

 

Sub Test 5: 47 miss, 12 hits

 

Test 1 Total: 143 miss, 58 hits

 

p-value miss: .711443

 

 

 

Sub Test 1: 31, 12

 

Sub Test 2: 25, 12

 

Sub Test 3: 39, 12

 

Sub Test 4: 56, 13

 

Sub Test 5: 38, 12

 

Test 2 Total: 189 miss, 61 hits

 

p-value miss: .756000

 

 

 

Sub Test 1: 17, 11

 

Sub Test 2: 42, 12

 

Sub Test 3: 23, 11

 

Sub Test 4: 35, 12

 

Sub Test 5: 44, 12

 

Test 3 Total: 161 miss, 58 hits

 

p-value miss: .735160

 

 

 

Conclusion: Each of these are remarkably close to .75. After doing 918 hits, the ratio I got was .743

 

So, in the setup:

 

1 attack versus 1 defense, hitting in defensive mode, the probability of a miss is .75[/hide]

 

 

 

[hide=The second test... with curse]This setup was pretty difficult: Before attacking, I had to cast curse, then hit repeatedly until curse lost its effects, cast curse again, etc.

 

 

 

After three tests:

 

Test 1: 29 miss, 12 hit

 

Test 2: 31 miss, 11 hit

 

Test 3: 42 miss, 12 hit

 

which means the probability of missing is about .745 - the same as the first setup.

 

 

 

Conducting these tests were difficult because if I didn't notice the curse was reset, and I continued to record hits, all the data was null at that point. This happened to me once.

 

 

 

Anyhow, with no significant reason to believe that curse makes any difference at such a low level, I decided to abandon testing.

 

 

 

Conclusion: It didn't make a difference if it says your defense is 0 or 1. Curse could have been -5% on defense, but I'd never know without several thousand tests, to make the Standard Error lower than 5%.[/hide]

 

 

 

[hide=The third test]Alright, so after the "failure" with curse, I'd like to test and see if doubling your defense halves the probability of a hit.

 

The setup: 1a 1s vs. 2d, using defensive instead of aggressive.

 

 

 

What I'm looking for:

 

The probability of a hit in this setup should be .125, or 1 in 8, if doubling your defense halves the probability of being hit on.[/hide]

 

 

 

Why it matters:

 

Jagex gave a clue as to how hits are determined in postbag #38. I assume that two dice are rolled, one for a player's attack, and one for the others defense. Assuming a hit, a third dice is rolled to determine the size of the hit.

 

 

 

If you would like to help out with this testing, please PM me. If you see any flaws in my methods, post here.

[Enipeus]

i'll have to reread postbag 38 but i'd like to help out i'll PM you as well.

[compfreak847]

Your critical assumptions are wrong. I've done several hundred hours of this, and the numbers are not pseudo-random. They are generated with a complex but identifiable pattern, and you will draw completely erroneous conclusions if you assume they are simply random when the reality is far more complex.

[DF_Thomas]

The dice thing seems reasonable - first the 'attack' die is rolled. This outputs either 1 or 0, giving hit or no hit. Then the 'strength' die is rolled, giving a number 0 to value of hit. Note that both are rolled whatever; sara godsword can heal even if damage hit is 0 - this will be from strength hitting but attack not hitting, assuming the gs heals using only the number from str die.

 

 

 

Max damage calculators should help - they will show if relationship is linear or not.

 

 

 

Since there is a max. damage, we can assume that all calculations are done using discrete distributions for hits. (this is actually fairly obvious; there's no way to compute these things without using floating point arithmetic). It's an obvious generalisation that all levels must use integer values; there's no reason they would use levels like 0.95 just because curse was cast.

[sees_all1]

http://java.sun.com/j2se/1.4.2/docs/api ... andom.html

 

 

 

Compfreak, I know the numbers aren't purely random.If its done by a program, it can't be purely random. If its done in java and you don't know the seed, it might as well be random. That's what pseudo-random means - somewhere the number generator gets seeded, and then off it goes with your "random numbers."

 

 

 

Now, you might have done enough testing to get started with the same seeds, and that's why some numbers are more frequent than others. I don't know. What I do know is that I have absolutely no clue where the random nature of hits come from, and that you have absolutely no clue. Unless your name is Andrew, and you created Runescape.

 

 

 

What I do know is that somewhere there is a method / function called something along the lines of "calculateHit" - and that numbers are fed into it, and a number is spat out -- It isn't magic. I don't know if the numbers are generated server side or client side. I don't know if the servers are programmed in Java. Right now I can't tell you how the client or server communicates, or if it matters to this application.

 

 

 

What I do know is that the first probability I got for this first set of data - this .71, is incredibly close to .75 -> 2 numbers are generated, 1/0 for attack and 1/0 for defense, and then a 1/0 for strength, which means that 1/2 the time attack beats defense and 1/2 the time you won't hit a natural zero, and violia, .25 you'll hit, .75 you'll splash.

[sees_all1]

 

Since there is a max. damage, we can assume that all calculations are done using discrete distributions for hits. (this is actually fairly obvious; there's no way to compute these things without using floating point arithmetic). It's an obvious generalisation that all levels must use integer values; there's no reason they would use levels like 0.95 just because curse was cast.

 

I'd like to be able to confidently say that though... I have to start from scratch and work my way into a thesis. If I go out and say - Hey, it should be 0 instead of .95, and go out and my data indicates .95 instead, I'm hosed.

 

 

 

That's why I'm starting with level 3's - because I can control them much more than I can control level 126's, too many numbers to get accurate results.

[compfreak847]

First of all, why are you linking to a random number article on a 5 year old java operating system? Second of all, how do you know what's what' used to calculate it? Third, I know it isn't simply because I've tested and seen the results - you have not.

[sees_all1]

First of all, why are you linking to a random number article on a 5 year old java operating system? Second of all, how do you know what's what' used to calculate it? Third, I know it isn't simply because I've tested and seen the results - you have not.

 

http://java.sun.com/javase/6/docs/api/ <- this one is the most recent latest and greatest... and guess what, its exactly the same! And its not an operating system, its the API for the Java Programming Language. There are two natural ways to generate a random number in Java, one is using the Random class, or the other is using the Math class's Random method. I highly doubt that Jagex has made their own, it would be a waste of time. Even if they had, though, they would've used a similar algorithm to get good randomness.

 

 

 

What you've just described as patterns is just seeding.

 

Check it out - code up something that uses Java's random class. Run it. Observe the results.

 

Then, seed the random number generator to, 5. or 10. or any other constant.

 

Then run it, and run it again. and again. You'll get the same results. Why? Because it's seeded.

 

 

 

Probably 95% of all languages that offer random number generators allow you to seed the generator. Most of them probably use the same algorithm anyway.

[compfreak847]

Was in a hurry, it's obviously not a operating system :lol:

 

 

 

Now, as I was saying, seeding isn't going to produce a pattern like the one I mentioned earlier on the melee thread. Certain hits are very common, while others are virtually non existent. You'll see what I'm saying as you get through the first few thousand hit tests, and realize that your going to need far, FAR more then a few thousand hits to get accurate results.

[Kenshinjapan]

Why do you care so much about understanding the game engine anyways? Seems like a waste of time.

[la la la]

@compfreak

 

How many years of java education do you have? You seem very keen on attacking others for their java knowledge.

 

 

 

Kenshin, I guess they figure they can better use the system if they know exactly how it works.

[3hitm4g3u]

I've had people with 99 attack spec 0/0 on my 1 defence.

[compfreak847]

@compfreak

 

How many years of java education do you have? You seem very keen on attacking others for their java knowledge.

 

This isn't about java coding, it's about statistics. I'm discussing the very obvious patterns I've seen over tends of thousands of recorded hit spreads, which sees all is completely ignoring in favor of the 'it's a simple pseudo-random generator' theory, which is unquestionably wrong.

[la la la]

@compfreak

 

How many years of java education do you have? You seem very keen on attacking others for their java knowledge.

 

This isn't about java coding, it's about statistics. I'm discussing the very obvious patterns I've seen over tends of thousands of recorded hit spreads, which sees all is completely ignoring in favor of the 'it's a simple pseudo-random generator' theory, which is unquestionably wrong.

 

Actually, this thread is all about java coding. Answer my question, please. A scientific approach to a game engine would most definitely imply a study of java itself.

[compfreak847]

 

Actually, this thread is all about java coding. Answer my question, please.

 

Explain what part of interpreting hit spread results involves coding in java.

 

 

 

In answer to your question: I've coded many java applications over the years, and have become quite proficient at it. I studied Java in detail for 3 years, although my focus has shifted to other emerging languages in recent years. Anything else you'd like to know?

[la la la]

 

Actually, this thread is all about java coding. Answer my question, please.

 

Explain what part of interpreting hit spread results involves coding in java.

 

 

 

In answer to your question: I've coded many java applications over the years, and have become quite proficient at it. I studied Java in detail for 3 years, although my focus has shifted to other emerging languages in recent years. Anything else you'd like to know?

 

Nope, I have studied three years myself and now tutor it.

 

 

 

Correct me if I am wrong, but the OP was referring to the fact that there are multiple randoms generated (3 i believe) each time you hit, and it is based on some kind of combination of the three that your hit is determined. However, you are saying that something like random.nextInt() is not used in generation of hits, and that there is a recognizable pattern? If so, what is this?

[compfreak847]

I'm saying that there is a pattern and the data does not exhibit characteristics attributed with a randomly generated number. The sample size I have is far too large to write it off as a characteristic of randomness. I have not, however, been able to identify a pattern standard to different setups, stats, and weapons. Feel free to peruse the thread on Rants, it'll give you a background in this.

[la la la]

I'm saying that there is a pattern and the data does not exhibit characteristics attributed with a randomly generated number. The sample size I have is far too large to write it off as a characteristic of randomness. I have not, however, been able to identify a pattern standard to different setups, stats, and weapons. Feel free to peruse the thread on Rants, it'll give you a background in this.

 

Oh, you mean the damage hit, not the chance of hitting, is not random? I think I can agree with that. Middle to higher hit values are more common than hitting 1s and 2s, if that's what you mean. It could be management of a random number generator something like

if (random.nextInt(3) == 1)

hit = random.nextInt(max/2)+1;

else 

hit = random.nextInt(max/2) + max/2 + 1;

 

That way, higher hits would be distributed more often. I know thats not the code used, just maybe around the idea.

[compfreak847]

There's several patterns. One is similar to that, with hits grouping at the higher\lower ends, but another one is that certain numbers are as much as 50% more common then others. This varies from weapon to weapon and doesn't seem to have a definite pattern.

 

 

 

Yes, you are correct about the chance of hitting. I am not debating that, though I have noticed odd hit streak or dry streak occurrences, as I only have a limited amount of data on accuracy. I'm talking about what number you will hit after it is determined that you will. He says it's simply a random number generated between your max and 0; I'm saying it's much more complicated then that.

[sees_all1]

 

To start, I'd like to know the following scenario:

 

What is the probability of hitting on a character with 1 defense, with 1 attack, 1 strength, and no bonuses?

 

...

 

Why it matters:

 

Jagex gave a clue as to how hits are determined in postbag #38. I assume that two dice are rolled, one for a player's attack, and one for the others defense. Assuming a hit, a third dice is rolled to determine the size of the hit.

 

 

Yes, you are correct about the chance of hitting. I am not debating that, though I have noticed odd hit streak or dry streak occurrences, as I only have a limited amount of data on accuracy. I'm talking about what number you will hit after it is determined that you will. He says it's simply a random number generated between your max and 0; I'm saying it's much more complicated then that.

 

That isn't what I'm testing now - I don't care about the number you hit. If I did, the strength level I would be testing would be higher, say 10, or 20, or 90.

 

Start at the beginning, make few assumptions, prove the assumptions, and then use what you know to find the other things.

 

I haven't gotten to the hit hypothesis, and I'm not going to argue it either way until I have statistics to back it up. I will get there eventually, but until I do, quit nitpicking about stuff I may or may not do in the future.

[compfreak847]

Notice the "to start". I'm referring to the rants thread where you planned on hits researching.

 

 

 

As for accuracy, I'm afraid that's doomed to failure too. Your testing a system with many, MANY variables and trying to reduce it to one ratio. Defence bonuses, attack bonuses, levels, prayers, potions, special effects, weapon special effects, various weapon characteristics, and several attack styles.

 

 

 

Good luck, but the amount of information needed to be collected would be unbelieveable - we're talking about quintillions of possible combinations that you want to reduce to one number.

[Makoto_the_Phoenix]

A scientific study on hits, ooh, I'm so in.

 

 

 

As a Computer Scientist myself with Java experience, the following three premises of "random" numbers associated with RS combat can be confirmed:

 

 

 

1) All numbers generated are pseudo-random, so if the same seed occurs again, a similar pattern will emerge.

 

2) Our stats (Defense, weapon power, etc) affect the potential of our hit (I think it comes out to the overall difference of your weapon's attack and your attacking stat versus their defense and defensive equipment).

 

3) It is probable to miss frequently or hit frequently on someone with either 1, 0, or 99 defense.

 

 

 

A pattern occurs with pseudo-random number generators, simply that's how a pseudo-random generator works. I don't like getting preachy about this, but software can't use a truly random generator, for debugging purposes - it's exponentially more difficult to debug a random-generated string of numbers (real, rational, integer, etc) than it is a pseudo-random string of numbers.

 

 

 

With Java's pseudo-random generator (both Random.nextGaussian() and Math.random() are equivalent), the range is from [0.0, 1.0). It's feasible that this number has been expanded on by 2 orders of magnitude, giving [0.0, 100.0). 100 is not possible, but it's feasible to believe that the higher number you get on a roll to determine if you actually hit (bounded on the initial range) plays a role on the number that you can hit. This number is then subtracted (multiplied, divided, whatever is done to it to make the number smaller than initially possible) by the defensive stats and equipment that the target has, and the end result is the hit.

 

 

 

I must insist that our stats and equipment have an effect on the chance of us getting hit. Re-run your test with one character wearing Bronze, one with a weapon, both with a weapon, and so forth. I'll even provide the Bronze.

[sees_all1]

 

I must insist that our stats and equipment have an effect on the chance of us getting hit. Re-run your test with one character wearing Bronze, one with a weapon, both with a weapon, and so forth. I'll even provide the Bronze.

 

Lets first figure out if its a straight scale with various attack/defense levels, and then we'll start looking at atk/def bonuses.

[sees_all1]

I've just completed another 3 tests, and updated the OP to reflect this. Not sure if I want to keep beating a dead horse, I suppose I can do another 7 or so. Anyhow, each value is remarkably close to .75, the value I estimated.

 

 

 

I just spat the numbers into a calculator, and the z-value that came up was .459.

 

Which means that we cannot be sure that the data that I got was different than the expected value. Unless someone does more tests and comes up with a significantly different proportion, I'm happy with the results.

 

 

 

So, the answer to the first test: What is the probability of hitting on a character with 1 defense, with 1 attack, 1 strength, and no bonuses?

 

The answer: you can expect to hit on average 1 out of every 4 hits. About 1000 hits has shown this to be quite accurate.

 

 

 

Now its on to the next test.

 

What is the probability of hitting on a character after casting curse, when their defense is lowered to 0?

 

 

 

I've just run some tests, and the results are as follows:

 

Test 1: 29 miss, 12 hit

 

Test 2: 31 miss, 11 hit

 

Test 3: 42 miss, 12 hit

 

This data indicates that your defense can't go lower than 1, or that curse only lowers your defense by 5%. To determine if it is actually 5%, the number of tests I'd have to run would be much more than I did for the first question.

 

 

 

Help! Is there any way to lower your defense in F2P?

[Makoto_the_Phoenix]

Help! Is there any way to lower your defense in F2P?

 

 

 

Strange... it might be that the level of defense is taking the higher value of the integer when calculating its overall defense stat. But there isn't any other way to absolutely reduce the defense of any character on freeplay to 0.