|
| Thread: Heroes' Stats and Skills Chances |  This thread is pages long: 1 2 3 4 · «PREV / NEXT» |
|
Xarfax111
Bad-mannered
Supreme Hero
The last hero standing
|
posted February 08, 2006 06:12 PM |
|
|
....first of all u already broke rule one and two...but u already mentioned that u will do so in the end and ordered us to cut of your balls if u do so...what a great sign of selfsacrificing!!
...second..if you say names in alphabetical order start with z and end up with a to not make me angry.
I would strongly suggest to make a pan box with exp like angelina already said...plus all hereos on the map with each having a pan box. I disagree with angelina ..youve always to take the very next given skill no matter what, to get the correct results. Also uve to turn of the game each time before restart. Then ull see the "pattern" behind them.
An easier way is to simply take the bin-data after each start open it with excel and let run a makro over it. Then u could do the testing on your own. 
..me concerning this? "stop" *hehe*
Xarfax1
|
|
LegendMaker
Promising
Famous Hero
The Metal Specialist
|
posted February 09, 2006 05:25 AM |
|
|
Reliable Results / Legendary Selection Method
To (and not "at" like you all mistakenly say ) the big X :
Suggestion of using a Pando Box was Russ', not Angel's.Quote:
if you say names in alphabetical order start with z and end up with a
That would be reverse alphabetic order, wouldn't it ?  Quote:
all hereos on the map with each having a pan box.
All heroes on the map ? I'm pretty sure at least that is a joke... 
Still, better safe than sorry : Is there an actual reason why you think it'd work better this way than simply selecting each one as starting hero for every new batch of 100 tests I launch ?  Quote:
Also uve to turn of the game each time before restart.
Uh hu ? Same here, O Tactical Wizard : what would I need to do that for exactly ? Does the pattern stay the same as long as the game isn't restarted ?!...  Quote:
An easier way is to simply take the bin-data after each start open it with excel and let run a makro over it. Then u could do the testing on your own.
Errr... Wow ! While not completely computer illeterate, I'm a total noob at programing, Xarfax ! BBcode, Html, that's it !
To all : Quote:
ALWAYS chose a skill you already have, until u r on expert level, but note down the other offered skill every time. This way u get more reliable results i guess.
You still have a good point, and yes, it does offer more reliable results to apply a fixed and general rule to the selection method instead of actually choosing like I originally planned.
But that is still not enough to get 100% reliable results. To achieve that, we need a perfectly fixed rule for the selection method, one that would completely eliminate the possibility of a subjective choice spoiling the accuracy of the results.
I gave it quite some thoughts (what you think kept me awake this late ? ) and ran some test samples.
Check this out :
Even worse, check that out :
Then, I considered and sample-tested Xarfax's selection method as well : Quote:
youve always to take the very next given skill no matter what, to get the correct results.
Gave it some more thoughts... Scratched my head...  Some more... Yet some more... ...And, finally... ....concluded that...
1) Choosing "A" everytime can't guarantee reliable results, because you'll always run into a subjective choice at Level 6 for Heroes starting with two basic skills and at Level 3 for those starting with one advanced skill.
2) Choosing "B" everytime is perfectly and unlimitedly doable. You'll always get all 8 Slots full without ever having a subjective choice to make. It will be at Level 7 for Heroes starting with two basic skills and at Level 8 for those starting with one advanced skill. 
So Xarfax was right. His selection method works. You get 100% reliable results.
Thanks and congrats for that, Mister X. !
HOWEVER...
That's still not enough for the goal I wish to attain with this project ! 
Because you'll never ever get to see what any Hero would be offered beyond Level 8 ! You'll never get to see if or when any particular skill would be offered again (since you never discard any new skill). You will simply focus on too low a range.
SO...
I came to the conclusion (slow, am I not ? ) that one rule couldn't be enough to gather a sufficiently wide range of results AND guarantee said results' accuracy.
Subjective choices MUST be completely eliminated for the results to be reliable, but they must be in a way that will still allow to observe the whole pattern.
I will neither watch the whole tree through broken glasses nor contemplate its roots through a microscope !
THEREFORE...
Alphabetic order ! Alpha ****ing beta order, for ****'s sake ! Eureka !
Nothing is more objective and reliable. Nothing except maybe... Numbers ! Moo hoo ha ha
HENCE...
1) NUMBERS
Always take the highest Skill Level between two skills.
Expert this > Advanced that > Basic **** !
2) LETTERS
Always take the Skill which has the higher rank in alphabetic order, when the Skill Level is equal.
Same Level AAA > Same Level AAB !
Now, check the technique, mere mortals !
Choosing according to the Legendary Selection Method everytime is perfectly and unlimitedly doable. You'll always get all 8 Slots full without ever having a subjective choice to make.
It will be at Level 21 for any and all Heroes, wether they're starting with two basic skills or with one advanced skill.
x x = ?!!!!
This will give any tester the best of both worlds. No subjective choice will ever be involved, and you'll get to see the full spectrum of the Skill Advancement Tree !
Two choices left now, folks :
1) Prove me wrong.
2) Call me a genius and bow before thee Might (& Magic) ! Moo hoo ha ha !!!!
The Legend ! For President ! 
____________
LM  
|
|
angelito
Honorable
Undefeatable Hero
proud father of a princess
|
|
posted February 09, 2006 01:07 PM |
|
|
Quote:
..youve always to take the very next given skill no matter what, to get the correct results....
Will this way lead us to the results we wanna have?
We donīt want to know which are the first 8 skills a hero gets offered, but we want to know which skills he gets offered during a complete level up.
In your example, if a barbarian gets air offered on level 5 for example, chances to get earth offered are very low. This would lead to the conclusion, this hero class has a low chance to get earth (if this happens with the following tests aswell....be it fire or air, which is offered first).But it is interesting to know, how often a barbarian hero gets earth offered till he reaches 7 expert skills and has the last chance on next level up.
So i would make the tests the other way around.
Chose a specific skill as goal, and let a hero level up till he gets this specific skill offered (thatīs why i said, always chose an existing skill till u get the one offered u want to have).
This way u get the possibilities of each skill for each class on one hand, and on the other hand, u also would notice if (and how often) u get a skill offered u denied the first (second...third...) time. This is a very interesting thing to know, coz sometimes u get the choice between 2 good skills, and it would be usefull to know which of them has the higher chance to be offered again.
What does it help to know, an overlord gets offered scouting and balistic during the first 8 levels, if u know for sure u will deny them and wait for better ones?
The chances of each hero class getting the best skills offered are interesting, not how often Crag gets mysticism offered....
____________
Better judged by 12 than carried by 6.
|
|
tigris
Supreme Hero
Supreme Noobolator
|
|
posted February 09, 2006 02:28 PM |
|
|
Seems like a very ambitious plan u have set here legend.
I still haven't understood one detail though: do you target to see a certain skill and it's chance of being offered to a certain hero class, or do you want to know the best skill selection you can have with one hero class?
I think variations amongst heroes from the same class are quite small(only concearn starting skills). I will love to test some as soon as i have more time(will be next week though). I hope you will get me some explanations on MSn the next time we mweet legend, it's kinad fuzzy to me.
____________
|
|
Russ
Promising
Supreme Hero
blah, blah, blah
|
|
posted February 09, 2006 03:31 PM |
|
|
|
It kind of sucks though that everyone skipped my entire post aside from the pandora box suggestion.
|
|
LegendMaker
Promising
Famous Hero
The Metal Specialist
|
|
posted February 09, 2006 05:28 PM |
|
|
Serious, Reliable, Complete Testings, NO less !
Quote:
I will neither watch the whole tree through broken glasses nor contemplate its roots through a microscope !
'Nuff said !
Now, if you gentlemen would please do me a favor and read my whole post again... (thoroughly and completely this time around ! Don't assume these screenshots are only there to show that I can use Photobucket, 'cause they ain't !)
No offense meant, pals, but it's pretty obvious to me you guys only browsed my last post carelessly, from your three replies.
1) Angel, your last reply makes sense only if it was adressed to Xarfax and if you did write it before you noticed my "Legendary Selection Method" post. Quote:
We donīt want to know which are the first 8 skills a hero gets offered, but we want to know which skills he gets offered during a complete level up.
That's not the whole truth. What we want to find out is as many clues as possible as to which pattern(s)/formula the comp follows in selecting the skills it offers throughout any given hero's complete level up. At least, it's what I aim for, no less ! Quote:
Chose a specific skill as goal, and let a hero level up till he gets this specific skill offered (thatīs why i said, always chose an existing skill till u get the one offered u want to have).
That would be an amateur's wild guesses disguised as statistics. Not at all what I want to do.
I must admit I was horribly wrong with my original protocol, in that respect. My whole idea of selecting the skills given a subjective rank from "crap" to "great" was nothing but CRAP itself ! I'm laughing at my own self in retrospect !
The reliable and effective way was in the back of my head from the beginning, I think (my subconscious probably sent me that "sort the names in alphabetic order" idea as a hint My subconscious really is a nice guy, I guess... hehehe).
However, it might never have seen the light of day (or night, in fact) if you and Xarfax hadn't pushed me back and forth to both extremes. You guys acted as my external conscience ! You know, when you have that tiny little Angel on your right, yelling "white" while at the same time, the tiny little Demon on your left keeps shouting "no ! black !" (mandatory pun, Xarfax, nothing personnal lol) ? That's really a great help to find out that the right answer is grey ! So thanks to both you guys a darn lot for that ! And now that I saw the light, here is what I'm doing there :
I'm doing serious, scientific, 100% reliable statistics, here ! **** the guesses, I want to KNOW !
If you want to find out how the AI "thinks", it's an absolute necessity to stop thinking as a human, let alone a player ! To understand the machine, become it ! Take a deep long journey into the abyss of its "mind" and analyse the things from its "point of view", my friend !
That's why all subjective criterias MUST be completely discarded for this project.
So I don't give a **** any more which skill is "good" or "best" or "crap" for this project, 'cause that criteria doesn't exist for thee comp ! It's all LETTERS & NUMBERS, for the comp !
And I can tell you, from the last preliminary testings I've made today, I'm already beginning to distinguish some of them patterns shape. Now guess how good we'll be able to see them when all 14500 tests will be completed ?!
2) Russ, first I noticed and appreciated your post. Second, I thanked you for it and replied to it, not just the Pando box part ! Though it's by far your greatest hint so far, it saved me much time (even though I had already wasted hours crafting this stoopid test map of mine, aligning those darn trees and mines for ****ing ever ! Lmfao). Reread my posts, dude. And thanks again, just to make sure ! Lol
Btw, you didn't say "stop", but I don't recall hearing you say "yes" neither... Would you mind making it clear wether you want in or not, man ?
3) Tigris ! Thanks a lot to you, dude. You're the first and, this far, only of the dream team line-up to not only join the project but also seem enthusiastic about it ! That's a relief.
4) lord_crusader... O God of All Testings, I beg of you, do appear in front of thy humble servants ! We need your godly skills for this Legendary Quest !
Just You Wait. Legend.
____________
LM
|
|
angelito
Honorable
Undefeatable Hero
proud father of a princess
|
|
posted February 09, 2006 05:34 PM |
|
|
Quote:
....That would be an amateur's wild guesses disguised as statistics. Not at all what I want to do....
Seems u donīt get my idea, so please let me out of these tests then.
____________
Better judged by 12 than carried by 6.
|
|
Russ
Promising
Supreme Hero
blah, blah, blah
|
|
posted February 09, 2006 05:43 PM |
|
|
Quote:
2) Russ, first I noticed and appreciated your post. Second, I thanked you for it and replied to it, not just the Pando box part ! Though it's by far your greatest hint so far, it saved me much time (even though I had already wasted hours crafting this stoopid test map of mine, aligning those darn trees and mines for ****ing ever ! Lmfao). Reread my posts, dude. And thanks again, just to make sure ! Lol
Btw, you didn't say "stop", but I don't recall hearing you say "yes" neither... Would you mind making it clear wether you want in or not, man ?
I read all of your posts. I haven't seen any of my suggestions in them. Neither did I see you saying "Russ, you are full of it!" My conclusion? You didn't read it. (I don't blame you personally because everyone kinda ignored it.) If noone wants to read it - it is np with me. I can't force my ideas on people.
I explained the way I wanted to do the testing if I had time and the reasons behind why I was going to do it that way. I belive I've also explained the kind of conslusions we could draw from analyzing it. I think I gave a pretty good reasoning, but well, it got ignored by everyone, what can I do?
And my answer is "no". No offense, but I see a lot of work, yet I don't see how we could get any useful results and conclusions out of all this testing. Correct me if I missed something, but I don't think you've made a post explaining how the results of this testing are going to be analyzed to yield some useful conclusions, instead of leaving us with a bunch of numbers and tables that don't help us in any way. An example of a useful conclusion is what we know about wisdom and magic school skills.
|
|
LegendMaker
Promising
Famous Hero
The Metal Specialist
|
|
posted February 09, 2006 06:37 PM |
|
|
Russ's Ignored Post / Legendary Maths Class
Quote:
If I was doing the testing, I would do a test for each useful skill (log, tactics, offense, armorer, path), plus maybe some other decent skills too such as archery or intelligence. There is no need to do any more tests for magic school skills and wisdom, and I would assume noone cares about crap skills like eagle eye.
The main reason I didn't answer specificly to that part is because Angel had made pretty similar suggestions, and I assumed my answer to it would have been also accurate for your suggestion.
I don't plan to analyse any specific skill. It's the overall patterns I hope to find. And I've planned to gather an enormous pannel of statistics as a starting point in that search for patterns.
So my project includes leveling up the same hero 100 times in a row, every time following the same exact objective protocol for the skill selection method.
Then you get statistics. Any statistics. Not just the 10% Angel cares about, neither the other 10% that Xarfax suggested to focus on, nor the 10% you, Russ, suggested in that post of yours. I'm doing 100%. So what are you all complaining about ? You will ALL get your favorite part of the results in the end !
The more results I'll gather, the more useful the stats will be. However, I won't be able to give you "an example" of the conclusions until the project is completed, Russ ! If I already knew, what would be the point of making this huge project to find out ?!
It's a mathematical pattern. To decypher it, you need to apply an objective, permanent protocol from the very first test of the whole 14500 batch to the very last one.
If I, the tester, always do the same thing, then what I do isn't a variable of the formula anymore. It becomes a CONSTANT.
From that, we can conclude that any and all VARIATIONS that will occur in the results will come from the comp pattern(s)/formula only.
Hence, with a titanic enough mass of results (ALL made following strictly the same protocol, unless the whole results would be spoiled by subjective variable)... It will more than likely be possible to observe the pattern(s)/formula.
Or at the very least get to know / understand a WHOLE lot more about it than ALL your partial tests gave, give and will EVER give ! No offense meant, either. I'm just 100% convinced of the accuracy of my theory. It's mathematic, dudes ! You just CAN'T resolve maths with subjective protocols !
At least, Xarfax understood that.
Even though he's wrong by assuming 7 or 8 Level ups for each test will lead to sufficient results to observe the actual patterns. They are based on a timefram MUCH longer than that, and my bet is you know it, Mister X !
With my selection method, I will get 3 times as many results from any given test. And said results will be 100% reliable statistics, with a wide enough range of observation to draw conclusions from.
Furthermore, another advantage of my method is that ALL heroes will go as high as Level 21, no matter how many skills they start with and at which level the skills are. That's a lot more reliable and convenient for future analyse of the results than any method you all suggested, IMHO.
Yours faithfully. Legend.
____________
LM
|
|
Xarfax111
Bad-mannered
Supreme Hero
The last hero standing
|
|
posted February 09, 2006 07:23 PM |
|
|
Testing and getting correct Results.
It was about 4 years ago, when Andiangelslayers apprentice Motorschaff and me had a small conversation about "which main hero" to take. He noted that he "always takes Barbarians as main", because it seemed that they "always get masshaste and massslow" and the fight is over before it began. I also experienced that, while all fanpages and handbooks and all others in theory ment that there are low chances for a Barbarian to get the skills Earth AND Air.
Experiencing this with about 3000+ games played, some good players decided to get the correct skill table on their own, so we started the Drachenwald-Statistic-Projekt with about 12 players. To put one result to the beginning: Barbarians do have an about 80% chance to get Earth AND Air, and NOT a 2 times 3 out of 112 lmao. So all known skilling tables are plain wrong.
There were discussions about how to do this testing. Some of those are mentioned here in this thread too. All of them will lead to the wrong results in my opinion.
1. One player suggested only to take only one skill, only the first skill and then restart. This looks like the statistically science correct method.
But the numbers ull get out of that are plain wrong. The result is that u get the skilling chance for the first slot, which means the results will give u a VERY low chance of getting wisdom, earth and air for a Barbarian. Thats just plain wrong as EVERY hero has a 100% of getting wisdom offered.
2. One player suggested to bring all skills to expert to see how the complete level up is.
But the numbers ull get out of that are plain wrong.
Ull get a mixture of two independant probabilities. Like when uve 4 skills on basic level, the chance that one of these skills will be offered again is about 1 out of 4, while the first one are preferred.
3. One player suggested only to take the wanted skills, to see the percentages of the wanted skills.
But the numbers ull get out of that are plain wrong.
Thats a mixture of point 2 and the unknown fact that the offered skills are NOT independent to each other. If u want to "judge" the skilling of a hero class, make it after getting the correct probabilies. Make a poll amoungst players form 1-28 of preferred skills, take the average and multiply that with the probabilies and u get smthn like "qualified probabilities".
4. One player suggested to just skill a hero until he gets a certain skill offered. Then just write how many tries he get that skill out of the complete tries.
But the numbers ull get out of that are plain wrong.This is a mixture of point 2 and 3.
So what i suggested was to take always the next offered skill in basic until all slots were filled. Many player favorised point 1, but i could convince them to try that one first.
With doing so ull get a clear pattern of the skilling, which proofes that the results out of that are 100% correct. If u test it on your own that way, ull first note that wisdom is getting offered at a certain skill slot. Then u might see that a magic school is offered at a certain skill slot. And finally u may find some other "skill groups" per slot.
This will lead you to the correct conclusions, and despite some other veteran players say it wont help you in the game, it is very good to know how it works.
One Vet just noted to me that "he changes his main as his Barb already got offered air instead of earth". One Vet noted that as soon as his main gets offered the wrong magic school he goes for another main hero. Both are wrong.
First a Barbarian gets offered one of the magic schools at a certain slot and then again at a certain slot. So as there are 4 Magic Schools it is 1 out of 4 at that slot for a Knight (thats why he sucks) and as a Barb CANT get Water it is about 1 out of 3 at that slot.
Additional to this the magic school have posibility numbers 3 or 4, which sums up to 12, which means in case of a Barb Earth 4 Air 4 and 3 Fire.
So there are only 3 posibilities at that certain slot of Earth 4/12times Air 4/12times and 3/12times.
But what will happen if u reject the wrong Magic School offered the first time? Well, the chances will STAY the same the very next time ull get a Magic School again...means the chance to get the wrong School offered AGAIN is still there.
Lets say u accept Air. Now the next time there can be only 2 possiblities while Earth has a better chance then fire.
Conclusion: If u get a Magic School offered with a Barb that is not Earth still take it, to advance your chances the next time.
Even worse with the Knight: 1 out of 4 with a huge chance of Water Magic. Each time you reject Water it is still in the line with very low chances for Earth.
Xarfax1
|
|
Russ
Promising
Supreme Hero
blah, blah, blah
|
|
posted February 09, 2006 07:31 PM |
|
|
Well, before scientists make an experiment, they have an approximate idea of what patterns they might be looking for BEFORE doing the tests. They never go the tests 100% clueless. They don't do thing like "hey guys, I wonder what's gonna happen if we put together 10 kg of Uranium - wanna try it?" It goes more like this: "Ok, I have this theory that when you put 10kg of Uranium together it will make a very big boom - so, lets test it in a special location 1000 km away from us because I don't know exactly how big of a boom it is going to be."
I can see a point in your approach though and I'll be more than happy if you come up with some results. I still recommend that you only test each hero class instead of testing each hero, otherwise you'll never finish and this project will be forgotten.
|
|
Russ
Promising
Supreme Hero
blah, blah, blah
|
|
posted February 09, 2006 07:40 PM |
|
|
Quote:
1. One player suggested only to take only one skill, only the first skill and then restart. This looks like the statistically science correct method.
Are you talking about my method? This is what I was talking about when I said noone read it. You've read the 1-st line and now you are flaming what you think I said because you didn't bother to read past the first line. WTG.
Quote:
But what will happen if u reject the wrong Magic School offered the first time? Well, the chances will STAY the same the very next time ull get a Magic School again...means the chance to get the wrong School offered AGAIN is still there.
Are you saying that you'll be offered another magic school again in 6 levels even if you don't reject the first one?
|
|
Xarfax111
Bad-mannered
Supreme Hero
The last hero standing
|
|
posted February 09, 2006 08:24 PM |
|
|
Quote:
Quote:
1. One player suggested only to take only one skill, only the first skill and then restart. This looks like the statistically science correct method.
Are you talking about my method? This is what I was talking about when I said noone read it. You've read the 1-st line and now you are flaming what you think I said because you didn't bother to read past the first line. WTG.
Quote:
But what will happen if u reject the wrong Magic School offered the first time? Well, the chances will STAY the same the very next time ull get a Magic School again...means the chance to get the wrong School offered AGAIN is still there.
Are you saying that you'll be offered another magic school again in 6 levels even if you don't reject the first one?
First i want to state that i was not talking about you.
Yes dont reject the magic schools and youve a 100% chance of getting earth with a 3 schools might hero!!!!
|
|
angelito
Honorable
Undefeatable Hero
proud father of a princess
|
|
posted February 09, 2006 09:43 PM |
|
|
Just out of curiosity, i took Gurni and Crag, placed a box with 150.000 exp.points (enough for level 23) and tried 2 different methods.
First test (i did it 30 times), i did it like Xarfax said. I always took the new skill showing up on the new level.
During this test, Gurni got offered navigation 9 times, logistics 2 times.
Crag got offered Eagle eye 8 times, tactics 3 times.
Second test, iīve always chosen the skill i prefered (offense, armorer, tactics, air, earth, logistics, artillery....8th skill differed, coz sometimes u have to chose between 2 new skills).
Gurni got logistics 29 times.
Crag got tactics 28 times.
With the second method, i could also proof Xarfaxīs theory about the magic skills. I always took the first magic skill offered, but still, both guys got earth 27 times.
I know, 30 times is not much, but which method leads to the result a player takes into account making the decision about his main hero?
____________
Better judged by 12 than carried by 6.
|
|
LegendMaker
Promising
Famous Hero
The Metal Specialist
|
|
posted February 10, 2006 12:13 AM |
|
|
Just A Note
First of all, thanks for the valuable insight you shared from your personnal experience as a tester, Xarfax. I had already read an older thread in which you mentionned that Drachenwald-Statistic-Projekt you just exposed further here. It's obviously of high interest to learn more about your team's results for the similar Quest I'm currently on.
I have the impression you guys already succesfully decyphered the patterns 4 years ago. But you won't tell me until I've achieved an equivalent success... Just a wild guess though, so it's most probably not a reliable result ! Rofl
Also thanks to you all for keeping the thread active and interesting, even inspiring in parts ! Wrong or right in the end, all suggestions made in a constructive way are more than welcomed, if you ask me !
Angel, I don't know what to say. It seems pretty clear there has been one more missunderstanding. I understood pretty clear your point the first time, honest. You understood my original protocol, which was completely wrong, and you were perfectly right to point that out to me. Thanks again for that move.
But it seems we ceased to understand each other (on that topic) near the moment I exposed my current theory (which ruled out the original one, if anyone still wondered).
And now it seems to me you're simply ignoring whatever I might say in this thread. You tested your method and Xarfax's method, while you didn't say a word about the similar "Angel's method vs Xarfax's method vs Legend's method' test I did and posted before yours.
To me, my preliminary testings prove that both methods are unable to lead to the isolation of the whole pattern of skill advancement selection.
If you chose the skill you prefer, then your actions as a tester is a VARIABLE of the formula. Because subjective choices cannot be permanent throughout a whole batch of test.
If you chose 6 or 7 basic skills in a row, your actions as a tester ARE a Constant of the formula. Yet, since the experience only lasts up to the 8th Hero Level at best, the amount of data gathered will be too partial to lead to the whole pattern, IMHO.
So, I discarded both methods of my protocol, FYI.
Ahem. One more try at explaining my method :
1) NUMBERS
Always take the highest Skill Level between two skills.
Expert this > Advanced that > Basic **** !
2) LETTERS
Always take the Skill which has the higher rank in alphabetic order, when the Skill Level is equal.
Same Level AAA > Same Level AAB !
With these two simple and very easy to apply rules, I will be able to test any given hero (regardless of his number of starting skills and their skill level) up to Level 21, at which time he will always get his 8th and last slot filled, everytime.
(EDIT : Ulland is actually the exception that confirms the rule : he will always achieve that at Lvl 20, one Level before all the others. But that's only because Ulland starts at Level 2, at least secondary skills wise.)
Without adding any new variable in the process (meaning there is NO possible conflict between the two rules of my protocol, and so it will always be possible to follow it without ever being forced to make a choice, which would spoil the whole result by adding a new variable).
Try it if you still think I'm out of my mind !
Russ, I've taken your advice into account. At least I think I will start the experience this way. So when I'll have finished Lorelei and Jeddite, I will switch to a DeathKnight and a Necro instead of another Overlord and another Warlock. Maybe this will prove sufficient to isolate the formula, and spare me a lot of time.
If not, I'll still have the possibility of doing the 12700 remaining heroes afterwards.
Concerning your remark about scientists... lol There are actually two kinds of researchers. You described the first one only. What I'm doing here is relevant to the second. Ever hear about Fundamuntal Research, Russ ? That's when you don't make suppositions, you just observe something very very thoroughly until you start to understand how it works.
That said, my wild guess is, for the moment :
1) That the comp selects which skills it will offer at which time following a certain formula wich includes several distinct constants (some have already isolated two examples of that : Magic Schools and Wisdom) and also several distinct variables, which are yet to be found.
Suffice to say for now that Xarfax is absolutely right when he states that the known tables couldn't possibly be farther removed from the truth.
2) That all 28 Skills are "labelled" by the comp, and that it has to do with Letters and/or Numbers. I'm not sure yet wether it's based on alphabetical order or not, but my conviction is that it's quite likely from what my first results showed me. The very fact that my alphabetic order based protocol gives contant results in terms of at which Level any hero gets his 8th and last skill (Level 21. Not proven yet, but it's always been the case from the almost 200 tests I completed this far) strongly suggests that.
3) In any case, the kind of formula I expect to find out would look roughly like a Basic program text. A succession of "IF", "OR", "AND", "THEN" with variables and/or constants in between.
Xarfax, now that you could hardly pretend you didn't notice that a 6th player (me) has suggested a brand new method / theory, would you mind please telling me wether or not you think it has a chance of leading to "correct results" ? I'd be thankful if you could.
More on that soon. Legend.
____________
LM
|
|
angelito
Honorable
Undefeatable Hero
proud father of a princess
|
|
posted February 10, 2006 11:08 AM |
|
|
Legend....i understood your method pretty well.
But Russ posted it already above: "What do we want to prove?"
What results will your method bring? What conclusions will we make of it for our games?
If i will chose "archery" before "earth" coz of the alphabetical order, it may fit in your rules, but will NEVER happen in any game, so how reliable can these results be?
If i wanna test if a Ferrari wins in a race vs a Porsche, i wonīt drive races vs a Fiat 500 and a Toyota Starlet first, neither will i throw the Ferrari into a lake to see how far he can swim. Sure i will get results from the races vs the Fiat and from that lake drive, but no1 is interested in these results, coz they donīt tell me anything about my goal..."will i win vs the Porsche?".
Thatīs why i say, this method is an enormous work, which brings non important results.
As u said in your very first post here:
Iīm tired of hearing "This hero levels up crap, this hero levels up good". I wanna know it exactly.
If u have a hero in a game and want him to be your main, you ALWAYS chose the skills you prefer, and after finishing the level up to level 21, u can say "He leveled up good" or "He leveled up crap".
You will NEVER be able to judge that by chosing the skills in alphabetical order. Thatīs why, in my opinion, your results wonīt help anything in practice.
____________
Better judged by 12 than carried by 6.
|
|
LegendMaker
Promising
Famous Hero
The Metal Specialist
|
posted February 10, 2006 04:09 PM |
|
|
Now I See
1) Everything you (Angel) just said is perfectly right and makes perfect sense in itself.
2) Your doubts are perfectly justified.
3) I know exactly what the missunderstanding is.
4) It is :
5) You assumed that the raw results would be the final stage of my project.
6) If it was the case, again, you're 100% right : the whole thing would be completely useless. None will ever use alphabetical order selection method in a game.
7) But it's definitely not the case. The case being :
8) As I stated, my protocol is set as an additionnal Constant in the formula. One that I know for sure, because it's me who added it.
9) When I'll have sufficiently numerous raw results at hand, the second stage of the project (and of course the most crucial one) will begin.
10) In the second stage will start the observations, the hypothesis, the calculations... all based on my Constant. All meant to try and isolate the original formula.
11) My hope is it will lead to important conclusions.
12) Even if it's only say, Earth's position in the whole pattern that you care about...
13) That doesn't change the fact that Earth is but one in many elements in the pattern/formula.
14) So, understanding how the whole pattern works will lead to understand how Earth position works, eventually.
15) While, IMHO, observing countless times Earth and just Earth will lead neither to the whole pattern nor even to Earth's position in that pattern.
Regards. Legend.
____________
LM
|
|
tigris
Supreme Hero
Supreme Noobolator
|
|
posted February 10, 2006 11:17 PM |
|
|
what makes you think that alphabetical order is less subjective than other method,Legend?
U once said to me that you want to put yourself in the mind of the AI. what makes you belive that the Ai would be tempted to pick exp wisdom instead of basic mistycism at lvl 2 for jeddite, or pick basic earth in the same situation?
u think this depends on the alphabetical situation? i really doubt it.if you really want to understand how the AI thinks you do have to find patterns as u said, but starting a search based on a simple alphabetical criteria isn't right.
____________
|
|
bjorn190
Responsible
Supreme Hero
Jebus maker
|
|
posted February 11, 2006 05:07 PM |
|
|
|
When Homm5 comes this will not matter so much anymore.. What little benefit comes out of it will fade away.
|
|
Binabik
Responsible
Legendary Hero
|
|
posted February 11, 2006 09:03 PM |
|
|
I thought of suggesting "gang tests" before but thought there wouldn't be any interest. I also knew it would be difficult to get people to agree on how the test is done. But since there appears to be at least some interest, I'll put in my 2 cents.
First, testing is very tedious and time consuming. Many people, including myself, may be willing to do it. But ONLY if we agree the test is valid and will give usable results. Otherwise it's just a waste of time and a lot of work.
Legend: The reason people are dropping out of your test is that it is way too broad. What you want to do is test the entire level-up system. You suggest 100 tests per hero. Statistically, 100 tests per hero is not even close to being enough tests for that. IMO it would take 100s of thousands, if not millions of tests to do what you want. Xarfax's group already did 81K tests. You suggested Xarfax wouldn't tell you the results. He DID tell the results. He basically said it was inconclusive and he doesn't know how the system works.
Even if enough tests were done to be statistically significant, there would be way too much data to analyze.
Bottom line, I agree completely with Russ that you can only test for specific things. That not only requires MUCH less testing, it's easier to analyze the results since you're only looking for one thing.
So the question is, what to test for.
The tests I've done show that the table in Russ's first post (and other posts) is approximately correct. But, only as a default....it doesn't tell the entire story. There are exceptions and special rules. We know about wisdom and magic at levels 4 and 6. But we don't know how many other exceptions there are. We don't even know *if* there are any other exceptions (at least I don't).
The goal is to find those exceptions....things like wisdom/magic that are predictable.
Always taking the new skill. Always upgrading existing skill. Always taking good skills and turning down bad. All three of those make valid tests, depending on what you are testing for. You can't design a test unless it's known specifically what you are testing for.
And you must do your best to eliminate any other factors. Like you can't use restart or reload game. You must start a new "game" from the main menu after each test. We all know that random is not random in HOMM....there are things that cause repeating patterns. We just don't know what those things are. It's even been suggested before (by Hex???) that the grid location on the map might make a difference.
To summarize: Tests need to be for specific things. The test needs to be designed for that specific thing. As much as possible, all other factors be eliminated. Properly analyze the results.
I'd suggest the following tests:
Find other things like wisdom/magic that have a fixed pattern. Especially look for logistics, offense and armorer.
Find out if individual heroes have specific level-up rules. Same for hero class. (there is some evidence that individual heroes MIGHT have their own rules, but that evidence isn't very strong and would probably require a LOT of tests to prove either way)
Find if specialists are treated differently. (My tests suggest they are treated the same, but I don't have enough data to be conclusive. The 81K test would probably show this if it tracked individuals and not just totals)
Find out if past skill selections affect future offerings (my guess is "no"). Find out the opposite, if turning down a skill affects future offerings.
One last thing.
Quote:
When Homm5 comes this will not matter so much anymore.. What little benefit comes out of it will fade away.
I'll still be playing HOMM3. There's been a lot of good games come out over the last 6-7 years, and none have replaced HOMM3. They are just something to play while I take a short break from H3.
____________
|
| |
|
|
