Heroes of Might and Magic Community
visiting hero! Register | Today's Posts | Games | Search! | FAQ/Rules | AvatarList | MemberList | Profile


Age of Heroes Headlines:  
5 Oct 2016: Heroes VII development comes to an end.. - read more
6 Aug 2016: Troubled Heroes VII Expansion Release - read more
26 Apr 2016: Heroes VII XPack - Trial by Fire - Coming out in June! - read more
17 Apr 2016: Global Alternative Creatures MOD for H7 after 1.8 Patch! - read more
7 Mar 2016: Romero launches a Piano Sonata Album Kickstarter! - read more
19 Feb 2016: Heroes 5.5 RC6, Heroes VII patch 1.7 are out! - read more
13 Jan 2016: Horn of the Abyss 1.4 Available for Download! - read more
17 Dec 2015: Heroes 5.5 update, 1.6 out for H7 - read more
23 Nov 2015: H7 1.4 & 1.5 patches Released - read more
31 Oct 2015: First H7 patches are out, End of DoC development - read more
5 Oct 2016: Heroes VII development comes to an end.. - read more
[X] Remove Ads
LOGIN:     Username:     Password:         [ Register ]
HOMM1: info forum | HOMM2: info forum | HOMM3: info mods forum | HOMM4: info CTG forum | HOMM5: info mods forum | MMH6: wiki forum | MMH7: wiki forum
Heroes Community > Library of Enlightenment > Thread: On the internals of offered skills when leveling-up a hero.
Thread: On the internals of offered skills when leveling-up a hero. This thread is 17 pages long: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 · «PREV / NEXT»
dimis
dimis


Responsible
Supreme Hero
Digitally signed by FoG
posted June 27, 2006 12:42 AM
Edited by dimis at 01:53, 27 Jun 2006.

Quote:
So is my theory complete nonsense (), or could it happen like this?
Just for the record: I am with you on this one.

And another thing I want to verify is whether Already Gained Skills have higher priority over Not-Already Gained Skills; meaning that the exception on every 4 levels "hits" first your magic skills that your hero already possesses and only if they are not available (for upgrade or ... not at all ) then you are presented a new magic skill as basic.
If we are able to verify this up to a logical extent of certainty, then ANSA chart is still OK because all 5 assumptions in list on "Groups" post on page 1 are not contradicted under the shed of this new light (along with the assumptions in "Model-Related-Options" paragraph in ANSA post).
____________
The empty set

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Binabik
Binabik


Responsible
Legendary Hero
posted June 27, 2006 02:18 AM

It's been said there might be other factors involved in the skill tree that we don't know about, especially the possibility of skill groups. I can't seem to find my test data, so the following is from memory.

There is evidence (but FAR from proof) that there are at least 2 broad groups of skills. And possibly 1-2 more small groups or sub-groups. The data shows some skills being offered MUCH more than expected from the skill table.

When looking at this stuff, it helps to put yourself in the designer's shoes. Does it make sense from the designer's point of view? Is there some reason why they would do it a certain way? You also have to consider that the experienced players have changed their priorities over the years....the designers didn't have 1000's of hours of experience to go by when making their decisions.

Anyway, it appears the designers tried to balance the skills for a hero between the broad groups of map skills and fighting skills. Some skills were specifically given higher priority which is obvious from the skill table. Other skills are lower priority and have 1's and 2's in the table. Individually these are low priority, but as a group they seem to be  given higher priority to provide some balance to the hero. In other words, no single skill from this group is important, as long as some of them are offered.

Modern players create a super hero whose only job is to train for the final fight. But did the designers design for this? Probably not. It appears that they intended for different heroes to be used for different things, but not as specialized as the way modern players use them, especially the use of a single super hero.

So there is not only evidence of skill groups, but it also makes sense from the designers point of view to balance the heroes with a variety of skills types.

I wish I could find my test data, but from memory, there appeared to be a group of about 5-7 low priority skills and they were all map skills. Adding up the value of these skills might total 8-10. An individual skill with an 8-10 rating would usually be offered, but not always. The skills from the low priority group were ALWAYS offered. I tested both ANSA and AGS. When using AGS, skills from this group were not only always offered, but they were offered at least 2-3 times.

These low priority skills were the most obvious, but it seemed like there may be other groupings as well. Patterns of battle skills for a might hero are harder to see because they have high values. If there is some kind of grouping, it doesn't stand out because the odds of getting them are already high. But there is some WEAK evidence that battle skills may be balanced between "direct" combat skills (like armorer, offense and archery), and indirect or special purpose skills (like tactics and ballistics).

____________

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 27, 2006 01:00 PM
Edited by Ecoris at 14:13, 27 Jun 2006.

@ Binabik
Did you do your tests manually (i.e. reloading, level up, note skill offers, reload etc.) ? You mentioned having done about 1300 tests with knights from level 1 to level 2.

Is there any way to make the computer do such tests?

@ dimis
Does the "ansa" program compute the skill probabilities like I computed the spell probabilities, i.e. compute the probability for each permutation of selections (skill #3, skill #4, magic, skill #6,wisdom, skill #7, skill #8) for heroes that start off with two skills, and then add the probability for this permutation to the probabilities of the selected skills?

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
dimis
dimis


Responsible
Supreme Hero
Digitally signed by FoG
posted June 27, 2006 01:26 PM
Edited by dimis at 14:09, 27 Jun 2006.

Quote:
Does the "ansa" program compute the skill probabilities like I computed the spell probabilities, i.e. compute the probability for each permutation of selections (skill #3, skill #4, magic, skill #6,wisdom, skill #7, skill #8) for heroes that start off with two skills, and then add the probability for this permutation to the probabilities of the selected skills?
If I understand correctly what you are saying, the answer is yes.

Just a clarification though:
Quote:
.. the probabilities of the selected skills
I suppose you mean the skills that appear in the corresponding permutation, where one has just computed the probability for this permutation to occur, exactly as on the "per-level" permutations you encountered on spell-probabilities problem.

If this makes it easier I'll try to rephrase what I write on "The idea of the solution" paragraph, since this might be a good oppurtunity for others to clear some things up:
Quote:
... In fact, the tree where computations take place has weights on probabilities for each offered skill at each level. As a result, the computed probability for a skill is the sum of "partial probabilities" of the specific skill through all possible paths; where "partial probability" is the overall probability for the specific instance of the path (multiplication of probability weights from root note to leaf node).
Ofcourse a path in the generated tree is a legitimate permutation of the skills (meaning taking into account Wisdom and Magic exceptions but no other group exceptions so far).

In any case, the probability for skill_i on non-exception levels is computed by the formula:
prob (skill_i) = weight (skill_i) / SUM ( weight (skill_j) )
for all skill_j skills that are not already selected by the hero.
On exceptions:
Assuming no wisdom-offer prior level 6:
prob ( Wisdom ) = 1.
and regarding Magic School on level 4 (under the assumption you were not offered a magic school earlier):
prob ( skill_i ) = weight ( skill_i ) / SUM ( weight (skill_j) )
where skill_j varies only among the Magic Schools.

The reason I use the above formulas is because I believe (99+% convinced) that programmers used their most common technique when they had to select randomly; meaning implement the mod () function on random number returned by rand () and that's it. The only other thing I would like to check is a Boltzmann-Gibbs distribution like here. But this will have to wait since I find it slightly unreasonable this selection method was used. However, in the end with slight experimentation on temperature-variable we might reach better results (closer to actual game-play). But this is way ahead on our search ... I just place it here as a future note.
____________
The empty set

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 27, 2006 02:12 PM

Quote:
Just a clarification though:
Quote:
.. the probabilities of the selected skills

I suppose you mean the skills that appear in the corresponding permutation, where one has just computed the probability for this permutation to occur, exactly as on the "per-level" permutations you encountered on spell-probabilities problem.

Yes that is what I meant; the probability of the permutation is added to the total probability of every single skill of the permutation, then the next permutation is selected, etc. etc. Thus many different permutations contribute to the total probability of any given skill.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
dimis
dimis


Responsible
Supreme Hero
Digitally signed by FoG
posted June 27, 2006 02:21 PM
Edited by dimis at 14:22, 27 Jun 2006.

These additions take place:
GMP-lib-version: lines 1260-1263.
double-version.: lines 1226-1227.
All above refer to mighty_classes.cpp file. Hope this helps in understanding the code better.
____________
The empty set

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Binabik
Binabik


Responsible
Legendary Hero
posted June 28, 2006 12:46 AM

Quote:
@ Binabik
Did you do your tests manually (i.e. reloading, level up, note skill offers, reload etc.) ?

I do all my tests manually. The test I referred to in my last post, I Wrote down the entire skill tree. I went back to the main menu and started the test new each time. Occasionally I quit HOMM comlpetely and started it again.

Quote:
You mentioned having done about 1300 tests with knights from level 1 to level 2.

I've done a lot of tests and I'm not sure which one this is. It's definitately not the one referred to in my last post. Where did I post it? Did you have a specific question about that test?

Quote:
Is there any way to make the computer do such tests?

Legend has several posts about automated tests in the Heroes'Stats and Skills Chance thread (currently on Library page 4). I didn't follow everything he said, but you can read it or ask him about it. Personnaly I don't trust something automated. But I hate making blanket statements like that because there may be times it can be done and be completely valid tests.
____________

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
dimis
dimis


Responsible
Supreme Hero
Digitally signed by FoG
posted June 28, 2006 12:50 AM
Edited by dimis at 00:50, 28 Jun 2006.

1300 tests on Knights

Quote:
Quote:
You mentioned having done about 1300 tests with knights from level 1 to level 2.
... Where did I post it?
It is on page 1 of "Heroes' Stats and Skill Chances"; right here.
____________
The empty set

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
dimis
dimis


Responsible
Supreme Hero
Digitally signed by FoG
posted June 28, 2006 01:26 AM

@Xarfax - Logistics matter.

I took this from the "Heroes' Stats and Skill chances" thread:
Quote:
If the first 4 upgrades did come u could be 100% sure that Logistic will come next.
This seems as an exception to me.
Xarfax can you please clarify if this is generic, or should you have a certain number of skills? Moreover, do you mean to have upgraded all 4 slots in your hero at advanced level? Or is it expert? Or you want those advances (and at which level) have been offered so far for your hero so that you are offered Logistics?

Thank you in advance.
____________
The empty set

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 28, 2006 01:38 PM

I think Legend's "automated tests" only consist of letting the computer do the choosing. I.e. letting the computer do the level-up, reloading the game as a hot-seat game where you take control of all sides and then looking at how the computer chose.

This was not was I was thinking of. I was wondering if there is any way to automate the process to save time and be able to do large samples.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 28, 2006 03:05 PM
Edited by Ecoris at 15:12, 28 Jun 2006.

RESTARTING contra MAIN MENU

I thought I'd do a test when leveling up from level 1 to level 2. Starting hero was Pyre with NO skills. I wanted to note down 1st offer and 2nd offer in a 28x28 table in order to search for dependence. The offers started looping for certain after 26 restarts. I relaunched the test scenario from the main menu and got a completely new combination the first time...
Thus what some might have expected but not known for sure is now a fact (I've never done tests where this was important):

RESTARTED SCENARIOS ARE NOT INDEPENDENT

The results obtained from large sample test are therefore useless unless the scenario is restarted from the main menu every time .

I guess we'll never find any more exceptions (if they even exist) unless we know exactly what we're looking for.

Is there any way of making the computer do the testing itself?

EDIT: I read something more interesting i Xarfax' post that dimis mentioned:
Quote:
The "Skilling Tree" will be generated right at the start of the game for every hero. Even for the heroes that are not bought and available in the tavern. All this data will be fixed and packed into a file. Period.

If uve a look into this data it looks approx. like:

Hack Offense Pathfinding Resistance Air Mysticism Logistic Artillery Wisdom Navigation Earth Blacksmith ...and so on.
Can anyone confirm this/describe how to access these files?
I always thought that the skill tree was fixed because it was associated with a seed, thus the random numbers generated would be the same.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Xarfax111
Xarfax111


Bad-mannered
Supreme Hero
The last hero standing
posted June 28, 2006 11:41 PM
Edited by Xarfax111 at 23:44, 28 Jun 2006.

Quote:
But to keep the chances high getting earth offered, u refused Log, Tactics and Armorer. Call me wannabe, but i can´t think of any situation where i would refuse Log or Tactics in the first 10-15 levels of my main.


Thats exactly what i say. Im 100% sure to get Earth when you accept every magic school u get offered. If you prefer other skills like Logistics the chances are getting lower. I think its a matter of taste. In my opinion on most maps, especially randoms a simple Mass Slow is the Game Winner. With Mass Slow you are faster then Logistics will ever make you. So I personally often did bite into Water/Air/Fire Magic instead of Logistics to ensure to get Earth. But thats just a matter of taste. I dont think that Tactics is that important, only if i play or face Stronghold/Fortress i take it. When i expect my opponent to take Tactics anyway i dont take it at all. But thats not the story we are talking about atm.


Quote:
I took this from the "Heroes' Stats and Skill chances" thread:

quote:
--------------------------------------------------------------------------------
If the first 4 upgrades did come u could be 100% sure that Logistic will come next.
--------------------------------------------------------------------------------


Well with my bad English i couldnt explain it better. So as an example. If you level up a Barb like this:

Offense Logistics Defence Earth EagleEye Wisdom Pathfinding
..you can be sure that if you see this next time a Barb leveling up..

Offense Logistics Defence Earth EagleEye Wisdom...

to get Pathfinding next.

So lets say youve about 80 complete level ups, and 3 of them look identical, the possibility of this happen with "real probabilities" with that much skills is sub zero. Means, in my Theory,that the level up follows a "stripe" that is fixed in the data transfer file.


Well that actually how a hero look like. Its from the data transfer file. There is everything fixed for the game you play:


Ajit                                                                                                                                                                                                         ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ  
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ
ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿ                       l   ðl                 ÿ   ÿÿÿ ÿT
Tÿÿ‹


  m       c      €    
 •                @                                           Y   T   [   V   a   ]   ÿÿÿÿ-   ¬      P  
    

Some of those signs are fixing the skills that Ajit gets.


My Theory is that each Hero class has a certain amount of those "stripes" to ensure a heros class skill balancing and to let them level up in the hero class typical way.

In real life those "stripes" are used for "one-armed-machines"(or lottery machines or however they are called). Thats because the law wants to be 100% sure that the chances for a win are on a regular and high enough basis.

Some criminals used this knowledge with computerprogramms. They watched those machines for about 1 hour and could calculate when there will be a win on the stripe.

Against my Theory is the txt-file which was used by the developers to adjust the game. There you can influence the skilling of the
hero to. So in the end it might be a mixture of both. But who will know?  

Im a bit out of business, so some numbers might be lacking a bit, but the main things stay right.
____________

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Binabik
Binabik


Responsible
Legendary Hero
posted June 29, 2006 09:56 AM
Edited by Binabik at 10:17, 29 Jun 2006.

Xarfax, when you use the word "stripe", I think the English word you want might be "sequence" (if you trust babblefish "reihenfolge" in German). When I did the tests on groups of skills that I posted earlier, I did some other tests about the sequence of skills. I was planning on posting this anyway, so I guess I'll do it now. It sounds like you were looking at the same things I was. And I bet you didn't figure it out.. I didn't figure it out either and came up with theories similar to what you did.

---------------------------

First, all my tests were done with a hero starting without any skills, so all 8 slots were open.

Consider the following sets of skills (real data, not made up):

1) learn, dip, path, art, wis, wat, off, int
2) bal, nav, arch, lead, wis, wat, myst, log
3) myst, art, earth, path, wis, dip, off, sorc
4) off, path, air, learn, wis, dip, est, log

OK, lets say you start a game and use an event to level up your hero. Then you start doing restarts. What will happen several times at the beginning is you will get a different set of skills each time. The first several restarts will appear to be random and will not repeat.

Now you get the ones listed above. A few things can happen. After set 4, you go back to set 1 and repeat the 4 sets in order. You can repeat these 4 sets several times in a row. If you keep restarting long enough, you MIGHT break out of the loop and start getting random skill sets again. If you continue to restart after that, you might go back to the same 4 sets. Or you might get 4 NEW sets that repeat themselves.

There are some variations of this. That's what makes it hard to figure out. You definitely go into loops, but it doesn't always happen exactly the same way. I only posted 4 sets, but in reality it's usually more like 5-8 sets that repeat.

Here's one thing I think Xarfax is saying. If I see the first three skills are learning, diplomacy and pathfinding, I KNOW I have skill set #1. Therefore I KNOW the next skill will be artillery. In other words, it becomes entirely predictable which skill tree I'm in after the first 2-3 skills are offered.

I only posted 4 skill sets, but there are a bunch of them. I see this and start to think. I found this by doing restarts, but what if there are a bunch of skill sets that are used in EVERY GAME??? If I test enough and record all the trees, can I use this in other games to predict which tree I got after seeing the first few skills offered? In other words, is there a relatively small finite number of trees? Instead of generating a new tree at the start of the game, does it only randomly pick from a bunch of pre-defined trees?

Or maybe they aren't trees, maybe it's one long SEQUENCE of skills. And the game just picks a random starting point somewhere in that sequence.

Picture this sequence of skills:

arm, scholar, log, learn, dip, path, art, wis, wat, off, int, bal, nav, arch, lead, wis, wat, myst, log, myst, art, earth, path, wis, dip, off, sorc, off, path, air, learn, wis, dip, est, log

The 4th skill in this sequence is the beginning of skill set #1, followed by sets 2,3,4. Is there really one big long sequence and the AI just happened to pick the point where skill set #1 started? And in another game it picks some other starting point?

If I understand Xarfax right, this is where his analogy to gambling comes in. (they're called "slot machines" Xarfax). You pull the arm on the machine and all the symbols start rolling. Then they stop at some random place. But is it really random? If it's pure random, the odds favor the casino. But with pure random it's POSIBLE for someone to hit the jackpot 100 times in a row and bankrupt the casino. They don't want this to happen even if the odds are extremely small. So they can make it semi-random, with certain things that will never happen, like 100 jackpots in a row.

Let's say the odds of a jackpot are 1:100. They don't want pure random because they can go bankrupt. They also don't want it to happen once every 100 times because that's predictable. So they can make it so the jackpot happens once every 75-125 times.

If HOMM is pure random, it would be possible for the worst 8 skills to appear next to each other in the sequence. But the game would be more balanced if it was done so the skills are more evenly distributed in the sequence. But you also don't want them exactly even and predictable. You want it to appear to be random, even if it's really only semi-random.

So you can create a long sequence of skills. Within any given section of the sequence, there is a variety of skills. The skills could be grouped. In each section, there are some skills from each of the groups. The specific skills from each group could be random, as long as each section of the sequence has skills from every group.

According to the skill table, armorer and offense will appear more often than mysticism and luck. But in the long sequence of skills, armorer and offense will never be bunched together in one place, and not appear at all in another place. They would be more evenly distributed, with skills like mysticism and luck in between.

I want to stress that I don't KNOW if there's really a long sequence like this. Or if the sequence is only a symbolic way to explain the concept of semi-evenly/semi-randomly distributing the skills.

____________

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 29, 2006 02:05 PM
Edited by Ecoris at 14:07, 29 Jun 2006.

The idea that a sequence determines the offers for each hero is not a bad one.

Xarfax, could you be a little more precise on where you found that txt file? Is it included in the savegame file (xxx.GM1) or in one of the .lod files from the "Data" folder?

The GM files are encrypted like I think all game files are. ModHomm3 can only access the .lod files. Does anyone know how to decrypt/access the .GM files?

I did a few tests myself. Leveling up from level 1 to level 2 (no starting skills). Noting which two skills were offered. In all cases the offers started looping after a few selections (I ran the loops through several times). There were tails leading to the loops (i.e. rho-structure):



In my tests the length of the loop was 3 in the first one and 5 in the second one.

Each restart + level up brings one to the next node. However I noticed that reloading twice would 'break' the loop.
The source of the XP gained or the skill selected had no effect.


The question is, what do such results suggest? The mentioned loop-structure reminds me of recursively defined integer sequences with an upper limit. Such sequences can be represented as finite graphs with a very restricted structure (but I can write a lot more about such graphs if they become important).

It seems as if restarting just selects the next 'random' integer from the sequence the random number generator generates, and uses that integer to determine the skill tree. But the 'double-restart' contradicts this: The only difference is that I otherwise actually did the level up. I had suspected that double-restarting would just skip one node in the loop, but the fact that I didn't level up actually affected this. Is it possible that the skill tree is not generated until 'needed'?. That would explain this odd behaviour.


Fianlly I have a question that might be important. Consider the situation with a level 1 hero with no starting skills. When leveling up to level 2 you're offered two skills A and B. Suppose you select A. Now when you level up to level 3 you'll be offered advanced A and a new skill C. Is it possible that B = C?
If new skills were selected at random with probability (skill value)/(112 - value of selected skills), then the probability that B = C (the same skills is offered twice in a row) would be > 0 and should occur once in a while. If it never happens, well, it doesn't work like that. Please let me known if you experience a counter-example.

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
maretti
maretti


Responsible
Supreme Hero
posted June 29, 2006 08:02 PM

It is possible to be offered the same new skill 2 times in a row.
____________
Crag rules, Orrin and Ivor suck

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 29, 2006 08:56 PM
Edited by Ecoris at 20:56, 29 Jun 2006.

In the situation I mentioned?
How many skill slots were occupied when it happened?

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
maretti
maretti


Responsible
Supreme Hero
posted June 29, 2006 09:10 PM

I started with a normal hero (2 starting skills) and was offered the same new skill 2 times in a row as level 13 and 14 or so. After reading your post I wanted to check it out and it happend with the 2nd hero i tested. So its probably not very rare.
____________
Crag rules, Orrin and Ivor suck

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 29, 2006 09:21 PM

Good, it might be important in determining what the cause of loops and other exceptions is.

1. A predefined sequence determines the skill tree.
2. Recursively defined 'random' integers cauese loops etc.
3. Other alternatives?

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Binabik
Binabik


Responsible
Legendary Hero
posted June 30, 2006 01:21 AM

Yes, being offered the same skill twice in a row is quite common. I don't remember if I've ever seen one offered three times in a row.

What I mentioned in my last post about preventing skills from being bunched up in one area still applies. It could be viewed as a rule about the "density" within a section of the sequence. A density map might show "on average" the skill is evenly distributed, but that doesn't prevent a skill from appearing twice in a small area.

I'm not a programmer. Some of you guys with programming knowledge might be able to explain what types of things are easy/hard for a programmer to do. It seems likely to me there would be known algorithms for this type of thing. And for sure there are mathematical models for generating sequences with various types of distributions.

Quote:
Each restart + level up brings one to the next node. However I noticed that reloading twice would 'break' the loop.
Do you mean reloading a saved game, restarting, or main menu? The tests where I got the sequence in my last post were done a couple months ago. I saved at the start of that test, and I could reload that save today and go into the same loop. My tests showed going back to the main menu as the only way to break the cycle.....and I'm not convinced even that does it.

Ecoris: Just for the heck of it, when you are in a loop, try the following: Go into your town and build a guild and note the spells offered, and note the heroes in the tavern. If you have an artifact starting bonus, note that as well. My tests have shown they are all part of the same loop. I'd be curious if others get the same result.


I have more info from the tests I've run, but I don't want to post it all in one post. I also have a small amount of info on branching. The branching tests did nothing more than create new questions. But I've thought of some good tests to run. For example doing enough restarts to identify the skill trees inside the loop. Then see what happens when you branch before going into the loop. And also what happens by branching inside the loop. Are there imbedded loops?

____________

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Ecoris
Ecoris


Promising
Supreme Hero
posted June 30, 2006 01:42 PM
Edited by Ecoris at 13:44, 30 Jun 2006.

Quote:
Each restart + level up brings one to the next node. However I noticed that reloading twice would 'break' the loop.
It should have been "restarting twice".
The test was quite simple. A level 1 hero with no starting skills levels up to level 2. Write down the two offers, select one, restart scenario (from the options menu) and do the same. The loop started quickly and was not very long. Then I restarted the game without doing the level up (i.e. double-restarting). I had expected to remain in the loop and just skip a node like this, but the loop was actually broken, and the only difference was that I didn't level up. (I never reloaded or went back to the main menu).

 Send Instant Message | Send E-Mail | View Profile | Quote Reply | Link
Jump To: « Prev Thread . . . Next Thread » This thread is 17 pages long: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 · «PREV / NEXT»
Post New Poll    Post New Topic    Post New Reply

Page compiled in 0.1301 seconds