|
| Thread: SKILLING TREE: Hack the Hack |  This thread is pages long: 1 2 3 · «PREV / NEXT» |
|
Binabik
Responsible
Legendary Hero
|
posted August 03, 2006 12:01 PM |
|
Edited by Binabik at 12:03, 03 Aug 2006.
|
This is something I've suspected for a while, but couldn't prove until now. I came to the same conclusion as Maretti. You must completely quit the game to get good test results. I've already found an easy way to do it though and it only takes a few seconds longer.
1. Set up a hotkey for homm, like Ctl-Atl-H
2. Move all maps except the test map to a backup folder
3. Use the hotkey to start homm
4. Click New Game, then Single Player, then Begin (hotkeys N,S,B in the English version)
The test map will automatically be used, so you don't have to pick the map.
I did 100 tests and entered all the data into Excel. I went to the main menu each time. I started seeing the repeat patterns. Then I quit homm completely and started the game fresh to see if I got any of the same skill trees. I did about 30 level ups and never saw a tree I had before. See my next post for more info on the tests.
------------------------------------------------------
We've discussed the idea that the skill trees are created at the beginning of the game. We've also decided the trees are not pre-defined and stored in a data file because hacking the skill table changes the trees. I mentioned in another thread that when you hack the data file, the change is not read unless you completely quit the game and start again. A restart, or going to the main menu does NOT read any changes to the data file.
My theory is that the game creates skill trees as soon as you start homm, not when you start a new game. If this is true, a good test would be:
1. Create two different "hack the hack" test maps.
2. Start homm.
3. Record enough data from the first map to find patterns in the tree.
4. Don't quit homm, but use the second test map to look for the same patterns as the first map.
It's 6:00 AM for me and I haven't been to bed yet. So I'll let the Europeans who've already had some sleep play with it.
____________
|
|
maretti
Responsible
Supreme Hero
|
|
posted August 03, 2006 12:12 PM |
|
|
Funny that you mention it, I allready did that but you posted before I had edited my post.
____________
Crag rules, Orrin and Ivor suck
|
|
Binabik
Responsible
Legendary Hero
|
|
posted August 03, 2006 01:01 PM |
|
Edited by Binabik at 13:25, 03 Aug 2006.
|
I'm going to post my test data from when I went to the main menu instead of quitting the game completely. I don't think this is wasted data. We don't know exactly what it is we are looking for. We are just looking for patterns, so any data is usefull. We just need to be aware of how the test was done.
Here's an example:
Out of 100 tests, 16 of them offered Luck as the first skill. There were 3 different trees that started with Luck.
8 out of 16 were the following tree:
luck,resistence,fire,leadership,wisdom,armorer,scouting
When I quit the game completely and re-tested, I never saw the above tree exactly. But I saw trees that were similar. Like 5 or 6 of the skills were the same, but in a different order. Why? And the 1-2 skills that were different, is there some connection? Like do they belong to a skill group?
I'm thinking it might even be a good idea to gather data by going to the main menu and finding the patterns. Then we can compare the results from different people and try to find things that are similar. And look at the things that are different.
Dimis questioned if we should test other things besides "always click right". I agree, but not right now. I think testing something small and easy is the best way to go. Then we have data that can be analized easy. If we do something complicated, we'll have a bunch of random data that will be extremely difficult to analize, maybe impossible. I think the best way to go is to take one small step at a time. Then decide after that what is the next step.
In my opinion, a good way to test is:
1. Everyone test Crag Hack and always click right.
2. Do the test going to the main menu, not quitting homm completely.
3. Look for repeating patterns.
4. Post results.
5. Compare results from different testers.
6. Compare results when skill "A" is the first skill offered.
7. The trees will be different when skill "A" is offered first, but how?
Testing and comparing results like this is something small and easy enough that we can analize it without getting totally confused. In my opinion, we MUST be systematic about it....one simple step at a time.
-------------------------------------
Anyway, here's my results. It's comma delimited. The first number is from the odds table, just for reference. To fit on the screen, I broke it down into groups of 20.
Xarfax, I exported from Excel to 4 different text formats. Comma delimited seems to be the best one for posting at HC. It should be easy to import to excel.
Test: Crag Hack, always click right, go to main menu after each test.
Odds,Crag Hack,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20
3,Air,4,,,,,,8,,,,,8,,,4,,4,,4,
7,Archery,,,,3,8,,,,,,,,,,,4,,,,
6,Armorer,,3,,,,,2,7,,5,,,,7,,,,8,8,5
8,Artillery,7,,,,,,,,,7,,,,,,8,,7,,
8,Ballistics,,5,,2,5,,5,2,7,8,,,2,,2,,,,3,
1,Diplomacy,,,,,,,,,,,,,8,,,,8,,,
2,Eagle Eye,,8,,7,,,,,,3,,,,,,,,5,,
3,Earth,8,,7,4,4,4,,4,3,4,4,,3,,5,3,,4,,4
2,Estates,,,,,,,,,2,,,,,,,,,,,
2,Fire,,4,4,,,,3,,5,,,4,,4,,,7,,,
1,First Aid,,,,,,,,,,,,,,,,,,,,
1,Intelligence,,,,,,,,,4,,,,,,,,,,7,
5,Leadership,5,,2,,,,4,,,,,,5,5,8,,,,,
4,Learning,,,,,,,,,,,,7,,,,,5,,,
7,Logistics,,,,,,7,,,,,8,,,,7,,3,,6,3
3,Luck,,,,,7,,,,,2,5,,,2,,,,,,
3,Mysticism,,,8,,,,,,,,,,7,,,,,,,
2,Navigation,,,,,,,,,,,,,,,,,,,,
0,Necromancy,,,,,,,,,,,,,,,,,,,,
10,Offense,,,,,,,,,,,,,,,,,,,,
8,Pathfinding,,2,3,,3,2,,5,,,6,5,,,,7,,3,2,
6,Resistance,3,,,,,5,,,,,2,3,,3,,5,,,,7
1,Scholar,,,,,,3,,,,,,2,,,,,,,,2
8,Scouting,,,5,5,2,8,,3,8,,7,,4,8,3,,,,,8
1,Sorcery,2,,,,,,,,,,,,,,,,,,,
8,Tactics,,7,,8,,,7,8,,,,,,,,2,2,2,,
0,Water,,,,,,,,,,,,,,,,,,,,
2,Wisdom,6,6,6,6,6,6,6,6,6,6,3,6,6,6,6,6,6,6,5,6
Odds,Crag Hack,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40
3,Air,,4,,,4,,,,,8,,,4,8,,,,,8,
7,Archery,,,,,2,,3,8,,,,,5,,,,,,,
6,Armorer,,8,2,7,,,,,,2,7,,,2,7,,5,,,
8,Artillery,,,7,,,,,,,5,,,,5,,,7,,,
8,Ballistics,,,,,,,2,5,,,2,7,7,,2,7,8,,,2
1,Diplomacy,,,,,,,,,,,,,,,,,,,,8
2,Eagle Eye,,,,,,,7,,,,,,,,,,3,,,
3,Earth,4,,4,,,,4,4,4,,4,3,,,4,3,4,4,,3
2,Estates,,,,,,,,,,,,2,,,,2,,,,
2,Fire,7,5,,4,,4,,,,3,,5,,3,,5,,,4,
1,First Aid,,,,,,8,,,,,,,,,,,,,,
1,Intelligence,,,3,,,,,,,,,4,,,,4,,,,
5,Leadership,,,,5,5,,,,,4,,,,4,,,,,,5
4,Learning,,,,,,,,,,,,,,,,,,,7,
7,Logistics,8,3,,,,,,,7,,,,3,,,,,8,,
3,Luck,2,,,2,,,,7,,,,,8,,,,2,5,,
3,Mysticism,,,,,,,,,,,,,,,,,,,,7
2,Navigation,,,,,,,,,,,,,,,,,,,,
0,Necromancy,,,,,,,,,,,,,,,,,,,,
10,Offense,,,,,,,,,,,,,,,,,,,,
8,Pathfinding,3,2,5,8,3,3,,3,2,,5,,,,5,,,6,5,
6,Resistance,,7,,3,,,,,5,,,,,,,,,2,3,
1,Scholar,,,,,,,,,3,,,,,,,,,,2,
8,Scouting,5,,8,,7,2,5,2,8,,3,8,2,,3,8,,7,,4
1,Sorcery,,,,,,7,,,,,,,,,,,,,,
8,Tactics,,,,,8,5,8,,,7,8,,,7,8,,,,,
0,Water,,,,,,,,,,,,,,,,,,,,
2,Wisdom,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,3,6,6
Odds,Crag Hack,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60
3,Air,,4,,4,,4,,,4,,,4,,,,,8,,,
7,Archery,,,4,,,,,,,,,2,,3,8,,,,,
6,Armorer,7,,,,8,8,5,,8,2,7,,,,,,2,7,,5
8,Artillery,,,8,,7,,,,,7,,,,,,,5,,,7
8,Ballistics,,2,,,,3,,,,,,,,2,5,,,2,7,8
1,Diplomacy,,,,8,,,,,,,,,,,,,,,,
2,Eagle Eye,,,,,5,,,,,,,,,7,,,,,,3
3,Earth,,5,3,,4,,4,4,,4,,,,4,4,4,,4,3,4
2,Estates,,,,,,,,,,,,,,,,,,,2,
2,Fire,4,,,7,,,,7,5,,4,,4,,,,3,,5,
1,First Aid,,,,,,,,,,,,,8,,,,,,,
1,Intelligence,,,,,,7,,,,3,,,,,,,,,4,
5,Leadership,5,8,,,,,,,,,5,5,,,,,4,,,
4,Learning,,,,5,,,,,,,,,,,,,,,,
7,Logistics,,7,,3,,6,3,8,3,,,,,,,7,,,,
3,Luck,2,,,,,,,2,,,2,,,,7,,,,,2
3,Mysticism,,,,,,,,,,,,,,,,,,,,
2,Navigation,,,,,,,,,,,,,,,,,,,,
0,Necromancy,,,,,,,,,,,,,,,,,,,,
10,Offense,,,,,,,,,,,,,,,,,,,,
8,Pathfinding,,,7,,3,2,,3,2,5,8,3,3,,3,2,,5,,
6,Resistance,3,,5,,,,7,,7,,3,,,,,5,,,,
1,Scholar,,,,,,,2,,,,,,,,,3,,,,
8,Scouting,8,3,,,,,8,5,,8,,7,2,5,2,8,,3,8,
1,Sorcery,,,,,,,,,,,,,7,,,,,,,
8,Tactics,,,2,2,2,,,,,,,8,5,8,,,7,8,,
0,Water,,,,,,,,,,,,,,,,,,,,
2,Wisdom,6,6,6,6,6,5,6,6,6,6,6,6,6,6,6,6,6,6,6,6
Odds,Crag Hack,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80
3,Air,,8,,,4,,4,,4,,,4,,,4,,,,,8
7,Archery,,,,,,4,,,,,,,,,2,,3,8,,
6,Armorer,,,,7,,,,8,8,5,,8,2,7,,,,,,2
8,Artillery,,,,,,8,,7,,,,,7,,,,,,,5
8,Ballistics,,,2,,2,,,,3,,,,,,,,2,5,,
1,Diplomacy,,,8,,,,8,,,,,,,,,,,,,
2,Eagle Eye,,,,,,,,5,,,,,,,,,7,,,
3,Earth,4,,3,,5,3,,4,,4,4,,4,,,,4,4,4,
2,Estates,,,,,,,,,,,,,,,,,,,,
2,Fire,,4,,4,,,7,,,,7,5,,4,,4,,,,3
1,First Aid,,,,,,,,,,,,,,,,8,,,,
1,Intelligence,,,,,,,,,7,,,,3,,,,,,,
5,Leadership,,,,5,8,,,,,,,,,5,5,,,,,4
4,Learning,,7,5,,,,5,,,,,,,,,,,,,
7,Logistics,8,,,,7,,3,,6,3,8,3,,,,,,,7,
3,Luck,5,,,2,,,,,,,2,,,2,,,,7,,
3,Mysticism,,,7,,,,,,,,,,,,,,,,,
2,Navigation,,,,,,,,,,,,,,,,,,,,
0,Necromancy,,,,,,,,,,,,,,,,,,,,
10,Offense,,,,,,,,,,,,,,,,,,,,
8,Pathfinding,6,5,,,,7,,3,2,,3,2,5,8,3,3,,3,2,
6,Resistance,2,3,,3,,5,,,,7,,7,,3,,,,,5,
1,Scholar,,2,,,,,,,,2,,,,,,,,,3,
8,Scouting,7,,4,8,3,,,,,8,5,,8,,7,2,5,2,8,
1,Sorcery,,,,,,,,,,,,,,,,7,,,,
8,Tactics,,,,,,2,2,2,,,,,,,8,5,8,,,7
0,Water,,,,,,,,,,,,,,,,,,,,
2,Wisdom,3,6,6,6,6,6,6,6,5,6,6,6,6,6,6,6,6,6,6,6
Odds,Crag Hack,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100
3,Air,,,,,8,,,4,,4,,4,,,7,,,4,,
7,Archery,,,,,,,,,4,,,,,,,,,2,,3
6,Armorer,7,,5,,,,7,,,,8,8,5,,8,2,7,,,
8,Artillery,,,7,,,,,,8,,7,,,,,7,,,,
8,Ballistics,2,7,8,,,2,,2,,,,3,,,,,,,,2
1,Diplomacy,,,,,,8,,,,8,,,,,,,,,,
2,Eagle Eye,,,3,,,,,,,,5,,,,,,,,,7
3,Earth,4,3,4,4,,3,,5,3,,4,,4,4,,4,,,,4
2,Estates,,2,,,,,,,,,,,,,,,,,,
2,Fire,,5,,,4,,4,,,7,,,,7,5,,4,,4,
1,First Aid,,,,,,,,,,,,,,,,,,,8,
1,Intelligence,,4,,,,,,,,,,7,,,,3,,,,
5,Leadership,,,,,,5,5,8,,,,,,,,,5,5,,
4,Learning,,,,,7,,,,,5,,,,,,,,,,
7,Logistics,,,,8,,,,7,,3,,6,3,8,3,,,,,
3,Luck,,,2,5,,,2,,,,,,,2,,,2,,,
3,Mysticism,,,,,,7,,,,,,,,,,,,,,
2,Navigation,,,,,,,,,,,,,,,,,,,,
0,Necromancy,,,,,,,,,,,,,,,,,,,,
10,Offense,,,,,,,,,,,,,,,,,,,,
8,Pathfinding,5,,,6,5,,,,7,,3,2,,3,2,5,8,3,3,
6,Resistance,,,,2,3,,3,,5,,,,7,,7,,3,,,
1,Scholar,,,,,2,,,,,,,,2,,,,,,,
8,Scouting,3,8,,7,,4,8,3,,,,,8,5,,8,,7,2,5
1,Sorcery,,,,,,,,,,,,,,,,,,,7,
8,Tactics,8,,,,,,,,2,2,2,,,,,,,8,5,8
0,Water,,,,,,,,,,,,,,,,,,,,
2,Wisdom,6,6,6,3,6,6,6,6,6,6,6,5,6,6,6,6,6,6,6,6
____________
|
|
dimis
Responsible
Supreme Hero
Digitally signed by FoG
|
|
posted August 03, 2006 01:17 PM |
|
Edited by dimis at 13:48, 03 Aug 2006.
|
review
Quote:
to 1.: The rules how the level advance is else to the skilling probabilities.
to 2.: There is no randomness in the upgrades after reaching level8. All skills are given, so senseless to write them down.
1: this is not necessarily correct, but I can agree on ignoring them while testing ANSA.
2: that is correct only if you are interested in the probs for simply gaining skills; which of course is the most interesting question with respect to actual gaming. Anyway, if you are interested in skills only, I believe it is better to ignore them, since this speeds up the process.
Quote:
Quote:
I think we should keep the data for tests where wisdom/magic are in the wrong place. I hate throwing away test results. We can figure out how to interpret the data later.
I agree. We could always leave out those results afterwards.
If you want ease of manipulation, use DB-Handler. I am afraid you 'll soon regret it.
Quote:
Another thing is, it seems that if you pick a new skills, like we do in this test, you will get wisdom as lvl 6 every single time. This wont happen every time if you sometimes pick an existing skill. Therefore it seems this test cant stand alone to determine all the patterns in the skilling tree. For example this test wouldnt have found the "some kind of magic at least every 4th turn" pattern.
Sorry, maretti, but this is wrong. For 1. you obviously didn't make many tests, or you were plain unlucky, cause you can be offered Wisdom earlier than level 6 under ANSA. As for 2., wrong again. If you check out the sequences you generate and track down magic skills only, you will see that on N out of N episodes you generated (100%) a magic skill is offered on level <= 4.
As for the pre-made sequences conjecture here is mine ():
There are many rules that define the offers on each level. This has as a consequence severe pruning on the truly "random-tree" (weighted-die, whatever ...) and as a result one can experience much less sequences than expected (meaning what you estimate in the absense of rules).
Quote:
Dimis questioned if we should test other things besides "always click right". I agree, but not right now. I think testing something small and easy is the best way to go. Then we have data that can be analized easy. If we do something complicated, we'll have a bunch of random data that will be extremely difficult to analize, maybe impossible. I think the best way to go is to take one small step at a time. Then decide after that what is the next step.
I just want to be honest, so I 'll leave a comment on this one as well. First of all, I am a great supporter on the idea of generating more episodes on ANSA policy. However, I think it would be better (at the moment) to go for the Always Upgrade Skill If Still Not Upgraded policy (perhaps for simplicity the Always Accept Left Offer policy) since most of the results you obtain with ANSA-tests are pretty much predictable (likelihood > 50%). However, as I have already stated, I am currently very busy and I sleep about 5 hours a day so that I can manage to have a good pace on my obligations. So, I don't know how much help I can offer on any thread relating skill-advancing no matter how interesting it is.
____________
The empty set
|
|
maretti
Responsible
Supreme Hero
|
|
posted August 03, 2006 01:47 PM |
|
|
Quote:
Quote:
Another thing is, it seems that if you pick a new skills, like we do in this test, you will get wisdom as lvl 6 every single time. This wont happen every time if you sometimes pick an existing skill. Therefore it seems this test cant stand alone to determine all the patterns in the skilling tree. For example this test wouldnt have found the "some kind of magic at least every 4th turn" pattern.
Sorry, maretti, but this is wrong. For 1. you obviously didn't make many tests, or you were plain unlucky, cause you can be offered Wisdom earlier than level 6 under ANSA. As for 2., wrong again. If you check out the sequences you generate and track down magic skills only, you will see that on N out of N episodes you generated (100%) a magic skill is offered on level <= 4.
1. You are right. I noticed that, when I saw Binabiks results. From about 40 tests I made I got wisdom every time as lvl 6, so I asumed that was a rule.
2. You are right that you can see that you are offered magic as lvl 4 the latest but what I ment was, you cant, from this test, see that you will be offered magic again at least once in the next 4 lvls as either adv or basic. If we didnt allready know this, we wouldnt find out with this test.
____________
Crag rules, Orrin and Ivor suck
|
|
dimis
Responsible
Supreme Hero
Digitally signed by FoG
|
posted August 03, 2006 01:52 PM |
|
Edited by dimis at 14:26, 03 Aug 2006.
|
modular
Quote:
2. You are right that you can see that you are offered magic as lvl 4 the latest but what I ment was, you cant, from this test, see that you will be offered magic again at least once in the next 4 lvls as either adv or basic. If we didnt allready know this, we wouldnt find out with this test.
That's why I am saying this skill-recording technique has leaks.  If we stored what was the left offer as well, we would have observed that on 100% of the cases up to level 8 our hero is offered magic at least every 4 levels. Anyway, all I am trying to say is that you don't have to get burnt out testing, while you can reach the same conclusions by reading carefully what is presented here and of course try to analyze as much as you can. For instance, I already know, that no matter how many tests you make, it is not likely (probability > 99% ) to gather more than (about) 400 different sequences.
The idea is to apply modular arithmetic on the episodes we generate. This can easily refute groups of skills (and give indications for approval as well - but not certainty as in the refutation) and checking them. I thought this was clear enough, but seems I have to post more on these to explain it better.
____________
The empty set
|
|
Ecoris
Promising
Supreme Hero
|
|
posted August 03, 2006 02:27 PM |
|
|
Nice work Binabik. However I believe that we get better results if the tests we make are independent (quit H3 and restart it).
Even though the ANSA policy may uncover the number of possible skill trees and sequences it may not detect other "skill-group exceptions" (like the magic school exception and the wisdom exception). Why? Because we might be offered an upgrade of a skill from a certain group instead of a new one. Suppose Offense was part of a group, something like (Offense, Armorer, Tactics) and that a skill from this group had to be offered at least every 5th level (just like the magic exception). ANSA would never notice since it would be very likely that the "counter" would be reset too often because Crag starts with advanced offense. Get my point?
If not here's another example: Suppose Crag started with advanced Wisdom instead of offense. Then we would never notice the wisdom exception.
The point is that "skill-group exceptions" may become difficult to find because all the skills at basic level that Crag acquires will affect the "right" (B) skill offer.
I think we should edit Crag such that he starts with basic Necromancy (because it is the only non-magic school skill which has odds 0). I guess you may claim that in a regular game Crag always starts with advanced offense, but the point is that we're studying general rules of how any hero not a specific one levels up.
Unfortunately we can't create e.g. a level 6 hero with 7 empty skill slots and 1 skill at basic/advanced level. But this would be the ideal way of testing. If this was possible we could easily detect if a skill from some skill-group had to be offered at level 7.
BTW: does anyone know what happens to the wisdom/magic school counter if you acquire one such skill through other means than a level up (Univeristy, Magic University, Witch's hut, event, pandora's box, etc.)? E.g you're level 3 and haven't been offered a magic school yet. You visit a witch's hut and you get fire magic. Will you be offered a new magic skill or adv fire when you reach level 4?
|
|
maretti
Responsible
Supreme Hero
|
|
posted August 03, 2006 02:36 PM |
|
|
As far as i know (this might be wrong!!!) the skilling is uneffected if u get a magic skill from a witch hut. So if you get air from a hut as lvl 1 you will be offered a magic skill as lvl 4 the latest, either adv air or basic something else.
____________
Crag rules, Orrin and Ivor suck
|
|
dimis
Responsible
Supreme Hero
Digitally signed by FoG
|
|
posted August 03, 2006 02:46 PM |
|
|
ANSA properties
Quote:
Nice work Binabik. However I believe that we get better results if the tests we make are independent (quit H3 and restart it).
Totally agreed since this reflects also actual game-play.
Quote:
Even though the ANSA policy may uncover the number of possible skill trees and sequences it may not detect other "skill-group exceptions" (like the magic school exception and the wisdom exception). Why? Because we might be offered an upgrade of a skill from a certain group instead of a new one. Suppose Offense was part of a group, something like (Offense, Armorer, Tactics) and that a skill from this group had to be offered at least every 5th level (just like the magic exception). ANSA would never notice since it would be very likely that the "counter" would be reset too often because Crag starts with advanced offense.
Just a clarification. Following ANSA and recording both offers still allows you to catch exceptions that occur every <= 8 levels. I agree though that it is more difficult to track them down since you need way too much testing - but it is not fair to say that it can't give you exceptions that occur every <= 8 levels.
Quote:
The point is that "skill-group exceptions" may become difficult to find because all the skills at basic level that Crag acquires will affect the "right" (B) skill offer.
Exactly. That's why we'll end up testing another hero as well for verification/model improvement.
The good thing with ANSA is that it allows you to finish fast a large sample size on skill-advancing. The problem is that sometimes it is harder to draw conclusions, especially if you are looking for group-exceptions. But on the other hand, it's the fastest route on refuting candidate groups of skills (because obviously a single episode is enough to prove that a candidate group / level-exception combo is not correct).
____________
The empty set
|
|
Ecoris
Promising
Supreme Hero
|
|
posted August 03, 2006 02:57 PM |
|
Edited by Ecoris at 14:58, 03 Aug 2006.
|
Quote:
Quote:
Even though the ANSA policy may uncover the number of possible skill trees and sequences it may not detect other "skill-group exceptions" (like the magic school exception and the wisdom exception). Why? Because we might be offered an upgrade of a skill from a certain group instead of a new one. Suppose Offense was part of a group, something like (Offense, Armorer, Tactics) and that a skill from this group had to be offered at least every 5th level (just like the magic exception). ANSA would never notice since it would be very likely that the "counter" would be reset too often because Crag starts with advanced offense.
Just a clarification. Following ANSA and recording both offers still allows you to catch exceptions that occur every <= 8 levels. I agree though that it is more difficult to track them down since you need way too much testing - but it is not fair to say that it can't give you exceptions that occur every <= 8 levels.
Yes, that is true if we record the left offer as well. But it is still likely that we won't notice other skill groups because skills from these groups could be offered the "natural" way before the exception applies.
But above all I think that our ANSA test should also follow these two rules:
1. Crag should be given basic Necromancy instead of offense.
2. Always exit H3 after a test. restart/main menu is banned.
Otherwise our results will become biased.
|
|
Xarfax111
Bad-mannered
Supreme Hero
The last hero standing
|
|
posted August 03, 2006 05:43 PM |
|
|
Just let me say 2 things:
This thread shouldnt go into another "ANSA Discussion". There is already a thread like this, please write there.
Otherwise it would keep to many peeps away that are "not that far into statistics".
To make this little project a succes i want to keep things simple.
So please concentrate on the testing. Please send your results in an Excel file to my email address: Xarfax1@t-online.de
I know some peeps did already a lot of tests. Lets just collect all
the results FIRST, before giving analyses here. The collected data
will be given to everyone if someone wants to work with it.
Open for discussion:
Should the game be completely restarted before a game or only go to the main menue. This should be agreed by us before further testing.
____________
|
|
dimis
Responsible
Supreme Hero
Digitally signed by FoG
|
|
posted August 03, 2006 09:33 PM |
|
|
problem
I wanted to start testing, but seems that the version of the map I have is corrupted since I get the following picture when I try to open the map with the map-editor:
I also tried to download again the map, but the link seems to be broken Olaf. Can you please send me a copy when you can at:
diochnos@gmail.com
Thank you,
- dimis -
____________
The empty set
|
|
angelito
Honorable
Undefeatable Hero
proud father of a princess
|
|
posted August 03, 2006 10:26 PM |
|
Edited by angelito at 22:26, 03 Aug 2006.
|
The file is not corrupt. It just shows it's made with a different Heroes version than yours..
____________
Better judged by 12 than carried by 6.
|
|
Binabik
Responsible
Legendary Hero
|
|
posted August 04, 2006 08:45 AM |
|
|
It would sure be nice to find a way to extract the data from the game files. I've been experimenting with a macro recorder and have it working now. I estimate I can generate 1500 saved test games by letting the macro run all night. I gave it a test run by putting it into a loop for 50 tests, which took about 15 minutes. Basically it does the same thing as a human tester by starting homm, leveling up hack, saving the game under a different name each test, quitting homm, then repeat loop xx number of times.
I have another macro that speeds up recording the data a little, but the actual data entry still must be done manually. The macro just automates opening the saved games, opening the hero screen, then moving the cursor to the next column in Excel.
The bottom line is, I can generate thousands of saved test games this way. But how to get all the data? If we want to make use of this I see two options.
Option 1. I can send saved games to people so they can help enter the data. I have dial up and the files are 16k each. I could probably email several hundred of the saved games, but I wouldn't want to send more than that with dial up.
Option 2. Other testers can download the recorder and generate their own saved games. The macro I created will not work on other machines because it plays back mouse clicks by pixel location on the screen. (it's similar to the old win3.1 recorder.exe program). I could send others the macro and they could modify it to work on their machine (it's a text file). The macro recorder comes with a utility that shows your mouse location on the screen. You could use that to enter the proper coordinates into my macro and it "should" work.
What do you guys think? Are either of these options worth messing with? The macro can easily be modified to click left or click right during skill selection....it's just a matter of entering the cursor location into the macro.
____________
|
|
Ecoris
Promising
Supreme Hero
|
|
posted August 04, 2006 11:32 AM |
|
|
Could you make the macro make a screenshot of Crag's 8 skills when the level up is done and save those screen shots instead?
If you could somehow cut out the eight skills only and save it as a bw/grayscale/jpeg image.
That would certainly speed up testing.
|
|
Binabik
Responsible
Legendary Hero
|
|
posted August 04, 2006 01:10 PM |
|
|
|
That's a good idea. I'll look into it, but it's too late tonight. I might be able to create a thumbnail sheet with several screen shots per page.
|
|
Xarfax111
Bad-mannered
Supreme Hero
The last hero standing
|
|
posted August 05, 2006 10:38 PM |
|
|
...mmh looks like we cant agree on simple things (testing procedure, how to write them down). Also there seems to be a huge lack of interest and time with some peeps.
So for me the project is already over.
Xarfax1
____________
|
|
angelito
Honorable
Undefeatable Hero
proud father of a princess
|
|
posted August 06, 2006 12:10 AM |
|
|
I knew that would coming
But i have to admit, i'm with Xarfax on that point.
HE opened this thread and invited everybody to join HIS idea of a hugh testing row. HE explained how he wants to have it done.
Why can't we (the participants) just follow his idea and see what results we get?We can do a different kind of testing technique afterwards aswell, if needed. But don't kick an idea from the very beginning by trying to change the whole process.
Let's give it a try and complain afterwards...too much theory ruins the practice.
____________
Better judged by 12 than carried by 6.
|
|
Binabik
Responsible
Legendary Hero
|
|
posted August 06, 2006 01:51 AM |
|
Edited by Binabik at 02:09, 06 Aug 2006.
|
Here's what happened.
I ran 100 tests exactly like Xarfax wanted. That took several hours time. Maretti also ran tests like Xarfax wanted. We both discovered going to the main menu gave repeating skill trees. Out of 100 tests, I only had 27 different trees. Xarfax asked if we should change test methods. Discussion about that was not off topic.
Some people suggested several other test methods. I agreed we need different methods, but we should start with the test Zarfax wanted. Then decide later what other tests are needed. Only one person can be in charge, and that's Xarfax.....at least in this thread.
Maybe you guys didn't understand what I was saying about using macros to do the testing. It does the tests exactly like Xarfax wanted. It just automates it and speeds it up.
I was getting ready to email Xarfax my excel file from my first 100 tests. Then I read the post from Ecoris in the other thread. His test results change everything. I put everything on hold.
I don't mind doing more tests. But first I think we need to verify Ecoris' results and decide what to do. That can be discussed in the other thread.
Xarfax, you KNOW people will get off topic and question your test methods. You just have to take charge and not abbandon the project. If people agree with your test, they can help. If they don't agree they can start another thread.
Edit: I just sent you my excel file from the first 100 tests. There are several sheets. The first one is all the original data. All the other sheets are the same data that has been sorted in different ways to look for patterns. I don't have a way to post graphics files, so you can post it if you want.
____________
|
|
dimis
Responsible
Supreme Hero
Digitally signed by FoG
|
posted August 11, 2006 12:32 PM |
|
|
Quote:
...mmh looks like we cant agree on simple things (testing procedure, how to write them down). Also there seems to be a huge lack of interest and time with some peeps.
So for me the project is already over.
Xarfax1
I am disappointed. I came back today from a mini-vacation (and leaving again tomorrow ) and dropped by to see the results of this thread. However, it seems that the testing process slightly fades out. Are other people away on holiday?
Anyway, can I upload tests that I will make on another map Xarfax?
As you can realize I can not help you right now because I have a problem with the version of the map (see my post above).
By the way,
If it is possible, can you send me the results you have up to now Xarfax? Thank you in advance.
- dimis -
____________
The empty set
|
| |
|
|
