Yaksha | 2011-04-13 00:59:23 |
I know the existing information about draw rules comes from someone who scripted lots of hotseat games, recorded what cards were drawn, and then calculated draw rules. That information is now a little outdated, since the game just got updated. It seems the guy who did this isn't around anymore.
I'm a programmer and pretty interested to try and do this again, and see what the current draw stats are like.
Such a script requires: 1. Automatically starting hotseat game, then a cycle of "skip turn --> surrender--> rematch" 2. For each hotseat game, record what cards both players were given 3. Analyze the results statistically
I think I know how to do #1. #3 is trivial. But #2 stumps me. I don't suppose there are any other techies here who has any ideas on how to automatically 'extract' the list of cards a player has been given, based on what's displayed on the screen? The most practical solution I could come up with is take screenshot + analyze picture, which isn't really that practical at all.
HeadphonesGirl | 2011-04-13 01:32:59 |
That does seem like it would be the hard part. Can you have the script "check" each card (the equivalent of hovering the mouse overtop of it) and program it to record the card off of that?
You might want to ask at a programming forum. The goal is simple and I'm sure they wouldn't need to be intimately familiar with the game to help.
And thank you for looking into doing this, it would be really neat to get this updated. Kaylee | 2011-04-13 04:03:05 |
I love how engineers say something is trivial when what they really mean is "I know how to do that". Part #1 is the part I don't know how to do. I think I know how to do the rest, though it certainly wouldn't be trivial. My idea for part #2: Inside your spectromancer folder (for me it's C:\Program Files\Spectromancer) there is a file called Game.txt. Information is written to this file as you play a game. I'm not sure all of what goes in there, but I know if you start a new hotseat game then towards the bottom of the file you should find a section that says something like the following: 03:45:13.089 Local cards are: 03:45:13.089 ^Goblin Berserker^ 03:45:13.089 ^Wall of Fire^ 03:45:13.089 ^Minotaur Commander^ 03:45:13.089 ^Fire Elemental^ 03:45:13.089 ^Meditation^ 03:45:13.089 ^Ice Golem^ 03:45:13.089 ^Ice Guard^ 03:45:13.089 ^Mind Master^ 03:45:13.089 ^Faerie Apprentice^ 03:45:13.089 ^Faerie Sage^ 03:45:13.089 ^Chain Lightning^ 03:45:13.089 ^Air Elemental^ 03:45:13.089 ^Forest Sprite^ 03:45:13.089 ^Elf Hermit^ 03:45:13.089 ^Giant Spider^ 03:45:13.089 ^Earth Elemental^ 03:45:13.089 ^Healing Spray^ 03:45:13.089 ^Sacrifice^ 03:45:13.089 ^Ritual of Glory^ 03:45:13.089 ^Disintegrate^ 03:45:13.089 Local cards are: 03:45:13.089 ^Fire Drake^ 03:45:13.089 ^Orc Chieftain^ 03:45:13.089 ^Flame Wave^ 03:45:13.089 ^Armageddon^ 03:45:13.089 ^Sea Sprite^ 03:45:13.089 ^Merfolk Elder^ 03:45:13.089 ^Acidic Rain^ 03:45:13.089 ^Astral Guard^ 03:45:13.089 ^Griffin^ 03:45:13.089 ^Lightning bolt^ 03:45:13.089 ^Lightning Cloud^ 03:45:13.089 ^Titan^ 03:45:13.089 ^Nature Ritual^ 03:45:13.089 ^Rejuvenation^ 03:45:13.089 ^Stone Rain^ 03:45:13.089 ^Master Healer^ 03:45:13.089 ^Insanian Berserker^ 03:45:13.089 ^Doom Bolt^ 03:45:13.089 ^Insanian Shaman^ 03:45:13.089 ^Insanian Catapult^ Note that information for both players' draws are captured here back to back. You should be able to write a program that can parse this text file and pull out the information you need. The file seems to hold information from several games at once. I don't know what its limitations are. I'm not sure that's the best way to get the job done, but it may give you a starting point. Or a place to begin exploring, at least. Modified by Kaylee on 2011-04-13 04:12:56 Yaksha0 | 2011-04-13 06:37:23 |
You might want to ask at a programming forum. The goal is simple and I'm sure they wouldn't need to be intimately familiar with the game to help. Yeah well I'm sure there's someone out there who would know how to read the info off the cards, but I was hoping for something more simple. Something like what Kaylee suggested. So hence, better off asking people who are familiar with spectromancer. I love how engineers say something is trivial when what they really mean is "I know how to do that".
Part #1 is the part I don't know how to do. I think I know how to do the rest, though it certainly wouldn't be trivial.
My idea for part #2:
Inside your spectromancer folder (for me it's C:\Program Files\Spectromancer) there is a file called Game.txt. Information is written to this file as you play a game. I'm not sure all of what goes in there, but I know if you start a new hotseat game then towards the bottom of the file you should find a section that says something like the following: Haha well trivial is relative I guess. You're probably right. Well reading the info out of that file should definitely be doable. I'll give this a try I guess. Part #1 - you can use software to simulate mouse clicks at coordinate (x, y) of the screen. It's easy to do for spectromancer because the things you need to click (e.g. the skip turn button) are always on the same spot on the screen. I just need to script it so clicks happen at the right place, with the right time interval. Then keep a computer on overnight (or longer) and let it happen. And of course, take a copy of game.txt after the match. I just have to be sure that a new hotseat game, the cards will reliably appear on the Game.txt file. Anyway thanks for telling me this. I'll give it a try. wiggin | 2011-04-13 08:16:25 |
Hey, that's great Yaksha. If you'd like, I can help with 3.
Yaksha0 | 2011-04-13 09:47:35 |
Sure, thanks for the offer.
one of the things I imagine I will have problem is, is that once I have all the data (e.g. a list of card draws for X games), what do I look for?
Things like banned combos are easy to spot. But more complex things like "Each player should have at least 2 healing cards from this group of healing cards" are easy to test (i.e. I can look at the data and tell you if the statement is true), but difficult to find. i.e. if I'd never known that there's a group of healing cards, it would be hard to spot that pattern in all the data. It is easy to test proposed 'draw rules', but hard (or impossible really) to truly 'derive' draw rules from the data.
Vorrivan | 2011-04-13 10:40:52 |
Sure, thanks for the offer.
one of the things I imagine I will have problem is, is that once I have all the data (e.g. a list of card draws for X games), what do I look for?
Things like banned combos are easy to spot. But more complex things like "Each player should have at least 2 healing cards from this group of healing cards" are easy to test (i.e. I can look at the data and tell you if the statement is true), but difficult to find. i.e. if I'd never known that there's a group of healing cards, it would be hard to spot that pattern in all the data. It is easy to test proposed 'draw rules', but hard (or impossible really) to truly 'derive' draw rules from the data.
Rule "Each player should have at least 2 healing cards from this group of healing cards" doesn't work now. Twice i have draw where only healing source was Earth 11. And twice this was Water 10. And in one draw i doesn't have any heal spell. P.S. Sorry for my English, this is'nt my native language
... Rule "Each player should have at least 2 healing cards from this group of healing cards" doesn't work now. Twice i have draw where only healing source was Earth 11. And twice this was Water 10. And in one draw i doesn't have any heal spell.
P.S. Sorry for my English, this is'nt my native language
No heal spells? What class were you using at the time, and did you have any card in that class that healed you? Yaksha6 | 2011-04-13 10:54:55 |
Also, do you realize that water6 is considered a healing card, even if it does not 'add' to your HP.
Vorrivan | 2011-04-13 11:45:46 |
Also, do you realize that water6 is considered a healing card, even if it does not 'add' to your HP. Hmm. You are right, in draw that i mean "no heal spell" i was have Water 6. I was played Time. But i'm not sure that i have it in other draws.
Modified by Vorrivan on 2011-04-13 11:47:47 HeadphonesGirl | 2011-04-13 11:47:52 |
I believe the possibility of only earth 4 has been known for a long time. People have been claiming to get one healing card draws forever, but nobody's ever produced a screenshot. I think it's just that sometimes it's really easy to overlook your second one.
Vorrivan | 2011-04-13 11:50:30 |
I believe the possibility of only earth 4 has been known for a long time. People have been claiming to get one healing card draws forever, but nobody's ever produced a screenshot. I think it's just that sometimes it's really easy to overlook your second one. Ok. I'm play at least 6-7 draws every day, so if I had that draw once more -- i'm make screenshot.
HeadphonesGirl | 2011-04-13 17:03:52 |
... Ok. I'm play at least 6-7 draws every day, so if I had that draw once more -- i'm make screenshot.
Please do, it would be good to get confirmation. darkweaver | 2011-04-13 19:49:46 |
Sure, thanks for the offer.
one of the things I imagine I will have problem is, is that once I have all the data (e.g. a list of card draws for X games), what do I look for?
Things like banned combos are easy to spot. But more complex things like "Each player should have at least 2 healing cards from this group of healing cards" are easy to test (i.e. I can look at the data and tell you if the statement is true), but difficult to find. i.e. if I'd never known that there's a group of healing cards, it would be hard to spot that pattern in all the data. It is easy to test proposed 'draw rules', but hard (or impossible really) to truly 'derive' draw rules from the data.
one way to find the 2 healing cards within the 5 is by "brute force". You cycle through every possible combination of 5cards among 48 (4 basic houses with 12 cards each), now for each combination you cycle through all the draws and stop if there is not at least one cards from the draw in the combination. If you cycle through the draws succesfully it means that the combination (or set) always provide at least one card (you can check later if it provides more cards etc...). You can repeat the step for sets of 2,3,4,6,7,... cards. The problem is the huge increasing amount of those possible sets.(for 5 cards in the basic houses (48 cards) its 1712304 possible combinations). Also the sample data must be large enough to provide somewhat valid results. You have luck that the new spectromancer has the draws in a log file, in my previous version i couldn't find the draws listed in the file. Also i did a quick check with a sample of 1000 hotseats games (i've downloaded the latest version of the game but i dont have the latest expansion activate) and there was always 2 to 4 healing cards, the 2 players never shared a card, and i never found a orc chieftain + forest sprite nor a meditation + stone rain.(healing cards being:Ice Guard,Water Elemental,Faerie Sage,Elven Healer,Nature Ritual,Rejuvenation,Master Healer) I really wish that the draw rules will one day be part of the rule index in the game.
Yaksha0 | 2011-04-14 01:29:04 |
one way to find the 2 healing cards within the 5 is by "brute force". You cycle through every possible combination of 5cards among 48 (4 basic houses with 12 cards each), now for each combination you cycle through all the draws and stop if there is not at least one cards from the draw in the combination. If you cycle through the draws succesfully it means that the combination (or set) always provide at least one card (you can check later if it provides more cards etc...). You can repeat the step for sets of 2,3,4,6,7,... cards. The problem is the huge increasing amount of those possible sets.(for 5 cards in the basic houses (48 cards) its 1712304 possible combinations). Also the sample data must be large enough to provide somewhat valid results.
As you already pointed out this quickly becomes unfeasible. There 6 healing cards just in the basic houses. I'm hoping to get a lot of results. But even for a thousand games, cycling through combinations for 5 possible cards is far too much. We haven't even talked about special cards yet. The additional problem is that brute force approach is pretty dumb. If the rule is still "at least 2 healing cards, except for <one exception>". That is still a significant trend from a player point of view, but hard to find that if you didn't know you were looking for it. You have luck that the new spectromancer has the draws in a log file, in my previous version i couldn't find the draws listed in the file.
Were you able to find game.log? If you are on windows vista, default installation of spectromancer into the Program Files folder. If that's the case, you need to run spectromancer with administrator rights. Otherwise it cannot write into the Program Files folder, and hence cannot write the log file.
Angryonion | 2011-04-14 12:40:13 |
5 Heals, 6 heals...? There is seven. W6, W10, A4, E1, E2, E4, E11. No wonder some of you think you aren't getting 2.
Yaksha0 | 2011-04-15 03:26:32 |
Okay well data collection is up and running.
@Kaylee it turns out that game.log stores the cards for ALL hotseat games for that session. i.e. the information is kept until you restart spectromancer. Which is absolutely excellent. Because it takes only 3 seconds to surrender / rematch (~1000 matches an hour). I left my script running overnight last night as a test and woke up to see a lovely 60MB game.log file with about 11k hotseat card lists ^___^
Reading the game.log files should be trivial (yes, I'm an engineer :P). The problem is, initially I was aiming for 100k games, and putting the data into excel.
Unfortunately, 100k games is going to generate more data than excel can handle, unless I can find a smarter way of processing all that game info. I'm asking around a few programming forums about this. But I guess worse cast scenario, I'll look at just 25k games, which is the amount that excel could hopefully handle.
You're awesome, Yaksha <3
wiggin | 2011-04-15 07:54:03 |
That sounds great Yaksha. If you can upload the file somewhere, I would like to look at it too. As many games as possible. Modified by wiggin on 2011-04-15 08:55:31 Vorrivan | 2011-04-15 10:31:50 |
Unfortunately, 100k games is going to generate more data than excel can handle, unless I can find a smarter way of processing all that game info. I'm asking around a few programming forums about this. But I guess worse cast scenario, I'll look at just 25k games, which is the amount that excel could hopefully handle. It's really no need in saving data in file -- because you already have it. I'm developer too, so if you upload file i'll write parser and workshell for it. Without restrictions that's have excel. Modified by Vorrivan on 2011-04-15 10:33:55 darkweaver | 2011-04-15 17:55:54 |
if you need some other sample i can provide some (but not for the new 3 classes). If you're using autoit i can provide my code (should be adapted to unlocked version), i process 1 game in about 1.3sec ^^
Also in the game.log you can find those lines : Starting duel creation duelseed=#somenumber Trying to choose a card Local cards are:
The thing is that those lines can be repeated for a same draw, maybe the game redo a draw in case of banned combo etc... What matters is the last duelseed number before the 'Local cards are:' , it will happen that some draws share the same last duelseed number and end up with the exact same hands. To do correct statistics you should keep only one of those draws. Note that i checked this on a big sample with only clerics as players, i dont know if a same seed for a duel cleric vs demon and cleric vs beast will give the same hand for the cleric.
Yaksha0 | 2011-04-16 01:37:16 |
It's really no need in saving data in file -- because you already have it. I'm developer too, so if you upload file i'll write parser and workshell for it. Without restrictions that's have excel.
I like excel, because i was hoping that if i could do one row = one card, a lot of the information we want (e.g. banned combos), I could get easily with pivot tables. Then again, given the state of my computer, trying to work with a spreadsheet that's 1mil rows long is probably going to freeze it :S The parse should be easy, all the lines with cards are start with "<timestamp stuff> ^" So can just grep for it. if you need some other sample i can provide some (but not for the new 3 classes). If you're using autoit i can provide my code (should be adapted to unlocked version), i process 1 game in about 1.3sec ^^
When you say not in the three new classes, do you mean you got the samples from a previous version? Or from the current version? The draw rules could have (well rather likely to have been adjusted) changed in the new expansion release. So samples from the old version aren't useful. I am using autoit. I'm doing "click on surrender button", 0.5sec pause, "click on yes button", 1 sec pause, "click on rematch button", 1.5 sec pause, "click on surrender button". I find if I make the pauses significantly shorter sometimes the clicking goes out of synch. Like the script would miss a 'yes click' or something. Although i guess that's not a huge deal. The thing is that those lines can be repeated for a same draw, maybe the game redo a draw in case of banned combo etc... What matters is the last duelseed number before the 'Local cards are:' , it will happen that some draws share the same last duelseed number and end up with the exact same hands. To do correct statistics you should keep only one of those draws.
Yeah I have wondered about the order that draw rules are implemented in. Like the rules about which card goes into which slot, that's obviously 'part' of the draw. But are banned combos 'filtered' out afterwards? What about draws with no healing cards, I guess they may be filtered out afterwards too. Regarding the seed number, I disagree that they should be removed. Let's ask a practical question - "If I draw dragon but no tornado, what is the % chance my opponent has tornado". Is it 5%? 10%? 50%? If we filter out same seed games, the question becomes: "if I have a hand with dragon and no tornado, what % of seeds will result in the opponent having tornado". What we want to know is "if I have a hand with dragon and no tornado, in what % of games does my opponent have tornado". I suspect the number of same seed games will actually be insignificantly low to affect any statistics. But hypothetically we discover that certain seeds appear very often (say, 100 times in the 100k games). Then players would encounter those seeds more often right? Therefore, the hands produced by those seeds would have a stronger influence on questions like "what's the chance I'll get a hand with only earth4 as healing next". Point being, I don't see any draw rules where we should be concerned about how often spectromancer's card generator generates. For all draw rules, we are only concerned with how often the end player actually receives them. Yaksha0 | 2011-04-16 01:38:13 |
That being said, I guess I'll check and see how often duplicate seeds come up. But I would predict that number of duplicates to be small in a statistically insigifnicant way.
Yaksha0 | 2011-04-16 02:58:49 |
wiggin, vorrivan, darkweaver, anyone else who's interested, I'm happy to email around the data for the games. Here's my email address: http://img600.imageshack.us/img600/3118/emailp.jpgSend me an email and I'll send you the data I have. darkweaver I wouldn't mind seeing your autoit code. I'd show mine, but i'm not at the right computer atm. darkweaver | 2011-04-16 09:42:42 |
When you say not in the three new classes, do you mean you got the samples from a previous version? Or from the current version? The draw rules could have (well rather likely to have been adjusted) changed in the new expansion release. So samples from the old version aren't useful.
I've the latest version (v1.21) but i dont have truth and beauty unlocked (i'm waiting for steam version), so i cant play with the three new classes. Also like i said in the previous version (1.11) the draws where not given in the Game.log file.
I suspect the number of same seed games will actually be insignificantly low to affect any statistics. But hypothetically we discover that certain seeds appear very often (say, 100 times in the 100k games). Then players would encounter those seeds more often right? Therefore, the hands produced by those seeds would have a stronger influence on questions like "what's the chance I'll get a hand with only earth4 as healing next". Point being, I don't see any draw rules where we should be concerned about how often spectromancer's card generator generates. For all draw rules, we are only concerned with how often the end player actually receives them.
You're right, we dont really need to take care of same seeds, especially with random vs random matches. However on my 24k game sample with cleric vs cleric there where only 18k really different seeds draws. Anyway i'll send you my code, but i need to clean it a bit first and explain the methodology so you can adapt it. Yaksha0 | 2011-04-19 13:25:01 |
Seeds: About the duelseed, it seems in different sessions of the game, the same seed can generate different games. So there's something else, not just the immediate duelseed, that determines a game. (Yes, i've taken different specialty classes into account. I consider two deals, with different specialty classes, but same mana cost of specialty cards to be 'identical'.)
That being said, for 100k games, there were about 430k 'extra' "duelseed=x" lines. This leads me to believe that draw rules are implemented in two stages;
1. First stage deals cards based on rules about which card can go into what slot. And the fact that two players can't have the same card. 2. The game then checks for banned combos, checks to make sure each player has at least one mana gen, etc. 81% of deals are scrapped because they are invalid. The game just keeps generating 'deals' until a valid one is reached.
Single card appearance rate:
Out of 100k games, I count the number of games that card X appears in Player 1's hand. I also count the number of games card X appears in Player 2's hand.
Classic element cards, in % F01: 33%, 34% F02: 33%, 34% F03: 39%, 30% F04: 31%, 34% F05: 31%, 32% F06: 34%, 32% F07: 32%, 33% F08: 31%, 32% F09: 30%, 36% F10: 45%, 25% F11: 29%, 41% F12: 33%, 38% W01: 38%, 36% W02: 37%, 37% W03: 36%, 38% W04: 33%, 28% W05: 30%, 36% W06: 32%, 30% W07: 30%, 32% W08: 31%, 31% W09: 33%, 33% W10: 33%, 34% W11: 33%, 33% W12: 34%, 33% A01: 39%, 37% A02: 37%, 39% A03: 38%, 37% A04: 36%, 35% A05: 35%, 34% A06: 34%, 36% A07: 13%, 16% A08: 36%, 32% A09: 32%, 34% A10: 34%, 33% A11: 33%, 33% A12: 33%, 33% E01: 34%, 33% E02: 33%, 34% E03: 33%, 32% E04: 32%, 32% E05: 32%, 34% E06: 33%, 30% E07: 30%, 33% E08: 31%, 31% E09: 30%, 33% E10: 39%, 33% E11: 33%, 40% E12: 39%, 35%
So for most cards, both players are equally likely to get it.
Fire: - P1 is more likely to get F03, F10. - P2 is more likely to get F09, F11 and slightly more likely to get F12.
Water: - P1 is more likely to get W05. - P2 is more likely to get W04.
Air: - P1 is more likely to get A08. - Both P1 and P2 are very unlikely to get Pheonix (A07). More on this later.
Earth: - P1 is slightly more likely to get E10 and E12. - P2 is more likely to get E11.
All the other differences are less than ~3%. This small difference must also include impacts of banned combos (i.e. impact of banned combos on the rate the cards are drawn overall is actually very small)
Anyway, this all makes sense. For mana gens, we have: -----P1---P2-- F03: 39%, 30% W05: 30%, 36% E05: 32%, 34% As you can see, the numbers add up nicely.
For sweeps, we have: -----P1---P2-- F06: 34%, 32% F09: 30%, 36% A08: 36%, 32% Numbers also add up nicely.
P2 getting W04 more often should be a direct result of P1 getting W05 more often.
The differences in the high fire and high earth cards - not sure what is causing it yet. I think it's to do with what cards can go into which slots. Fire/Earth behaves differently to Water/Air. I haven't looked at slot occurence numbers yet. However I would theorize that all of P1's cards are dealth first, then all of P2's cards. Tight restrictions on the last slot for Fire and Earth cause these differences.
That leaves us with one last anomaly - Pheonix. I have no idea what's going on with this guy. Other cards have a total appearance of about 63% to 76%. Which is more or less what you'd expect (roughly appears around 2/3 games). But Pheonix only appears in 29% of games.
Why does Pheonix appear so un-often? Do you think this is intentional - if so, why? Pheonix isn't exactly overpowered to a broken degree. If it's not intentional, then what could be causing this one card to appear so un-often. As far as we know Pheonix isn't in any banned combos (although I haven't tested for them yet)
wiggin | 2011-04-19 13:53:18 |
This is interesting. In the old version, the distribution of the mana generators was roughly 50% chance of drawing Priest of Fire, and for sweeps there was roughly 50% chance of drawing Flame Wave. Now that I think about it, I'm pretty sure I've seen more W5 and E5s than I used to.
I think they made Phoenix appear less often by design. Because they felt that too many games turned into "Phoenix game", especially on the lower levels. I think this was a good decision.
The Armageddon / Fire Elemental differences between player 1 and 2 are crazy! How did you shuffle the special houses?
Modified by wiggin on 2011-04-19 13:56:08 Wavelength | 2011-04-19 13:59:51 |
That's awesome analysis, thanks Yaksha! Next, I'd love to see two-card appearance rates; that would take us beyond anything we've collected so far.
Go ahead and re-check your "P1/2 is more likely to get X" section; you mixed a few of them up (such as W4 and W5).
Yaksha0 | 2011-04-19 14:31:45 |
"The Armageddon / Fire Elemental differences between player 1 and 2 are crazy! How did you shuffle the special houses?" hmm... Not sure what you mean by shuffle. I ran hotseat games on random vs. random. Recorded 100k games. In those games, I basically went through every game and asked "Does P1 have F01?" and "Does P2 have F01?". Did that for all 48 cards and tallied results. Are you suggesting that maybe, the imbalanced distribution with the high Fire cards (and maybe high Earth too) is due to some cards appearing more often with some houses? I think that is unlikely. Because although I have not counted, in 100k random vs. random games, I'd expect just as many "P1 of <some particular house>" hands as "P2 of <some particular house>". That's awesome analysis, thanks Yaksha! Next, I'd love to see two-card appearance rates; that would take us beyond anything we've collected so far.
Go ahead and re-check your "P1/2 is more likely to get X" section; you mixed a few of them up (such as W4 and W5). Yes. that's next, after I do the single card appearance rates for special house cards. Regarding mistakes... yes... staring at too many percentages-that-are-in-the-30s-and-40s numbers i think. I can't edit my posts due to not logging in when posting >.< Well when I'm done with this I'll make a new thread with all the interesting findings. Rather than long walls of numbers. Will have to log in to post that one. Yaksha0 | 2011-04-19 15:10:26 |
I remember re-rolling archmage games over and over again until I got a good phoenix hand... took *sooo* long. Now I know why. I like playing with pheonix :(
Single card appearance rate:
From a total of 100k games, there are 200k hands. Rougly 8.33% of hands for each specialty house, as expected.
Specialty element cards. For a P1 hand of <house>, what's the chance a particular special card is there. For a P2 hand of <house>, what's the chance a particular special card is there.
Death S01: 66%, 56% S02: 34%, 44%
Holy
Mechanical
Illusion S03: 53%, 56% S04: 47%, 44%
Control
Chaos
Demonic S07: 49%, 53% S08: 51%, 47%
Sorcery S03: 52%, 48% S04: 48%, 52%
Beast
Goblin S02: 52%, 45% S03: 48%, 55% S04: 47%, 50% S05: 53%, 50%
Forest
Time
--------------------
So I've removed all the rows that are 50%/50%. Because those rows are not very interesting. So what I mean is, for Time, all 8 cards are 50%/50% For Death, all cards except Death1 and Death2 are 50%/50%
--------------------
Let's look at Illusion:
S03: 53%, 56%
S04: 47%, 44%
If we looked at only Illusion vs. Illusion games, then the numbers would be different. This is because in an Illusion vs. Illusion game, Illusion3 and Illusion4 both must appear once per game.
What those numbers mean, is that on average if you are playing Illusion vs. a non illusion class, you are more likely to get Illusion3 than Illusion4.
This makes sense. Because illusion4 is involved in a banned combo, where as illusion3 is not. So when the game is generated a random hand, and then filtering out invalid ones. Hands with illusion4 are going to be filtered out more than hands with illusion3.
Why does P1 get illusion4 more often and P2 get illusion3 more often? I have no idea... maybe I will when I start looking at pairs of cards. For now I guess what I've presented is mostly trivia for those of us who spend too much time on spectromancer :P
Lastly, look at all the cards that are less likely to be drawn: Illusion4: known banned combo with E02 and F11 Death2: known banned with W01 and W04 Demon8: greater demon, known banned with F11 Sorcery4: There was some talk about it being banned with high starting-water hands (to prevent people rushing out W12). I'm not recording starting mana, so I can't prove or disprove this.
Goblin 2 and 4 - possible banned combos????
wiggin | 2011-04-19 15:11:07 |
Are you suggesting that maybe, the imbalanced distribution with the high Fire cards (and maybe high Earth too) is due to some cards appearing more often with some houses?
Yes, that was what I was suspecting. Maybe there is low chance that player 1 gets Armageddon due to many banned combos. There is high chance he gets F10 or F12. Then that fire card is not available for player 2, so there is higher chance he draws Armageddon. Modified by wiggin on 2011-04-19 15:23:14 Wavelength | 2011-04-19 15:21:23 |
Why does P1 get illusion4 more often and P2 get illusion3 more often? I have no idea... maybe I will when I start looking at pairs of cards. For now I guess what I've presented is mostly trivia for those of us who spend too much time on spectromancer :P
Probably has to do with the Banned Combo with Armageddon. Since Player 2 is must more likely to have Armageddon than Player 1 is, they are also much more likely to not be allowed to draw Wall of Reflections. Therefore, it seems reasonable that Player 1 gets the Wall more often. This brings about the much harder question of which cards are drawn first, but it seems to be a good start nonetheless.
wiggin | 2011-04-19 16:29:19 |
I've read the data too now. Thanks to Yaksha for the files.
In the 200.000 hands, not once was there a hand with less than 2 healing cards.
The distribution is like this: number of healing cards: 0 percentage: 0.0
number of healing cards: 1 percentage: 0.0
number of healing cards: 2 percentage: 0.687875
number of healing cards: 3 percentage: 0.26691
number of healing cards: 4 percentage: 0.045215
number of healing cards: 5 percentage: 0.0
number of healing cards: 6 percentage: 0.0 Modified by wiggin on 2011-04-19 16:53:15 wiggin | 2011-04-19 16:45:50 |
This is the distribution for immediate damage cards (F10,F11,A2,A3,A6,A11,E6):
number of damage cards: 0 percentage: 0.11%
number of damage cards: 1 percentage: 10.6%
number of damage cards: 2 percentage: 42.935%
number of damage cards: 3 percentage: 36.39%
number of damage cards: 4 percentage: 9.525%
number of damage cards: 5 percentage: 0.44%
number of damage cards: 6 percentage: 0.0%
The amount of draws with 0 direct damage cards is tantalizingly close to 0, but it doesn't seem to be a hard rule. ozkavo | 2011-04-19 17:00:58 |
wiggin y not include F4 and A8 into the immediate damage cards ?
wiggin | 2011-04-19 17:20:37 |
Some Stone Rain statistics:
If you are the first player, and don't have Stone rain, these are the chances that the opponent has:
Stone rain: 47%
Stone rain, where you have meditation: Yes, Troll & Earth elemental: No : 57%
Stone rain, where you have meditation: Yes, Troll & Earth elemental: Yes 65%
Stone rain, where you have meditation: No, Troll & Earth elemental: No: 29%
Stone rain, where you have meditation: No, Troll & Earth elemental: Yes: 34%
If you are the second player, and don't have Stone rain, these are the chances that the opponent has:
Stone rain: 45%
Stone rain, where you have meditation: Yes, Troll & Earth elemental: No : 56%
Stone rain, where you have meditation: Yes, Troll & Earth elemental: Yes 60%
Stone rain, where you have meditation: No, Troll & Earth elemental: No: 29%
Stone rain, where you have meditation: No, Troll & Earth elemental: Yes: 33%
Conclusion: Risk of stone rain: 30% + 27% if you have meditation + 5% if you have both Troll and Earth Elemental Modified by wiggin on 2011-04-19 19:22:17 wiggin | 2011-04-19 17:34:58 |
wiggin y not include F4 and A8 into the immediate damage cards ? Right, I forgot Air8!
I tried with just this list: Fire11, Air3, Air6, Air8, Earth6, that is just the damage spells, not the creatures.
number of damage cards: 0 percentage: 0.0%
number of damage cards: 1 percentage: 39.9585%
number of damage cards: 2 percentage: 46.2735%
number of damage cards: 3 percentage: 13.768%
number of damage cards: 4 percentage: 0.0%
number of damage cards: 5 percentage: 0.0%
So that is actually the rule: You always get at least 1 basic direct damage spell. Modified by wiggin on 2011-04-19 17:37:27 Wavelength | 2011-04-19 17:41:44 |
You always get at least one creature sweeping spell, and you always get at least one direct damage spell. Unfortunately, Chain Lightning can be the one you get for one of them, and in rare occasions it's both.
I'm not sure it's really relevant, but it would be interesting to see how often you get Chain Lightning with no other sweeps, Chain Lightning with no other direct damage, and Chain Lightning with no other sweeps OR direct damage.
wiggin | 2011-04-19 17:49:08 |
Well, if you get Chain Lightning you never get another sweep.
Chain only sweep: 34.1%
Chain only dmg: 8.195%
wiggin | 2011-04-19 18:13:30 |
The chances of drawing both adjacent cards, as a percentage of the time you draw at least one of them:
Modified by wiggin on 2011-04-19 18:33:49 wiggin | 2011-04-19 20:04:50 |
, so I can't prove or disprove this.
Goblin 2 and 4 - possible banned combos????
Not those two, but there are other banned combos in the Goblin house:
Goblin 0 + Earth 5 Goblin 3 + Water 4 wiggin | 2011-04-19 20:10:10 |
Time has one banned combo:
Time 7 + Earth 2
Forest has one banned combo:
Forest 2 + Fire 5 Wavelength | 2011-04-19 20:26:18 |
[Banned] Goblin 0 + Earth 5
This one seems kind of dumb; the cost for the two to four extra mana you're collecting is that you "waste" a turn and whatever the Hermit was in front of gets a free direct attack. Rushing an opponent who plays such a parlay is a nearly guaranteed win! [Banned]
Forest 2 + Fire 5
This one seems unnecessary given how crappy the Wolf is, in general, without Orc Chieftain (except in Forest vs Forest mirrors). Even if the current balance is considered good, this banned combo could be worked around by changing Wolf's functionality to "Forest Wolf's attack power is equal to the number of turns Magic Rabbit has been in play", which would mean that Minotaur Commander and Orc Chieftain's boosts would only apply once. I've always liked the strategy involved in using those boosters with Wolf, though, and it's certainly no more powerful than the sometimes-broken Goblin Hero + Orc Chieftain. wiggin | 2011-04-19 20:29:20 |
This one seems unnecessary given how crappy the Wolf is, in general, without Orc Chieftain (except in Forest vs Forest mirrors). Even if the current balance is considered good, this banned combo could be worked around by changing Wolf's functionality to "Forest Wolf's attack power is equal to the number of turns Magic Rabbit has been in play", which would mean that Minotaur Commander and Orc Chieftain's boosts would only apply once. I've always liked the strategy involved in using those boosters with Wolf, though, and it's certainly no more powerful than the sometimes-broken Goblin Hero + Orc Chieftain.
I think it is a necessary ban, given how the wolf works now. I do think that your change in functionality is a better solution though. darkweaver | 2011-04-19 22:37:40 |
i checked all the duo cards in the basic houses (from my sample + yaksha's) and found 10 possible banned combos (that means i've at least found a game with all the other possible duos): F6 F9 F6 A8 F9 A8 (=> one sweep rule) F3 W5 F3 E5 W5 E5 (=> one generator rule) F5 E3 W1 E9 F9 F11 E5 E6
Zannoland | 2011-04-20 00:55:28 |
e5 e6 would fall under 1% rule which apparently is actually a 0.1% rule
i think the draws are so infrequent that wiggin's table can only go so far, it's likely that some stats ended up higher due to plain old luck
Zannoland | 2011-04-20 01:00:02 |
... This one seems kind of dumb; the cost for the two to four extra mana you're collecting is that you "waste" a turn and whatever the Hermit was in front of gets a free direct attack. Rushing an opponent who plays such a parlay is a nearly guaranteed win!
... This one seems unnecessary given how crappy the Wolf is, in general, without Orc Chieftain (except in Forest vs Forest mirrors). Even if the current balance is considered good, this banned combo could be worked around by changing Wolf's functionality to "Forest Wolf's attack power is equal to the number of turns Magic Rabbit has been in play", which would mean that Minotaur Commander and Orc Chieftain's boosts would only apply once. I've always liked the strategy involved in using those boosters with Wolf, though, and it's certainly no more powerful than the sometimes-broken Goblin Hero + Orc Chieftain.
one opening that i noticed was pretty successful from some players in the beta was earth 5 goblin 0 goblin 4 if earth 5 is sweeped, resummon it's counterintuitive and luck based, but it kept working as for the wolf ban, well, you were doing really well with it in the beta, and perhaps you did too well :P remember that time illusion got nerfed because me and finalslayer kept making alts and using the class as a counterpick to bad cleric play
Modified by Zannoland on 2011-04-20 01:03:37 Yaksha0 | 2011-04-20 03:01:54 |
Yes, that was what I was suspecting. Maybe there is low chance that player 1 gets Armageddon due to many banned combos. There is high chance he gets F10 or F12. Then that fire card is not available for player 2, so there is higher chance he draws Armageddon.
I'll take a look at this later today and see if there's any difference in the single card draw rates for f10, f11 and f12 if i'm only looking at illusion and demon. @Wiggin - since you are already looking at groups of cards, could you also take a look for the following groups: 1. Global destruct cards: F11, acid rain, earthquake 2. Elementals. I don't think there's any rule here but i'm curious to see the chances for getting 0 elementals and chances for getting all 4. i checked all the duo cards in the basic houses (from my sample + yaksha's) and found 10 possible banned combos (that means i've at least found a game with all the other possible duos): ... F9 F11 ...
This is interesting. I don't see this being a banned combo. Could this be a quirk resulting from which fire cards are allowed in which slots? Any theories on this one? ozkavo | 2011-04-20 06:45:32 |
after reading wiggin stats. now it is no longer possible to get earth 4 as ur only healing same for fire4 or fire10 or air2 as ur only direct damage if so, then things really changed from prev versions. Modified by ozkavo on 2011-04-20 06:49:17 wiggin | 2011-04-20 06:49:36 |
F9 F11 E5 E6
This is surprising.
I just verified it on Yakshas data (200.000) draws. F9 F11: 0 W9 W11: 25459
E5 E6: 0 W5 W6: 242
So there is something going on, it's not just small sample. wiggin | 2011-04-20 06:58:51 |
@Wiggin - since you are already looking at groups of cards, could you also take a look for the following groups: 1. Global destruct cards: F11, acid rain, earthquake 2. Elementals. I don't think there's any rule here but i'm curious to see the chances for getting 0 elementals and chances for getting all 4.
number of elementals: 0 percentage: 19.69%
number of elementals: 1 percentage: 37.815%
number of elementals: 2 percentage: 29.885%
number of elementals: 3 percentage: 10.92%
number of elementals: 4 percentage: 1.69%
number of global mass destruction: 0 percentage: 26.75%
number of global mass destruction: 1 percentage: 51.205%
number of global mass destruction: 2 percentage: 19.05%
number of global mass destruction: 3 percentage: 2.995%
Modified by wiggin on 2011-04-20 07:02:10 wiggin | 2011-04-20 07:41:38 |
I'm getting some surprising results regarding the high card.
Inferno can never be the high fire card, and Stone Rain can never be the high earth card. But Merfolk Overlord and Lightning cloud are each the high card 11-12% of the time. Modified by wiggin on 2011-04-20 07:48:25 wiggin | 2011-04-20 07:53:04 |
Armageddon risk.
If you are the first player, the chance of opponent having:
Armageddon, where you have both Fire Elemental and Dragon: 100%
Armageddon, where you have only Dragon: 46%
Armageddon, where you have only Fire Elemental: 59%
If you are the second player, the chance of opponent having:
Armageddon, where you have both Fire Elemental and Dragon: 100%
Armageddon, where you have only Dragon: 56%
Armageddon, where you have only Fire Elemental: 33% Modified by wiggin on 2011-04-20 07:55:34 Yaksha0 | 2011-04-20 08:19:54 |
I'm getting some surprising results regarding the high card.
Inferno can never be the high fire card, and Stone Rain can never be the high earth card. But Merfolk Overlord and Lightning cloud are each the high card 11-12% of the time. I think this was already known. The old draw rules listed what cards could be in what slots. I was just going to run over the data and check that info. Yaksha0 | 2011-04-20 08:23:23 |
wiggin can you also check tornado risk for if you have dragon?
wiggin | 2011-04-20 08:45:35 |
wiggin can you also check tornado risk for if you have dragon? Generally the risk of the opponent having a card if you don't have it is ~50%. I just checked, and that is also the case here.
Modified by wiggin on 2011-04-20 08:48:52 Yaksha0 | 2011-04-20 09:27:34 |
Slot statistics for classic element cards.
Count the number of times each card appears in a particular slot. This is counted separately for the 100k P1 hands and the 100k P2 hands.
Under each slot, is a list of cards and how often they appear in that slot. If a card is not listed, then it means it never appears in that slot.
Note that most of the differences between P1 and P2 % values are a direct result of overall single card occurrence rates. So for example, F03 appears in P1's hand more often than in P2's hand. F03 therefore appears in Slot Fire-1 of P1's hand more often than Slot Fire-1 of P2's hand.
Fire Slot 1: M01: 32.6%, 33.9% M02: 27.3%, 28.4% M03: 25.4%, 20.5% M04: 14.7%, 17.1% Slot 2: M02: 05.4%, 05.8% M03: 13.1%, 10.0% M04: 16.8%, 16.5% M05: 26.8%, 29.0% M06: 20.8%, 20.3% M07: 14.2%, 16.2% M08: 01.9%, 00.8% M09: 00.9%, 01.4% Slot 3: M05: 03.8%, 02.9% M06: 13.6%, 11.7% M07: 17.7%, 16.3% M08: 28.8%, 31.3% M09: 28.7%, 34.3% M10: 07.4%, 03.4% M11: 00.0%, 00.1% Slot 4: M10: 37.6%, 21.6% M11: 29.4%, 40.7% M12: 33.0%, 37.7%
Water Slot 1: M01: 38.2%, 35.5% M02: 36.9%, 36.6% M03: 24.9%, 27.8% Slot 2: M02: 00.1%, 00.1% M03: 10.9%, 09.8% M04: 32.5%, 28.2% M05: 27.7%, 33.4% M06: 18.3%, 18.2% M07: 07.2%, 07.4% M08: 03.2%, 02.8% Slot 3: M04: 00.1%, 00.1% M05: 02.2%, 02.2% M06: 13.9%, 12.2% M07: 22.7%, 24.3% M08: 28.1%, 28.5% M09: 22.0%, 21.3% M10: 10.9%, 11.3% M11: 00.1%, 00.1% Slot 4: M09: 11.2%, 11.4% M10: 21.7%, 22.7% M11: 33.3%, 33.2% M12: 33.8%, 32.7%
Air Slot 1 M01: 39.3%, 36.9% M02: 37.0%, 39.1% M03: 23.7%, 24.0% Slot 2: M02: 00.1%, 00.1% M03: 14.3%, 13.2% M04: 36.0%, 35.1% M05: 30.6%, 30.7% M06: 13.8%, 15.2% M07: 02.1%, 02.9% M08: 03.1%, 02.8% Slot 3: M04: 00.1%, 00.1% M05: 04.2%, 03.6% M06: 20.6%, 20.9% M07: 11.0%, 13.0% M08: 33.0%, 29.5% M09: 20.5%, 22.3% M10: 10.4%, 10.5% M11: 00.1%, 00.1% Slot 4: M09: 11.5%, 11.8% M10: 23.2%, 22.0% M11: 32.4%, 33.2% M12: 32.8%, 33.0%
Earth Slot 1: M01: 33.7%, 33.1% M02: 33.1%, 33.5% M03: 19.5%, 21.2% M04: 13.7%, 12.2% Slot 2: M02: 00.1%, 00.1% M03: 13.4%, 11.3% M04: 18.6%, 20.0% M05: 31.5%, 33.8% M06: 23.7%, 22.2% M07: 10.2%, 10.8% M08: 02.5%, 01.7% M09: 00.1%, 00.2% Slot 3: M06: 09.8%, 07.6% M07: 19.9%, 22.4% M08: 28.9%, 29.7% M09: 30.1%, 32.6% M10: 11.2%, 07.7% M11: 00.1%, 00.1% Slot 4: M10: 28.2%, 25.8% M11: 33.2%, 39.5% M12: 38.6%, 34.7%
Which cards common appear in slots: 1-F: 1, 2, 3, 4 2-F: 3, 4, 5, 6, 7 3-F: 6, 7, 8, 9 4-F: 10, 11, 12
1-W: 1, 2, 3 2-W: 3, 4, 5, 6 3-W: 6, 7, 8, 9, 10 4-W: 9, 10, 11, 12
1-A: 1, 2, 3 2-A: 3, 4, 5, 6 3-A: 6, 7, 8, 9, 10 4-A: 9, 10, 11, 12
1-E: 1, 2, 3, 4 2-E: 3, 4, 5, 6, 7 3-E: 7, 8, 9, 10 4-E: 10, 11, 12
For Fire and Earth, it's possible to have 4 cost card as your lowest card. However you are guaranteed a 10/11/12 cost card. For Water and Air, you are guaranteed a 01/02/03 card, but it's possible to have a 9 cost card as your highest card.
Earth5/Earth6 never appearing together. I think it's possible but just very rare. Notice how Earth5 is an unusual card in that it can ONLY ever appear in slot 2. So it's possible they can appear together but just VERY rare.
F09/F11 not appearing together is still a bit of a mystery.
wiggin | 2011-04-20 09:30:44 |
Earth5/Earth6 never appearing together. I think it's possible but just very rare. Notice how Earth5 is an unusual card in that it can ONLY ever appear in slot 2. So it's possible they can appear together but just VERY rare.
How is that unusual? The other 5s can only appear in slot 2 too.
Modified by wiggin on 2011-04-20 09:38:41 wiggin | 2011-04-20 09:44:51 |
Here is an alternative list counting all distributions with a prob above 0.5%:
1-F: 1-4 2-F: 2-9 3-F: 5-10 4-F: 10-12
1-W: 1-3 2-W: 3-8 3-W: 5-10 4-W: 9-12
1-A: 1-3 2-A: 3-8 3-A: 5-10 4-A: 9-12
1-E: 1-4 2-E: 3-8 3-E: 6-10 4-E: 10-12
My guess: Earth 5 and 6 not appearing together is an accidental result of the 1% rule, Earth 5 being a mana generator, Earth 6 being a damage spell, and the slot distribution. Fire9 and Fire11 not appearing together is by design.
Modified by wiggin on 2011-04-20 09:48:35 Yaksha0 | 2011-04-20 09:46:29 |
How is that unusual? The other 5s can only appear in slot 2 too. Ugg yeah the summarized list was meant to be cards that appeared often. I did 10%. I guess that was a bit misleading. And yeah, I'm guessing you've noticed now the other 5s can appear in slot3, just very rarely. wiggin | 2011-04-20 09:53:08 |
And yeah, I'm guessing you've noticed now the other 5s can appear in slot3, just very rarely.
Alright. But this can't really be the reason that we don't see E5 + E6, when we see W5 + W6. Since if W5 appears in the third slot, W6 won't be in the 4th slot anyway. Yaksha0 | 2011-04-20 09:56:53 |
Yeah, I guess so.
It's probably as you say with E5 being a mana gen and E6 being in the direct damage group.
We can only spot patterns, can't *really* derive how the draw rules get implemented. E5-E6 not appearing is probably just a quirk of what order the draw rules are actually applied.
I forgot you can edit posts here, I notice you've been updating the second post of the pinned thread. I don't suppose you can add in the info about slots (your list). Right now people have to scroll down a few posts to see the old data.
Also this is kind of unclear: "The chance of drawing each card is about a third, and the chance of
opponent having a card if you don't, is about half. Exception: Phoenix
is only about 15%."
Was there anything else we wanted to check draw rules wise?
wiggin | 2011-04-20 09:58:21 |
Yes, I will put that too. What do you think is unclear about the quoted part?
Yaksha0 | 2011-04-20 10:03:25 |
Could read that Phoenix is an exception because the chance of drawing it is 15%, not a third.
Or it could read that Pheonix is an exception, because the chance of opponent having it is 15%, and not a half.
I would say
The chance of drawing each card is about 33%, except for Phoenix (only 15%). If you don't have a card, the chance of the opponent having it is about 50%.
wiggin | 2011-04-20 10:12:38 |
I just looked into the actual stats. Interestingly, the chance of opponent having Phoenix doesn't go up by much if it's given that you don't have it.
14% chance you have it ~17% chance opponent has it if you don't Modified by wiggin on 2011-04-20 10:14:34 darkweaver | 2011-04-20 13:33:37 |
Earth5/Earth6 never appearing together. I think it's possible but just very rare.
Very very rare, we are talking of 1/100000 also i've found the same banned combo for special houses as those listed by wiggin, no more no less. And for each special houses you can never have 1-2, 3-4, 5-6 or 7-8 Yaksha0 | 2011-04-20 13:47:14 |
Yes, it's known that with special houses, the cards are paired. You always get one of each pair.
So for forest, the pairs become 0-1, 2-3, etc. For mech the pairs are 0-2, 3-4, etc.
Wavelength | 2011-04-20 14:00:42 |
Very, VERY interesting statistics, guys! Amazing how everyone's doing such great work in just a couple days!
The Inferno + Armageddon ban is really interesting, because that used to be an extremely common combination to draw. While it was a slight screw in that you'd have to choose one or the other (and Inferno was usually the weaker option), it was never something I really hated drawing, so this one was a really interesting surprise.
Interestingly, this means that if the opponent has played Inferno in Slot 3, there's about a 94-95% chance that he can also play Dragon if his mana gets high enough (and a 100% chance if he played the Inferno in Slot 2). Use common sense, of course - if they are laying down cards that complement Fire Elemental (but not Dragon) well, it's worth assuming he ended up with that uncommon draw and playing around it. Which means, for example, saving for a Titan rather than laying down an Ice Golem.
Zannoland | 2011-04-20 15:22:50 |
that's an interesting idea ban a combination not because it's overpowered but because it's too weak ...
How is that unusual? The other 5s can only appear in slot 2 too.
water 1/3/5? it happens wiggin | 2011-04-20 15:25:12 |
water 1/3/5? it happens
Right, I misunderstood. That can't be the explanation anyway, I think. Zannoland | 2011-04-20 15:37:41 |
here's a thought i bet what they did is flip the order that mana points and cards are drawn the reason that the first player would have priest of fire half the time, was because you always have the mana to summon your accelerator on turn 1 (barring weakness) if the cards are distributed, then the mana points are adjusted to suit the mana requirements, that makes sense they did something similar in the last patch with the order house cards are drawn, to even out the draw % on cards like wall of reflection, which used to be very low one thing i've always wondered is whether or not any of the other cards have mana min/max requirements as well in particular one of the best things you could do is prevent astral from being drawn if starting water is 7, because that's a common auto-win hand with control and cleric Modified by Zannoland on 2011-04-20 15:40:50 Yaksha0 | 2011-04-20 16:21:18 |
About f11/f9, I was thinking the opposite. If that's an intentional ban, it's like this. You are saving up big mana for an arma. You are just waiting to get enough mana, or chip down the opponent's health enough, that you can arma and win. While you are doing that, you also have the option at any time to throw out a F9 and clear the opponent's monsters. Your own monsters do direct damage, and you arma to finish the game. I think it's preventing someone from being able to horde fire for arma, but also have the option of throwing out the most powerful sweep. they did something similar in the last patch with the order house cards are drawn, to even out the draw % on cards like wall of reflection, which used to be very low
Can you elaborate on this>? I don't quite understand what you mean. one thing i've always wondered is whether or not any of the other cards have mana min/max requirements as well
Yeah that'd be interesting to find out. I don't suppose anyone knows of an easy method to record starting mana from the hotseat games screen? Starting mana doesn't go into game.log file, unlike cards. in particular one of the best things you could do is prevent astral from being drawn if starting water is 7, because that's a common auto-win hand with control and cleric Hmm... I don't know about control and cleric, but I have heard ti said on forums that there's some condition between starting water, W12 and S4,. So as to prevent someone from rushing out an early W12. I think it was that, if starting water is high, you can't draw S4. I don't remember what thread though, who said it, etc. Made sense to me though. wiggin | 2011-04-20 20:33:28 |
About f11/f9, I was thinking the opposite. If that's an intentional ban, it's like this. You are saving up big mana for an arma. You are just waiting to get enough mana, or chip down the opponent's health enough, that you can arma and win.
I also think it's banned because it's too weak darkweaver | 2011-04-22 16:33:46 |
... one way to find the 2 healing cards within the 5 is by "brute force". You cycle through every possible combination of 5cards among 48 (4 basic houses with 12 cards each), now for each combination you cycle through all the draws and stop if there is not at least one cards from the draw in the combination. If you cycle through the draws succesfully it means that the combination (or set) always provide at least one card (you can check later if it provides more cards etc...). You can repeat the step for sets of 2,3,4,6,7,... cards. The problem is the huge increasing amount of those possible sets.(for 5 cards in the basic houses (48 cards) its 1712304 possible combinations). Also the sample data must be large enough to provide somewhat valid results.
So i did something like this for sets up to 9 cards, and obtained a list of what I'll call minimum available sets (if you can find another better name ). The list can be found here: http://www.editgrid.com/explore/user/hdoe/Spectromancer_SetsAs example take the set {F10 F11 F12}, it means that every hand will have at least one card from this set and it is called a minimum set because for every subset there is at least one hand that have no card from the subset. So from the example every hand will at least have F10 F11 or F12, and there exists at least one hand without F11 and F12, one hand without F10 and F11 and one hand without F10 and F12. By definition you will never see a set in the list that contains {F10 F11 F12} such as {F9 F10 F11 F12 }. In the list you can see: F|06 F|09 A|08 (= one local sweep) F|03 W|05 E|05 (= one mana generator) There are more complicate patterns to spot: W|06 W|10 A|04 E|01 E|02 E|04 W|06 W|10 A|04 E|01 E|02 E|11 W|06 W|10 A|04 E|01 E|04 E|11 W|06 W|10 A|04 E|02 E|04 E|11 W|06 W|10 E|01 E|02 E|04 E|11 W|06 A|04 E|01 E|02 E|04 E|11 W|10 A|04 E|01 E|02 E|04 E|11 In those 7 sets you can notice very little change from one set to another. In fact there is exactly only one changing card for every 2 sets. For each of those 7 sets you have at least one card, lets say its W6, you now have fullfilled 6 of the constraint. You still have to satisfy the set {W|10 A|04 E|01 E|02 E|04 E|11}. But this is valid for every card in those sets. In fact you have discovered the two healing cards rules. Having At least one card from those 7 sets is equivalent to have at least 2 cards from {W|06 W|10 A|04 E|01 E|02 E|04 E|11}. The reason why it is plitted this way is because i list only minimum available sets. Those 7 sets are all the 6 cards subsets from the healing pool, and having at least one card from each subset is equivalent to have at least 2 cards from the bigger set. For the rest i dont know how to analyze it. Its getting complicate to spot things with all the banned combo and such. Still i've provided the list for those who wants to check it. wiggin | 2011-04-22 20:15:01 |
This was quite clever.
If I understand it correctly, I can see that I didn't find the minimum direct damage set. I thought it was F11, A3, A6, A8, E6. But actually it is just F11, A6, A8, E6. Wavelength | 2011-04-22 20:30:20 |
the minimum direct damage set...is just F11, A6, A8, E6.
This is one of the reasons Dragon is a powerful card in nearly every single game he's drawn in :-)
sweetgab | 2011-04-22 22:58:33 |
Aha, now it makes sense to me. F11, A6, A8, E6 is the real deal "at least one direct damage" rule, while F6, A6, A8, E6 (also on the list) is a coincidence as follows:
Assume you have none of F6, A6, A8 and E6. Then you have F9 (since you don't have F6 and A8), which in turns imply you don't have F11 (since it was banned with F9), but then this hand would've gotten reshuffled because it has none of F11, A6, A8 and E6.
I can see that this sort of situation happens with a majority of the listed sets, making direct analysis quite difficult (and I wouldn't be surprised if we've already found all the "draw at least X of these" rules). But nice find, darkweaver & wiggin.
wiggin | 2011-04-23 02:12:06 |
Yes, that was exactly what I was thinking about that sgab.
Yaksha0 | 2011-04-23 02:49:57 |
For the rest i dont know how to analyze it. Its getting complicate to spot things with all the banned combo and such. Still i've provided the list for those who wants to check it.
Hmm... let's look at this one: "F|06 F|10 F|12 A|08" (d) This is a result of "F06 F09 A08" (a) and "F10 F11 F12" (b) being minimum available sets. However "F09 x F11" (c) is a banned combo. So group (a), (b), (c) logically implies group (d). In fact, any minimum available set with F09 in it is equivalent to a minimum available set with F09 replaced by "F10 F12". Since the presence of F09 dictates that F11 is not present, which dictates that F10 or F12 is. Therefore (b), (c), and "F|05 F|06 F|07 F|08 F|09" logically imply "F|05 F|06 F|07 F|08 F|10 F|12" You could filter for any set containing "F|10 "F|12", so long as there's a smaller set where those two are replaced by "F09", since sets like that give no new information. Theoretically, every set on that spreadsheet should be result of: - Opposition pairs (banned combos) - Slot rules (i.e. resulting in sets like "F10 F11 F12" - Basic groups (healing, mana gen, etc.) Sets like this: "W|01 W|02 W|04 W|05 W|06 W|07 W|09 W|10 W|11" Are obviously result of only slot rules. If we try to 'decompose' all the large sets, they should all break down nicely (I'm assuming all the significant 'groups' of cards have already been discovered/known). Would be interesting to try never the less. darkweaver can you email me a copy of those minimum set results, but: - printed without the "|" (so write "F01 instead of F|01") - printed with a single space between each card, one set per line. Yaksha | 2011-04-29 14:47:19 |
Okay... last thing I'm doing with this. 40k Spectromancer games with starting mana included: https://rapidshare.com/#!download|142l35|459773033|Spectromancer_40k_games_w ith_mana.rar Format should be pretty self-explanatory. Starting mana for the four classic elements are recorded. I didn't bother with starting special mana, because it is always 2 for both players. Notes about starting mana distribution:- You can start with 3, 4, 5 or 6 mana in each of your classic houses; and 2 in your special house. - For Player 1, starting with 3 mana in a classic house is rare, occuring about a third as often as starting with 4, 5 or 6 mana. - For Player 2, starting with 3 or 6 mana in a classic house is rare, occuring about half as often as starting with 4 or 5 mana. - Player 1 starts with a total of 19 classic mana and 2 special mana. - Player 2 starts with a total of 18 classic mana and 2 special mana. This means on Player 2's first turn, Player 2 has a total of 22 classic mana and 3 special mana. So the 'starting mana advantage' for Player 2 is actually 4 mana (not 5, as often believed). There are only two 'rules' (that I have found) between starting mana and cards dealt:1. You are always able to play your mana gen on turn 1, unless Player 1 does something to change Player 2's mana levels (e.g. Control 1) --> Following combos banned for P1: W05+3WStart, W05+4WStart, E05+3EStart, E05+4EStart --> Following combos banned for P2: W05+3WStart, E05+3EStart 2. You cannot have Sacrifice, W11 or W12 and start with high water mana --> Following combos are banned for *both* players: Sorc04+W11+5WStart, Sorc04+W12+5WStart, Sorc04+W11+65WStart, Sorc04+W12+6WStart, Note that there is no similar banned combo for Holy5 with W11 or W12. So the earliest turn to get out W11/W12: - Player 1, with Sorc4: Can get W11 out on Turn 5, and W12 out on Turn 6 - Player 2, with Sorc4: Can get W11 out on Turn 4, and W12 out on Turn 5 - Player 1, with Holy5 and 6 starting water: Can get W11 or W12 out on Turn 5 (by playing Holy5 on turn 4) - Player 2, with Holy5 and 6 starting water: Can get W11 or W12 out on Turn 4 (by playing Holy5 on turn 3) In my sample of 40k games, there were 3331 games where Player 2 was Holy. Out of them, in 105 games, Player 2 had W12, Holy5 and a starting mana of 6 water. So that means if you play Holy, you have roughly (50% chance of being P2) * (3.17% chance of those conditions) = 1.58% chance of a hand where you can do a W12 on turn 4, if you are playing Holy. Now with the sorc ban, the earliest you can do W12 is turn 5. If there was no sorc ban, the earliest you can do W12 would be turn 3. The chance of someone playing Sorc, being player 2, with Sorc5, W12 and starting mana of 6 water I think would be also be ~1.58% So i'm thinking... if W12 early is really so harmful, that we need a ban to prevent sorcery players from getting W12 out on the board until turn 5/6. There really should also be a ban to prevent holy players from getting W12 out on the board until turn 5. (You'd only need one 'banned combo' for this: Holy5+W12+6WStart. The combo only needs to be banned for P2) Then again, there's only a very small chance (~1.58%) a Holy player would get a hand capable of doing a T4 W12 anyway. And we get more complaints from people who play against (what they think are) un-winnable dragon hands, than people who play against super-early W12 hands. So umm... more experienced players: just how bad do you reckon a T3/T4 W12 out on the board really is? What about the ban on W11, Sorc4 and high starting water. Is getting a W11 out on the board early really as bad as getting a W12 out early? Angryonion | 2011-04-30 04:52:14 |
Just wondering, were their any instances of fire 9-10-11? Just wondering since I had a 10-11 deal yesterday so thats probably still under the 1% rule.
Yaksha | 2011-04-30 05:45:31 |
You mean hands that had all F9, F10 and F11?
No, F9 and F11 are a banned combo.
A hand that has F9, F10 and F12 is possible.
Angryonion | 2011-04-30 06:25:07 |
Yeah, reread the posts. Was thinking f9+f11 was a 1% rule. And I know about 9+10+12, I got it 2x on the first day of expansion.
Wavelength | 2011-10-31 03:18:22 |
Sorry to bump an old topic, but the link Yaksha posted (even after I fixed the spacing) didn't work. Does anyone have the raw data on their computer, that they'd be able to either post or send?
Thanks so much. :-)
wiggin | 2011-10-31 06:02:30 |
Maybe it's changed again though?
daifei4321 | 2011-11-04 15:18:58 |
... So i did something like this for sets up to 9 cards, and obtained a list of what I'll call minimum available sets (if you can find another better name ). The list can be found here: http://www.editgrid.com/explore/user/hdoe/Spectromancer_Sets As example take the set {F10 F11 F12}, it means that every hand will have at least one card from this set and it is called a minimum set because for every subset there is at least one hand that have no card from the subset. So from the example every hand will at least have F10 F11 or F12, and there exists at least one hand without F11 and F12, one hand without F10 and F11 and one hand without F10 and F12. By definition you will never see a set in the list that contains {F10 F11 F12} such as {F9 F10 F11 F12 }.
In the list you can see: F|06 F|09 A|08 (= one local sweep) F|03 W|05 E|05 (= one mana generator)
There are more complicate patterns to spot: W|06 W|10 A|04 E|01 E|02 E|04 W|06 W|10 A|04 E|01 E|02 E|11 W|06 W|10 A|04 E|01 E|04 E|11 W|06 W|10 A|04 E|02 E|04 E|11 W|06 W|10 E|01 E|02 E|04 E|11 W|06 A|04 E|01 E|02 E|04 E|11 W|10 A|04 E|01 E|02 E|04 E|11
In those 7 sets you can notice very little change from one set to another. In fact there is exactly only one changing card for every 2 sets. For each of those 7 sets you have at least one card, lets say its W6, you now have fullfilled 6 of the constraint. You still have to satisfy the set {W|10 A|04 E|01 E|02 E|04 E|11}. But this is valid for every card in those sets. In fact you have discovered the two healing cards rules. Having At least one card from those 7 sets is equivalent to have at least 2 cards from {W|06 W|10 A|04 E|01 E|02 E|04 E|11}. The reason why it is plitted this way is because i list only minimum available sets. Those 7 sets are all the 6 cards subsets from the healing pool, and having at least one card from each subset is equivalent to have at least 2 cards from the bigger set.
For the rest i dont know how to analyze it. Its getting complicate to spot things with all the banned combo and such. Still i've provided the list for those who wants to check it.
Thanks for the work. However, maybe your data is not enough, so some of your results are not validated by my data. Here is a link to my data, 83123 duels, just in case if anyone needs it http://daifei4321.com/spectromancer/83213ForumRst.zip in csv format, can be opened by excel. These results of 3,4,5,6,7 cards passed my data. In combination of 8 cards, There are 19 results that cannot pass my data. In all your 1442 of 8or9 cards combination, 1329 of them cannot pass my data. By the way, my data is from 1.22. I got 3,4,5 data by myself, and my result of these matched the above. |