The Old God has left the world and the pretenders are awakening and coming out from hiding. You start the game by designing one of the pretender gods that will compete for true ascension to godhood. The type of god can range from a magically powerful arch mage to an ancient kraken or a mystic monolith that people pray to. Your pretender controls one of over sixty different nations and with the help of that nation he will spread his word and battle the other pretenders. Dominions 3 is a turn based strategy game. You can play single- or multiplayer (1 - 23 players) with simultaneous turns. There are more than 1500 different units, 600 spells and 300 magic items in the game. The game also features a medieval musical score by Erik Ask Uppmark and Anna Rynefors, both awarded the title of Musicians of the Realm by the Swedish Zornmärkeskommiten. Dominions 3 is a highly detailed game and a 300 page pdf manual is included in the download.
| Posts | ||
|---|---|---|
| Site Searching Strategies | Locked | |
| Thread Options | ||
| Dec 17 2012 Anchor | ||
|
There have been many discussions on the optimal site searching strategies on various Dominions 3 forums and communities for a long time and it is a question that often rears its head. This thread is intended as a guide to the discriminating site searcher who desires the most efficient return on investment on the mage turns and gem usage that he puts into finding magical sites. THE BASICS
THE RADAR SPELLS
SITE FREQUENCY & DISTRIBUTION Site frequency is set during game creation and can vary from 0 to 75. Site frequency can also be forced to a specific value regardless of game creation choices if the map file has the command #features [0-100]. The exact algorithm that determines how the sites are assigned is not known, but it has been theorized that it is a binomial equation, weighted by the site frequency. It is known (from tests) that if a map has #features 100, all provinces on the map have four sites unless modified by terrain. The following terrain modifiers are applied to site assignment:
Only the most beneficial modifier counts. If the roll to determine site is a success (i.e. a site is assigned), the exact site is determined through the following process:
There are also certain special sites that are assigned by special scripts in the game and which only appear rarely. These sites always have special guardians. They do not appear through the normal site determination process (as far as anyone knows at this time). Some examples are The Mine of the Cyclops and The Holy Crypt of Anre. It should be noted that random site determination is performed at game creation. Any sites assigned by map file commands are only applied AFTER the random distribution. If a province affected by site assignment map commands is already full, the map file script is ignored. If there are more sites assigned by the map file than there are free slots, the sites from the map file commands are assigned in order until all free slots are filled. If a map file absolutely needs to assign all four slots, use the #killfeatures map command on that province to clear all randomly assigned sites before using the site assignment map commands. It should also be noted that unique sites can appear more than once if they are scripted in the map file, because map file scripting does not follow the random site determination rules and is applied after random site generation. HOW MANY RANDOM SITES ARE THERE AND WHERE CAN THEY BE FOUND? So, where should an intrepid godling direct his or her minions to search for sources of power? The Dominions 3 DB (available from here) has a full list of sites, including level requirements, rarity and what terrain they appear in. Out of the 709 sites, 149 are national start sites or sites that only appear by event, special script or in special scenarios, so they can be disregarded. That leaves 560 sites that can appear randomly. All paths have at least one lv 4 site. Astral has 4, Holy, Fire and Death have 2, Nature has 3. All told there are 17 lv 4 sites out of 560 possible random sites and all lv 4 sites are rare. Meaning that if they get rolled on the first try, the site is rolled again and if you also get it on the second try, only then will it be placed in the province. In actual games I've found 10 of them all told and that's including all of my Dominions 2 games. There are 66 lv 0 sites (meaning automatically visible), 16 rare (3 unique), 22 uncommon (8 unique) and 19 common (0 unique) and 16 of them are unique. There are 252 lv 1 sites, 65 rare (9 unique), 119 uncommon (35 unique) and 68 common (1 unique). There are 158 lv 2 sites, 70 rare (29 unique), 86 uncommon (25 unique) and 2 common (1 unique) There are 68 lv 3 sites, 42 rare (30 unique), 18 uncommon (17 unique) and 7 common (no unqiues). The remaining one is a special site with special defenders and appears very rarely if it gets selected for in the determination of special provinces. Another note about unique sites is that they can only appear once on the map, no matter their frequency, unless they are scripted in the map file, because map file scripting is run only after random site determination has been done. MAGIC SITES BY PATH There are some 574 non-national sites, around 560 of which are truly randomly determined. Effective number means that special sites (frequency 5, appears from event or invalid locmask) are not included. FIRE, 55 sites (53 effective)
AIR, 57 sites (56 effective)
WATER, 89 sites (87 effective)
EARTH, 112 sites (103 effective)
ASTRAL, 58 sites (57 effective)
DEATH, 73 sites (72 effective)
NATURE, 86 sites (83 effective)
BLOOD, 39 sites (39 effective)
HOLY, 11 sites (10 effective)
And if that is not enough and you want breakdown by terrain type, level, rarity and more, the Dominions 3 Site Searching Matrix has all the answers you need. read the instructions page first and you should have no trouble navigating the hieroglyphs on the actual matrix page. However, as the matrix may be somewhat difficult to decipher for everyone, it is worth opening up the numbers a little bit and making things clearer with regard to where you should search what paths. MAXIMUM PATHS TO SEARCH DIFFERENT TERRAIN Of the handful of level 4 sites, only a single Holy site is non-unique (Temple of Time). The unique level 4 sites are:
If you have found one of these and there are no other level 4 sites of that path (or there are but they have specific terrain requirements), you can skip further level 4 searches for that path (at least in the terrains that do not apply). For level 3 sites, the math is more complicated. Most level 3 sites are unique. In fact, there are only a handful of level 3 sites that are both non-unique and common. Each one of them gives one gem of its own path and can appear in any land except caves, unless noted by exception. They are:
There is no common level 3 Blood or Holy site. Almost all other level 3 sites are unique and many of them are rare (and specific rare sites do not necessarily appear at all in any given game). So level 3 searches are not nearly as likely to get you anything as lower level searches, but if they do find something, it is almost always very useful. Just not done to excess, especially on large maps. For example, in he Air path, there are four unique level 3 sites, two of which are rare (Valley of the Roc, found only in mountains, and The Strom Underneath, found only in caves). The other two (The Rainbow Shroud, The Inverted Tower) are uncommon. So while these two sites are likely to appear, there can only be one of each. If you find them, the only places you should look for Air 3 sites are in mountains, wastes and caves, which are the only terrain types where the remaining possible air 3 sites can appear. There are similar restrictions for the other paths as well, but air is the most limited of them. Sea and deep sea provinces are something of a special case. Sites that can appear in normal seas can also appear in deep sea, but deep sea sites cannot appear in normal sea. That said, the maximum paths to search seas are:
QUICK REFERENCE OF SCALE AND UNREST EFFECTS BY SITE (originally collated from Dom3 DB by vfb) Data is presented like this: Order: That means you should search at least E3 in any province with extra Order scales for no apparent reason (not your or a neighbor's dominion). If you can't search E3 then at least search E2 if possible. But if "The Statues of the Overlords" is visible in the province, then it's responsible for the Order, so don't bother searching. Any special effects from the site are listed like (ench 20%): 20% off Enchantment rituals. (F1,H2) means you can recruit F1 mages or H2 priests at the site. Any other effects you should notice in the province are listed like [Productivity]. So if you see both abnormal Order and Productivity in a province, it's The Statues of the Overlords. You don't even need to own the province to notice these effects, so you can beline there for some additional benefits. Any terrain restrictions are listed like {!sea,!cave}, meaning no sea or cave provinces, or {sea,deep}, which means only sea or deep sea. Unique Sites have a '*' after their name. Capitol sites and sites which must be hardcoded on the map are not listed. Sites that only do bad stuff (disease, horror mark) and have no good stuff (no gems, gold, recruitables) have an 'X' after the name. Here's the list: Order: Turmoil: Productivity: Sloth: Heat: Cold: Growth: Death: Luck: Misfortune: Magic: Drain: Unrest 5: Reduce Unrest: Disease: Horror Mark: HolyFire (being struck by holy fire): HolyPower (being killed by a holy force): Curse: Edited by: Edirr |
||
| Dec 18 2012 Anchor | ||
|
I want to thumbs up this so badly. |
||
| Dec 21 2012 Anchor | ||
|
In CBM EA Ermor has what is basically a land version of Voice of Tiamat, Greater Augury. It is a Thaumaturgy 3 spell, requires 3F1S to cast, costs 10F gems, and reveals all elemental sites (level 9 search in AEFW) in a province. It is a nice way to kickstart a broadbased gem income for EA Ermor. (From memory, I think it is literally based on Voice of Tiamat, but made national to EA Ermor and only able to be cast on land.) Edited by: jimbojones |
||
| Jan 7 2013 Anchor | ||
|
The DB lists common, uncommon, and rare sites. Yet the algorithm you laid out only makes a distinction for rare sites, everything else appears to be treated as equivalent. How does the process take into account common vs. uncommon? |
||
| Jan 8 2013 Anchor | ||
Edi is away for the next two weeks, so you might have to wait a while for an answer (just a quick FYI) |
||
| Jan 8 2013 Anchor | ||
I asked Edi about this when the thread was first made but he didn't have an answer. |
||
| Jan 21 2013 Anchor | ||
The reason I don't have an answer is that the answer has never been posted anywhere. The terrain modifiers and site determination mechanics have been gleaned from a forum post by Kristoffer way back in the Dominions 2 days before Dominions 3 was not even in development and everything else has been found out by analysis of the Dom3 DB (where site info is based on source code) and testing. I have asked for clarification on that particular issue, now it's time to wait and see if I get a more detailed answer. If I do, I will update that section. One other note: When people look at the DB, remember that the first possible site that can appear truly randomly is Tar Pits. Everything before that (including copper mine, iron mine, silver mine and gold mine) is special stuff that can appear only by special scripting or events even if it has a valid locmask. This is due to some issues and coding that dates way back to Dominions - PPP. |
||
| May 6 2013 Anchor | ||
|
What about situations where there's more than one terrain type in a province, how is the chance to having sites calculated then? |
||
|
|
May 6 2013 Anchor | |
| May 7 2013 Anchor | ||
As explained, only the highest modifier applies for probability. Any site that can appear in any of the terrains can appear there. |
||
| May 30 2013 Anchor | ||
That's actually not true. I ran some tests with the following results, Acashic Recording all observed provinces. There's 17 tested provinces of each type that I tried, and counted how many total sites were in them. Early era so 45 base site probability. Capitals were not included in these counts. Mountain - 36 sites. Of course the sample sizes are small, but it seems clear that if a province has a farm plus anything but the "many sites" ticked, then it'll only yield farm probability sites or very close to that. Further testing can be done if someone's interested, but based on this I'd assume that there's a hierarchy between terrains of many sites > farm > the rest in some order, deciding which probability is used. |
||
| May 30 2013 Anchor | ||
|
It is possible that there is a hierarchy. However, the guide has been written based on what Kristoffer (mostly) and Johan have said regarding the issue and the list of modifiers plus the quote "only highest modifiers applies" is from them. So it was taken at face value. It would not be the first time that the actual mechanics of it were more complex than Illwinter remembered. In case there is a hierarchy, it is probably this (lowest first):
This is the ascending order of terrain types when they are determined in the map file (by terrain mask). As you can see, everything but manysites (1024) has a lower terrain mask than farm (256). Terrain mask 512 is nostart and higher terrain masks than 1024 are all special stuff except for cave and border mountain. In this case, "only highest modifier applies" would still be true, but it would be more accurately worded as "only the modifier for the highest terrain mask applies". Thank you for testing this. You can probably do a control test by running a test of mountain vs farm vs border mountain (same sites as mountains).
If 1 and 3 get the same number of sites, which is higher than case 2, then the same modifier probably applies to both mountain and border mountain (and it should). If 1 + 2 gets significantly lower number of sites than 2 + 3 (despite these being supposedly identical setups in terms of frequency), the hierarchy listed above is more or less confirmed. |
||
| Jun 7 2013 Anchor | ||
Maybe it is already mentioned in the guide, just I don't realize it. Here is a simplified sample. Now I cast radar spell of this not-yet-explored magic path on each of these provinces. The point is that when I decide which provinces to search (when access is gained to some new magic path) there is some 'natural' urge to put provinces where nothing was found as prioritized ones (I omit terrain type here for simplicity, and I assume that all sites are searched equally so far - for simplicity as well). The same feeling tells, that provinces where many sites are already found should unlikely contain some not discovered magic site. However this 'common sense' consideration is probably misguiding (I don't know). Edited by: chelubey |
||
| Jun 7 2013 Anchor | ||
|
There's a larger chance you'll find a site in province B, where you have searched all paths but one without finding a site. This assumption is an effect of probability theory, which is probably beyond the scope of the guide. But it's right of you to prefer sites where you haven't found sites yet, and the automatic casting of site searching spell also prefers provinces where sites haven't been found yet. It's a useful basic guideline along with "prioritize swamps/wastes, then mountains and forests", and "some paths' sites are more common in certain terrains". The site searching spells don't choose optimally for terrain types. The difficult part occurs when you have searched for many sites and still have empty slots. E.g., it's better to further search a province where you have searched for 7/8 possible sites yet still have at least 1 slot (25%) remaining as compared to searching an empty province. When you fall behind the bell curve for the amount of sites you would have expected then finding sites when searching for the other paths gets more likely. One must realize though that the end goal is to search every province fully anyway. What theory teaches us is that it is better to search many provinces for different paths and then rotate your searchers to get full coverage in all provinces than it is to focus your searching on single provinces to search for all paths in one turn. It is possible to make passable calculations about the expected payoff of searching, but unfortunately there isn't enough information about how the system works to give definite numbers, such as how we lack knowledge about how rare and uncommon sites differ from each other. Edited by: Bоо |
||
|
|
Jun 7 2013 Anchor | |
| Jun 7 2013 Anchor | ||
That's quite an extraordinary claim as as long as we don't know how the game allocates sites in the first place. You would have to prove that the allocation of sites to each slot in a province is not a statistically independent event.
By and in itself, this sounds to me like 'man, I've rolled a die 25 times without getting a three, that's way behind the bell curve, there must be a three coming up soon', which is the sort of thinking that gets you into playing roulette. I assume you mean to say something different though - care to explain? Edit: Wait. Let's say you've searched for all paths except Air. When you search for Air now, you get four shots at finding a site, and finding one has become more likely because all possibilities except Air and Blank have been ruled out. Is this what you mean? Edited by: tenuki |
||
| Jun 7 2013 Anchor | ||
The Socratic method apparently.
All you need to know is that you can only have up to 4 sites in a province; the logic behind sites being statistically dependent follows from that.
That's the basic principle. The bell curve idea shows that you always are more likely to find sites on your next search if you don't find one on the prior search. The opposite side of the coin is that you become less likely to find sites if you get one. The middle point would be, purely as an hypothetical example, if you had found exactly 1 site after somehow having searched for exactly 1/4 of the possible site outcomes, as then your expected payoff for additional searching would be the same as in an identical but empty province. Edited by: Bоо |
||
| Jun 10 2013 Anchor | ||
|
Do you mind if I port a lot of this information to the wiki edirr? |
||
| Jun 10 2013 Anchor | ||
|
As long as I get credit, no, so go right ahead. I've already credited vfb for his contributions in the post above, so everything should be in order in that regard. |
||
| Jul 3 2013 Anchor | ||
|
Is there anyway to determine if a province is the manysites type? |
||
| Jul 3 2013 Anchor | ||
|
No, unless the map file graphic has something to indicate this. The Cradle of Dominion map had special glowing white ball thingys in provinces that were manysites, some other maps may have used other indicators and most have none at all. |
||
| Jul 3 2013 Anchor | ||
|
Or opening the map with the editor. But in a PBEM game you might not always have the exact .map file used. |
||
Only registered members can share their thoughts. So come on! Join the community today (totally free - or sign in with your social account on the right) and join in the conversation.