Middle Earth: Battle For Arda - Official Thread
-
@cernel Everything is going to be fine once it is released in-game. Till then you have to deal with the way GitHub packages stuff.
Btw, having arrived home from the last vacation, I'll have more time for development once again. I hope it will be downloadable ingame before long.
-
@alkexr
I would like to help with balancing your game, are you at this stage?Some of your units need their PUs reviewing, but Im sure you know that.
Thanks to you and all who have brought it this far, it looks and feels awesome.
I did the coding and balancing for The Tyranids for the Original Dawn of War, so Im used to the repetitive cycle of testing.
-
@thedog said in Large Middle Earth - official thread:
I would like to help with balancing your game, are you at this stage?
A couple of unit images need a facelift, sea units need to be added, placements picked (all in progress), and game notes written. After that, it's balancing time! (i.e. playing a ton of PBF with everyone, I suppose... I've never done serious balancing before)
Some of your units need their PUs reviewing, but Im sure you know that.
They looked good when I set them. When I started games, they didn't
Turns out that the value of the unit is not only determined by it's effectiveness in large battles (yeah, I know, obvious... in hindsight).But I'm curious. Which units do you think are too cheap/too expensive?
Actually, the unit cost/stat balancing can start right away; everything relevant about them is done.
-
@thedog Looks like you are in the right place. Most TripleA maps have never been seriously balanced (NWO etc. are few major exceptions, and that doesn't even comprise a good half of the high quality ones, like WAW and TRS).
-
@alkexr
My caveat, Im no expert at Triplea balancing and have come up with my own formula, it maybe naive, but here it is. It is based on spending about 100 TUV a side in the Battle Calculator and includes movement.(Attack + Defence + Movement - Average_infantry_movement + Fudge_Factor) x Hit_Factor
Average infantry movement on your map is 2, so its a constant of -2.
Giving most infantry a movement cost of 0pu, a unit moving 3 costs 1pu, flying move of 5 gives 3pu.Where Fudge_Factor also includes personal tastes/goals.
On your map its focus is a land battle so Air & sea units are +1
Most units have a bonus of some sort, gives or receives support, so this is the norm, where a unit does not have a bonus then -1, eg. bat, balchoth_tribesman, trollman, raven all have -1.
Play testing will increase or decrease the Fudge Factor.
As your map has terrain this adds a whole level of complexity that is is difficult to account for, so Im ignoring it.
For units with 1 Hit, the above is fine.
For 2 Hits x by 1.5
For 3 Hits x by 2
Modified by how they are repaired. These hit values are probably lower than most other mods, one reason is the AI does not try to protect its 2+ hit units and the other for players they are rarely cost effective.The above 2+ hit modifiers would be higher say for ship units as there is fewer units in a battle and so hits are worth more.
To your question, "But I'm curious. Which units do you think are too cheap/too expensive?"
ent 7 5 2 with 2 hits (7+5+2-2=12) x 1.5 = 18 your value 24
bear 7 5 3 with 2 hits (7+5+3-2=13) x 1.5 = 20 your value 16
warg_rider 5 2 4 (5+2+4-2=9) x 1 = 9 your value 6There are more units but you have my base formula what do you think?
As an aside here is worked example of a bat unit to show how the points build up. Flying units are really hard to balance as only a humans can really exploit their capabilities. There is a no bonus because it has no bonuses against ground units.
bat 6 2 4, flying +1, no ground bonuses -1, (6+2+4-2+1-1=10)@Cernel
Thanks! -
@thedog One thing to remember is that units don't necessarily need to be balanced across players just mostly within a single player. For example, you can have some small players that have low production but better units and having their allies help them grow could be part of an overall strategy.
Also the formula really won't work on this map because of all the targeted attacks, support values, and terrain. This complex of a unit set really requires playing through some games and see what players are tending to build and not build. You might be able to compare some units that are similar but the variety is pretty large.
-
@redrum
I agree, the High Elves backup your points, they have no chaff and have specialised units so should be considered for a discount to allow for this.But you need to baseline your units and then move from there..
-
@thedog Yeah, having some sort of baseline is useful. I just tend to look to apply it player by player rather than all units at once. As at the end of the day, the key is units being balanced within the players unit set so they have interesting choices to make and build a variety of units. Balance across the map as a whole can be achieved in many different ways even if some players have stronger unit sets.
-
@thedog Yes, that is somthing along the lines of what I've done, and once you start playing, it will become obvious why this isn't what units are worth.
For example, a recurring tactical move is to attack a weakly defended territory with minimal forces to block the main force of the enemy (only makes sense because units move 2+). You can do this to slow down the advance of a superior army, or to screen your main force if it is heavily offense-oriented and bad at defense, or if it is on plains (favourable for the attacker), etc. In many cases, heavily armored units can achieve a 100% win ratio without numerical advantage. Or take the snaga skirmisher, which defends plains with a power of 1. Thanks to their shield, spearman can kill them without losses, 1v1. What all of this means is that
- armored / shielded units are very well suited to screening
- units with very low defense values are very exposed to armored / shielded units, and therefore not well suited to "weakly defend" territories, regardless of their cost-effectiveness (i.e. discourage the attacker from wasting forces on taking it / from trying to screen of your main force)
- and as a consequence, it doesn't matter how these units perform in massive battles, because maybe you never intended to send them into massive battles in the first place. Armored units could even be absolutely thrash in large battles, and have a terrible power-to-cost ratio, I would still buy them for tactical purposes other than winning large battles.
And when you have considered all of this, then comes the hard part. Because, for example, elven cavalry is pretty overpowered... but where on (Middle) Earth do you intend to send your cavalry to fight as the High Elves? Into the mountains? The caves? The High Elves have no strategic objective for which cavalry would be useful.
As for the Ents... they have siege. 5-6 Ents with some escort can tear down the otherwise impregnable fortifications of Moria. And suddenly, the defense power of the defenders falls from 200+ to 80. Of course this only works for a very limited number of tactical goals (the list of them being "taking Moria", and... yeah, that's about it); and that's basically why they only cost 24 and not 40.
BTW multiple-hit units essentially generate free hits when they participate in battle. The amount of free hits generated is independent of how strong the unit is, so it shouldn't be a multiplier. The only thing that affect the value of extra hits is how often you can bring that unit into combat, which is higher for units with higher movement, and higher for air units.
(I agree that warg riders and bears are underpriced, but for different reasons.)
TLDR: there are way too many tactical situations, and in a great number of them raw attack/defense is next to irrelevant. The value of the unit is measured by its tactical utility, which is not reducible to a formula due to the diversity of tactical objectives
Anyway... I tried to use similar heuristics, but it's just... not that simple. I think the costs will be fine-tuned during the balance playtesting period. Then you can try buying the units your formula calculates as most cost-effective, and prove me wrong by stomping me into the ground with those units.
-
@alkexr
What you have done with armour/shield/charge/trample /terrain is clever and will make a great game for players to play and to min-max if they wish.Im suggesting to put to one side all the clever bits above (respect and admiration intended), these variables for now give them a straight 0 PU as its the norm, this is just for now, as their interaction with each other is too complicated without play testing. I will predict that different armies will have different values for similar stats, depending on their army composition, starting terrain and army objectives etc.
But now I would like to try to point all the other variables that can be pointed.
Can the Battle Calculator be trusted to give an accurate outcome compared to in game?
If so, to what degree would you trust its outcome 80%-ish or more?The reason for asking is, I take the formula, Trolls (have no clever addons so have -1 PU) and 2 hits have a 15TUV x7=105 in the Battle Calculator fight their enemies with infantry and cavalry with 100TUV-ish on the plains, they perform in a balanced way.
But add in enemy Wizards, Ents or other special units the Trolls tend to get slaughtered, so Im reasonably happy with the multiplier for now.
So how do you calculate 2+ hit units?
alkexr said “Then you can try buying the units your formula calculates as most cost-effective, and prove me wrong by stomping me into the ground with those units.”
I don’t wish to prove a point, just to understand how best to baseline all Triplea units, but especially your units. Im a wargamer see here for my Middle Earth interest.
[http://www.forum.specialist-arms.com/index.php?board=6.0](link url) -
@thedog said in Large Middle Earth - official thread:
Can the Battle Calculator be trusted to give an accurate outcome compared to in game?
If so, to what degree would you trust its outcome 80%-ish or more?Depends. The percentage is likely accurate, and so is the average TUV swing. But the variance tends to be really high, especially compared to other TripleA maps. In some cases, when the opposing forces are closely matched (20% - 80%), and especially during sieges or if a duel is present, the outcome can vary from you having 20/25 units left to your opponent having 20/25 units left out of less than 20 attempts! (Probably an extreme case, but still.) I sometimes do go for small battles even below 80%, but for large or otherwise crucial battles, not a step below 95% (variance drops dramatically as you overwhelm your opponent).
The reason for asking is, I take the formula, Trolls (have no clever addons so have -1 PU) and 2 hits have a 15TUV x7=105 in the Battle Calculator fight their enemies with infantry and cavalry with 100TUV-ish on the plains, they perform in a balanced way.
Trolls counter infantry + cavalry, in some sense. They are not affected by armor, shield, charge, flank, etc.
(have no clever addons so have -1 PU)
You very seriously underestimate the significance of clever addons.
But add in enemy Wizards, Ents or other special units the Trolls tend to get slaughtered
And that's exactly why you don't send in 7 trolls into a battle without other units helping them out. In fact, trolls are cost-ineffective in battle, and so are all multiple-hit units to an extent. Their utility lies not in fighting well, but in generating free hits. (They still perform in a balanced way vs some infantry + cavalry compositions despite their cost-ineffectiveness, due to quasi-countering them as mentioned above.)
So how do you calculate 2+ hit units?
The heuristic I used is that being 2 hit adds roughly 8 PUs to the cost, meaning that (calculating with 4PU fodder) after 2 battles they start returning on investment, except that you missed out on the power of the potential fodder in those 2 battles. Also the opportunity cost is high (you're basically trading the short term for the long term).
-
@alkexr
Thanks for the replies and insights.Here is a list of 1 hit units with suggested PUs for you to review and consider.
halforc 7
uruk_pikeman 11
ranger 10
tower_guard 10
warg_scout 10
wainrider_chieftain11
noldorin_warrior 10
trollman 9
dwarven_halberdier 10
raven 9
helming_warrior 8
kings_company 13 -
@alkexr said in Large Middle Earth - official thread:
Anyway... I tried to use similar heuristics, but it's just... not that simple. I think the costs will be fine-tuned during the balance playtesting period. Then you can try buying the units your formula calculates as most cost-effective, and prove me wrong by stomping me into the ground with those units.
Genenally, I don't think that the trial by combat is reliable, since there are so many other more important factors in play, the one with the better understanding of units' values would hardly get the upper hand. Nor do I believe that being better at evaluating units theorically traduces into being a better player, but for a small part of it.
I very much doubt that units balance can be fine-tuned with playtesting. Still just talking in general; I know this one is much more complex than what I've experience with, but I think this is just going to make experience even less insightful.
-
@thedog said in Large Middle Earth - official thread:
wainrider_chieftain11
helming_warrior 8Dude! Try sending in 8 wainrider chieftains (88 PUs) against 22 helming warriors (2x 88PUs)! And then you are not even using the 8 leadership of the wainrider chieftains!
Wainrider chieftains have
- 6 attack
- 3 defense
- 10 charge
- 2 armor
- 8 leadership
This amounts to a total potential combat effect of 29, of which only 9 is the raw attack / defense! The rest you're ignoring!
But ok, then, here is roughly my line of reasoning. I don't remember the exact formula, but it was something very similar.
First, we calculate the normalized combat effect. This is the sum of
- 50% of the attack power (halved because attack-only)
- 50% of the defense power (halved because defense-only)
- 50% of the charge (halved because attack only; it only fires in the first battle round and it can't always be used, but targets are common enough that it is counterbalanced by being a first strike)
- 70% of the flank (not 100% because targets are not common)
- 80% of the leadership (not 100% because you only get the total effect in large armies)
- 90% of the terror (similar to leadership, but malus to enemy is better than equivalent bonus to you)
- 120% of the armor (again, malus is better than bonus)
- 80% of the shield (ranged units are much rarer than melee)
- 10% of formation (soo very situational)
- 20% of antiair (targets are rare, but they are an effective deterrent)
- X% of the abilities I missed
Then we add
- 40% of the normalized combat effect we calculated above
- +4 per siege roll
- +8 for 2 hit, +16 for 3 hit
- +3 just for existing and having a hitpoint
- +1 for having a favourable terrain preference category
We further modify this by
- +15% if unit has 3 movement
- +25% if unit has 4 movement
- +10% if unit is mountaineer
Aaand then you get a value that seems very reasonable with the battle calculator but is quite off when you start to actually play.
So here is how the balancing is going to go. If there are units that people are abusing the hell out of (and winning), then those will be nerfed. If there are units which no one is buying, then those will be given some love. We simply have no hope of actually finding a balance - but we don't have to. So long as players, playing to the best of their knowledge, see interesting strategic choices, we can call it balanced.
-
@cernel
I have been testing using the FastAI for all players, is it the most appropriate AI for repetitive testing?My first objective for many reiterations has been no major capital to fall before turn 10, again is this a suitable objective?
-
@thedog Not really sure that testing with the AI would produce any usable results. Given that many of the things going on within this map would be incomprehensible to the AI.
-
@thedog The Fast AI will perform very poorly on this map because of all the complex units (targeted attacks and support). You really need to use Hard AI to probably get anything within reason even. But player vs player testing will be key either way. And some of the small nations can probably lose their capital before turn 10.
-
@thedog said in Large Middle Earth - official thread:
I have been testing using the FastAI for all players, is it the most appropriate AI for repetitive testing?
My first objective for many reiterations has been no major capital to fall before turn 10, again is this a suitable objective?I have already ran many hands-off AI games, and they slowly but steadily revealed how irrelevant they were with respect to balance.
-
@alkexr I believe the AI should handle most of the features in this map. Balancing large & complex maps to any reasonable degree will always need player vs player games though. If you notice particularly poor AI behaviors, let me know.
-
@redrum said in Large Middle Earth - official thread:
I believe the AI should handle most of the features in this map.
There are ways in which a player can abuse the AI's lack of understanding, e.g. countering its units without the AI realizing, or that the AI only uses siege accidentally, so building walls on places you want to defend for longer periods is a no-brainer. But it's tactically competent and does pull off some impressive operations sometimes.
If you notice particularly poor AI behaviors, let me know.
Well I did notice for example (not map-specific) that Hard AI is more likely to take a territory if it's defended than if it's not. I understand how this behaviour can be effective in cases, but on this map it really hurts the mobility of the AI (with armies moving 2+, controlling territories of whatever little value can extend the range of threat/pressure of a main force, or simply limit the options of the enemy). Because on this map advantage in mobility is worth more than one cheap unit. (Of course, on maps with small stacks this might not be the case.)
This is especially weird in cases like Age of Tribes, where sometimes the AI decides to stay in their capital and not conquer anything ever.
