dood, Healer is a pick that enable an army led by a hero to participate in many many many battles in a row as long as the army can manage to stretch the battles out.

While a healer less army will only particpate in 1-2 battles before losing units or be forced to retreat and sit for lots of turns.

There isn't quite a way to counter something like healer because it's a passive skill, not active. And it is put into action each time a turn ends on strategy map To blow a coronary over trying to counter healer is just silly. And there is already a way to counter healer.. just keep on attacking the army until it gets wiped out in a single turn.

And.. AI already sent like eight armies after my single sov and failed to kill even a single unit despite throwing 81 units after my 9 units in a single turn.

Off topic --v

Where the AI is weak at is in tactical battles, not strategy map. AI pretty good, its the tactical battles that really screw the AI over humans.

If I autoresolved, I would lose lots of units to moronic calculations that have units bypass the frontlines magically and attack the rear lines. Autoresolves love to disregard zoc. That's why you need to use tactical battles.

I'll say it again, AI is screwed when it comes to tactical battles, they just don't stand a chance. Only way the AI can defeat you in tactical battles is through brute force of superpowered units. If it haz 100 atk/def vs your 20atk/def then your gonna die, hard.

90% of my defeats in tactical battles is due to brute strength of enemy army overpowering my army, nothing else.

90% of my defeats in autoresolve battles is due to it being moronic, six air shrills with 10 hp each isn't simply capable of doing over 800 damage to my sovereign army but they continue to pull it off somehow in autoresolves thus shrills must be only killed in tactical mode where they only do like 80 damage to you and die.

Only use autoresolve if you have brute strength advantage.

Ai only plan for autoresolve battles, never plan for tactical battles, armies that look great in autoresolves, is outright horrid in tacticals.

Urgh sleep calls.

Little Willie may have some serious personality issues, but he is at least in the ballpark on one thing as much as I hate to say it.

There are a couple different game plans that stand out as clearly better than the rest.  The game is just designed to reward players that play according to those plans.

For instance, having a lot of cities is really good in this game.  The AI picks on people that are the lowest on score first and the buildings inside cities add greatly to your score.  Lots of cities times lots of buildings will keep the AI off of you even if your army sucks.  Army just isn't given enough score weight in relation to the weight that buildings have.

The takeaway from that is to try to expand if at all possible.  Plenty of things reinforce that.  More cities generally means more research, more money, more mana, and so on.  Having more of all those things helps you win more.

Additionally, if you have more cities then by definition the enemy has fewer, and keeping the AI from expanding everywhere is another way to keep their score from getting out of control.

So clearly the game actively rewards having lots of cities and actively punishes having few.  It doesn't matter which race you play, you should generally try to have a lot of cities rather than only a few.  That part doesn't change based on what race you choose.  The race you choose will make a lot of the micro things getting to the end result a little different, but in general the same macro is going to remain in regards to expanding your empire as much as you can safely do so.

Research wise there are really only two serious macro plans.  The micro of exactly which order you research things in may vary slightly, but the macro of your overarching goal is probably not going to differ that much.

The two plans generally are, 1) Research the cheapest stuff to get the most number of useful techs the quickest, or 2) Try to beeline straight to some powerful tech and abuse the heck out of it before the opponents can get anything that can remotely match it.  One of them is going to be a jack of all trades and the other a master of one.  The one that is decent at everything will try to use its versatility to counter the raw power of the other and vice versa.  There isn't really a 3rd plan here.

While I may defend one of those plans over the other, I would definitely say that one of those two must be the best.  Every player probably uses one or the other and they might switch between games, but they probably don't use a 3rd strategy.

The implementation of all of these and other overall game plans will vary, but at the core there are a very small number of strategies that are rewarded by the game and everything else is clearly below those when it comes to end results.

This game is better in that regard that most of the other games I have seen, but it isn't the utopia of gaming that little willie and his ilk are searching for.  In dreamland where he lives, a game should have 20 or 30 viable overall strategies and if you selected two randomly and pitted them together the win rate would be close to 50-50 with equally skilled players.  Also in dreamland where he is from the AI is somehow programmed to think equal to or better than humans can.

Maybe 50 years into the future we will have AI capable of meeting his needs and it would be really cool if time travel existed and we could just teleport him forward in time to a time where he can be happy (and be rid of him ourselves), but sadly that is too much to hope for.

For now we are stuck with games that have AIs that are far below human level and games where you can't just pick research randomly and hope to win.

I don't known how many seasons it takes me since it really varies. But when i get the first champion i make him a warrior and baby him, i sen him traveling and attack weak targets (i end up spending a lot of time running from stronger groups lol) that i believe i can kill, after 4 or 5 battles he levels up and than i choose my warrior line and repeat. i don't in the mean time i'm alternating my cities to build troops and buildings. So by the time the champion is either killed (sent back to closest city) or returns i have a few units waiting for him. 

But i agree that it's a little underwhelming in beginning, as the first part of the game shouldn't be focused so much on baby sitting your champions.

It depends what you mean by "decent", and to give a definite answer I'd have to check my saved games, but I guess I aim to have a stack capable of taking on tougher battles from about turn 70, and that probably includes one mage and one commander of level 8 or so. I normally expect to be on a roll taking AI cities, Deadly lairs, maybe Wildlands, from about turn 100, and then turn 130 onwards is mainly mopping up, game is generally over somewhere between turn 170 and 200. Main factor in how quickly I level up my heroes is movement; if I have plenty of mana and can cast Tireless March, and if I can research and train mounted units early, then I'll be levelling up faster because my main stack will be moving twice as far each turn.

Around the same timeline for me. Although I might get my first tough stack out a little earlier. I also tend to play relatively peacefully. Once I  havetaken a few AI cities and am clearly dominating I don't really bother to expand more.

Let me ask a new clearer question. Do people feel non sovereign combat champions, not generals, are powerful forces in tactical combat that are noticeably tougher then normal troops and make a difference. Both your own and AI champions.

my second hero stays w sov, learns roads and army leadership.  Seems a good trade off.  Others just go adm cities, and reinforce garrisons.

I make all but one of my champions into mages.  I'll make sure that each of my "main" armies has two champion mages in it - at least one of which can heal and shrink, and one of which can cast resist fire (after leveling up a bit).  I also try to get one of them fireball and one teleport.  I always play the custom race trait that gives you spellbooks so I can make sure this happens.

I make one hero a commander for building roads and helping out in conquered cities.  

I've found defenders/assassins/etc, to be pretty useless in combat.  I'd rather have a fully upgraded combat unit in the army than even a well-upgraded single champion.  

My custom sov is always a mage versed in death and fire magic for serving up arcane annihilation later on once I can get the casting time down.

I've generally made my first champion into a commander because you're not guaranteed to get one later on and he's pretty useful just sitting around in the beginning while I level up my mage sov.  A little later on I like the commander for making early units a little more accurate, as a mobile healing platform, and when not campaigning for unrest reduction in a city that's crucial at the moment (in a conclave for a decreased unrest research penalty or in a captured city for reducing unrest production penalty).  Finally, if Pariden are in the game I like to buy the air and water books for him if possible for slowing and hasting, and then try and get down the initiative boosting branch of the tree, though I've never gotten to war cry before.

My second champion has often been a warrior, unless I can choose someone to give me access to schools of magic that I don't have.  A warrior early on with a bit of luck in the gear drops can be pretty good at clearing and collecting loot from low level lairs across the map.  With decent armor, a good sword and shield, upgraded counterattack, a few levels in reave, and maybe those tears of cyndrum (3hp for killing something) he can wade into foes and recover hp lost with killing.  One place where my later warriors always get squishy is when facing three or more enemy militias while taking towns.  With later blunt weapons, if two or more of the militias connect with a bash I tend to get carried off the field.

I've never used assassins because leveling them early was always really unappealing, but WhiteKnighted makes a pretty compelling later game case for them.  I might not use them in lieu of the commander which is hard to compare to, but I might try one next time to see how they work instead of a warrior.  They might end up dying less when facing units who can't really be tanked.

Responding to the OP, I would agree that in many games my sov has generally been my go-to unit for taking on anything challenging and I've often just left some of the later champions (whose stats weren't that spectacular or I didn't take to just because of their looks) to philander in my cities.  I also think that many of the posters in this thread are correct that champions can't really compete in terms of raw damage output with later units.  What they do bring to the table is the different unique skills and magic abilities that normal units don't get, and unless you've got henchmen there's no other way to really get that.  Facing off against a tough threat with a commander, mage, warrior, and a defender, in addition to a couple tough units gives you a wealth of tactical options especially if they have all the magic schools between them.  There's also the roleplay aspect.  It's so much more fun to do a tactical battle with a well balanced party then the clickfest that is swarming things with overpowered units (which admittedly does work fairly often).

Why don't you give "War of the Lance" a try and Centurion Defender of Rome on it's two upper difficulties and tell me that dribble. I've seen good playing ai's in my lifetime and apparently you haven't. Even Civ I had a better ai than this, at least it was aggressive and came after you in spades if you were weak. I don't know how many games on just the normal difficulty I lost on Civ I, but it was challenging as hell. I don't live in a fantasy world....you do by thinking an ai can't beat a human. Plenty of chess games can and plenty of games from the 80's and early 90's can, producers and developers just stopped putting too much effort into ai's and started doing "graphics crapola" upgrades. It's like what is so great about HOMM IV-V-VI over II & III (two of the best HOMM in the market). It's lazy programming...that's what it is and fanbois like you that keep it coming. Well, I will never bend my ways or fall to my knees or let my mother dress me funny...like some people I'm aquainted with lol. )

Willie: Someone tries to vaguely stand up for you and you call their views dribble? Nice.

"Plenty of chess games can" beat a human? Well, yes, because some of the best minds and best computers were put on the problem for decades. Do you think IBM is going to spend a few million dollars on improving the AI of Legendary Heroes? And for an example of a very simple game which computers currently struggle with, try http://arimaa.com/arimaa/, which is played on a chess board. Players choose how to set-up their pieces, which is one of the things computers struggle with.

If computers struggle with that, how much do you think they struggle to beat a human in a game which lasts around 200 turns, in each of which there are many different choices to make as to what to build, what to research, where to go, what spells to cast, where to attack, where to defend, etc. etc.

At the point where you can show me a game with several hundred spells and skills, with multiple possible winning strategies, and that AI consistently challenges human players without being given big advantages, I might accept you have a point.

Cool thanks.

Chess programs don't use AI.  

All chess programs do is direct a huge amount of computational power at "trying" every possible combination of moves and counter moves.

There is a formula built into the program that gives the computer a basic understanding of how much each piece is "worth" compared to each other piece so when it is done calculating the first long string of moves it saves the "state" of the game as the best possible result.  It converts the pieces on the board into a value and remembers that value along with all the moves it just tried to get there.

Then it tries a different set of moves and if the resulting value is higher than "state" then it will replace "state" with the new value and replace all the original moves with all the new moves.

Then it tries another different set of moves and another and another and another.

How good your processor is and how optimized the program is determines how long it takes to "Think X moves ahead".  Generally speaking, the chess programs will have various skill levels like 1, 2, 3, 4, 5, 6, etc.  Those generally correspond to how far ahead the computer will be thinking in half turns.  The slang term for that is "ply".

So 1 ply means it just tries all it can do and does whatever has the highest "state".  2 ply means one of its moves and one of your moves.  3 ply means one of its moves, one of your moves, and another of its moves.

The computational power required for each higher level is exponentially higher than the previous level.

At 1 ply there might be 30 possible states.  At 2 ply that number might rise to 600 possible states.  At 3 ply it might be 10,000.

This is all very brute force.  The faster your processor the faster these calculations get done.

Chess AI is very simple and straightforward and only limited by how much processor power you can direct at the problem.  Every time computers get 100x more powerful, they can think another 1 ply ahead.

All humans have on computers is the ability to discard the worst 95%+ of moves and use their mental effort to try and delve deeper than the computer could ever hope to.

As the game goes on, more and more pieces leave the board.  Computers don't change their difficulty level based on number of possible moves remaining (usually) so if it thinks at 4 ply on turn 1 with 32 pieces on the board it will think at 4 ply on turn 50 with only 8 pieces on the board.  OTOH, humans do change their difficulty level based on the number of possible moves remaining.  A chess grandmaster will envision a dozen turns ahead in an end game situation while they may only envision 3 or 4 turns ahead in the beginning.  For a computer a 4 in the beginning is a 4 for the whole game.

This is why human grand masters tend to do better in end game situations than computer programs do and why the AIs tend to beat the humans when it comes to early and midgame play.

Programming TBS game AI is pretty much nothing like that.

It is 100x+ easier to make a chess AI than a TBS game AI and once you have that it is much easier to make the chess AI more difficult (add more processing power and calculate more ply) vs how hard it is to make TBS game AI better (add lots of new lines of code unlike anything else you created so far for every tiny improvement).  

Your explanation is more complete I was speed typing with 2 min to spare after a lunch break.