Defending by running away megathread

[Nelphine]

And while it may not be as common, I believe there's no reason for the AI to stay in the 15 tiles if there are no city walls. (Personally I've never built city walls in my life, so any time I need to reconquer them, there's no walls to deal with)

[Seravy]

There isn't - without walls, the entire "stay behind walls or leave" decision is ignored, as there are no walls to stay behind. At least one of the three walls is required for this decision to have any effect.

[Suriname]

No need for an algorithm, which wouldn't be possible, because I've added 2 failsafes: the defender advantage and the obligation for each unit of the attacker to be in the city to count.
With defender advantage we ensure that what you said Nelphine doesn't happen - grabbing a few counters and then leaving again won't be enough to capture.
With the obligation for all the units to be in, we ensure that baiting doesn't work: if a bait exists, then the attacker's counter isn't raising.
Finally, with the ability for the defender to increase the counter even with some of the units in the city - the attacker needs all, don't forget - we ensure that even in cases of heavy spell inflicted slowdown, the situation needs to be absolutely crazy as all the units need to be prevented from coming back.

And that, it's not easy to make it happen. So, for the web+wall example, all the human units need to have crossed the AI units (on the same path in opposite directions: usually that means fighting) to the city, all the AI units need to be outside (and they leave cities so slowly usually, they take many turns to be all outside), all of this needs to take less than the defender advantage allows... You see? If during test we see this happening we can just increase the advantage. But I'd really be surprised to see it happen. 

Furthermore, the extreme situation you described Seravy is quite unlikely to happen for a reason on the human side: a human needs to learn the trick and perfect it. This is possible with cavalry, but if it only works with spell skill in the several hundreds (blocking several units for 15 battle turns) then it's never going to start. Even if somehow someone discovers and perfects it, well, it'd still need the ability to bait AI out, slow AI down, have all the units in the city and make it all work despite the high spell-casting skill present at that advanced stage of the game. At that point, is it still a trick? Because casting buffs and attacking takes less player ability...

One last option if you're still not convinced is to reduce the likelihood that AI units will leave cities, maybe randomising a little the decision. This makes tactical moves not so easy to foresee, which is a bit of a sore point for me.

[Nelphine]

No, with all of those conditions you have a problem: you play exactly as you are now. You use 5 bowmen and a cavalry to defend your city. You flee the city with your cavalry, your bowmen sit and shoot; since the defender doesn't need all of this units to increase the count, those 5 bowmen add to the count. 5 bowmen can cause decent damage; combined with spells, there might only be 3 attackers left. The bowmen blocked the gate, so it takes too long for the attacker to get in (defender is probably at a count of what, 50, and it's turn 10?); and now, by the time the attackers get into the city (3 attackers left, 15 turns, only gets them 45 points), you can still use exactly the same trick you're trying to fix right now - the cavalry is far away, and the defender has 'nothing' in the city, yet the defender still keeps the city. (Also note,assume the defender gets the same rule of needing all the units in the city, and you somehow avoid the other problems that brings, then you just don't build city walls. Your bowmen block the cavalry till the last moment, and only at the end does the cavalry go out, and you still end up with the defender at 50, attacker at 45, cavalry far away, and cavalry 'keeping' the city at the end despite not being anywhere nearby)

So now, i can abuse your proposal in the later game with certain unit types, AND you can still use your current 'problem' without any change.

[Nelphine]

Also, you haven't addressed: if there are no city walls. Currently, the AI has no algorithm to stay in the city at all on defense. If the AI has 9 ranged units, they'll all just back out of the city immediately in order to maximize time to shoot.

And, you haven't addressed: currently if I cast mana leak, all of the defenders immediately run out even when there are city walls. That means after turn 2, he has no defenders in the city tiles. So he has at most 19 counter. I can easily run in my dragon turtle, and get 20 counter with just that one unit, allowing it to win. All i have to do is make sure all my other units die by turn 5. (ANY rule that encourages suiciding my own troops is awful.)

[teelaurila]

Any "counter over the battle" system is just overtly complicated, let alone one with a handful of treshold and intial value and special cases for flying. Compare it to anything else in the game and it just looks obvious to me. (Most of the game is complex, which is generally good, but this is just convoluted and complicated, which 8 pages of tweaking around should show for one)

Suriname said yourself you generally need to sacrifice a couple bowman to suck the ranged damage allow the cavalry to start this dance. Which means this only slows down the enemy a turn - those bowmen are then gone. And the enemy can cause significant damage to your city even if you win the dance.

Also you likely could have sacrificed those bowmen to slow down the enemy a turn on the way to the city (but you can of course do both if you have enough units)

It's just painfully obvious to me that this is not worth it unless it can be done something like e.g. Total War did it: If the central square contains only attacking units for time X, attacker wins. Thus forcing a running-defending unit to end its turn in the city might work - only way attackers occupying the city cannot beat the cavalry that popped back into the city is if the cavalry flies and none of the attackers can reach. In which case it's not a dance and sort of fair enough. X needs be long enough, and/or AI needs to work with this rule well enough, so that you can't trick the AI with this. If that can't be done, then trying with a more convoluted rule is a fool's errand in my books.

[Nelphine]

Any one square rule fails of the defender just keeps summons in that square.

[Suriname]

No, with all of those conditions you have a problem: you play exactly as you are now.  You use 5 bowmen and a cavalry to defend your city. You flee the city with your cavalry, your bowmen sit and shoot; since the defender doesn't need all of this units to increase the count, those 5 bowmen add to the count. 5 bowmen can cause decent damage; combined with spells, there might only be 3 attackers left. The bowmen blocked the gate, so it takes too long for the attacker to get in (defender is probably at a count of what, 50, and it's turn 10?); and now, by the time the attackers get into the city (3 attackers left, 15 turns, only gets them 45 points), you can still use exactly the same trick you're trying to fix right now - the cavalry is far away, and the defender has 'nothing' in the city, yet the defender still keeps the city. (Also note,assume the defender gets the same rule of needing all the units in the city, and you somehow avoid the other problems that brings, then you just don't build city walls. Your bowmen block the cavalry till the last moment, and only at the end does the cavalry go out, and you still end up with the defender at 50, attacker at 45, cavalry far away, and cavalry 'keeping' the city at the end despite not being anywhere nearby)

So now, i can abuse your proposal in the later game with certain unit types, AND you can still use your current 'problem' without any change.

This example shows a minor limitation of the proposal - it's "good but could be better": you're going to have victims, so the city is going to change hands in the following turn. So, already better than now.

I'm not going to repeat myself as we'd be going in circles, but basically you say that it's either useless or abuse, I say that there's a sweet spot in the middle that can be found thanks to the variables that I've added to the proposal (delay, advantage) and obviously neither of us can prove it beyond doubt. So at this point the only way to know for sure is to try this thing and see.

Also, you haven't addressed: if there are no city walls. Currently, the AI has no algorithm to stay in the city at all on defense. If the AI has 9 ranged units, they'll all just back out of the city immediately in order to maximize time to shoot.

I don't think that this implies what you seem to think. I routinely see AIs go back to the "city" square even in outpost fights.

And, you haven't addressed: currently if I cast mana leak, all of the defenders immediately run out even when there are city walls.

Same - I've done quite a lot of mana leak fights with my undead.

Any "counter over the battle" system is just overtly complicated, let alone one with a handful of treshold and intial value and special cases for flying. Compare it to anything else in the game and it just looks obvious to me. (Most of the game is complex, which is generally good, but this is just convoluted and complicated, which 8 pages of tweaking around should show for one)

Actually the proposal is pretty simple for an algorithm, I've described it in pseudo-code above, initialising a variable to a specific value and another one to 0 and then increasing under some conditions is all there is to it. Then, check and decide.

Suriname said yourself you generally need to sacrifice a couple bowman to suck the ranged damage allow the cavalry to start this dance. Which means this only slows down the enemy a turn - those bowmen are then gone. And the enemy can cause significant damage to your city even if you win the dance.
Also you likely could have sacrificed those bowmen to slow down the enemy a turn on the way to the city (but you can of course do both if you have enough units)

The city can produce another unit and your real army can come to the rescue, I've shown an example where I am calmly managing a 3 fronts war at lunatic: one the left I attack and on the right I just stall for time mainly using this trick.
On sacrificing troops I agree, it should be a bit less effective to sacrifice spearmen but that's another thread

It's just painfully obvious to me that this is not worth it unless it can be done something like e.g. Total War did it: If the central square contains only attacking units for time X, attacker wins. Thus forcing a running-defending unit to end its turn in the city might work - only way attackers occupying the city cannot beat the cavalry that popped back into the city is if the cavalry flies and none of the attackers can reach. In which case it's not a dance and sort of fair enough. X needs be long enough, and/or AI needs to work with this rule well enough, so that you can't trick the AI with this. If that can't be done, then trying with a more convoluted rule is a fool's errand in my books.

That might work too? It could be any of the city squares instead of the central one. Any solution that works is fine for me.

[Suriname]

More importantly you can block the tile the AI needs to go through to reenter their city with a flying (or webbed enemy) unit so they can't return even if they want to. As long as there is a city wall, they're stuck outside...

Right, this. Why can you move out to the 3 squares adjacent to the gate when leaving the city, you can only go in the opposite direction from the central one? It's inconsistent. If this could be fixed then the webbing issue would go.

On flying, I like it that you find it unfair to block with flyers...   I think that flying is not going to be an issue because AIs target them pretty early in their spellcasting priority. Material for testing.

Is there any point that you raised that I've not addressed?

[Nelphine]

You've dismissed my points and said 'i don't think it's a problem'. Ive provided exact numbers and specific examples, from a wide variety of tiers and realms where your proposal would be worse. You have said they aren't relevant.

I'd like numbers and specific examples of why they aren't relevant.

If the game coder says there is no algorithm to return to the city tiles, and my experience agrees with that, I have to conclude that when you saw it either the ai was attacking, or you had units positioned in such a way that the ai randomly returned to those squares.


In short, no, I don't think you've adequately addressed ANY of the counter examples I've brought up.