Why don't log rams have like a 50% chance to escape from combat aftre having attacked a city? The unit seems pointless now, as it is dead after attacking and have only a 10% chance of reducing the defence by 1% each?
1) Historically ram teams would not survive the attack.
2) If their withdrawal was made any earlier they may enter into a scenario where they cannot do enough damage to the walls simply because the opponent is too injured which is extremely counterintuitive, however, this is due to the withdrawal mechanism they are using being based on the same method as other siege weapons - it's not a chance but rather is automatic and based on how damaged the enemy gets. If we made it a 'chance' then we are imagining that at a point they give up and drag their ram back after they feel they are about to die which simply did not ever happen. The current method is far more like they've done all they can do and now more may join them in the effort. Otherwise, without reaching that point, they are killed first (which again, historically, often would happen.)
3) Promotions can vastly enhance both the chances and the defense reduction.
4) That chance and the damage if successful actually takes place each round of battle rather than once per battle. There can be up to 10 rounds during each battle or so and if the rams are roughly equivalent in power to their foe it can be much more extended. Once they get much more power than the defender the battle can end a bit frustratingly prematurely and if they are up against impossible odds they may die a little too fast to have too much impact. The trick is to use your distance attackers to wear down the defenders somewhat to give the rams the most time to fight. (Watch your combat log to see a round by round report so you can see what kind of round counts you're getting in different combat scenarios.)
Ultimately what this means is that rams, with a fair degree of promotions (most of which they get from proper training preparations made in the city from which they are trained in) rams can be absolutely devastating. It's just strategically tricky to find the sweet spots which means they are adding a layer of strategy never before seen in Civ franchises. If you find them frustrating to use, just remember how frustrated you may have been learning how to invade a city coming from CivIII experiences into CivIV (where siege does actually now matter.) It's just a second learning curve.
Also, I have found one of my neighbors having 29 41 log rams stacked inside his city. Maybe that is because the AI calculates that you need A LOT of log rams to do anything useful?
Interesting... while this may seem like a big problem for the AI, it's actually a statement of feedback that represents a major recent victory. I've been pushing and pushing to get them to build more than one of them and finally I've found the trick... at this point I need to now back off of some of the previous efforts made to get them to value rams more so that their attack stacks start becoming a bit more reasonable. So this is a major step forward to hear this!
EDIT: I moved a small stack near the city with all those log rams. I was thinking this would be a looong siege, with me killing off one unit at a time, slowly gaining xp and captives. But, to my surprise, the AI moved more than 40 units out of the city, allowing me to attack and capture the city on my next turn. As this was the only city of that civilization, all those units dissapeared...
Not the first time this has happened - fixing it is going to be a little tricky. A while back we instructed the AI on a devastating strategy to destroy incoming attack stacks. Unfortunately they don't yet realize they need to NOT use this strategy when it's their LAST CITY! Once I figure out where Koshling implemented this strategy I can probably quite easily tweak it to correct that flaw.
I don't want to say much more on this... perhaps you can guess what they were TRYING to do.
Also, a few turns later, another AI attacked my newly conquered city. Some of my units had not even had time to heal up. The city was not on a hill, there was no river between the attackers and the city, and I defended with 4-5 combat units (merged atlatls and mace men with combat and city raider promitons and around 10 xp each) against 20+ units, again mostly log rams, but a few were merged. Only one or two of my units died in the AI attack, that obliterated the attacking army. Something doesn't add up here, me thinks. The city should have been taken...
Some improvements to size matters stack building is known to be an issue to soon focus in more on. Sounds like what happened was they had a lot of units and thus many had merged but once those merged units had achieved some wins all that was left was all the ones that split after the merge so they had split units in there too that didn't have nearly as much capacity to deliver the pain. At the moment, the SM AI is extremely rudimentary, just enough to keep a player on his toes and to be generically proactive against a variety of potential player strategies with splitting and merging. Some more reading the situation and reacting to it appropriately is yet to be developed.