|
Author |
File Description |
WigWam |
Posted on 02/16/11 @ 02:52 PM (updated 05/06/11)
File Details |
Version: |
The Conquerors 1.0c |
A lot of the Blacksmith files are AI's and I think about 80% are "My first AI" scripts.
Well now it's my turn!
Here is my first AI designed for the Teutonic civilization.
I'm posting it here too hear what I can improve, so please comment! |
Pages: « First « 1 [2] 3 » Last » | Author | Comments & Reviews ( All | Comments Only | Reviews Only ) |
---|
WigWam
File Author |
Posted on 02/28/11 @ 04:08 PM
Updated to 2.3
hints are still welcome! |
Campidoctoris |
Posted on 03/01/11 @ 11:43 AM
I’ve tested your script this time against Saiyan, an old flushing script but stronger than the Standard Computer. You have won without ant problem, but with a few changes you could do it quite faster. Suggestions:
- Your Dark Age is quite good. I’d add a second lumberjack to avoid villagers crowding, although it can be built later. You make a good amount of farms, which is a good idea, but once you click the “Research Feudal Age” bottom, it would be a better idea to stop farm building to create your first barracks earlier, it should be constructed before you arrive to Feudal Age. It’s the way to start your feudal buildings (to flush or fast jump to castle) immediately.
- Training AI has a very interesting goal called g-age-status which could be very helpful to you. It allows you recognize in which age you are (the same than the condition current-age) AND if you are advancing or not to the next age. Other scripts like The Dragon do the same combining current-age with a goal which only says if we are advancing or not (with values “0” or “1”). I prefer the Training AI method, but it’s your decision, both should improve your AI. How? For example, Training AI increase its number of villagers on wood once it starts transition to feudal. Later it changes a few villagers on food, removing more woodies as it constructs more feudal buildings, and getting more foodies.
- If you test my AI, Crusade 2.0, you’ll see that it gets a quite good feudal times but arrives to feudal age without enough wood for archeries and other feudal buildings. It sometimes hasn’t finished its barracks. This problem delays my flush, giving to my enemy time to prepare its defense, advance to castle or even flush itself to me. Crusade 2.1 (uploaded in aiscripters.com) has “worse” feudal times, but it’s faster getting its feudal buildings and trying its flush. Well, you have the same problem than 2.0 but more exaggerated, last two paragraphs were written to explain you how avoid it.
- Man at arms upgrade. Man at arms are good feudal units, and can smash spearmen without problem. Militias are the worst feudal unit, with the same power of spearmen but more expensive and without the anticavalry bonus. I’ve seen your AI training many of them before upgrade them to mat at arms. Don’t do it, except in transition to feudal if you have an excess of food, and with the intention to upgrade them in feudal instead to create more. But if you make the other changes I’ve suggested, you’ll only have food for the m@a upgrade and the economy feudal technologies.
- You need a better balance resources in the other ages, also. You have too little wood in feudal and too much in castle age, delaying your blacksmith techs because you have too little food and gold, and your imperial time.
- You attack when you amount an specific number of soldiers, but without count the military population of your enemy. So you can attack to early or too late. Training AI has a goad called g-military-superiority (I think) which has values from -2 to +2, determining which player has a more numerous army. I have the same goal and with the same system, but more complicated and more rules, so study first Training AI. This way has the limitation that you could have a little superiority in number, but enemy gets knight, cavalry archers and elephant, or viceversa, but it’s better than wait to x soldiers without study your enemy. There are scripts which also have rules to attack even with a little numerical inferiority or not attack because the enemy has too many castles and we too few rams, or heavy cavalry inferiority, etc. But it’s more complicate, so let it for later and start with pure numbers comparison.
- You don’t research blood lines. It’s very necessary to your knights. See the tech tree of your see to be sure that you don’t forget any other important tech for your strategy.
Well I’ve suggested in the Ladder forum (http://aok.heavengames.com/cgi-bin/aokcgi/display.cgi?action=st&fn=28&tn=39043&f=28,0,0,30&st=1750) to create a new category for the new scripters. Leif Ericson (the organizer) seems interested in the idea. You could say in that forum if you are also interested.
I’ve only tested your latest update once because you are very busy. But I suggest you to test it next times (to mesure your level) against Tiger_RM (an old flushing script who is similar to Saiyan, but a little stronger), Zycat (other old who prefers attack in Castle Age and can play in every setting) and finally against Chong Wong Fei and Hispain Special. Stronger scripts will fall later. |
WigWam
File Author |
Posted on 03/07/11 @ 09:55 AM
Have some free time on the hand so I'll start going.
Where can I find the "Training AI"?
PS: Tiger RM got destroyed by WW - Teutons[Edited on 03/07/11 @ 12:23 PM]
|
Suriel |
Posted on 03/08/11 @ 10:28 AM
You can find it here, in the Blacksmith |
WigWam
File Author |
Posted on 03/08/11 @ 06:16 PM
Updated to 2.4
I'm not sure 2.4 is an improvement in regard to 2.3. I've changed the Attacking system and some other things and I'm still experimenting on it.
The question I ask you is again: what can I improve? [Edited on 03/08/11 @ 06:17 PM]
|
Campidoctoris |
Posted on 03/10/11 @ 08:14 AM
I’d need test your last versión against to your preview one, or both against others to know which is better (it’s posible A>B>C>A in AI wars). You could do it with your own time.
Anyway, I’ve tested 2.4 against Hispain Special and your AI is not bad. You have lost this game but you could improve a lot and take revenge very easily with only a few changes:
a) You are forcing to arrive to 30 villagers before feudal age. It can be a good idea if you pretend a fast jump to caste. But IMO (other scripters better than me could disagree), you could do it even with 28. Yes, you fear don’t get enough wood for instant feudal buildings. No problem. I’m working in a new Crusade, and to prevent it (I suffer the same with my Mayans) I’ve added two alternative gatherer percentages in late feudal. I use a goal who will be set to 1 if I have a lot of food and I’m near to feudal, and to zero if I have much less . For example, you could set to 1 when 450 food and return to zero if food problems down your food under 400 (of course test it to change your ideal numbers). If that goal is set to 1 you can use one gatherer percentages (with more woodies to get them prepared), set to zero you use a different one (with more foodies to accelerate feudal time).
This is only one way. For example, Phyrexx, if I’m not wrong, uses always the same gatherer percentages, adding more woodies at 28th villager. It often avoids idle TC, thanks to its very good dark age, and the Hun bonus. The risk is that if it has bad luck with sheeps, for example, it can get a few second of idle TC. Although even in this case its flush is really fast and incredible. Of course other scripts have different systems. Use the most adequate to yours. Anyway, your feudal time (14:38 in this test) is not bad. I often get something similar with the same villager number if I don’t find enough sheep and deer.
b) Feudal age. You should research villagers with no stop except if you pretend an instant jump to castle. But IMO it’s not a good idea stop villager production since early feudal if you create a good feudal army. You will delay your castle time anyway and with lesser villagers. Although I recognize that at least your castle time is better than mine.
c) Rams. You only create them with a wood excess and after mangonels except if you have enough gold for rams but not for mangonels. Well, it’s a clear mistake. Mangonels can be a good complement for TKs, but when your are attacking rams are more important, at least if your enemy has castles. So you should give to your rams a higher priority than mangonels when attacking, at least if enemy castles detected. I’ll add that with your rules, you will only see rams in your script in the real game if enemy has forward built you or if you have a wood excess. In any case with preview mangs or little gold. There are scripts which train rams even without attacking if they detect castles, to get them prepared if they decide to do it.
You had a clear advantage against Hispain Special, but your soldiers, after win the first battle, the only thing that they could do was suicide themselves against castleS with no stop.
Of course, once arrived to Imperial Age, you can decide stop ram production if you have castles and can train trebuchets. There are scripts which prefer trebs, other rams, other bombard cannons and others choose it depending of the civ selected.
d) Attacking time. I’ve seen that you attack in Castle Age if you have X soldiers and specific researches. It’s other error. You must study your strength and yours adversary. You can find in Training AI an algorithm which detect if you have ten or more soldiers advantage or disadvantage over your enemy. I use other very similar, copied from Goose AI, which is more complicated but detects differences of 5, 10, 15 or 20 soldiers. I’ve added a few rules to detect possible heavy cavalry advantage. Other scripts have different rules to prevent attacking with other conditions (a minimal military pop if enemy has a tower or a castle, wait for an important research, etc). Only you can decide the most adequate algorithm for your AI, but it should consider your military strength… and your enemy.
e) Warfare. You are using the attack-now command to smash your weak enemies. It’s a good option. But there are other two good options, attack groups which is similar but more complicated and TSA (town size attack). You can find an article about TSA in this web, and other explanation in the Training AI. Town size is the square area which will defend your soldiers, TSA consist that when you want to attack you are increasing slowly your town size until it reaches enemy’s buildings, then your soldiers will attack it because they think that the enemy has forward built in your base.
Every attack system have advantages and disadvantages over the others (explained in an article from Mabuse in the aiscripters.com), so only you can decide which is the most adequate to your AI. Anyway, it’s to decide it a few tests, so I recommend you correct first the thing you know you have wrong, and later you can consider if it would be better change to attack groups or TSA, or combine a different methods. It’s possible that your current attack-now is the best for your AI. |
WigWam
File Author |
Posted on 03/12/11 @ 09:05 AM
WW - Teutons 2.6 is now available at your electronic store! (or the AOK blacksmith of course :D)
I've changed the attacking system again and I hope it will work properly now...[Edited on 03/15/11 @ 05:09 PM]
|
WigWam
File Author |
Posted on 03/22/11 @ 02:57 PM
updated to 2.7
Have some little questions:
1. When my soldiers attack they walk into the enemy town and start chasing enemy bowman/skirmisher/other-range-units. That's because someone shoots an arrow at them, then they'll chase that one unit. My quite slow TK's start chasing skirmishers (who are a lot faster) and that wouldn't be such a problem because TK's have some pierce armor but the main problem is that, while they're chasing the ranged unit they get slaughter by other units who are a bit faster.(champions and paladins do a lot more damage as skirmishers). My question is: Is there A way to stop my units chase the enemy units? (or just their ranged units)
2. When one of my soldiers attacks the enemy he is helped by near units (that is some kinda brave but not always useful). The problem is that my soldiers keep attacking that way. I once give (attack-now) then a armie will attack but when they lose and there are just a few units left they are helped by units near them and those are helped by units near them. And that's how my AI just won't stop attacking. small groups attack and get killed - they get reinforced but the reinforcements are killed and they get reinforced etc. etc. etc. :( Is there a way to stop that?
-I say (attack-now) once
- my soldiers attack
- probably lose
- and if that happens my soldiers are in my town and there soldiers are in there town. And the fighting is stopped (for a while )
Is that posible?[Edited on 03/22/11 @ 02:58 PM]
|
WigWam
File Author |
Posted on 03/30/11 @ 01:52 PM
Updated to 2.8
has less idle TC time and will get to the castle age faster, so it has a chance to catch some relics. (in stead of getting castle-aged when all relics already have been captured)
-
Do you think you know what WW - Teutons is missing? --- post a reply! :)
-[Edited on 03/30/11 @ 01:53 PM]
|
offwo200 |
Posted on 04/09/11 @ 05:15 AM
@WW 2.7 question I think there is a sn sn-enemy-sighted-responce-distance that can be used to avoid this. |
Pages: « First « 1 [2] 3 » Last » |
HGDL v0.8.2 |
Statistics |
Downloads: | 327 |
Favorites: [] | 0 |
Size: | 12.09 KB |
Added: | 02/16/11 |
Updated: | 05/06/11 |
|