MFI (beta)

[scripter64]

Welcome to the world of .

[Campeador]

Hi dead_kniight:

Finally I've tested your AI, and although it's quite ok it has a few of the usual problems in new AIs:

1- It hunts boars when you have 18 villagers. You should do it earlier, to be able to hunt both boars in a good time, making you economy much stronger.

2- There are two main types of Fast Castle strategies.
a. Barracks + Blacksmith + Stable/Archery, depending of the units you want to create in the castle age.
b. Market + Blacsmith. You less wood because you just rush 2 feudal buildings instead of 3. You also use the market to sell stone, so you can delay gold mining and getting even more food gatherers. This type of Fast Castle is faster, although with a weaker economy and more vulnerable to possible enemy flushes.

But you don't do any of these FC rushes. You build barracks, blacksmith and market, but you don't sell any stone. This is only an option if you are sure taht you won't use archers nor horses, just pure infantry. And even then, you delay your barracks after the blacksmith and market. If after the market+smith you have wood for a barracks before the castle age transition, it means that you have too many woodcutters.


Military mistakes. Your AI sometimes train militias in the castle age. It's a mistake, because then they are so obsolete that they are useless until you upgrade them to m@a, I'd say taht even to lsm. You also train too many types of units. It delays your upgrades too much. You should train just a few types, considering to add a few units to counter the enemy army (add skirms just if enemy archers, etc). You also can consider to replace units that are being countered, but it's more complicated and sometimes and bad decision, it's because I say "consider".

You should improve your military superiority code. I watched in a test vs the Standard AI that you started TSA, to cancel because you had lost superiority before to figth. It happened because you set 5 soldiers as the minimal advantage to start, and it is not bad. But your AI got those 5 soldiers superiority, your enemy trained an extra soldier so you had 4. You trained another so had 5 again and restarted TSA, enemy trains another and 4 again, etc.

A possible solution is that you start an attack if you have 5 superiority, but stop it if you have equal number, just 2 superiority, etc. What a mean is that the limit is different to start an attack to cancel it. Of course be careful with this change, it could change the optimal conditions for your AI, etc. Consider to count possible enemy castles and towers, to have too enough percentage (it isn't the same 5 soldiers when 10 vs 5 than when 100 vs 95), etc. Although my own AI is not much more complicated than yours.

You also need a rule to attack when full population and full upgrades, adding possible exceptions. With your current code it is possible that both civs have full pop but your army is fully upgraded and enemy's one not. So if you delay your attack you can lose your advantage.

Last problem of your AI is that you need a trade code. I've watched it with +5000 food and gold problems despite that you could sell 100 food for 73 food or alike. Of course I recommend you that your first code is very light, doing it just in emergency cases. Experience will teach you to set better this code.


P.D: I imagine that you understand now why we call this the world of . Don't worry, we can help you with any doubt that you could have with your new neighbour to the block.

[dead_kniight]

P.D: I imagine that you understand now why we call this the world of . Don't worry, we can help you with any doubt that you could have with your new neighbour to the block.

Actually that is the only thing i understand perfectly about scripting.

By the way thanks for your inputs, although the version you have tested is very old and i have updated it since then. I had thought of uploading the new file and have at last accomplished it without procrastinating any more.
I would love it if you tested out the new version. Waiting for your suggestions.

Now on to another important point. I have no idea about using the more advanced(complex) features of UP like search etc. It is all hit and trial for me.
I would be grateful if someone could tell me about it or point out the pages where such things are mentioned(please link the specific page if you can and not the topic).
Also I would like to know some basic things given that I am piss poor at AOE :
-Gatherer percentages.
-Basic strats and optimum no. of villies etc.
-Escrow

Other suggestions and advises are also welcome.

[_II2N_]

Hi,

tested MFI 1.43 in some games, and it is decent. In addition to what Campeador said, there were some mistakes I noticed:

-Some gold was mined at dark without a mining camp.
-MFI goes up to feudal for FC too early. The optimal is that there would be no idle TC at any point and that 2 villagers are trained at feudal age.
-It trained 2 spears early to counter enemy's starting scout (since opponent was Franks) and took tracking for them. They delayed castle age time.
-Mining stone was started at early castle age although stone wasn't necessary for anything. MFI dropped a castle at 27min. Usually it's better to either add more TCs or invest even more in army.
-General blacksmith research order is armor upgrades first for melee units and attack upgrades first for ranged units. MFI took forging early for knights, though scale barding is usually more beneficial.
-MFI delays bow saw which is very important economy upgrade while taking less necessary technologies like town watch somewhat early.
-For food-costing military units a food buffer is good so that they don't block villager training. When MFI was training eagles, it had difficulties to afford villagers at early castle age. Villager training shouldn't be stopped if it isn't necessary.
-Villager limit at castle age could be higher than 100.

Most strategies can be split into FCs and flushes. For AI FCs usually attempt to reach castle age with around 29-31 dark age villagers (depending on start, with more villagers if resources are missing. It's easier to go with more first and then lower the count as you optimise the AI). It's important that there wouldn't be idle TC and there are enough resources to afford intended buildings and researches. Too fast FCs will hurt economy. I'll list some example FC strategies and AIs you could observe:

Knight rush (krush) is typically done with stable+blacksmith FC with 2nd stable added while upping to castle age. It aims to have a decent number of knights early and take atleast scale barding and bloodlines relatively early. It is most common pocket strategy in team games, atleast in human games. Horde's franks and Juggernaut's huns use it frequently.

Crossbowman rush (xbow rush) ups with range+blacksmith and is usually done with 3 early ranges. It's important to take bodkin arrow and crossbowman upgrade early, since they provide more firepower and range. Xbow rush can hit very high numbers quickly, and it can be hard to counter. UnfairSteel's mayans have good xbow rush.

Plain boom uses market+blacksmith as feudal age buildings. Instead of training military boom's idea is to drop more TCs quickly and invest all resources in economy. Once there are enough villagers, it's easy to train a large army. Plain boom is rare, since it's weak to rushes. The_Unknown's britons are good boomers.

Flushes attempt to reach feudal age as quickly as possible with still enough resources for military buildings and unit production. Most common flush in AI games is probably archer+skirmisher flush that tries to drop 2 ranges immediately at feudal age and get fletching early. Most civs can do it with 24-25 villagers with a good start, although fast ups require strict economy. M@a+archer+skirmisher can up 1-2 villagers later since militias can be trained while advancing to feudal age and m@a require more economy. Promi's aztecs do a dangerous flush.

It's also good to check expert records from aoczone.net for build orders and strategies (although AIs may not be able to directly use them).