| 06-12-2004, 01:25 PM | #196 |
@Strategy Master: Did he build the barracks ? @kasimir: In the GetUnitStrength function the life of the units is already computed in, but you are right, sleeping units should be excluded there. About the flee_health_percentage I am not sure what you want to achieve. Is the intention that since there is not so much damage dealt in the beginning so one can flee later and in the end there is more so flee earlier? The formula you wrote needs to be optimized though. Since you convert the time to an integer it will use integer division in /2200 which will likely result in 0. Just remove the R2I there. Also I guess it is good to add a maximum too. BTW, that is a computation that could be done with the LinearInterpolation function. Changing how often MICRO_HERO is called is probably a good idea but then the differential values in there like hp_loss need to be adapted to the time. There are many additions and optimizations that can still be added to AMAI and I have less time than I'd like to, so if you'd like to help contact me. |
| 06-12-2004, 02:11 PM | #197 |
@AIAndy: Thanks for your reply! your viewpoint about flee_health_percentage was right, when the game running further, the forcus fire strength of the enemy will become higher, and flee_health_percentage should be risen. But the formular needed to be optimized, and R2I is wrong :) To my opnion, there're a lot optimizations we can do to AMAI, for example, the AI is a UD and enemy is human, when human uses hero paladin and his mana >=65, then ai's hero should teleport earlier to avoid HolyLight, etc. I'm tring to add some code about those improvements, and i'll paste and discuss them here. I'm from China, sorry for my poor English ^_^ |
| 06-12-2004, 02:24 PM | #198 |
@AiAndy: Yes they do build the barracks and other units as well but just sometimes not the heros. I have also made all profiles favourite units those heros, and it slightly removed the problem but just happens even less often. |
| 06-12-2004, 11:58 PM | #199 |
hmm now here id definitly too much to quote everyone but I'll try to clear things up a bit at least. @Goose: We don't really have any control over how that hardcoded AI is using the burrow ability it's the same both for normal war3 AI and AMAI as you have noticed, we might be able to force burrow units for the computer if we want it to heal them but usualy the hardcoded AI already done that so there is no fiends that need to burrow to heal in base between battles. The reason to why I check for 4 archers is becuse I want to make sure the computer not only is able to make hippo riders but also that it atleast has a small force of archers to defend it self with before it spend money on that upgrade instead of making archers. Well spring priority is ok as it is I think, I will adjust the healer merc priority before next release. @Strategy Master: Nice of you to be the fiirst one to upload a custom made AMAI strategy. it's even mostly ok, no big errors but I got some tips for you. firstly the main problm with your strategy is that you only build 8 head hunters, why don't you fill upp with enough units to use all food? there is nothing hindering the towers anyway as long as you build all units with lower priority. But this takes us to another problem caused by this, I can't blame you since this is not in the manual yet but you got this distraction group: call AddHarass(2, 4, HEAD_HUNTER) set distraction_group = 2 We made AMAI only use the distraction group if the remaining force after the distraction left to distract is larger than 2x the distraction group. In your strategy the problem is that you add 4 head hunters to distraction and only build a maximum of 8 head hunters, you need atleast 12 of them if it shall be able to use the distraction attack. This was just a small advice for you. @FieldMedic: That FFA behavior sounds odd, I haven't tested FFA much lately but I will try to replicate that problem and see what we can do to fix it for the next release. @kasimir: I think I sent you a mail but im not sure if you got it or if you have answered to it. just wanted to check with you if you answered any mail from me becuse my spam cleaner might took that answere by accident in that case. @thenonhacker: Nice to hear that AMAI helped you, thats what AMAI is made for. :) Sadly I don't get that help I want in my online playibng by playing against AMAI, just becuse I make AMAI and do testruns with it very mch beating AMAI 1on1 is not hard at all for me since I know all of it's behavior. :/ But sometimes it can even suprise me. :) |
| 06-13-2004, 01:34 AM | #200 |
@Zalamander: I've write you an email several days before . And the program attached to it had lots of bugs. I got a intractable problem, when i and my roommates playing W3 with 3 AIs,{3V3 in (12)DevideAndConqure.w3m}. After a random time (usually when the first hero build complete), we all dropped (disconnected) from the game!Even the worse, when i installed the original AMAI to the maps, after the language select menu, we all dropped! i don't know the reason(maps or our net problem or os problem), the ai runs perfectly in 1.14B (now i'm using 1.15 TFT official, and all maps updated through BN after update) |
| 06-13-2004, 11:58 AM | #201 |
@Zalamander: So thats how distraction groups work. Alright thanks for the info, i was wondering how they work. I can fix that right up now. |
| 06-13-2004, 08:32 PM | #202 | |
Quote:
maybe your friend still had the map with your version installed on it in the download folder so it used that version for him and you used the official version. I sent back a mail to you about what we think of your version, comments and such on your improvements. You didn't get that? |
| 06-14-2004, 08:13 AM | #203 | |
Quote:
Your reply mail was surely lost :( , i didn't recieve it. i admit that version was pretty bad. i added too much overlapped "TQAddTimeJob" and lagged the computer. i'll make a better one, after my final-exams. ^_^ |
| 06-14-2004, 11:08 AM | #204 |
Nice, but best so your work is not lost, contact us first so we can talk about what to change and how the change is applied. Just changing code at many places in the code makes it hard for us to actually merge the code in the end. |
| 06-14-2004, 01:33 PM | #205 | |
Quote:
ok, when i finished i'll email you and attach commom.eai, i will mark all changes with "//" :D |
| 06-14-2004, 08:52 PM | #206 |
That is not quite what I meant. I could easily find out what lines you changed by doing a file compare. But that does not help when you change stuff everywhere like in the common.eai you sent me. The important thing is to talk to me before doing any changes. That way we can coordinate so I do not change things at the same place in common.eai like you and I might know more about side effects of certain changes that might not be obvious when reading the code (I do not put in comments that much). That is an important step in keeping the code maintainable. If two people work on the same code without coordination then it will either make the code age faster or the effort to merge the changes is higher. As an example in the code you sent me you added many RemoveLocation and similar stuff at locations that do not need that all over the file. Or you added a sleep in the RetreatControl function that belongs to the job thread. Doing so delays all other jobs (they run all in the same thread). Best you contact me on ICQ or IRC. Then we can talk better than on the forum. |
| 06-15-2004, 09:00 AM | #207 |
Question. How would i go about disabling units from micromanaging themselves when they are hit. I.e. when they walk about 2 steps when hit then try to attack then get hit so walk 2 steps away again. This continues and just gets them killed. Is there a condition somewhere i can set or is there a AMAI code causing this micromanagement to occur. |
| 06-15-2004, 04:33 PM | #208 |
That is actually a hard coded behaviour but luckily it is one that can be switched on and off. That is the native to control it afaik: native SetUnitsFlee takes boolean state returns nothing |
| 06-15-2004, 08:30 PM | #209 |
Great. Thanks, at least somethings gone well today. |
| 06-16-2004, 12:23 AM | #210 |
@AIAndy: My icq is 57443335, but i havn't use it a long time, i prefer to use msn :) |
