DyDo skirmish AI: ver 2.1.1 (21 Apr 2010)

[KukY]

When I save game,and then load it again,DyDo doesn't does anything for few minutes,and then responses "Playing with..."(Or whatever it says...)
I just don't have enough time to really test it(Skirmish game until AI gets LasSat...)
I'll try it later,if I get time.

[DylanDog]

When I save game,and then load it again,DyDo doesn't does anything for few minutes,and then responses "Playing with...".

Yes, this messages are to ensure that DyDo AI is loaded.

[KukY]

But when i load game,DyDo doesn't do anything for few minutes.It just sits in his base,and after few minutes he starts playing,but then i'm much more advanced than him,I captured all oil resources,I'm ready to destroy it...

[DylanDog]

But when i load game,DyDo doesn't do anything for few minutes.It just sits in his base,and after few minutes he starts playing,but then i'm much more advanced than him,I captured all oil resources,I'm ready to destroy it...

When a game starts DyDo uses an "hack" to reveal himself where the oilResources are, until this even has not been cempleted, DyDo is not starting harvesting (but it builds structures). On big Maps the oilHack event can takes up to 120 seconds.

I can try to improve this, added on the to do list.

Are you sure that when you load a saved game DyDo does not do anything? possibly it is researching and you do not notice it...

Thanks!

[DylanDog]

New test version for download on DyDo webpage. It is the DyDo 1.0.5 (20 Nov 20009). I have left the AI play few full games on the 8 players map ThePit 192×192...no bugs appeared. I played personally in sk-Rush2 map (1vs1vs1vs1) but I fear I am too weak as after 30-40 min my position is always too bad. I need better players to test DyDo.

I am quite satifsfied of the improvements made on General 1 (standard behaviour) and General 2 (VTOL AI). Many tank templates have been added and the research paths have been modified to get the necessary new bodies.

I still need to develop better General 3 (turtle AI) and General 4 (Cyborg AI), this does not mean they do not work, it is just that I have to enhance the turtle and cyborg features of those AIs.


Enjoy it!!!!!!!!!

[DylanDog]

I would rather add a script function that adds info about all oil resources to a given player. Seems a cleaner way to do it.

This is not feasible I fear, I have not found any function which would allow to become aware of oilRes not discovered by any AI. The only one available is:
initGetFeature(FEATURESTAT, PLAYER, BUCKET) but only features visible to PLAYER are returned.

[Per]

This is not feasible I fear, I have not found any function which would allow to become aware of oilRes not discovered by any AI. The only one available is:
initGetFeature(FEATURESTAT, PLAYER, BUCKET) but only features visible to PLAYER are returned.

That's why I wanted to add a new script function for this purpose. If you do not feel like going through the C code, I can add it to my todo list.

[guciomir]

i played vs gen1 on startup

1. I think that the research path for DyDo is not optimal. At the moment AI is trying to get everything, it uses cannons, flamers, machineguns, minipods etc. It uses various bodies (project + new paradigm). This is the main reason it is so weak. If DyDo specializes in chosen technologies, it will be much more powerfull. I think DyDo should stick to 2 max 3 weapon technologies and do not build 2 different body technologies but just 1.

Maybe some better than me players could suggest the best in their opinion research paths, so Dydo could use them?

2. Vehicle speed. When you added tracks templates to Dydo, it became weak. The reason is that when there is 1 slow vehicle in a group of fast vehicles, those fast vehicles will slow down to meet the speed of the slowest one. So entire army is extremaly slow all those hover tanks are slow and they die very easily. I propose not to have hover and tracks in the same attack groups. But maybe other players / developers will have better idea.

[Buginator]

I don't really got the time to test this, but how far along is this AI?
Just wondering if this could make it into a beta release, or are there still many things left to be done?

[guciomir]

To Buginator:
I think Dylan is making an awesome AI, but it is not ready for release yet. It will be eventually, but after Per helps him with OilHack and after there are better research paths up to T3.

To DylanDog:
3. DyDo speed research is quite slow (gen1). I was playing only with 5 oil derricks while the enemy had around 15. By the time i was attacked by vtols, I had already lasers and avenger SAMs.

I think there are two reasons for this. Dydo does not use research facilities fully, and the reserach paths are not optimal (as explained in previous post). Dydo should specialize in chosen technologies so it can develop them much much much faster and in a more effective way,

4. I still have to play vs gen2 to see how vtols attacks work, but for gen1 they are weak. Attack should be made by more than 4-5 Vtols, something like 10-20 I think, with multiple targets etc.

5. This one is to experienced players -> please write some tips to DylanDog, his work is awesome and great, but the AI will be much better if it uses the best templates and research paths.

[DylanDog]

If you do not feel like going through the C code, I can add it to my todo list.

I am sorry, my knowledge of "C" equal to zero.

I don't really got the time to test this, but how far along is this AI?
Just wondering if this could make it into a beta release, or are there still many things left to be done?

Let`s say that I need to work a lot on on Generals 2 & 3 and to rebalance Generals 1 & 2. I have no idea how moch it will take, this is depending also on the help from other players like guciomir is doing, he is really of great help. If I let DyDo play against BP or the standard AI it always wins. The AI is at a good stage and has no bugs on 1.0.5 (as for now). I get personaly messages of players satisfied already of this AI but I am going to work on it for the next few months. The objective is to get a strong AI, an AI against which you can play the same map many times without being bored, every game should be a new experience. I have so many ideas in mind on how to improve DyDo...you are also continously improving the game warzone iteself since years, aren`t you? Do not expect me to say in 3-4 weeks: I am happy with DyDo, I stop develop it...unless it comes soon out that no one enjoys my work.

Last thing...I changed the licence to allow DyDo be included in the WZ package but I am not aiming to this really, I leave this decision up to you developers guys, but anyway, DyDo needs some code cleaning before this. But this part is a bit boring and I prefer work on script improvments now.

i played vs gen1 on startup
1. I think that the research path for DyDo is not optimal. At the moment AI is trying to get everything,

Well not really everything but yes, it is researching a lot. I am going to put on the website the researches I have chosen for each Generals so that everyone can see and advise what could be improved. You can obviously also look at the script code, the researches are in the "VLO" file.

An idea could be, for General 1, to create 3-4 possible good research paths which are chosen randomly at the start of the game. What the others Generals concerns, they are tied to certain research paths due to the AI personality itself (VTOL for General 2, Defence for General 3, Cyborg for General 4)

Maybe some better than me players could suggest the best in their opinion research paths, so Dydo could use them?

This would be of really great help, as I wrote, I would need suggestion to ideally have 3-4 research paths/strategies for General1 and the best research path for the other Generals.

2. Vehicle speed. When you added tracks templates to Dydo, it became weak. The reason is that when there is 1 slow vehicle in a group of fast vehicles, those fast vehicles will slow down to meet the speed of the slowest one. So entire army is extremaly slow all those hover tanks are slow and they die very easily. I propose not to have hover and tracks in the same attack groups. But maybe other players / developers will have better idea.

I think I can improve this, actually DyDo builds up also teams (groups) of mixed units between cyborgs and tanks, I am using in the script a function which gives the order to the team to move but I could also give the move/attack order to each single member of the team to achive the best speed for each unit (In some cases DyDo already uses this). The only problem is that sometimes you want to have all units together to keep the team strong. I will change the AI script such that:

- The members are moved as team when they leave the base to attack
- As soon as some enemy is in range, then they act as sinlge unit using the max speed they can reach.

@ guciomir:
1) I have the impression you just test DyDo in 1vs1, please try to test him also in maps with more players and let me know.
2) please keep on testing "Tier 1" Generals 1 and Generals 2. Let`s work to improve those two personalities first.
3) I hope you understand we will never reach the stage of having DyDo acting like an human player, therefore do not get disappointed if I am not going to work on each advice you are giving but as you see, I have already modified DyDo since the 1.0 using your advices.

Thanks a lot for the help

[Buginator]

Fair enough.

I hope you are working with a VCS (version control system) like svn or something like that? Really easy to set up on windows, and then you have a history of all your changes...

[DylanDog]

Fair enough.

I hope you are working with a VCS (version control system) like svn or something like that? Really easy to set up on windows, and then you have a history of all your changes...

No I am not using any but I think it could be a good idea but I have no experience of VCS systems. I will search one and will give it a try. Hopefully I fond a portable one (I keep everything on USBs...)

[DylanDog]

Now version 1.1 released as stable.

It can be downloaded from DyDo website or from the first post of this topic. New stable version to follow not before 2010 but you can anyway download the most updated "nightly build" from the DyDo website anyway...please do it for test purposes and let me know your thoughts.

I have also posted herethe research paths for both General 1 and General 2, please let me know your thoughts.

As always enjoy it.

[guciomir]

Nice! Great work!

I played another 1-1 game, when I have some free time I'll try 2-2.

AI was attacking my base from 2 directions (nice)
When testing it is good to see which general is playing and now the messages are gone.
If you feel that only two generals are balanced, than do not use 3rd and 4th general yet.
Hovers are still slow when in group with tracks. (my artillery was destroying them easily) I do not think you fixed it succesfully
I played turtle , AI was controlling all the map but its only tactic was to send waves instead of 1 giant attack.
VTOLs appeared late in the game, after 1 hour, but it was devastating (I lost )
I think that research paths need to be rebuilded (but lets wait for better players). One thing i am sure is to make power researches come faster, especially if DyDo is dominating the battlefield.

Anyway, it was a very nice game With every week I feel that Dydo becomes stronger.