DyDo 2.2.2, AI for skirmish games and challenges

[Sonsalt]

Your AI is incredibly good and I love playing against it.

However there is one thing I noticed, that could be improved.
The AI is mostly using the same routes to engage, I was wondering if you could add a function for the AI to determine the Strength of defenses along a path and make the AI choose another one if the threat level is too high.

Lets assume a base has 2 entries but 1 of them is well defended while the other one is mostly unsecured. The AI should Calculate the Primary route to target and if there are defenses on the way, it should initially disable this route and check for another route. On the second route there are also defenses, but less than on the first route, again the AI is also closing this route and is checking for another (one with no defenses). However in this scenario there are only 2 routes available. Now the AI goes back to the second path as it's threat level is the lowest and moves its units to this position.
With this script the AI will always try to avoid defenses for as much as it can.

the best way to check this out is gridlock

Another idea on how to deal with defenses is to have the AI calculate the type of units it needs to overcome them. However this requires a more dynamic research and build order. If the enemy is relying heavily on Defenses the AI must go for Arty (when concrete) and flamer (on bunkers).

Engaging a defense building with the wrong type of weapon can be quite deadly and is at least for the moment the most common reason the AI looses against human players!

Apart from that I noticed that the AI can't collect research boxes nor oil barrels. If possible you could add a script that automatically adds a research item from the opponent to the AI once such a building has been destroyed if it can't collect the items.
A workaround for the barrels could be 100 oil per destroyed oil platform. This is not exactly collecting but compensates the disadvantage and evens the ties.

[Rman Virgil]

.

Love the work you have done with both DyDo & DyDo Scav a.i.s and I am building a project around them called 4c-ManGodAi4X.

I'd like to share a behavior anomaly I've noticed just recently which does not seriously impact what I'm doing but which may be of interest and could very well effect others game play detrimentally.

When testing my work me vs. 3 DyDo in a locked team, the Path-walking gets to stalling, hurky-jerky and Borgs slipping in place going nowhere for extended periods (along with Tanks but since wheels and threads are not animated you don't see the "slipping" like you do with Borg Legs... they are just "stalled").

This does NOT happen with No locked team,... The path walking is as smooth and optimal as can be. Odd. I do not have a clue as to why this would be.


- L8r,

.

[Sonsalt]

After some tests I can confirm that the upgrade behavior for the power plant and the research building is indeed related to the limitation of the research buildings.

If I limit the research buildings the AI won't upgrade its research facility any more even though it still researches the upgrades. It also rarely upgrades the power plants.

Since I very much like to limit the research buildings for gameplay reasons, I am wondering what can be done in order to solve this bug?

Do you think it will help to limit the total number of research buildings via mod?

Update: Limiting the research building number via Mod does not work either.

[DylanDog]

Why crashs the Game, if i set the AI Strengh in the VLO to random (0)? If i set the personality to random, the Game will work and the AI work faster with Strengh 5.

DylanDog, pls check your Mails, i have you send some interesting Data

Sorry I have not been able to replicate this, my DyDo is not crashing with level=0...I will run some more tests.

[DylanDog]

However there is one thing I noticed, that could be improved.
The AI is mostly using the same routes to engage, I was wondering if you could add a function for the AI to determine the Strength of defenses along a path and make the AI choose another one if the threat level is too high.

Lets assume a base has 2 entries but 1 of them is well defended while the other one is mostly unsecured. The AI should Calculate the Primary route to target and if there are defenses on the way, it should initially disable this route and check for another route. On the second route there are also defenses, but less than on the first route, again the AI is also closing this route and is checking for another (one with no defenses). However in this scenario there are only 2 routes available. Now the AI goes back to the second path as it's threat level is the lowest and moves its units to this position.
With this script the AI will always try to avoid defenses for as much as it can.

This is not an easy thing to develop. It is almost impossible to discover via AI script the base entries. One way could be to loop through gateways point and check defences around them but not always the gateways are placed on base entrances.

Another idea on how to deal with defenses is to have the AI calculate the type of units it needs to overcome them. However this requires a more dynamic research and build order. If the enemy is relying heavily on Defenses the AI must go for Arty (when concrete) and flamer (on bunkers).

Another difficult thing to implement mate! The problem with arty is that, yes I could let DyDo build some arty to defeat the enemies defenses but the arty has to be defended itself from both VTOLs and droids counterattacks (or Anty-arty attacks using CB towers). Defending artillery away from AI' s base is difficult. In the 2.2. version I am working on DyDo builds more artillery in General (expecially in T2 games) also to defend derricks, in small maps artillery near the derricks could also attack enemy defenses. I see if I can implement something in the direction you suggested.

Engaging a defense building with the wrong type of weapon can be quite deadly and is at least for the moment the most common reason the AI looses against human players!

DyDo 2.2.1 is very intelligent against enemy droids if not enemy defending structs are in range: this means that DyDo flamers attacks hovers first, then cyborgs then tanks just to give an example but if a struct is in range DyDo orders ALL droids to attack it first. I will change this and make it more effective. this required a big change in the droid micro management but should not be difficult to be implemented.

Apart from that I noticed that the AI can't collect research boxes nor oil barrels. If possible you could add a script that automatically adds a research item from the opponent to the AI once such a building has been destroyed if it can't collect the items.
A workaround for the barrels could be 100 oil per destroyed oil platform. This is not exactly collecting but compensates the disadvantage and evens the ties.

I am not sure other players would like this to have in DyDo as default because this could be considered as a cheat, but I could add a chat command like "/set cheat on" and then DyDo activate this new feature.

[DylanDog]

.
Love the work you have done with both DyDo & DyDo Scav a.i.s and I am building a project around them called 4c-ManGodAi4X.

at which stage of development are you with this map?

.
This does NOT happen with No locked team,... The path walking is as smooth and optimal as can be. Odd. I do not have a clue as to why this would be.
.

I am sorry but I have no clue neither...maybe a warzone bug and not a dydo one? is this happening with the standard AI too?

[DylanDog]

Since I very much like to limit the research buildings for gameplay reasons, I am wondering what can be done in order to solve this bug?

I think I would need to modify the AI code A LOT for this. The problem is that I cannot check the limits with the AI script, if this would be possible then the bug fix would be much easier.

[Rman Virgil]

.
Love the work you have done with both DyDo & DyDo Scav a.i.s and I am building a project around them called 4c-ManGodAi4X.

at which stage of development are you with this map?

Due to re-tooling I'd say 90% done with cartography & about 30% done with my new BOSS stageings all over were I am in part combining some Nexus forces with Player 7 Scavs.

.
This does NOT happen with No locked team,... The path walking is as smooth and optimal as can be. Odd. I do not have a clue as to why this would be.
.

I am sorry but I have no clue neither...maybe a warzone bug and not a dydo one? is this happening with the standard AI too?

Because I had some other anomaly come up running it as a command line mod I went ahead & replaced the stock scripts in base.wz with yours.... then this other behavior came up.

Tonight I'll put the stock scripts in base.wz & retest to see if this behavior comes up with 3 players locked as a team.... and I will let you know the results one way or another.

The initial reason I ran your scripts direct out of base.wz was because when I was doing it commandline 1 of the 3 A.I. players was not developing it's base after the first min - just stopped cold. I was never able to get a confirmation from anyone else one way or another so I just assumed it was local, did a clean install & replaced stock scripts with yours in base.wz and that particular anomaly vanished... and this other cropped up.

In any case, I'll do some more tests tonight with the stock scripts.

Thanks for your good work & response to my peculiar issue.

- Regards, rv
.

[Rman Virgil]

.

Arrgg... obviously I have yet to get back to this testing. My apologies. Should be able to make time over the next couple days and then report my findings.

- RV

.

[Rman Virgil]

.

K.. did another clean install, ran all stock A.I., same config, 3 A.I. in locked team, plus Scavs activated vs. me and the anomalous behavior described did NOT manifest.

I then replaced all stock scripts with Dydo (both Ski & Scav) in base.wz without any Mods, same config, and the anomalous behavior described manifested again.

I still have no clue but I will keep testing with it and observing to see if anything else comes to the fore that could help explain what's going on and I'll report anything here.

BTW: Any thoughts on HOW I could lock Player 7 to a locked Team of 3 A.I. Players (like with 1,2, & 3) ? Or is that not at all possible without a source code change ?

Thanks, in advance.

Regards, RV

.

[DylanDog]

Hello, just to let you all know I am on holiday since 2 weeks and will be back in 6-7 days. I will then be able to spend some more time on the forum.

[DylanDog]

DyDo 2.2.2 has been released. Enjoy it.

[DylanDog]

.

K.. did another clean install, ran all stock A.I., same config, 3 A.I. in locked team, plus Scavs activated vs. me and the anomalous behavior described did NOT manifest.

I then replaced all stock scripts with Dydo (both Ski & Scav) in base.wz without any Mods, same config, and the anomalous behavior described manifested again.

I still have no clue

me neither...

.
BTW: Any thoughts on HOW I could lock Player 7 to a locked Team of 3 A.I. Players (like with 1,2, & 3) ?

At the moment it is not possible to create an alliance between AI and Scavengers (if this is your question), or at least it did not work when I tried on few challenges.

[guciomir]

Wow this AI is great! 2 research paths are great! I have a few suggestions based just on 1 game (it is best for me to test 2p maps because it gives me a lot of info how AI is working). Advanced bases on. AI flamers rockets.

1. Dydo is using new paradigm bodies. I think it is not necessary because they are similar to Project bodies and researching them takes time.
2. Flamer droids should be do or die (not retreat). In many situations they retreat even though they did not engage enemy and die few seconds after. This is because my weapons have bigger range.
3. Dydo is grouping droids in a strange way. They go from point to point, group there, go back a few steps then procede. Everything is happening under my artillery fire so Dydo is losing units when they are just moving from point to point. It could have fewer grouping points and fewer go-retreat-go-retreat actions.
4. Flamers/Rockets AI builded tons of machinegun tanks and mortar cyborgs at the beginning of the game (no flamers nor rockets). Is it on purpose? I think it was easy for Dydo to develop minipod and flamers (it was advanced bases game) but it chose not to use it in the early part of the game.
5. Dydo had very big defence VTOL group but no attack VTOL group. When I engaged its base with my armored tanks (like 25 tanks with full armour, scourges and twin assault guns), they got destroyed by VTOLs. If Dydo attacked me with such a big VTOL group, my defences would be shattered. I suggest attacking player defences with VTOLs prior to sending ground Dydo forces. Especially attacking player artillery would be nice. Dydo attacked my base with VTOLs only once during the game and attacked only 1 Oil derric. I just rebuilded it so no loss for me (apart for some energy that would be produced)
6. When does Dydo build composite (more armor and hp) technologies?
7. What is your current ToDo list for Dydo? I see most points on your website are already done and it is very nice to read what you are working on.
8. Will you introduce later Dydo paths Flamers+Cannons and Machineguns+Rockets? I do not advise for doing this now because it is of course better to focus on existing paths and make them perfect. But i wonder if this will happen in a distant future.

[NoQ]

they are similar to Project bodies

O rly?