Sorian vs DuncanE AI

I was wondering, when the Sorian AIs are pitted against the Duncane AIfixes, which AI tends to win the most games? I know this is a loaded question, with lots of variables (such as map size and style, as well as the specific AI being tested), but I would be interested if somebody has actually taken the time to test the AIs against each other under different circumstances (and if they are sufficiently nerdy, did enough tests to ensure that the results are statistically significant, with p values and whatsuch). Often I hear that one AI is better than the other, but if I were to see an excel spreadsheet for exampke detailing the results of several tests, it would be more convincing. Results of tests between the Sorian AIs and the stock AI (not the DuncanE AI fixes) would also be interesting.



I would test it myself, but I have a very limited amount of time I can devote to playing computer games in general, and I would rather use that precious time playing rather than testing. I'm just interested in finding the AI that is toughest to play against.|||Well the games i have played with both the Sorian and The Duncan Ai, has the Sorian easily destroyed the duncan ai, both in cheat mode and "normal" mode.|||Sorian's typically wins, but it's not really a fair comparison, Duncane's AI is mostly concerned with fixing bugs and design flaws, Sorian's delves into adding new functionality to give it an edge.|||Falcrack,



You ask a very loaded question



I have obviously watched many AI vs AI games while testing my AI fixes. Currently Sorians AI generally does better if the right AI type is selected, buts its very close. On small maps it often comes down to commander behavior (which is much better in Sorians AI)... on large maps it usually comes down to the speed that the AI sets up expansion bases.



I do feel that while Sorians AI is better at longer games with more players, I have play tested my AI changes more vs 1 human player for short games so the patched AI does a bit better in that situation.



Anyway... I would still be interested in any specific matches players report.



Please post with the following info:



Map name

AI type (Rush, Air, Adaptive etc)

AI mod and version (Sorian, Stock, Duncane)

Game type (Assassination, Supremacy etc).

Factions for the AI

Cheat settings

Result (i.e. which AI won).



e.g.



Williamsons Bridge

Sorian (1.3.4) vs Duncane Land (1.9)

1.5 Build speed, 1.5 resourses, Omni on

Assassination

UEF vs UEF



Cheers

D|||Map name: Fields of Isis

AI type: Sorian AIx (No suffix), Duncane AIx Adaptive

AI mod and version: Sorian 1.3.4, Duncane 1.9

Game type: Assassination.

Factions for the AI: Sorian UEF, Duncane UEF

Cheat settings: 1.5, Omni, unlimited expansion bases

Mods:

00_AIFix.scd

00_cheatoptions.scd

01_aisupport.scd

02_sorian_ai_pack.scd



Result: Duncane won by Nuke.



Details: Both AI's started off strong beating each other's score. Around 5 minutes, Duncane had a very nice air advantage, resulting in efficient bombing runs. Around 10 minutes, Sorian took over Air superiority. Around 20 minutes, neither had superiority. Around 25 minutes, Duncane had a Fatboy. Around 30 minutes, Duncane tried to build a Novax which was nicely being build under a T3 shield. Despite this, Sorian AI managed to drop about 6 (!!!) loads of transports consisting of T2/3 ground units. Disabled the shield (Not destroyed) and took out the Novax just before it was finished, Sorian transport army defeated. 36 minutes; Duncane launches it's first Nuke, and there wasn't a Anti-Nuke waiting for it, resulting it in killing Sorian AIx. It was a massive explosion, gametype wouldn't of mattered, Sorian was dead.



http://img503.imageshack.us/img503/9759/forgedalliance200806161hv6.png



Edit: Accidentally used [img] instead of [url], causing the forum layout to go boom.

Edit2: Funny to see at the scores that Sorian AI has a much higher score, even AFTER Sorian's supcom was dead.|||Yea sometimes Sorrian is a bit slow on the anti nuke.|||In my limited AI vs AI match ups, I have noticed that Sorian AI will go for TML snipes whenever it can, and as mentioned before, Duncane AI doesn't seem to react to this properly. Even with a fair amount of warning, the ACU just stands there and get sniped.|||Couple of questions... SyDaemon



What maps does the TML snipe work on? Mostly 5km maps I think?



Secondly... what do you do against it other that build TMD and shields which should be happening already? |||duncane|||SyDaemon,



You didnt sound hostile at all.. Im always looking for feedback and I pay particular attention to small map games as this is what I use the AI to play against most myself.



Theres nothing in the com behaviors to avoid TMLs because I assume the com is always well protected by TMD's... that's what I have seen in most games.



Do you have some example maps and AIs used where this commonly happens? Also can you confirm its the latest of both mine and Sorians packs?|||duncane|||I find that different AI types or even just different factions or start locations can make a difference on the same map.



Also sometimes one AI just gets lucky at the right time... the commander moves in the right direction or the the engineers get a PD up at just the right time... that kind of thing. And suddenly the whole battle turns the other way.



Then 10 minutes later it leads to one AI getting a T2 fac first or a TML quicker or a nuke up before the other has finished its SMD. And it all comes back to that lucky moment early.



I find that Sorians and my tweaked AI are fairly close now in the early game. Sorians does have the edge in experimental defense and group of T4 attack groups, but I would challange any player to vs either duncane or Sorian AIx at 1.5 cheat (or even 1.25) and win.

|||Yeah, Duncane and I have been sharing code for a while and now most of the base unit behaviors are pretty much the same. The main difference now is the overall features of each AI and some building priorities.