Notes on rev 6012

Describe your experience with the latest version of FreeOrion to help us improve it.

Moderator: Oberlus

Forum rules
Always mention the exact version of FreeOrion you are testing.

When reporting an issue regarding the AI, if possible provide the relevant AI log file and a save game file that demonstrates the issue.
Message
Author
User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 12802
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: Notes on rev 6012

#16 Post by Geoff the Medio »

The effect sets the structure equal to the max structure, based on the value for that max meter determined during the previous meter update. On the first turn a ship is created, the ordering of effects like this is tricky and error prone. An additional activation or location condition to only take effect the turn after a ship is produced might be helpful.

fixman
Space Krill
Posts: 1
Joined: Wed May 15, 2013 5:44 pm

Re: Notes on rev 6012

#17 Post by fixman »

Hello.
Suggestion change the calculation of a rating AI ships/fleet/threat.for Dilvish.
rating = sum(atk)*sum(def) - difficult to compare fleets. If we add/remove the ship/fleet again recalculate rating

my suggestion :
rating formula - the rating on ships not on the fleet, the overall rating - the sum of ratings of ships.
For shields coefficient multiplying shld:
1.0 no shield
1.1 defence grid
1.2 deflector
1.3 plasma
1.5 neutron
1.6 multi

ship rating=atk*def*shld
fleet rating = sum(ship ratings)
system=sum(planetratings)+sum(fleet ratings)

What for? easily:
compare fleets
allocate ships for missions
simply add or remove ships form fleets
consider shields

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: Notes on rev 6012

#18 Post by Dilvish »

fixman wrote:Hello.
Suggestion change the calculation of a rating AI ships/fleet/threat.for Dilvish.
rating = sum(atk)*sum(def) - difficult to compare fleets. If we add/remove the ship/fleet again recalculate rating
Actually, that rating system worked very well for comparing fleets; the most significant failing is that I haven't figured out a good way to deal with the new shield DR mechanic. The AI used some nonlinear relationships that were pretty decent approximations to allow for direct evaluation of a combination of fleets, and to evaluate the effective rating difference between fleets. It's true that for shields acting as DR it's a much more crude approximation which I need to update. I've given a fair bit of thought to that, but haven't yet finalized a new approach.
my suggestion :
rating formula - the rating on ships not on the fleet, the overall rating - the sum of ratings of ships.
For shields coefficient multiplying shld:
1.0 no shield
1.1 defence grid
1.2 deflector
1.3 plasma
1.5 neutron
1.6 multi

ship rating=atk*def*shld
fleet rating = sum(ship ratings)
system=sum(planetratings)+sum(fleet ratings)

What for? easily:
compare fleets
allocate ships for missions
simply add or remove ships form fleets
consider shields
What you describe, except for the shield multiplier, is the same as how the AI first started doing ratings this past fall. I found that a simple linear sum of ship (& planet) ratings does not work well. It would be nice if it did, but it just doesn't, and the changes since then are not going to make it work any better. Although you do try to take into account some measure of the new shield system here, I'm pretty confident that the existing rating system is still a better overall measure. Thanks for trying, though, and rest assured that I am still noodling away on a revision to the current AI combat planning.

Since you appear to know some Python programming, you can always experiment a bit yourself. Often when evaluating alternative approaches to things I'll have the AIs with an odd empire ID take one approach, and the ones with an even empireID take a different approach, and then see how much of an actual performance difference (if any) I can detect in play. If you try the same & find a rating approach that seems to work better than the current one, then certainly please report that.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 12802
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: Notes on rev 6012

#19 Post by Geoff the Medio »

It might now be necessary to consider a fleet not as having a single power / threat rating, but rather a power / threat against a particular opponent fleet. If a fleet has significant damage reduction, it will be much more resilient against opponents with many low-damage weapons. If the opponent has fewer higher-damage weapons on their ships, then the damage reduction will be less beneficial.

I'm not sure the degree to which the current content and AI ship designs support this situation-dependent utility of ships, but it's part of the intent (for me at least) of having a damage reduction mechanic... Ideally it should be adding some additional strategy to ship design and deployment, which hopefully makes doing so more interesting, at least for human players.

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: Notes on rev 6012

#20 Post by Dilvish »

Geoff the Medio wrote:It might now be necessary to consider a fleet not as having a single power / threat rating, but rather a power / threat against a particular opponent fleet.
I agree that to have decent accuracy in predicting a battle outcome, the AI will probably need to assess a specific fleet against another specific fleet. Even just doing a battle single assessment like that is going to be a great deal more involved than it used to be. Among other things, I've been reviewing some "variable sized bins & weighted balls" type statistical analysis to try coming up with an approach that will be less costly than doing a full battle simulation twenty to a hundred times to get a consensus prediction. Regardless, I think the individual matchup is unlikely to be very useful as the primary mechanism for deciding how to allocate a large set of ships/fleets across multiple different deployments. I'm expecting to take a two stage sort of approach, first pass being a more coarse grained assessment to rapidly allocate out most of the available force, and then followup with the more detailed fleet vs fleet analysis to decide where it seems additional reinforcements should be allocated.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

markpi
Space Floater
Posts: 35
Joined: Tue Jan 01, 2013 11:27 pm

Re: Notes on rev 6012

#21 Post by markpi »

hi,
sry for being a bit late with this comment,
and i´m not 100% sure which version i played but it was around this one.
i had 2 games i left for being too powerful ( boring).
in the first one i got an early kraken nest ( around turn 18 ).
i let the first kraken grow in the system and turn 45 i had a great kraken with 25 attack 120 structure.
15 turns later first ai was annihilated and heading to the next with 96 structure left xD
next day i played a game with ancient ruins next to my homeworld, no monster to defend.
i found plasma shield and in turn 60 i had an organic that is close to immortal at this stage.
i´m not sure about the usefulness of the ai ships with 1 weapon 2 platings.
it is a trade of 50% firepower for 33% structure with just a few pp to spare.
actually they harmonize well with ( lvl2) planetary defenses but in space they are inferior.
( i like to mix in scouts with 2 armour- no weapon,no shield - as meat shields they are really cheap )
concentration camps feel to cheap both in rp and pp.
in the german version the player isn´t allowed to build them for obvious reasons^^
these are just minor complaints, and i want to point out that i like very much how this project develops

markpi
Space Floater
Posts: 35
Joined: Tue Jan 01, 2013 11:27 pm

Re: Notes on rev 6012

#22 Post by markpi »

hm i`m sry but i´m uncertain if it was a large snowflake not a large kraken as i didn´t store.
one or both too mighty after the changes^^

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: Notes on rev 6012

#23 Post by Dilvish »

markpi wrote:concentration camps feel to cheap both in rp and pp.
in the german version the player isn´t allowed to build them for obvious reasons^^
?? A german version that eliminates Concentration Camps? Where is that found? What rev is it based on?

Also, re concentration camps -- have you tried them much since they got revamped a month or 2 back? They're really playing with fire now since scrapping the building doesn't automatically stop the population drop, and this also means that the production boost drops sharply also. At the very least they're much more close to being balanced than they had been. I haven't really seen any player feedback on them since then though -- I'd be interested if you have any.
hm i`m sry but i´m uncertain if it was a large snowflake not a large kraken as i didn´t store.
one or both too mighty after the changes^^
Well, if you get either one by turn 45 you've been quite lucky (the odds for that are very very low). It's still a fairly temporary boost though, since they don't heal and drydocks don't repair them. If you're only playing against two AI's then a temporary boost like that can ensure the game for you, but in a small game any lucky break can be like that. Perhaps try a larger galaxy with more AIs.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

Post Reply