My first giant NITPICK on the game (Military Power)

I've kinda noticed the same thing but I kind of think the generic military rating is just that - a generic rating. I would suspect (and hope) that the AIs use more in-depth analysis to determine on-the-fly power ratings for decision making.

There are other factors to consider beyond tech.

For ex, in your situtation you had the ideal attack and defense types to slaughter your opponent - something that doesn't reflect in the military power ranking at all (and it'd be hard to include as a global measurement since it varies).

An empire could have a high military rating but if they're actively at war with multiple enemies their forces are likely stretched and split up, meaning less powerful opponent could successfully strike them.

(Interstellar) geography is a factor for me. If I'm in a corner I can get more out of my ships than if I was in the center and had to defend a wider perimeter. I might rate lower but my forces are better concentrated and better poised to be shuffled to thwart invaders, etc.

And I'm sure there's other things to consider such as risk vs reward. The AI might decide to attack a theoretically stronger opponent if it has ideal attacks vs defenses, the enemy is already at war, and there's a fat planet with marginal defenses in reach. The AI might determine that it's worth the risk to attack such a target if obtaining the target would significantly boost its position.

Anyways - I kinda think the AIs go off of more than just military ranking.

The AI does take more into account than the graph. They will even adapt to your ship design if they see that they can not beat you reliably.

isn't the power based on its total attack, defence and hp, or some combination and calculation of such? not how high tech the ships are or how many you have I think.

I had an enemy civ declare war on me because the graph told it it had 10+ times as much power as me.

Needless to say I shredded its fleet and started marching from world to world.

Then it crashed with corrupted saves >.<

But yeah this is a big problem - you can be utterly destroying them and they can still think they're much more powerful than you and be unwilling to come to terms.

The real problem is that it doesn't take into account Logistics. IMHO Logistics is perhaps one of the largest factors in winning any given battle. Higher logistics fleets allow lesser tech ships to defeat much higher tech ships. Sure you take casualties but your ships are a lot less expensive.

I think Dearmad is pretty much on point. The AI will rely on the graphs in the absence of other info. I think the only factor taken into consideration is number of ships, with hull size perhaps factored in to some extent.

I know the AI does espionage but I am not sure how they use it beyond stealing tech. The AI will continue happily building ships that have no chance until they are shown up in battle. In the situation described above you could finish them off before they could react.

If they are not destroyed they will come back with the right attack and defense to counter.

Do the other AI factions learn from what you did to another faction? I am not sure about that

These points are all true and valid, but... I think it isn't that easy to find a solution.

One thing you're not considering, Dearmad, is that you are speaking about his military value compared to yours. Roc, Sissor, Papers... what I am trying to say is:
Imagine three civs A, B and C which all have a military rating of 100, but because of the different approaches (ship sizes, logistics, weapons, shields...) A could beat B, B could beat C while C is superior to A - it's not possible to calculate some absolut military power ratings.
Without making the 'military power' graph and the calculations behind it endlessly complex, it will always be more or less incorrect. We (and the AI) have to live (and die) with it.
I agree that there is room for improvement, but only to a certain extend. Making the formulars more complex and trying to consider every variable is usually not a good idea. Maybe just weighting big ships more or something like this could help.

The real problem is that it doesn't take into account Logistics. IMHO Logistics is perhaps one of the largest factors in winning any given battle. Higher logistics fleets allow lesser tech ships to defeat much higher tech ships. Sure you take casualties but your ships are a lot less expensive.

I'd have to agree... the best way to compare military power (As someone who does not yet have the game) would be to compare effective fleet strength. 4 laser 1 vs 2 (missile 4 shields 2) and then compare total # of fleets. It may also be a peacetime problem - most of the examples I've seen from Brad are at-war AIs. Perhaps an at peace AI has trouble figuring out who his biggest threat is and adapting it's strategy accordingly.

It also seems that this leads the AI to build hordes of lousy ships, and is happy because its military power rating is high.

Just wanted to add that even before I saw this post, I was thinking of the same thing in my game. In the one I'm currently playing, I was deathly afraid of my neightbors despite having decent units because the charts showed them with huge millitary scores. Likewise, they were haughty and dismissive of me, and a few eventually declared war on me based on these same ratings. I was able to take several on at once and wipe the floor with the, because I had fewer, but much better ships in good fleets. Even as I was overpowering them my millitary score remained considerably lower until I wiped out most of their forces, and they wouldn't hear of peace until their score fell low enough.

I have no real ideas on a fix, I'm afraid, but I'm sure you guys will find the best course of action. Great game overall.

I've noticed this as well.

If your opponent has 100 ships, each with 0 defense and 1 attack (beam) and 10 hps, and you have 10 ships with 2 (beam) defense, 3 attack and 26 hps a piece (and a logistics rating of say, 12) you're going to have about 30% of your opponent's "military strength".

You could of course hammer your opponent into the ground with your 3 fleets. The Vor Collective found that out the hard way last night when they decided to double team me with the Drengin. I sent one fleet to Vor, one to Drengi, and kept one at home.

Needless to say if you knock their military strength down a few notches, they will sue for peace...even the Drengin.

Still, I think the military strength gauge should have a heavy logistics component, as in the difference between your logistics scores should be a multiplier. If you have logistics of 12 and your opponent the starting 6, you should get a 6x multiplier on your "military strength."

This of course would make the AI pursue logistics a lot more than it does currently, which I think it needs to do. It's good at production, optimization of attack vs defense type, but lousy at logistics.

Imagine three civs A, B and C which all have a military rating of 100, but because of the different approaches (ship sizes, logistics, weapons, shields...) A could beat B, B could beat C while C is superior to A - it's not possible to calculate some absolut military power ratings. Without making the 'military power' graph and the calculations behind it endlessly complex, it will always be more or less incorrect. We (and the AI) have to live (and die) with it. I agree that there is room for improvement, but only to a certain extend. Making the formulars more complex and trying to consider every variable is usually not a good idea. Maybe just weighting big ships more or something like this could help.

i did think of this- that is precisely why I made the suggestion I did in my solution part of my first post.

In the cxase above all THREE fleets would average out to about 50% military power rating, since A kicks B that means he gets a 100%, but A vs. C = 0% for him.

The crux of my idea is you take a statistical look at a player's fleet (I used an average to make it simple, but there are other ways of quickly stat analyzing a fleet for even more accuracy in the representative sample). Then you match off that "typical" ship with the other statistically typical ship of other fleets and get a rating of how effective it is. You plot THIS number as the fleet strength. The calculations are trivial, and if done at the start of every turn or whenever would hardly any CPU overhead.

An even better way would be do keep the data race-specific, so that fleet A compares itself to B, C, D, E, F. The AI would then have a MUCH clearer idea of it's strength vs. it's specific rivals. you could graph your fleet strenght *in relation* to a rival over time this way. Truly trivial calculation given that they are done only once at the start of a turn.

remember, I'm not talking about pitching a pseudo battle between every fleet, but looking at a fleet in a slightly more sophisticated statistical manner than SUMMING it. SUMMING = no information in this case. Worse, it means the AI has the WRONG idea about its rivals. This is a huge Achille's heel for GC2's AI.

Arg. I just noticed one more thing about this. I think someone above *may* have mentioned this already: The AI will not sue for peace in a realistic fashion now as well because it has an inflated idea of its own military might. It makes the AI very vulnerable to the gang-up part of the other AI's too and just cranky.

Was this the intention of the Ai designer? I can't believe it is so- this seems like a leftover piece from GC1 when ships were much much less tactical in design and so this sort of situation never occured.

I have to say the more I play the more disappointing this element of the game is. While a LOT of the diplomacy cheese has been fixed (like BC/mo. deals where you had a BUNCH of low credit/mo. deals crippling the AI's and bankrupting them), THIS new rub is about as bad for me- it really crippled the AI. I'd almost call this one a design flaw..

I don't think we can assume the AI bases all its military decisions on the military graph. But an official response on it would be good. You especially want the AI to have an accurate view of your military in trade negotiation so they will be properly frightened even if you have fewer ships yet more powerful ones.

Based on some of the journals where the AI difficulty is high, I believe they do take into account a lot more information when making ship design choices and tactical choices.

Yes, but those "tactical" choices won't make a whoop of difference if I'm halfway through destroying an empire with my clearly superior ships which the AI didn't properly take into account. Without an economy to build his ships, his blueprints will sit unused.

The AI NEEDS to know when it is truly outmatched and then try to design ships that help avoid this. Right now- the fact is it can't.

Brad, or anyone at SD, you have no thoughts on this? Really!? Even to say "take off you hoser," maybe?

It really does make the AI a terribly bad judge of how to play the game.

Hundreds of 1/0 ships vs ships with defense sounds like an extreme case scenario. You should play a few more games before you jump conclusions, I won my first game (tough=intelligent difficulty) were the yor controlling much of the galaxy with hordes of crappy ships and me using better ships + military starbases in a "surprise" attack to beat him and thought, hey the ai isn't that hard... But I noticed he had some ships out in the end that could easily beat me...

My point is, you sometimes actually underestimate the ai (see the story in the "this sucks" thread), that's very cool.
https://forums.galciv2.com/?ForumID=162&AID=103225#805164
So you shouldn't be so hard on the ai for sometimes doing the same and underestimating you.

It is true that the ai should use some innovative creative algorithms for doing decisions other then just looking at the military graphs, like comparing enemy ships he can see to his own. Running them in combat simulations in the background while you play, this is what an ai would do best. I'm sure the ai alredy does this, but you have to realise that the hardest part of programming the ai is giving it a sense of the big picture. Everything is just a bunch of if this then that statements that has to have precisely defined criteria, you can't just tell him in plain english some general guidelines like a human. Maybe you could take the whole thing a step further and have the ai try to see what would happen in a few turns by having a parallell copy of the game running with the human player replaced by another ai, and then compare results from different ai decisions. Similar to save/loading by a human player.