Players Beyond Tough?

In his recent journal entry, Brad presents a lot of interesting information about the AI, his approach and philosophy. Apparently, Stardock set a very ambitious target: that only 5% of players can reliably beat the 'intelligent', no-bonus AI.

One of the reasons people like Galactic Civilizations II's AI is because it doesn't cheat. But that generalization is only true to a point -- it doesn't cheat until AFTER "intelligent" (aka "Tough" in the overall difficulty slider). Few players can beat it at that difficulty level and that's where meticulous care was put in to make sure that it's playing exactly the same game as the player. Beyond tough, all bets are off. At that point, we're talking about players who have either found an exploit or are simply incredibly good at strategy games and our goal then is to make sure they have a challenge.


Now, maybe I have a very skewed perspective, but I would have thought at least 25% of people who have some experience with the game play at above Tough. My guess would have been 50%. Maybe I am way off here? Are there actual statistics on this sort of thing? I just seems to me that the AI, while good in several respects, is eminently beatable on intelligent just using standard strategy game tactics.

How many people on the forums play above tough? Is it fair to say that they are exploiting, or just incredibly good at strategy games?

What is exploiting anyway? Is it getting Eyes of the Universe? Picking up the government techs? Setting up a military invasion to quickly capture several worlds? All those are strong advantages that the AI doesn't seem to duplicate.

I fully appreciate the non-cheating nature of the AI, and how difficult it is to make a machine do something that people think is 'natural'. But surely that means that we acknowledge that the AI has many weakness still. It seems a bit off to say that someone who can beat the AI consistently must be some kind of strategy game savant.

22,266 views 24 replies
Reply #1 Top
So, any takers? How many people can reliably beat 'tough'?
Reply #2 Top
I play on tough. I'm not sure what my win/loss ratio is. It ends up being I either completely stomp the comp or get stomped very rarely do I end up with a really fun match where its really a competition between equal powers. When I play above tough I also get stomped very soundly yet when I play below tough its freakishly easy to win.
Reply #4 Top
I've played 3 games with all nine computers set on Intelligent or Gifted in large universes. Seemed to be winning all of them based on population and planet totals before I'd quit. Practicing my strategy for Release 1.1.

Realize. I used to consistently win on Challenging in Gal Civ 1. Initial start location and abilities can play a large factor in long term success.
Reply #5 Top
Before the beta patch, I was able to reliably beat tough (about 80% success rate.) This was without using the galaxy reset button over and over again until the perfect setup came up. I believe a lot of the people that beat it on the harder levels keep reseting the galaxy until they start in a favorable position. For me, fighting from an unfavorable position is the fun part.

Oh yeah, and my success rate for the beta is 3 losses to 1 win. That might be just cause I'm still getting used to the new dynamics, or that might be because it's a lot harder. Personally, I think it's a mix.

-Dewar
Reply #6 Top
Just have a look at the Metaverse, people are playing above tough.

Given a not too unfair starting position, at least crippling difficulty isn't much of a problem. The 1.1 betas seem to be more difficult than it was before though.
Reply #7 Top
you know what would be great? a way to customize the diffculty. EG: sub-normal too easy, but normal too hard? give the ai and extra 25% econ or give him basic algos. It would reduce teh frustration level so much and it'd help us go up at out own pace.
Reply #8 Top
I win consistently on tough, but then again I'm a total restart whore. If I don't like the look of things I instinctively regenerate the map.
Reply #9 Top
you know what would be great? a way to customize the diffculty. EG: sub-normal too easy, but normal too hard? give the ai and extra 25% econ or give him basic algos. It would reduce teh frustration level so much and it'd help us go up at out own pace.

If I remember that right, Brad himself would like to have that too. So it's only a matter of time, I guess.
Reply #10 Top
The highest difficulty I have won is painful, I tried a game at suicidal and lost big time.

The game I like to play is one that is difficult enough to be somewhat challenging but winable without using the tricks like cargo laser interceptors and constant war. That is just not my game style. After playing the suicidal game I have strong doubts that I could ever beat the AI because to win you must jump on them early and stay on them. I don't see how anyone could beat the highest difficulties as a builder/culture player. If anyone is winning at the highest level without using a wrecking ball I would like to hear about it

I want the AI to be tough enough to not roll over and I want them to force me to fight them but not so tough that they overwhelm me with huge financial advantages. The game at painful was just that, painful and I won't want to work that hard very often.
Reply #11 Top
I have won at the higher levels. I'm not playing the beta but according to the release notes for 1.1 the strategies I was using are going to have to change. Nothing wrong with that.
Reply #12 Top
You could always to to the metaverse and filter on difficulty. That'd give an indication. Not definative, but an idea.

So out of 1447 players that have submitted with a score above zero at any difficulty...

244 of them have played at Tough,
163 at Painful,
119 at Cripplling,
53 at Maso,
6 at Obscene,
41 at Suicdal.

For a total of 626. Or 43.26%

Lies, damn lies and statistics...

The above reperesents players who have submitted at least one game to the Meta with a score greater than 0.
Reply #13 Top
I don't see how anyone could beat the highest difficulties as a builder/culture player. If anyone is winning at the highest level without using a wrecking ball I would like to hear about it

Actually in my experience playing successfully on the very high difficulty levels isn't about jumping on the AI as early as possible. It's the contrary: You keep them off your back with high military and bribes and build up your assets. Then you either sneak- and blitz-attack them or take them over with cultural resources and starbases. (This is without the alternative victory options, by the way, it'd be easier with them.)
I won on crippling today (1.1-Beta2) and had only a single war that lasted just one turn (caused by a random event) - other than that, it was building up and culture flipping everyone. I'm not saying that it is easy, but it certainly is doable.
Reply #14 Top
Looking at people who post on the boards or submit scores to metaverse is clearly selecting for the more experienced strategy gamers. I guarantee you that among the general game-buying population the numbers are nowhere near as high as metaverse indicates.
Reply #15 Top
My first game was a small universe on beginer. Increased the difficulty every game. Now I am almost done with a large map on crippling. To be honest, I haven't found the AI tough to beat yet. I don't restart, and don't use the cheesy transport raider either. If the final version of 1.1 isn't available before I finish this game, then I will go to masochistic.

The tactics I use are nothing special.

1. Tech whore like crazy, especially early on. The AI rarely refuses to trade a tech for something. And if it wont, then there will be a minor that ends up with it in a few turns. I personally don't care if I get a bad deal from the AI. Eventually, it will settle out and I will be No.1 in tech. Happens every time. In Civ, the computer guarded techs very carefully. I guess this could be considered an exploit and perhaps that is why there will be an option in 1.1 to dissable tech trading.

2. Always attack a weak neighbor as soon as possible. In Civ III, this could be risky. An unlucky string of attacks could throw everything off. And if you ran out of attacking units, there is nothing stoping the AI from comming back at you. Basically, it didn't need transports. In CalCiv II I will always kill a minor if it ends up in my territory. Once I have Planetary Invasion it is all over. And there really isn't any chance to it either. For example, a 10,000 population planet may take three transports. So go one higher for good measure. And this is just using conventional warfare. If you get unlucky, use a different form in a desperate situation. Also consider advantage factor when determineing No. of transports needed.

3. Suprise the computer. It isn't hard. Just mass up your forces. concentrate on transports, because they are really what you need. Make them fast, and make their escorts very offensively powerfull. Go charghing in, bypass any enemy fleet and go right in for the planet kill. Once the planet is yours, it is relatively easy to keep because you will most likely be able to pick off the counter attack transports. Again, speed is important here. Several times, I have completely killed an unsuspecting AI in just one turn. It takes a lot of planning, but it isn't hard. And then when there last planet is dead, all there forces vanish. So again, why bother attacking them. Just another little pointer: sell all your available techs to the unsuspecting target right before you kill him. That way, you get a lot of his money too. Ya, it is cheesy.

4. By always attacking the weakling, you get a critical mass. Don't really know what else to say here. Bigger is always better in 4x games IMO.

5. Use the diplomacy to your advantage. Definately don't overestimate the computer. He is not that dangerous. If a powerful empire declares war on you then even the odds with diplomacy. Do whatever it takes to get other AI's to attack the aggressor. I have found it really isn't that difficult. Weak ones will probably refuse, but you can most likely at least get one to join in on your side.

6. Speed is a killer. The current AI is too slow. I recognise this is being looked at.

7. In my oppinion, basic planetary imporvements are the best. Capitols are good, but I don't bother with galactic wonders or trade goods. The goods I can get by tech whoring or conquest, and the wonders I can get by taking them by force. All that production power is better spent on ships. Just my oppinion.

8. This isn't my idea, but it works great. Pack a transport with as many sensors as possible, then put one next to a planet in every system. They are cheap to make and keep you from getting suprised. Ya it is a transport trick, but at least I am not killing things with it.

9. In my oppinion, economy, moral and diplomacy skills are the best. I concentrat all my points on these three. Then I use the populists as a government. Someone did a thread and rated the governments. I was suprised when I saw that the populists had the highest numerical rating.

I think that with practice, anyone could beat this on suicidal. I have been playing these civ type games for over ten years, and I haven't found this one to be horribly difficult. These are just my takes on some possible strategies and are difinately not to be taken as gospel. I am also not doging this game in any way. Just a friendly commentary really.

I am confident that 1.1 will be an improvement.
Reply #16 Top
I've consistantly beaten the AI on Crippling in 1.0x without a lot of effort before switching to the beta. In 1.1 I've been playing on painful, and I think it is somewhat more dificult than the crippling in 1.0x, but have not actually lost a game either, so... the jury is still out. I definitely don't think that the "intelligent" AI is hard enough to beat 95% of the player in its current form though. Maybe 80%...

I should also mention that I have beaten the game on suicidal, but that was because of abusive restarting and save/reload deal, and wasn't a lot of fun except just to do it once for kicks. Normally I only restart so that everyone is spaced fairly far apart and give everyone a chance in the game. I have not bother with the metaverse, mainly because it seems like a fancy scoreboard for bragging rights, but also because I'm also rather annoyed at the advantage that the player have because of how powerful attack first is in the game, and have modded to increase the HP 3x to give defenders somewhat of a chance.
Reply #17 Top
So out of 1447 players that have submitted with a score above zero at any difficulty...

...

For a total of 626. Or 43.26%

Lies, damn lies and statistics...


Ok, this is not a very accurate.

Firstly you get a score whether you win or not.
Secondly what you are calculating there is the percentage of submitted games that where played above intelligent difficulty, not the percentage of games that where won when that difficulty was played.
Thirdly the metaverse isn't an accurate sample space, as the people who play the metaverse are generally better than those who don't.
Fourthly, Brad actually said:

In the development journals, we talked about the goal of having the "intelligent" AI being able to beat 95% of players. I think we succeeded.


So that is *able* to beat, not *will always beat* 95% of players. So basically means for those who have, for example, an 80% success rate on intelligent (or perhaps used to, but don't anymore after some practice) then they are a part of that 95%.

I don't mean to shoot you down, but it was important to note Especially the fourth point. We can't just ask "have you ever beat AI at intelligent?", but rather we need to ask "Has the AI ever bet you, when you where trying pretty hard?" And we need to ask a fair sample of people, not just the ubers on the metaverse
Reply #18 Top
I played 3 games on tough and it has been anything but. This is without research whoring, but I do use scout cargo ships and single laser cargo ships, but that is so obvious it's hard to imagine it's really abuse. The AI needs to do a MUCH better job of protecting and escorting transport ships. I'm hoping 1.1 will up the difficulty because I hate games that force the computer to be given an artificial advantage in order to defeat them.
Reply #19 Top
I've beaten all intelligent foes several times. Honestly it wasn't that hard (although I will admit I purposely start away from any civs because my strategy makes me weak early game). Haven't tried it on 1.1 though; will let you know.
Reply #20 Top
I always play on at LEAST Tough. I'm also a little surprised by the statistic, but I'm also relatively high on the intelligence scale and I asume many of the posters on this board are as well. Strategy enthusiasts usually are. So we might be a skewed sample.

If you include the general public I'm sure the success rate would be much lower.
Reply #21 Top
I play only on tough. I hate when the AI cheats and I hate the AI having hanicaps. I'll either get stomped or stomp the computer, but if I get waste its just bad tactics on my part at the start.
Reply #22 Top
I found the computer easy to beat at all levels. Attack first, go after planets not fleets, and target all his planets at once. My wars usually last 1 turn. Then it's on to the next victim.
Reply #23 Top
It's still all lies, damn lies and statistics...
Reply #24 Top
The ai rarely uses more then 1 type of weapon so i get some defense components and a 2nd weapon type and it becomes really easy from there.

suicidal is only rarely hard, like when my colonys flip to some minor race like the I-league