Greetings friends, today I want to discuss the AI in the latest versions of ISG as well as some select thoughts. Suffice it to say, there is quite a lot to talk about First let me give you some background, I’ve done multiple sessions in 1.05 and used the Impossible/Emperor settings exclusively, the only difference between my sessions is that each one had different production/research settings. The one constant in all these sessions was the marked improvement in the war ships put out by the AI. Though there were some slight variances, in all cases the AI fielded much more robust ships and for the first time in what seems ages, I lost a couple of ships in combat, (as well as some combats). Is the AI now designing ships comparable to a veteran human player? no, of course not, but at least you will no longer die laughing at their designs, they are credible and far more durable. If they come at you in numbers, or you don’t take them seriously, you will have problems. Now let me share some thoughts on how the AI performed under specific conditions. In the first setting, I gave the AI maximum cheats in production and research, 200%, which is more than the default setting for this difficulty level. I countered this by creating a slightly better custom race for myself than my standard Terrans, retained my own production/research settings at normal, (100%) and changed my strategy. I played against the AI as if it were human. Understand that at these settings, the AI starts out with around 8 times, your base production and 4 times your research… still, before 100 turns had elapsed, my own HW was at over 1000 base production and my research was allowing me to pump out good quality ships much sooner than previous versions. In this session the AI was formidable but victory would have been mine anyway if I hadn’t quit from boredom. Here the AI was able to field good ships and a lot of them, it rarely committed under-strength forces to raids or attacks. The problem with this setting however is in how you counter it. You can give yourself production and research bonuses, optimize your race, or a combination of the two, it doesn’t matter, as the result is the same. You will breeze through both research and production, effectively reaching the late game very quickly. If you expect to win, it will be a short session. If this is what you prefer and like in a 4X, rejoice, this setting is for you. Next up I played the same settings with everyone at normal (100%) research and production. I reverted back to my standard, less optimized race and also relaxed my strategy. This session allowed me to directly compare the actual changes to the AI rather than how the AI, aided with cheats, can play. As you may have guessed the end result was not much different from 1.04 with the exception that the game was longer, I did sweat some battles, and lost a ship or two. Of note, I did notice that the AI reverted to its single ship attacks more often at this setting. Perhaps without the production bonus, it could not field the numbers required for potent attacks or defense. This conclusion was partially inspired by the fact that I caught masses of un-attended support and assault ships a few times during my operations and this helped me cripple the AIs quicker. Finally I tried the same settings as above but cut everyone’s research to 50% and production to 75%. This aided the AI in that their initial advantages in technology and structures were far more difficult to overcome. Ahh, this was my comfort zone (was, 1.06 tilted the balance too much in research with these settings). I had to avoid conflict for a long time and my first ships were far less advanced than what I put out in the other sessions. The game also went far longer, with the issue still being in doubt well past turn 350. The negatives here were that this made for a slow paced game and the AI once again could not field the numbers it could with cheats enabled in its favor, the design of those ships however continued to be impressive, especially in relation to my own. So what does this all mean? A game like chess is focused and finite, there are only so many pieces, so many squares, so many possible moves. An AI can be programmed for chess that can kick 99.9999% of humanity’s behind. A game like ISG is full of subtleties and synergies with multiple combinations possible, racial picks, random events etc. It is not focused. A veteran player will simply focus and defeat the AI, which by the way is actually quite good. The way that the AI designs ships now is about focus. Have this tech by turn so and so, have that tech so many turns later, build ships with X component by turn Z… that’s why it is effective. The idea now is for everyone to find the custom zone and settings that suit them best, give the game a try and draw their own conclusions. Between the ability to customize research and production, as well as the better designed ships and operations from the AI, the game has been dramatically improved.
Only level 7 techs and above have seen their tech costs increased, so the level 7 techs only cost around +15% more in 1.0.6 (which probably translates into 1 or 2 extra turns only than before). Tech level 8 a bit more, etc until tech level 10, which got a +60% increase in cost. We're talking about Plasma Torpedoes, Class X shields, Expert Targeting algorithms, Death Ray here. In the future we're considering adding late techs in every field where you can invest and keep researching again and again for more bonuses, like more armor points, increased weapon damage, more money per POP, more planetary engineering bonuses or more production, etc, depending on the field in question. We're still working on the details but we should have something along these lines at some point. When that is in we may consider reducing the cost of the current end level techs above level 7.
Yep this is common in the 4X realm. I would advise increasing the costs each time that happens up to a (huge) maximum limit.
Forgive the delayed response, it was unavoidable. Increasing late game tech was implemented in an attempt to respond to players that ran through the tech tree quickly. This happens mostly when a player colonizes heavier than normal and/or sets research sliders above 100%, it was not an issue in a normal game environment or when research was set at less than 100%. As a solution it will not be adequate to solve the problem where it actually exists and had negative consequences where the problem does not exist. Techs do not exist in a vacuum, it is not about a few extra turns to research level 7 techs, it is the cumulative effect of these increases. You need so many more turns to get level 7, and more for level 8, etc. It is even worse if you research a second tech in a field. Again, if you set research to 200% the effects will be negligible, at 50% the effects degrade the balance of playing on that setting. I can’t help but feel that a better, more thought out option was available that would have focused on where the problem was rather than where it was not. Additional comments on the AI In retrospect, I can’t help but feel that the approach taken, where higher difficulties mean the AI gets a flat production and research boosts, could have been done differently. Right now if you set the AI to impossible, the AI produces and researches at 175%. If you conquer some AI pops this boost magically disappears, in other words the AI no longer plays by the same rules as you, and I am not certain that I would have given up such a unique and attractive selling point about the game so easily A possible solution could have gone like this and stayed in line with the original vision of the AI just being more advanced than the player yet still playing by the same rules Hard would have stayed the same but with six extra techs instead of three Severe would have become what impossible was with the mine and econ capital to start Impossible would have gotten two extra techs and a starbase to start. In addition, AI pops at this setting would receive a +1 to research and production, (in addition to their racial traits). The main difference here being that if you conquered and assimilated some AI pops, those pops would keep that bonus. An AI, putting out better designed ships and having the ability to crank out Capital ships from turn one could have been quite potent, as a bonus, that starting starbase would make an early rush against the AI less likely to succeed. Production and research sliders could have been separated and implemented similar to SEV. They would set the rate for everyone. Further to that, they could have been capped to offer something like 50% rate, 100%, and 150%. In this manner the players that want a faster or slower game would still have the option of getting one, without upsetting the balance of the game to the degree that 200% does. (Years ago, we did quite a lot of work on finding the proper balance for growth and research, toning down the upper rates would at the very least mitigated the need for finding solutions to research ending quickly. With all due respect to the devs, on the one hand we allow players to research faster and on the other we increase research costs because research is going too fast… tail meet dog lol). Anyway, in conclusion the changes have made for a better game, but that doesn’t mean it couldn’t have been better yet. One of the things that absolutely floored me a few weeks ago was when I lost ships against a numerically inferior force, something that has never happened before, and indecently, the AI was set at the same research and production rate I was using when that happened. TLDR Good changes, could have been better, poor solution for tech research going by too quickly that had negative effects elsewhere. 1.06 has been removed from my machine. 1.05 is still there.
I'm not sure I understand your point here. You can always manually set the AI sliders the same as yours on any difficulty if you wish--it just provides an option to boost them for the AI if you wish--I pump them to max to give the AI some chance of keeping up with my empire-building play style. I haven't really even noticed the increased cost of late game techs--but by that point my chance to instantly discover techs is so high the cost doesn't really matter (which could use a bit of a nerf). Personally I'd still like to see tech and/or wonder victories added. An extremely expensive end-game tech to win or open up an very expensive to build winning wonder is a good way to keep research points useful after the normal tech tree is complete, and in games that is usually my preferred victory path.
Let me explain it concisely. I know what you can do with the sliders, and I understand it is intended to help the AI if the player gives them a production/research boost. I also understand and respect that every player has different tastes, (and they are all equally valid). That said, I find the method used to give the AI a boost, (note the AI, not the AI races, there is a difference), unimaginative and simplistic. The same effect could have been attained by giving the AI races additional racial picks similar to MoO2, one or two additional key techs, and a starting Star-base on impossible. The additional racial picks could easily have given them the needed production/research boosts, but, (and it is a big but for me), these bonuses would have gone to the races themselves rather than the AI. Right now, if I conquer AI pops, the bonuses are gone as they belong to the AI, not the AI races. Similarly, if the AI conquers some of my Pops, those conquered pops receive the bonus. As I said, I respect and understand that this is ok with a lot of players, bu it is not ok for me and i also find it somewhat immersion breaking. Finally, just as some effective adjustments were made to the AI behaviour in designing ships and combat related matters, the same could have been done with a few other behaviours that would insure that on the harder levels the AI is far more competent, (I hope to illustrate this further in my next AAR). You did see where I specifically said that it is an issue when research is set to 50% of normal? Have you tried that setting? if so you should see that instantly discovering techs is not an issue. Increasing tech costs for normal rates and higher is understandable, at 50% there is very little chance that you will exhaust the tech tree during a game or reach a point where high level techs are researched "instantly" . Granted, a player that colonises everything possible, and takes the game way past the point where he/she would have won, might experience fast research even at 50%, but that is an outlier. Added victory conditions would be most welcome, perhaps as you suggest, perhaps even a cultural victory as it is very hard to complete all three branches without a serious and determined effort
By "instant discovery" I mean the "Early Breakthrough Chance"--i.e., "Creative". By mid-game I've usually got around 16 leaders, 3-4 of which have the secondary skills that boost it, so it's over 50%--which seems to get checked more than once, regardless of my tech output I seem to discover everything in 1-2 turns. I understand your point about conquered races losing the production and/or research bonuses you've given their empire, and your opposition to that in principle. But pragmatically, for a challenge, the last thing the AI needs is for you to get even a larger boost once you've started conquering it. I really do understand your points--but people like you and me are edge cases, if on different edges. The game needs more players--and while "the AI plays by the same rules" will appeal to a few experienced gamers, I don't think it's going to have a significant affect on marketing. I also understand your point about extra racial picks. But even +2 to production and research per pop isn't going to come close to the benefits of 150% after the multipliers. I'm not trying to argue with any of your points--I just see the current system as a pragmatic way to make the game more challenging *for now*, while they focus on things that might broaden the market. Ultimately they can't continue to endlessly devote time to a black hole if sales don't justify it--sadly the best use of their time at this point may well be on useless eye candy.
Great discussion Plot, you raise solid points, I raise solid points. The sad truth is this game is one of the best 4xs I've seen in decades, it's like being in school, taking a test and scoring in the 90s... then along comes the teacher and bitches at you for not doing even better. Don't mistake my critique for lack of support, and I can and do tweak my own difficulty levels as ISG is about more than the setting you give the AI, there are multiple factors involved. The AI in general is quite solid, but it can never interpret the mechanics and rules like a skilled player can. We can leverage events better, execute leaders (yes there is a way), use monsters as free garrisons at strategic systems, put leaders in suspended animation (yes there is a way for that too), etc. This doesn't even touch on the player that will reload saves to get desired outcomes on ruins or leader advancements. None of this is meant as criticism of the talent behind the game, it's just that the AI is bound by more constraints than the player. P.S. Have you ever played Imperium Galactica 2? That AI was a nightmare on normal, but I believe it was aided by the campaign setting. It's my favourite RTS 4x of all time, ISG is sitting right next to it in my book.
