The AI never selects the right techs !

In general I've found that since AHD 2.1, it's not necessary to always have the best Philosophy techs. In fact, I've found that you can gimp yourself pretty bad (in certain scenarios) if you're always grabbing the next +% research tech. In the year or two that it takes to get the next Philosophy tech, you could have gotten 2-4 industry (if you want to temporarily monopolize certain goods), commerce (if you want to fix your economy), naval (if you want to out tech the AI on the sea *this is a really good reason*), or army techs.

Mathmatically, if you want to have a broad range of techs, it makes sense to stay 1-2 levels behind in the Philosophy tree. Being a huge country with a ton of divergent priorities means that it makes less and less sense to sacrifice valuable army/navy/economic tech that you could have _right_now_ for a Philosophy tech that doesn't even do anything in and of itself. Although, it's kinda silly to have 5 levels of Political Thought tech. If you're going to go "high" as opposed to "broad" in the tech tree, you really do need to have more than 1 level of Philosophy.

Wminus said:

Nobody is researching a philosophy tech ever year - but every 15 years. It is also perfectly possible for the Russian AI to be up-to-date on both philosopy and economical techs at once. Not that it ever is, but it's possible.

I agree with you perfectly that the philosophy tech shouldn't always take priority, like if you need money fast or you're in the middle of a war. But otherwise, it always should.

A philosophy tech typically boosts research by 1/3 and takes 2 years to research, which means that only 8 years after you starting researching the tech, it has paid you back every RP and will start making profit after that. That's a damn good investment in peacetime, and quite crucial if you want to play efficiently.

Industry isn't really that important early on because you can easily catch-up later in the game. That is in stark contrast to technological progress, which is almost impossible to play catch up against a competent opponent.

Click to expand...

Now convert tht into the technology.txt....

But Russia doesn't just need to be up to date in commerce techs. It also needs industry techs for the economy, needs Political Thought for the NFs, needs army techs obviously, and naval techs because it's capital is easy to blockade. A Russia that obsessively researches Philosophy is going to be worse off than it is now.

Also, 8 years is a long time to break even in Vic2. I generally run games with a horizon of 5-10 years, and unless I'm absolutely going to avoid war in all those years I skip the Philosophy techs. 8 years is especially a long time when you have to do it every 10 to 20 years. You're spending 40 of the game's 100 years "paying back" the time you spent researching Philosophy, for a supposed benefit in the future, when there's nothing better at predicting future performance than current geopolitical position. I'd much rather have 3-5 extra regions after 8 years, than to break even on RP investment.

Of course, I also research Philosophy techs, and prioritize them when I do, and there's prestige bonuses from Philosophy techs that make the tech worthwhile in and of itself, and in your example in the OP Russia really should get at least 1 more level of Philosophy if it's going to grab level 4 techs. But I don't think that this is a big problem with the AI.

Wminus said:

Russian AI is useless. Are you kidding me? It didn't even research idealism by 1878. This is a HUGE problem which needs allot of attention, because it makes Russia really backwards for no good reason.

Click to expand...

increase the weigting of those techs after the 60's then,

n.b machine guns and other colonise techs are very heavily weighted after the 60's

Wminus said:

nasaeulus: isn't it possible to just use scripts, like in HoI3? How does this system work anyway? Is modifier additive to the factor, or multiplicative? Or does it overrule the factor?

How does AI tech selection work?

Click to expand...

The factors are multiplicative. It works pretty much like this:

The AI calculates the factors for every tech that is available to pick from.
It then puts these into a weight table, and rolls on it.
Wherever the roll 'lands' is the tech it researches.


So let's say that Russia has 25 possible techs to research (as it nearly always does). 12 of those techs have factors of 20, 1 has a factor of 50, 6 more have factors of 100 each, and the remaining 6 have factors of 500.

That's a total of 3890.

Effectively, the AI rolls an imaginary 3890-sided dice. the numbers 1-20 will be the first tech, 21-40 the second tech, etc etc, with the last techs having 500 numbers each, as that's their weighting.

Obviously, the 500 factor techs are much more likely to be picked - but if you look at the whole thing, then even with 500 weighting, the chance of that tech being selected is still only 1 in 8.

So let's say we give one 5000 weighting. Now it's chance of being picked is about 2 in 3. Even with ten times the chance of any other tech, it's still not going to be picked 30% of the time. Moreover, you've also made the AI unresponsive to other factors - so it'll go for that tech 2/3rds of the time even if it's ludicrously inappropriate for it to do so, like when it's massively outgunned by a hostile neighbour. So then we have to start adding in situational code.

But we're adding that to all 25 different available techs. In fact, it's worse than that, since this situation will keep occurring all the way up the tree, so we're doing that for 150 technologies. Suddenly, because you're increased the weighting on Education, Russia can't colonize and Prussia takes that province in Archangelsk every game. Or Russia stops teching military and it's army gets curb-stomped by Sweden. Or Russia stops researching commerce or industry, and goes bankrupt every five minutes - and worse, since it's not teching the RGO improvement techs, a lumber mill in Germany can't get inputs, and so the American ship industry crashes, and suddenly anyone playing an unciv can't build a navy.

So yeah, that's why we don't say things like 'it's not difficult to fix' or 'it's very simple to change' in relation to, well, almost anything you do in V2. Almost everything turns out to be zero-sum in nature, and ties into everything else in the game, so you end up cocking up 20 other things whenever you change one.

The first NF tech is a much better choice for Russia to go for first anyway, since it adds the same RP bonus from the pluralty inventions and it gives you an NF as well - which Russia rather needs. And I do see it getting the philosophy techs sometimes, anyway.

Modifiers multiply the factor in AI chances, always.

Well, just to be sure, i checked a 1905 AI Russia tech choice (with all AIs getting a 20% boost to research).
Was playing as Austria in this game.

I admit, some choices are strange...note the pityfull 8000 or so gold too and the unused NF.

http://hpics.li/40d4b98
http://hpics.li/33b1279
http://hpics.li/c7e244a
http://hpics.li/ee6046f
http://hpics.li/3c985fb

well, it looks like russia wants to be able to field a lot of resilient and fast reinforcing but shitty in doing actual damage armies and fleets and get most out of their RGOs

...which would actually be pretty realistic, considering Russia at the time. For reference, the huge army and industrial modernization project that so worried Germany and Austria took place 1910-1915.

Going screenshot by screenshot:
*The choices for Army techs are basically what I would choose. The artillery choice is a little weird, but understandable, and the focus on defensive techs fits well into the combat mechanics.
*Naval techs are a little fubar. There's no need to focus on the +damage/-damage techs unless you have a large and very diverse navy, which I never see Russia have. Even more concerning is that most of those bonuses don't even apply to ships that Russia can build. the Naval Leadership techs are worthwhile techs, though, so not everything is screwed up.
*Commerce techs. About what I would expect. A lot of those techs aren't going to do a lot of good unless you have good administration, which Russia doesn't. I'd get some of the +% taxes techs, but whatever.
*Culture techs. We have another "Naval" phenomenon. There's very, very little reason to spend so much time on the +reinforcement techs when there's free slots in the Army tree (maybe there weren't?). The Political Thought techs are good choices for Russia. I'd get Empiricism, but that's not a huge problem.
*The choices for Industrial techs are good. The +supply techs are a must for large armies (which Russia has), the railroad techs are necessary for moving them around quickly, the the Power tree is obviously necessary for a country that relies on RGOs. The only suggestion that I would have is to get Mechanical Production (which has such large bonuses that I think it should be heavily weighted), and Cheap Iron (Russia has a lot of iron RGOs).

On balance, these are good tech choices as AI Russia. Almost all of the non-Army/Navy techs it chooses have *something* to do with their military, which is realistic considering Russia's reliance on its military. If there were something to tweak, I would suggest moving emphasis to the Ship Construction tree from the Naval Engineering tree within Naval techs, and moving emphasis from the Psychology techs to the Army techs. I don't think researching Psychology would have resulted in a better Russia, and considering that they're in the GP club on the strength of their military alone, it might even result in a worse Russia. It's 1905 and Russia is number 7 in the world with a huge military, a middling economy, and poor prestige. That's a pretty good simulation of what actually happened.

The only strange thing is that Russia starts with (and apparently keeps) the Academic Intelligentsia administration. Given that their tech choices are almost entirely based around their military, I think that AI should be better at switching academic administrations.

So, there's 3 things I would change in regards to that example: AI Russia should change its academic establishment, should research Ship Construction instead of Naval Engineering, and should leave the Psychology tree alone. None of those things have anything to do with researching Philosophy, and in fact if Russia DID research Psychology instead of some military techs, they'd probably drop off the GP list.

TLDR: The AI's choice of techs is okay, and while it isn't perfect I think it does the job very well.

I wouldn't call those tech choices 'completely useless', tbh. The heavy psychology weighting is a bit odd (tho probably down to mil fixation), but other than that they're pretty good choices aside from the lack of philosophy techs.

What I'd be more concerned with is the dreadful state of the culture tech school, which means the Russians have a lot of penalties and very little bonuses.

Wminus said:

They should've at least researched clean coal, mechanical production, philosophy techs and (the only useful?) naval techline, ship construction... meh is it only me who gets annoyed when the AI does such strategic mistakes?

I unfortunately won't have time to fix the tech files this week, so I'll call paradox on this thread and hopefully make them check out this problem.

Click to expand...

what order did they pick there techs, why would those be better for russia then the ones they picked