Page 1 of 1

[AI] Does the AI always switch to research focus on turn 1?

Posted: Sun Nov 26, 2017 5:22 am
by alleryn
Whenever i look at the graphs, it looks like all the AIs switch to research focus on turn one, which seems like the opposite of what they should be doing (at least in most cases), considering how important production is at the start to expansion.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Sun Nov 26, 2017 3:12 pm
by MatGB
The only race I don't switch to research early is Egassem, the early techs, especially the flat bonuses and growth techs, are so essential to be able to expand it's almost certainly the better strategy.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Sun Nov 26, 2017 4:46 pm
by alleryn
MatGB wrote:The only race I don't switch to research early is Egassem, the early techs, especially the flat bonuses and growth techs, are so essential to be able to expand it's almost certainly the better strategy.
I guess i'll have to get in some multiplayer games to see if that holds true, as i strongly disagree (though i admit i am not an expert and could be wrong).

Re: [AI] Does the AI always switch to research focus on turn

Posted: Mon Nov 27, 2017 9:57 am
by Jaumito
MatGB wrote:The only race I don't switch to research early is Egassem
Laenfa, too, depending on galaxy settings. Bad research, bad troops, superior detection and stealth are all compounding incentives to focus on industry early to get some colonists on as many planets as you can asap. Especially in multiplayer. Then you can still scrap some early research from unconnected worlds.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Tue Nov 28, 2017 2:22 am
by alleryn
I'm in the process of conducting a little test, playing a game with prod focus at turn one, and then i'm going to restart it with research focus and take a few notes (though i'm getting a little lacksidaisical in my note-taking).

But i still find it hard to believe that research focus start is a good strategy (excepting the species best suited to it). Admittedly if you are able to colonize multiple worlds and research nascent ai, the research start is quite strong. But, unless you get incredibly lucky and have 3+ adequate or better worlds in your home system, or maybe adjacent to home system, it seems like it will be really hard to defend your systems against a player rushing production. Even if you have planetary bunker and are set to defensive status, the 4 troop ships needed to take your planet are cheaper than the one colony ship (or outpost + migrate) it took to build it in the first place, and that's a 2 colony swing.

Maybe i'm thinking of it too much as a 2-player game. I suppose if i rush you someone else is rushing me, but i just find it difficult to see how you are going to survive until you research Adaptive Automation if you start with research instead of industry.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Tue Nov 28, 2017 3:12 am
by Oberlus
Jaumito wrote:Laenfa, too, depending on galaxy settings. Bad research, bad troops, superior detection and stealth are all compounding incentives to focus on industry early to get some colonists on as many planets as you can asap. Especially in multiplayer. Then you can still scrap some early research from unconnected worlds.
This.

In my games versus AI, when I start with a species with better production than research I always focus on production. As long as I see reachable planets to colonise and no enemies I focus production on outposters/colonisers and use this starting research queue: Algorithmic Elegance, Nascent AI, Physical Brain, Robotic Prod., Nanotech Prod., Adaptive automation, Exobots. But once I see the current outposter I'm building is the last one that will have a proper planet to land on unless I improve supply range, if physical brain is not finished then I make sure it is getting max RPs and once it's finished I place the Automated History Analyser next for production after last outposter and colonies under construction (in starts where you have very few planets to get the most out of Nascent AI, this AHA building makes a great difference, in the opposite case it is rather optional but nice to have it). Laser and Military Robotic Control can be put in the queue if I see enemies close enough.

Either with few planets to colonise and an early AHA or with more planets colonised earlier, in both cases thanks to the focus on production, I have the sense that the overall increase in RPs is bigger than from focusing on research, and the production also benefits from it later on when I can get Adaptive Automation way earlier and then everything snowballs (if I can still expand freely, that's why I no longer play games with more than 25 systems per player, too easy) or I can start building a big Star Armada to counter my newfound enemies.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Tue Nov 28, 2017 3:55 am
by alleryn
For context of my opinion, i do play a lot of 50 system/ 3 empire, but i do sometimes play 240/8 and still prefer a prod start. I'm testing on 180/8 for a 22.5 = (15+30/2) system/player, since the tooltip (on the single player start screen over number of systems) indicates balancing around 15-30 systems per.

I've been using random on all settings except monsters which i set to low (i get tired of dismssing sensor ghosts :p).

Also, this is just sort of my strategy or whatever, but when i play a species like human that starts with research focus, which i am forced to switch if i want to start industry, i research physical brain on first turn and start building AHA immediately after the scout finishes on turn 3 (if it seems super confined i might just do a part of a warship instead of the scout). Otherwise i'd research robotic prod into fusion generation first, usually, but the excess production is wasted if you are converting from a species with research pref, if you research robotic prod on turn 4, due to time taken to refocus. [Because of the time it takes to switch to industry, seventh turn robotic production hits pretty square on if you have to switch, which is when you get it if you go human brain into robotic(+fusion)]

Re: [AI] Does the AI always switch to research focus on turn

Posted: Tue Nov 28, 2017 8:54 pm
by MatGB
Ah, yeah.

I balance and test for games of 150 system plus. I rarely if ever play below 100 systems, overwhelmingly the feedback we get says people want bigger maps to be viable.

If playing just 50 systems, then pure production would be the way to go, agreed, but that's a very edge case unusual setup these days, 60 systems as the default start was changed up several years ago.

Size of galaxy and relative density makes a massive difference to balance, if playing 250 systems and up research wins, hands down, for virtually every species. But on a tiny map you can win by turn 50ish if you go for a speed blitz. So it might be good for the AI to be more game size aware.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Tue Nov 28, 2017 9:19 pm
by alleryn
MatGB wrote:Ah, yeah.

I balance and test for games of 150 system plus. I rarely if ever play below 100 systems, overwhelmingly the feedback we get says people want bigger maps to be viable.

If playing just 50 systems, then pure production would be the way to go, agreed, but that's a very edge case unusual setup these days, 60 systems as the default start was changed up several years ago.
Systems/empire seems like a more pertinent measure than just overall systems, to me. Either you die to the empire next to you, or you survive (and likely capture their cultural archive/historical analyzer). How many empires there are is less significant, in my eyes.

I do see planet density and native frequency as big distinguishers between whether production or research start is better, with higher density (somewhat paradoxically) and fewer natives favoring research. [In lower density galaxies you are forced to fight over "neutral" systems to colonize, whereas in higher density it's easier to have a relatively safe "core"].

I imagine galaxy shape has a fairly big impact too. Shapes which are more conducive to fighting a single player at a time would more strongly favor production starts, whereas those with highly-mulitple borders would encourage defensive play and research.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Tue Nov 28, 2017 9:32 pm
by MatGB
No. If playing 50 systems/3 players total, then you either win or die in the first couple fights. Mass Drivers plus Robotic hulls and a production bonus will win the day.

But if it's a larger map with more systems, if you ignore research early to get an advantage in low powered, low tech ships while, for example, the Gysache player 200 jumps away has got themselves up to Gravtic and Titanic hulls with plasma weapons, you're dead in the late game.

In a large map getting behind in research is late game doom. In a tiny map, getting behind in strength is early game doom.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Tue Nov 28, 2017 9:59 pm
by alleryn
MatGB wrote:No. If playing 50 systems/3 players total, then you either win or die in the first couple fights. Mass Drivers plus Robotic hulls and a production bonus will win the day.

But if it's a larger map with more systems, if you ignore research early to get an advantage in low powered, low tech ships while, for example, the Gysache player 200 jumps away has got themselves up to Gravtic and Titanic hulls with plasma weapons, you're dead in the late game.

In a large map getting behind in research is late game doom. In a tiny map, getting behind in strength is early game doom.
I disagree. Starting production, and taking a second homeworld on turn ~70-80 will let you outresearch an empire that starts research.

If you've started research, you can't defend any colonies, so your research output is just your HW, until it falls. Unless you got lucky and have multiple adequate+ worlds near your starting system.

Gysache starting research is basically suicide, they are an extremely attractive target to their neighbors, having bonuses to industry and research, along with bad ships and troops.

If your argument is just that research is better if everyone leaves you alone for 200 turns, well sure okay, just research all the +res +inudsty +pop and colonize your heart out. But that doesn't really happen. There will be conflict because it's cheaper to capture a colony than to build one (in the early game), and you get free population, and you get an extra species to provide greater flexibility of environment choices.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Wed Nov 29, 2017 12:58 pm
by Oberlus
MatGB wrote:If playing just 50 systems, then pure production would be the way to go [..] if playing 250 systems and up research wins, hands down, for virtually every species.
I've never played FO against a human (would be great, can't connect at the right times, trying to change that). But against AI I always win in almost any setting, and in the setting I can lose it's seldom. I can lose 1 out of 30 games in galaxies with around 25 systems per player with medium or high monsters, natives and specials (better high), because with such settings I can be very unlucky with monsters, specials AND natives maybe 1 out of 10 games, and if one or two (but not more) of the AIs get very lucky and relatively soon they find some fantastic techs and extincs species and pick it on me with special interest, then I can lose. What I mean with this is that my strategy is relatively well tested against AI.
The thing is I always win the AI exploiting the strengths of my starting species. If I am Egassem, or George, even Exobot, I will not focus research. I will get my extra RPs by expanding faster and getting faster more "+2 RPs" from Nascent AI and conquering more research species and planets for them.
If I switch to research a George planet at the start, I expand way slower and get a Pyrrhic increase in RP. If I don't, I need more time to start growing RPs at start but I get a faster increasing trend later on that overcompensates the initial loss.

I would only switch such species to research focus in the case of no colonisable planets in fuel range at the start (i.e. if I can not put my PPs to any good use, then I don't need them), but that has not ever happened to me, ever, with any species in any galaxy setting I've tried (maybe could happen with low density, few starlanes, no natives, no specials and a narrowed environment species).

Re: [AI] Does the AI always switch to research focus on turn

Posted: Wed Nov 29, 2017 1:16 pm
by Oberlus
MatGB wrote:if you ignore research early to get an advantage in low powered, low tech ships while, for example, the Gysache player 200 jumps away has got themselves up to Gravtic and Titanic hulls with plasma weapons, you're dead in the late game.
If you are massing up low tech ships while your main enemy is 200 jumps away safe and blooming (may I assume the map has very high number of systems per player?), instead of you expanding like crazy and only building the exact military force you need to clear your way to the next native to conquer or planet to outpost, then you are doing something worse than not focusing research. If you are massing up low tech ships because you are harrased by your nearby enemies and it is that or defeat, while nobody is harrasing the Gysache that can expand at ease, then you are in a difficult (and unlikely) situation no matter what you do with your planetary focus (but better if you keep it at production, IMO).
MatGB wrote:In a large map getting behind in research is late game doom.
That's why I don't focus research with a production-gifted species (as per above message, the extra PPs become many extra RPs later on, snowballs are cool).
MatGB wrote:In a tiny map, getting behind in strength is early game doom.
This isn't exactly true. Against AI, you can fall behind in military strength (to the point of not building a single military ship until the enemy stack of upped mass drivers are knocking at your outer planets with a tail of troopers behind) as long as your research powerful enough and well planned. You can keep on bay a relatively large enemy stack with planetary defenses and comsats for a very low production cost. You focus production on expansion (outposting in the free areas) until you get the right military techs and then you focus production to pumping your shinny robodestros with defense grid and laser 2 (3 on the way). Then you can remove the enemy from your space, even if he has three times more "rough estimation of military strength" at the moment of the clash. You may have lost some planets on the first fence system but you get them back with very few loses for your brand-new fleet in a few turns and your enemy gets stripped of its military force in the area so you get also one or two enemy systems (as payback) before he gets there the new batches of underpowered warships to try to stop you.

Re: [AI] Does the AI always switch to research focus on turn

Posted: Thu Dec 07, 2017 11:57 pm
by alleryn
I finally finished my test game. Unfortunately my notetaking was a little lackluster, but i think it was a useful enough test to post the results. I started with prod focus, played the first 100 turns, then loaded from turn 1 and started with res focus to compare:

Game Settings (best guess):
Systems: 180
Galaxy Shape: Random (? - Irregular or Cluster probably)
Galaxy Age: Random (Young)
Starlane Frequency: Random (Medium)
Planet Density: Random (Low)
Monster Frequency: Low
Native Frequency: Random (High)
AI opponents: 7 (all set to Human -- my first test was a smaller game vs Egassem and Trith, and it just didn't feel like a good test, since the research-focused Egassem got stomped and the Self-Sustaining Trith were out producing me even with them on res. focus and me on prod. focus)
I'm also playing Human.

Analysis: The random settings turned out to be quite favorable for the industrialists, with High Natives and low planet density. I would break the game down into 3 segments:
Early early game (turns 1-18)
Mid early game (18-61)
Late early game (61-)

Here's a picture of the map at around turn 18:
ProdResEarlyGeography.jpg
ProdResEarlyGeography.jpg (172.71 KiB) Viewed 1667 times
Early-early:
At turn 18, the industrialists have a third scout, their AHA, a warship (Large hull), Robotic Production, Fusion Generation, Planetary Ecology, and Subterranean Habitation.

The techers have an outpost ship, their AHA, Algorithmic Elegance, Nascent AI, Planetary Ecology, and Subterranean Habitation.

Aurora and Whitefall both have high tech natives. Zyrgon is guarded by a Sentry, but is a pretty attractive target. The only other adequate+ world is at Slipher (a tiny desert world), on the border of the Blue Empire. The industrialists, thanks to their extra scout, have also managed to identify an unprotected native world (Yoyo-beta) just past Slipher. Soon after this, i learn that Green has a terran (with a prod special) and an adequate world in their home system. If this were a real game i'd be getting blown out.

As the industrialists, i decide to amass ships, colonize Slipher and Yoyo-beta, and try to move against blue, whose borders i've noticed expanding.

As the techers, i adopt a more conservative approach and go for Zyrgon.
__________________________________________________________
Mid-early comparison:
Both industrialists and techers do some basic military tech (zortrium and organic hull) and then work towards Adaptive Automation.

The industrialists take a colony from Blue, as well as capturing Slipher, Yoyo-beta, and eventually Zyrgon.

The techers take Zyrgon, and eventually start building some exobot colonies on the barren worlds in their home system and adjacent system.

I believe the end of this phase, when the industrialists research Adaptive Automation, is the best point of comparison. At this point the techers advantage consists of Exobots, Active Radar, Orbital Generator, and, if memory serves, Symbiotic Biology. The industrialists, however, have a much larger fleet -- 16 ships with attack/defense totalling 96/421 compared to just 6 ships with 36/184. If the two empires came into conflict, it's clear who would be losing territory. Researching Adaptive Automation 15 turns earlier gave the techers a slight advantage during that time, but they still ended up with a considerably smaller fleet.

The result is the Industrialists are poised to expand now, while the techers have to be content with their current territory (there are hardly even any Poor Environment planets to colonize). At turn 61 the (research+production) of the two empires is nearly equal.
______________________________________
Late Early game:
The industrialists are able to rapidly expand, capturing the blue homeworld on turn 76, with minimal losses to their fleet. The techers still are amassing their fleet to expand with.

At turn 100, the industrialists have 19 colonies to the techers' just 7, doubling them in both PP and RP. However, the techers have amassed a considerably stronger fleet, largely due to researching Rock Armor, and since their PP weren't being diverted to colonization. At this point the techers are ready to expand, but there is a pretty big gap to close.
____________________________
Conclusions:

I made loads of mistakes in these games to be sure; i didn't feel like i was playing very close to optimally in either game.

Still, i feel like i'm convinced that at least in certain galactic conditions, and even in a largish map (180 systems/8 players), and even on an industry-tech-neutral species like humans, that a production start is a viable, and probably superior focus.

I look forward to testing in multiplayer after i've honed my strategy. Until then, my next test (when i have some time) will be similar conditions, but with Low Natives and High Planet Density. I expect this to be consdierably more favorably to a research focused-start.

Notes:
Production Focus:
Turn 1: 8.5 PP 7.4 RP
Turn 2: High tech beige goo natives scouted. (1pop/30shld/30def/50troop)
Turn 3: The Physcial Brain res., Scout prod.
Turn 5: Sslith scouted (10pop/0shld/0def/6troop) guarded by Sentry
Turn 7: Robotic Production res., high tech Ourbools natives scouted (14pop/30shld/30def/43troop)
Turn 8: Blue Empire discovered. Desert (adequate) planet scouted (in supply, on border with Blue)
Turn 9: Derthrean natives discovered, across Blue border. (19pop/14 troops, no defenses). Same system has terran (good) and
ocean (adequate/good for Sslith) 16PP, 5 RP
Turn 10: Blue expansion visible
Turn 11: Automated History Analyser prod., Yellow Empire sighted, common border with Blue
Turn 12: further Blue expansion
Turn 14: Fusion generation res., Warship prod.
Turn 15: Planetary Ecology res.
Turn 16: Green home star system identified, with 3(!) inhabited worlds.
Turn 18: 20.9 PP, 10RP, Subt. Hab. res., Warship prod.
Turn 42 25.6 PP, 14 RP
Turn 61 Adaptive Automation res.
Fleet: 17 ships: 1 troop ship, 16 combat vessels (~96/421)
Colonies/Outposts: 5 Colonies (2 colonized 2 captured)
45.3 PP 19.0 RP
Turn 76: Blue HW captured, 106 PP 24.3 RP
Turn 100: 220 PP, 93.7 RP
19 Colonies
5 Outposts building colonies
24 Mass Drivers with 735 structure
8 Lasers with 220 structure
_____________________________
Res Focus:
Turn 1:8.5 PP, 8.4 RP9:06 PM
Turn 3: Planetary Ecology res
Turn 5: Algorithmic Elegance res
Turn 7: Subterranean Habitation res
Turn 12: Naxcent AI res
Turn 15: 12.5 PP, 14.5 RP
Turn 18: AHA completed
Turn 20: Org Hull res
Turn 27: Symbiotic Biology res
Turn 40: Native colony captured
Turn 46: Adaptive Automation res
16 PP, 26.3 RP
2 Colonies (1 captured)
Fleet: 4 ships plus outpost ship and colony ship (24/120)
Turn 61: 4 Colonies (2 colonized/1captured) + 1 in queue.
Fleet: 6 combat ships (36/184) + 2 scouts
33 PP, 31.5 RP (already has exobots, active radar, orbital generator at least)
Turn 100: 111PP, 41.7 RP
7 Colonies
7 Mass Drivers with 224 Structure
18 Lasers with 1163 Structure