Beta Observations and Thoughts

[DarkWolf_V]

very interesting, i have just had a quick look at it and i see drastic changes in gameplay which i am not sure how they are going to work out. a massive amount of resources asteroids, ship prices adjusted to be much higher and a variety of planets. this looks to be great additions and cant wait to see how they work out

[Busser]

There seems to be a problem with the combination of the following elements;

 

1. I did mention before the AI tends to expand very slowly. When faced with neighbouring Volcanic planets (which require 8 research stations) AI expansion seems to grind to a definitive halt. They seem to be unable to 'skip' planets as human players would do.

 

2. When playing a game with active pirates, pirate fleets seem to accumulate very fast and linger around planets they've just raided. I have found this to be a verry large burden on my CPU and causing lagg early on in some games. More pirate-raids are launched but no pirates are destroyed (inactive AI) resulting in a considerable number of piratefleets.

 

==> [Combine slow AI expansion, pirate accumulation + pirate raids that cripple the enemy AI early on]

you have an enemy AI that is just incapable of expanding.

 

i have had this happen now in 3 large games (on all difficulty levels except unfair) and in my opinion the AI is unable to cope with the current gameplay changes.

[Lavo]

Number 1 is known; it's an issue I'm personally looking into, at the moment I am currently unsure how to deal with it. In the following update I'm going to attempt putting in a solution, setting the volcanic's tech priority from "high" to "normal". The AI should not colonize those planets until later on in the game anyways, once it can handle the income penalty, so assuming this fix works it will make the techs work as they were meant to.

 

On 2, pirates simply like to linger even if everything is destroyed. Why they do this is not known to me, other than it's an AI thing, which is hardcoded. I will however, increase the time between raids.

 

Thanks in particular for the lower AI settings feedback; it's quite useful as I have little experience with those AI levels.

[Busser]

I have not yet tested the mod on unfair but if these problems do not occure on that specific difficulty level you could possibly just advise players to stick to that difficulty. Several mods from other games I've played did this and it could be a huge timesaver for you guys. (some Hearts of Iron 2 and Rome/Medieval 2: Total War mods spring to mind iirc). Balancing the game for all difficulty levels with the amount of current gameplay changes that have been implemented could prove rather time consuming or even impossible.

 

Also, is there a way of playing the game in an observer/neutral-mode ? For example playing as pirates you could speed up te game and watch AI behaviour or interpret the post-game graphs and would speed up testplaying a lot. Tomorrow I plan to do some testing with inactive pirates and see how the AI handles.

[Lavo]

Sounds like a good idea... Though even on unfair the AI takes a long time to expand, in comparison to humans.

 

There is actually. You can hit "Surrender" in the game exit menu. Hit "Keep Playing" and you won't be able to control any units, however, you will be able to see the entire map and thus each player and pirates.

[Busser]

In retrospect I think pirates are not the deciding factor in AI slowdowns that I have made them out to be. Looking at the following figures from two almost identical test runs, one with revision 576 and one after revision 577. (I have an excel file in case you’re interested)

 

Testgame 1; revision 576.

After 13 minutes of gameplay all 5 AI players had already built 5 or 6 military labs.

After 1 full hour of gameplay only 3 succeeded in building 5 civic labs and it took the other two another half hour.

After 1hour and 30 minutes of gameplay 3 players had only colonized 3 extra planets. One was slightly hampered by pirates but the other two were stalled at planets that required the bureaucratic tech. After 1 hour and 30 minutes two AI players dominated the map with respectively 8 and 11 planets colonized while the others were still stuck at 3 additional colonies.

Testgame 2; revision 577.

After 15 minutes of gameplay all 5 players had built 6 military labs.

Afyer 45 minutes of gameplay one player had built 6 civic labs and the others had built none. Most of the players caught up at 1hour and 20 minutes of gameplay.

4 AI players showed comparable levels of expansion in the first hour, all colonizing 4 planets at the same pace. One did not colonize more then 3, again stopping after 48minuten of gameplay and idling at a planet that required the Bureaucratic tech.

 

 

1st: the AI tends to build a lot of military research labs early on in the game. I’m not sure why this is but it puts a significant burden on early development.

 

2nd: the AI does skip over enemy/uncolonizable planets so the assumptions I made in the earlier post were incorrect. The real problem however is that the AI tends to idle at planets that require volcanic exploration or bureaucratic efficiency research tech. I assume they wait for the relevant tech to develop and continue colonization. The fact that military labs have already taken up a lot of logistic slots slows down this progres.

Thus I am convinced that the real deciding factor in slow AI expansion is the proximity of planets that require high tech levels. Although the problem is not as big as I thought it to be, I think it is at heart a simple map issue. I suspect the Sins AI has been developed with constant colonization in mind and not with this selective type of colonization.

 

Maybe one of the following changes could be possible

- Lower the tech level of volcanic/bureaucratic techs and let the AI take the negative financial impact. (can the AI handle negative planet income ?)

- Prioritize civic lab building over combat lab building.

- Make Shipyard planets/High Terran Planets/ Volcanic planets appear more at the core of the maps, and not in the outer sectors.

 

Again, I have tested this in only two games, both of them single star systems with 60 RANDOM planets.

[Lavo]

These are really good points.

 

1) While the civilian research does not surprise me, the military research absolutely baffles me. I noticed that the AI liked to research both trees after the implantation of NU, even though the AI did not do such military research to extent that it does so now, with a virtually identical military tech tree.

 

2) There isn't a way to prioritize which research structure the AI builds, nor what it researches. I am hesitant to lower the tech requirement of volcanic/bureaucratic planets,, as they are intentionally not easy to get. However, your suggest to make Tech 4/Volcanic planets appear further away from the outer sector of a map, even if just for a small amount of planets, is a very solid and feasible suggestion for Random maps. It wouldn't be too hard to make a planet grouping which only includes Terran/Desert/Ice/Asteroid type planets and sandwich that between the regular random groups and the extraPlanetGroup (aka. the planets that are ALWAYS near your starting world), which would allow for a small buffer that will allow for a decent start and more logistic slots. The Sins AI has indeed been created with the idea of "expand, expand, expand" no matter what. A small buffer would likely allow for a feasible, yet light, early expansion which would pave the way to higher tech planets.

 

Might I ask for the Random map(s) you used, did you start with extra planets, or not?

[Busser]

The buffer-idea is brilliant and it would solve this problem quite easy. My gut feeling tells me that as soon as 4 planets are colonized everything starts going smoothly so a buffer like you suggested would work as long as high-tech planets are at least 2 jumps away.

 

The maps I used were:

 

a) Random - Medium (Enlarged), 1 star, 4 players max, 37 planets. I played a 1vs.1 game. The first time I encoutered the AI getting stuck at 4 planets. [letting the game run at top speed for some time did brake the gridlock]

 

b) Random -Medium Large (Enlarged), 1star, 6 players max, 60 planets. (the revision 576 test)

 

c) Random -Medium Large (Enlarged), 1star, 6 players max, 60 planets. (the revision 577 test)

 

Would it be helpfull if I did a comparative test of the other random maps tomorrow ? Or some maps in particular ?

 

edit: Like you said, lowering the research requirements for the colonization techs would not be needed. They are good as they are now.

[Lavo]

Nah, this does wonders, and gives me exactly what I need. And the buffer idea is yours really; it's the best way to implement the core suggestion of yours.

 

I'll make the maps sometime today; will likely get the Medium/Medium Large (Single) maps up first.

[Lavo]

Added the buffer to one map as mentioned here; requesting some trial runs if possible.

[Busser]

I did some test runs on the map you suggested, all results are taken after +/- 1 hour and 30 minutes of gameplay (sped up offcourse ) .

The first three runs with 3 active Unfair AI. No human interference.

 

1. One player had colonized 6 additional planets, the other had colonized 2. High-Tech planets did not appear in unfavorable positions and did not hamper expansion. Some heavy bidding did result in a very large and crippling first pirate raid, which explains the struggling third player. The lack of expansion from the second player was due to the fact he was surrounded by planets with a very large defending fleet.

 

2. One player colonized 7 planets, the other 4. No players were halted. The difference n development was due to the fact the map heavily favored expansion for the first player.

 

3. All players colonized 5 or 6 planets. The last one was slowed down by consecutive pirate attacks and the bad luck of being surrounded by asteroidfields and a positron storm. This was perhaps the most balanced game of all.

 

Now for some test runs with 2 active Unfair AI. No Human Interference.

 

4. The first player colonized 6 planets, the second only 2. The second player did spawn very close to the pirate base and was targeted several times. By the time the first player encountered a high tier planet, he was able to quickly build the remaining civic lab and colonize it without a slowdown.

 

5. The green player spawned right next to a Large Volcanic Shipyard planet and had only two other neighbouring dead-end planets . One of which was an asteroid with 8 logistical slots. After colonizaition logistical upgrades and civic lab building followed. This took such a long time that the red player was able to take control of this industrial planet’s gravity well and effectively isolating the green player from the rest of the galaxy. Resulting in the red player having 7 colonies and the green player only 2.

 

6. First player colonized 6 planets, the second 3 (wormhole and ion clouds).

 

Seeing as Random maps tend to be . .euhm Random, it’s pretty hard to predict how the AI is going to react. The Medium – Random map with 33 planets and 3 active players is rather small, so the numbers I noted aren’t that bad. Especially when you subtract the uncolonizable gravity wells. On such a small map, small anomalies will have a large impact on the gameplay. I suspect that implementing the changes on larger maps will give more steady results. Perhaps the amount of ion clouds, worm holes, dead moons on small maps can be lowered ?

 

Only one game was doomed from the start due to bad planet placement (5). Slow starts in other games did not give the impression of giving a definitive advantage to the opponent. (the slower players in game 2 and 3 were hampered by bottlenecks that cut of their starting positions. These bottlenecks would become huge defensive advantages later in the game. Much like Australia in RISK)

On a side note, sometimes planets are placed right next/on top of each other. This results in gravity wells that run through each other and makes controlling units/zooming in awkward.

 

In conclusion I would say that there is improvement. The 3 previous tests all had at least one AI struggling with this problem. The 6 current tests show only 1 game so job well done !

[Lavo]

This is quite good to see. I'm glad to see some improvement... I'll look to implementing versions for the larger maps. It is indeed possible to decrease the odds/number of non-colonizables in smaller maps; would be as simple as making a new GalaxyScenDef grouping and putting that in place of the current groups in the map. For the placement, this was somewhat of a necessity; it's the only way for me to realistically "force" the rest of the map to not be atop the new Col 1-3 colonizables.

[magnus64]

I think Imperial Torpedo Spheres are bugged at the moment. They queue up but never start building. I had enough supply/crew, and could build any other capital ship. Can anyone possibly double-check and confirm?

[Lavo]

I think Imperial Torpedo Spheres are bugged at the moment. They queue up but never start building. I had enough supply/crew, and could build any other capital ship. Can anyone possibly double-check and confirm?

For the Torp. Sphere you also need the missile penetration tech (level 6) in order to build it for some reason. ...Come to think of it that's like forcing the Malevolance to require Heavy Ion Cannon research done before building it. I'll fix this.

[sloosecannon]

it has been coded to work for both friendly and enemy forces

Woah! Did I do that? I didn't even realize it did that lol... I thought it only worked for your units... wow.

[Lavo]

it has been coded to work for both friendly and enemy forces

Woah! Did I do that? I didn't even realize it did that lol... I thought it only worked for your units... wow.

In theory it should work for enemy units, judging by the buff code, though in practice it only effects your units. In fact, it doesn't even work for allied ships, even though they are counted as "Friendly" by the Sins AI.

[sloosecannon]

it has been coded to work for both friendly and enemy forces

Woah! Did I do that? I didn't even realize it did that lol... I thought it only worked for your units... wow.

In theory it should work for enemy units, judging by the buff code, though in practice it only effects your units. In fact, it doesn't even work for allied ships, even though they are counted as "Friendly" by the Sins AI.

 

Huh... that's interesting... that would explain sudden attacks tho... I'll have to look into that... but it's probably hardcoded

[Lavo]

In theory it should work for enemy units, judging by the buff code, though in practice it only effects your units. In fact, it doesn't even work for allied ships, even though they are counted as "Friendly" by the Sins AI.

 

Huh... that's interesting... that would explain sudden attacks tho... I'll have to look into that... but it's probably hardcoded

mmmhm, indeed it would. The reason for this is that in the Buff file it's coded so that Enemy, and NoOwner, planets also work with other Culture centers.