v0.4.2 Release Candidate #2 Discussion

[eleazar]

A played for a little over a hundred turns starting with the humans.

My closest opponent was a "Citadel" Cray empire. In spite of their natural tech advantage, i rolled over them pretty easily once i got to endomorphic. Probably could have done it sooner if i'd stopped assuming there was overwhelming force beyond my detection. Many planets were defended solely with Orbital shields-- some of these were on well established front-lines before i made my move. If i understand the mechanics, pure shield ships are worthless if there is nothing armed with them.

A yellow Laefna empire (beyond my reach, but in sight due to a handy Ourbools planet) continually behaved stupidly. There was a system with a native planet and a sentry. Every few turns they would send a single ship to that system -- sometimes an unarmed ship. It would immediately be destroyed, until around 100 turns when they took it

[Dilvish]

My closest opponent was a "Citadel" Cray empire. In spite of their natural tech advantage, i rolled over them pretty easily...

keep in mind, the Citadel planet is for "Turtle" empires, which are intentionally rather gimped.

Many planets were defended solely with Orbital shields-- some of these were on well established front-lines before i made my move. If i understand the mechanics, pure shield ships are worthless if there is nothing armed with them.

Well, not entirely useless. Recall the discussion about monster attacks; I believe the orbital shields can trigger battles with those biological-attack -only monsters and let the planet defenses finish them off. Also, Orbital Shields can help a little if there are other defenders present (such as an armed planet); without that, they are mostly worthless, yes, though could conceivably delay a weak attacker long enough for other defenders to arrive. I added them in primarily to further gimp Turtle empires, and it sounds like they do that just fine The more aggressive AIs build fewer and fewer of them, and perhaps in the current state of things I should simply stop having AIs make them for now, or only Turtle and Beginner.

A yellow Laefna empire (beyond my reach, but in sight due to a handy Ourbools planet) continually behaved stupidly. There was a system with a native planet and a sentry. Every few turns they would send a single ship to that system -- sometimes an unarmed ship. It would immediately be destroyed, until around 100 turns when they took it

I've seen something similar too and have been trying to figure out why -- even if the AI can't see the monster, it will infer its strength from the ships lost and not try again until it has a significantly stronger force to send. Was the system with the Sentry perchance immediately adjacent to the system the ships were being produced in?

[Dilvish]

I'm also starting to think that (at least in absence of changing the aggression scheme to no longer include significant "difficulty" aspects), that perhaps the "max aggression" setting and corresponding random distribution should just be shifted to a flat "aggression" and let all the AI's in a game have the same aggression level.

Alternatively, the random range could be narrowed to be the max setting and one or two levels below it. Especially if you play with few AI's, the distribution can go sour on you. With max aggression, about 60% of the AIs will be in [ moderate, aggressive, maniacal ] but if you only play with 3 AIs, then in about 1 game out of 15 all three AI's will be cautious or turtle; and in about 1 game out of 6 you'll only have one AI that's actually moderate or higher. That's not really what people expect or want.

[AndrewW]

I'm also starting to think that (at least in absence of changing the aggression scheme to no longer include significant "difficulty" aspects), that perhaps the "max aggression" setting and corresponding random distribution should just be shifted to a flat "aggression" and let all the AI's in a game have the same aggression level.

As another alternate, could allow one to set the aggression level of each AI individually.

[Karoushi]

That is pretty much all I have ever faced when attacking an enemy planet, orbital shields.

Their fleets were always either no where to be found or too weak to be concerned with.

I agree with the post below, organic ships are still way too easy to get and in my view, should be considered more advanced and thus just a bit harder to start the tech and to finish it.

[yandonman]

My playtesting results for RC2. 102 systems, 3 AIs, max aggression = manical.

Below are just my notes. I saw nothing that would block the RC.

• Had some rouge FreeOrionCA.exe's running, that the installer didn't kill, and therefor didn't replace during install. Installer should detect this case and kill (or - less user friendly - ask the user to kill) the open FreeOrion*.exe's.
• On Vista, the installer must run as Administrator, but as an semi security minded person would do, I don't use Administrator as my account for using the computer. So, at the end of the installer, when it asks to "Run FreeOrion", if I leave that checked, it runs under the Aministrator account, so any saves/modifications get saved under the Administrator's data folders. Then, when I close the game and launch it from the quickbar, it launches under my account and I lose any screen size modifications and I can't access my last (and first) game(s).
• The Galaxy Setup should read the windows/linux/etc current logged on user and pre-populate the Player Name field.
• Production and Research don't respect windows resize (solved by close and restart FO).
• Research starting position is not optimal - it's starts out centered on "Teir 4" tech. No matter up or down, the player always has to move the research window to the left.
• Yay for Active Radar icon in research screen.
• Sometimes I feel like I give a ship an order an it is ignored. Usually, if I look in the logs, I can see that it has skipped a turn (in this case, turn 34 was followed by turn 36. (see log)
• I also see cases where AI ships that were in a starlane and "moving to" a planet, not move when I hit the turn button.
• Yay for increased research costs. I actually thought about what I needed to research. Things that I noticed
  • Industry Centers (though they still need to be be Tier 2, 4 and 6 (as opposed to Tier 2, 3, 4) tech
  • The two early production ones (Fusion Generators and Robotic Production)
  • Orbital Generators
• Sometimes I swear I change the number-to-build on the production screen and it doesn't take; I'll find later a SitRep that says I finished building 1 ship, when I wanted x5.
• "Discovered New Empire: X" needed as a SitRep (with link to 'pedia). Pedia should list AI difficulty, known PP/RP production and known count of ships. Also, all known research.
• Outpost ships should count as "outposts" for supply lane calculations (maybe?).
• Bringing up Suitability report while in the Production window results in the production window elements hiding the suitability report.
• Battle report SitRep should be improved to "A battle occured at X. A/B enemy ships destroyed, C/D of your ships destroyed.
• Population shouldn't be a number - it should be a species inspired horizontal bar. Happiness should also not be a number - it should be either- or or or >:( (which follows the "don't needed to know the difference between 1.1 and 1.2 - aka: small differences). Hover should still work.
• Need a "Am I building this already" for outpost and colony bases in the same system indicator.
• "Empire X fleet has arrived at system Y" needs to not spam the SitRep when 13 fleets of one ship arrive at the same location. Better would be "<number> Empire X ship(s) have arrived at system Y".
• Organic ship line still way too easy to get to Tier 4 & 5 ships. Research costs need to double or triple for this line?
• Empire window needs a right-click 'pedia look up on individual empires.
• Just invaded colonies should have reduced (or negative) PP production for a few turns.
• Seeing a splattering of these errors:

  Code: Select all

  2013-02-12 00:04:29,424 ERROR Client : Variable<int>::Eval unable to follow reference: Source.Owner
  2013-02-12 00:04:29,424 ERROR Client : Variable<int>::Eval unable to follow reference: Source.Owner
  

• Performance was good. Didn't noticed turn button lag until turn 227. Turn 330 is unplayable due to turn lag.
• The AI was nice and hard.
  • I GOT PSYCHO-DOMINATED!!! (crafy AI)
  • ey... 10% chance? ... 10% of what??
  • Psycho-Dominatrix needs a SitRep!!! (lost a lot of ships unknowingly)
  • Losing 4 or more ships a turn
  • 10% of MY SHIPS!!!
  • 50 turns later, I have finally researched psycho-domination. AI is going to DIE _now_.
• AI cut off key supply lines - crafty (bastards)
• AI kicked my ass.

[Zireael]

Yandonman - the Admin thing might explain why I had weird stuff happen in first games after installing, like seeing long-deleted saves... Maybe there was a copy of them under Admin account, which I rarely use?

[Bigjoe5]

Why does FO need admin privileges?

[em3]

Installer should require elevated privileges, I don't think that running the game itself does.

[eleazar]

Yandonman: what are these tiers you keep mentioning?

[eleazar]

A yellow Laefna empire (beyond my reach, but in sight due to a handy Ourbools planet) continually behaved stupidly. There was a system with a native planet and a sentry. Every few turns they would send a single ship to that system -- sometimes an unarmed ship. It would immediately be destroyed, until around 100 turns when they took it

I've seen something similar too and have been trying to figure out why -- even if the AI can't see the monster, it will infer its strength from the ships lost and not try again until it has a significantly stronger force to send. Was the system with the Sentry perchance immediately adjacent to the system the ships were being produced in?

No, adjacent systems were empty. Does the AI know not to plan routes through dangerous systems? The sentinel system is part of the shortest route to much of the galaxy from yellow's homeworld.

I'm also starting to think that (at least in absence of changing the aggression scheme to no longer include significant "difficulty" aspects), that perhaps the "max aggression" setting and corresponding random distribution should just be shifted to a flat "aggression" and let all the AI's in a game have the same aggression level.

Alternatively, the random range could be narrowed to be the max setting and one or two levels below it.

For 0.4.2 a flat or flatter distribution sounds better.

My game is currently set to max aggression "aggressive" turned out to be against 3 turtles, which explains the ease.


EDIT: New game.
When i had a human and a scylior colony ship in system with an ocean planet, the auto colonizer wanted to use the humans -- who are in all ways inferior in this context.

[yandonman]

Yandonman: what are these tiers you keep mentioning?

In the research screen, the tech is lined up in neat columns. Each column I call a "tier". Tier I contains all of the first level research choices. Tier II contains any tech that has a requirement on only one or more Tier I tech. Tier III contains tech that only has requirements on Tier I or II tech, and so on.

There's no game play mechanic implied or inferred; it's just a common and easy way of talking about tech levels (reference: SCII) as the game progresses.

[Dilvish]

This is a bug that's been around for a long time, but I just had it appear three times in the same game before turn 10, and just in case there's an easy fix I figured I'd mention it now, in this thread, rather than starting a new thread in programming.

In 100+ system galaxies, especially 150+, I'll very often have some situations where star system circles either touch or actually overlap. Here's a case where I had both of those in the same vicinity:

overlap.png (155.3 KiB) Viewed 1450 times

The attachment overlap.png is no longer available

Brog and Bessel apparently touch, and retain a visible starlane connecting them; Teoblan and Po overlap and have no longer have a starlane rendered (I suppose its possible there is no starlane there, but in my experience there always has been). In this particular case there is no particular harm done beyond the uncertainty of Teoblan and Po connecting, which will be cleared up as soon as one tries to plot a course through them, and in any case one can be sure of which system one is entering first. Sometimes, though, the systems are overlapping almost 100%, so that it can be difficult to impossible to tell which system one would be entering first, and which starlanes connect to which system. If there's monsters/enemies at one but not the other, will one's scout run into it if one tries to go to one of the systems? How will supply behave as it tries to pass through the pair? Systems should not be tangled up like this.

The current star coord generating system uses an adjacency_grid to try to prevent things like this, it seems, but something is obviously breaking down. It's probably doing everything on a pairwise basis & if there are three or more systems crowding each other the process breaks down. Maybe that's not the problem & there is some other bug in there that can be fixed for RC3; I'll take a glance but I'm not especially hopeful.

A more robust solution, but perhaps not suitable for trying to get into 0.4.2 (Geoff?), would involve a bit more of a change to the galaxy creation sequence, but not a lot. Rather than creating a bunch of systems with locations & passing that off to a Delaunay Triangulation routine to get the list of all suitable starlanes, which then get pared down according to the starlane density setting, the Delaunay routine could include the process of setting system coordinates. The Delaunay routine builds a set of triangles-- it starts with three 'dummy' points (that get discarded later) making one big all-encompassing initial triangle, and then takes the system coords one by one, figures out which of the existing triangles that coord lies in, adds the coord to the incorporated coords and replaces that triangle by the 4 smaller triangles created by inserting the point into it. We could change this process to have the DT algo set coords -- it starts with the big initial triangle and one by one for each system : (i) randomly chooses an existing triangle to insert the system into, from all triangles greater than min size, (ii) randomly chooses coords for the new system within this trangle, taking into account min separation from the vertices, (iii) splits the triangle up into the 4 new triangles like in the regular process, and (iv) loops on to the next system. This would very robustly ensure a minimum separation between systems, and the size of the starting triangle could be chosen to guarantee that the desired number of systems could be fit in with the desired min separation.

[Geoff the Medio]

A more robust solution, but perhaps not suitable for trying to get into 0.4.2 (Geoff?), would involve a bit more of a change to the galaxy creation sequence, but not a lot.

Definitely not for 0.4.2. Medium-term, one of my would-like-to-dos is replacing the existing galaxy generation system with Python scripts.

[Dilvish]

A more robust solution, but perhaps not suitable for trying to get into 0.4.2 (Geoff?), would involve a bit more of a change to the galaxy creation sequence, but not a lot.

Definitely not for 0.4.2. Medium-term, one of my would-like-to-dos is replacing the existing galaxy generation system with Python scripts.

I've developed a decent amount of experience with the boost C++/Python interface routines while working on the AI; if you'd like to focus on other things I'm game to take this up as a medium term project.