Notes on rev 5465

[Dilvish]

I have no idea where in the log file it might indicate what color is used in game for the empire;and I don't really plan to try tracking that down.

actually, I suppose I might wind up tracking that down, but it would be quite low on my list. Even though it did somehow seem to derail the substantive discussion of interesting data.

[Dilvish]

... would solve the problem with endomorphic ships being spammed by the AI. Has the endo-nerf patch been submitted to the SVN?

What was the perceived problem? If the AI is doing decent it is going to try spamming ships at you, just as you will try to spam ships at it. The endo nerf has gone in, and the AI will now focus on different ships. Having more variety in ship design across AI empires will be in the pipeline at some point, but that's mostly a flavor thing.

[Geoff the Medio]

...necessary for the human player UI (though frankly I don't see why)

I'm not sure why this would be done in the gamestate code, but it is necessary to do an update of that sort at some point in the human client so that the player can see the results of changing focus.

I have no idea where in the log file it might indicate what color is used in game for the empire;and I don't really plan to try tracking that down.

actually, I suppose I might wind up tracking that down, but it would be quite low on my list. Even though it did somehow seem to derail the substantive discussion of interesting data.

I apparently misinterpreted the meaning of your claim that the chart indicated "that the Gyisache were actually doing fantastic"; specifically the "actually" which I thought meant this was contrary to what the map suggested. Regardless, I'm attempting to add a "colour" property to empires in the Python API.

[em3]

I will try to get to putting together a dependencies.py file or somesuch (anyone have a better name?) where I can try to highlight tech related or part related quantities that my algorithms rely on & that folks can check when they're making changes.

How about "assumptions.py"?

[Dilvish]

...necessary for the human player UI (though frankly I don't see why)

I'm not sure why this would be done in the gamestate code, but it is necessary to do an update of that sort at some point in the human client so that the player can see the results of changing focus.

A key portion of the focus order execution code is below. The call to planet->SetFocus triggers a signal that should, and I believe is, updating the sidepanel planet info as desired for the player. It's the additional calls to update the queues thereafter that appear unnecessary and wasteful to me. Based on some comments I saw somewhere, it sounded like at one point the queue projections were based off the projected PP and RP for the next turn, whereas it appears to me that as coded now the queue projections are based on the current turn PP and RP, in which case the queue projections would not change due to a change in focus. I think my current approach for the AI setting planet foci has given it very good control over its PP and RP balance, but for a large empire the cost of redundantly updating large queues, twice for each planet in the empire every turn, seems to really add up to a lot of time & I would like to reduce that.

Code: Select all

     planet->SetFocus(m_focus);

    Empire* empire = Empires().Lookup(EmpireID());
    empire->UpdateResearchQueue();
    empire->UpdateProductionQueue();

As for the maps & charts,

I apparently misinterpreted the meaning of your claim that the chart indicated "that the Gyisache were actually doing fantastic"; specifically the "actually" which I thought meant this was contrary to what the map suggested.

Sorry if I expressed myself poorly or if my unpolished charts caused any misunderstanding and sidetracked the discussion. Upon looking at the maps my initial impression fit totally with your commentary about their expansion, which was essentially

... the yellow AI ... empire size explodes between turns 75 and 100. It still seems like some AIs don't expand much despite ample opportunity to do so.

The actual magnitude of how well the Gyisache were doing was to my perception only roughly conveyed by the maps though and was still just a small part of what I thought very interesting in the charts (though I did also think it was pertinent towards better understanding why the AI slowed down so much at that point). To me at least, the charts indicate that for the most part, the other AIs did in fact expand at a reasonable rate competitive with how fast a human player can expand as the game stands today. The maps may give the impression that they expanded slowly, but 'slow' is hard to assess well in an observed game. I think it's primarily the comparison to the extremely strong showing by the Gyisache, in particular when assessed through the rough lens of territory control, that their performance might appear lacking. That's the key thing I found most interesting in the chart, and is why I started making the charts instead of just eyeballing territory control in an observed game.

That at least is my interpretation of the data based on my experience playing the game as currently coded. But I'm not at all sure, perhaps I'm misjudging the range of progression a human player can be expected to achieve, and misjudging what it takes for the AI to be pleasantly competitive. Regardless of any maps or charts of production, the ultimate test is the subjective experience playing against the AI, and I can't really only on my own limited play/test time for that. I hope to get more feedback from players. It was very helpful to have folks flagging that the AI had stopped performing invasions well even if it did nevertheless take me a little while to observe that myself.

BTW, if you do get around to exposing empire color for the log files, any of hex strings ('#008000'), RGB or RGBA tuples ((0,1,0,1)) would be compatible with the plotting library & I'd be glad to match up the colors.

[Geoff the Medio]

BTW, if you do get around to exposing empire color for the log files, any of hex strings ('#008000'), RGB or RGBA tuples ((0,1,0,1)) would be compatible with the plotting library & I'd be glad to match up the colors.

http://freeorion.svn.sourceforge.net/vi ... threv=5483
If you'd rather have tuples, I suggest / request you write a converter and expose that as the property and remove the GGColor class exposing.

[Geoff the Medio]

Something else I noticed in the latest SVN - so probably with a bunch of content updates since 5465 - is that I've several times had a monster with stealth ~15 move into my homeworld and start causing biological attack sitreps to show up each turn, with a slow decrease in population. The monster doing this is not detectable to (most?) empires starting with 10 detection strength, making the source of the message completely unknown and quite confusing and potentially frustrating for new players. Is this intentional, or the result of content changes to detection ability or monster stealth?

[eleazar]

Something else I noticed in the latest SVN - so probably with a bunch of content updates since 5465 - is that I've several times had a monster with stealth ~15 move into my homeworld and start causing biological attack sitreps to show up each turn, with a slow decrease in population. The monster doing this is not detectable to (most?) empires starting with 10 detection strength, making the source of the message completely unknown and quite confusing and potentially frustrating for new players. Is this intentional, or the result of content changes to detection ability or monster stealth?

Some monsters are supposed to be undetectable, to starting detection, but these are common monsters, so probably shouldn't be so troublesome.

EDIT:
revision 5487
"took bioweapons away from the floater, since the point is to secretly plant more dyson forests, not invisible bombardment."

[yandonman]

Would saying "A stealthed entity or entities bombarded PlanetX" or "A stealthed entity or entities destroyed ShipX at SystemY" make things more understandable?

[eleazar]

We definitely need clearer messages for stealthed blockades, bombardments.

[Zireael]

Would saying "A stealthed entity or entities bombarded PlanetX" or "A stealthed entity or entities destroyed ShipX at SystemY" make things more understandable?

This would be a good idea.

[Geoff the Medio]

Some monsters are supposed to be undetectable, to starting detection, but these are common monsters, so probably shouldn't be so troublesome.

EDIT:
revision 5487
"took bioweapons away from the floater, since the point is to secretly plant more dyson forests, not invisible bombardment."

Small Kill Swarm also has death spores and 15 stealth.

[Geoff the Medio]

I just had an invisible small snowflake show up on turn 10. No biological attacks, but still seemingly early for undetectable monsters to appear.

Edit: The biological attack sitreps keep appearing even after a planet has been depopulated, despite it not having any population to lose. The effectsgroup scope condition should probably check that a planet has some population before firing on that planet.