[Java MOO] Feedback on bugs, cheats & exploits

[Ray F]

I noticed that there's three circles around each colony. I'm assuming that the largest two are the regular fuel range and extended fuel range, but what about the small circles? I see that when they overlap, the area is more brightly colored, unlike the larger two circles. So what are they?

The small circles are one-light-year circles around a colonized system. There is currently some overlap between colonies because I have the minimum distance between stars set to 0.5 ly for testing purposes (for lots of races on tiny maps). This will not be the case for the finished game.

I like where your version is going with the interface. It's a big improvement over MoO 1's already good interface.

If it ain't broke, don't fix it.

[timoteob]

Ray F,

I have been away from these forums for a while so I am way late to the Java MOO party, but...

I still want to commend and thank you for your work on bringing MOO into the 21st Century! I love all of the improvements to the interface. You managed to make modest changes that will have a big positive impact on game play while still keeping the same classic look and feel of the original game. I am also glad to see that the new code base will get rid of the old bugs and exploits.

Take Care,

Timoteo

[Ray F]

I still want to commend and thank you for your work on bringing MOO into the 21st Century! I love all of the improvements to the interface. You managed to make modest changes that will have a big positive impact on game play while still keeping the same classic look and feel of the original game. I am also glad to see that the new code base will get ride of the old bugs and exploits.

That's the first time I've ever heard the word 'modest' used with anything I am involved in!

[RefSteel]

Some additional thoughts and notes:

Diplomacy is one of the more difficult aspects of the game for the AI to handle without "cheating." Original MoO doesn't even try, such that AI-AI relations do not even attempt to mimic remotely-possible AI-human relationships. AIs will go from war to alliance with one another in a single turn, break an alliance one turn (with or without human prompting) and restore it the next, etc. In spite of this, there are so many possible exploits relating to diplomacy that some veterans have gone so far as to stop using the "Audience" button altogether in order to avoid them. Among the core difficulties:

1) Correctly evaluating the strategic value of avoiding diplomatic repurcussions is very difficult for an AI. When is it safe to use bioweapons or become an "oath-breaker" in spite of the diplo penalty? Only when you're way ahead? What if you're way behind one rival but about to use it on them to make up some of the difference? And if the AI only does these things in specific circumstances, doesn't that make it too predictable? (Orion "solves" this by ignoring the diplo penalties for actions like sporing and xenocide when an AI does it, and never permitting the AI to break peace treaties.)

2) Correctly evaluating situational tech values, whether as bribes, incentives (e.g. to make peace, but see below) or (especially) trades. The MoO AI tended in general to overvalue its own techs and undervalue others' techs, which is a good baseline for a single-player game, as it means AIs rarely trade with each other (so their tech situation doesn't always look identical to one anothers') and the player will have difficulty achieving ridiculous profits as a tech broker since the return on each tech will be slight. What the MoO AI does not do very effectively (and what is extremely difficult to program) is recognizing when a specific strategic situation demands acquiring a tech even by overpaying (thus allowing some AI-AI trades to occur) or refusing to part with a tech at virtually any price (although the MoO AI refuses to trade its highest-level tech in any field regardless of what it is or the strategic situation in general).

One other diplo item that may have been missed: There is a rarely-seen diplo action that AIs do take in MoO under extremely unusual circumstances. I don't even know for sure what all the circumstances are, having only seen it I think three times in my years of playing Orion:



This 'Loks gave me this message during Imperium 37, at a time when we and they were both at war with the Meklar. I don't remember if this has ever happened to me when I was not at war with the target race - it's been ages. The one time I received the promised "reward" - much longer ago than Imperium 37; I don't remember if I bombed the target's planet or just slagged a bunch of their ships - it was one of their technologies: Along the lines of what AIs sometimes give to sweeten a peace deal or in response to a threat from an enemy they fear. I don't know if you want to duplicate this in Java MoO - no one is likely to miss it since it comes up so rarely - but I figured I should mention it at least!

[Ray F]

Surprisingly, that exact same dialogue showed up for me last week (first time I had seen it), when I was playing MOO1 to ensure I got the flow of the Galactic Council voting down correctly

I didn't follow up on the request, so I don't know what would happen if I had. There's not currently a way for a player to make a proposal like that to an AI, so I am currently inclined not to include it in the game. Feature parity between the player and the AIs is something that I am trying to be very strict about.

About understanding diplomatic repercussions --
I understand the concern about making the AIs seem predictable, but my first order of business is to make them not dumb! Once I cross that bridge, I can worry about predictability. However, I want to be sure that "too predictable" isn't really code for "doesn't make an obvious mistake that I can exploit"

About tech trades --
In Java MOO, each race (including the player) retains a list of enemy ship designs that he has scanned. Although this is useful for the player, the real intent is to give the AI some sort of data to better assess the value of technologies.

In Java MOO, the AI ShipDesigner currently tracks the best shield level of his enemies when determining what the "best" weapon is when designing a new fighter ship. However, I am hoping to eventually go beyond that very simple example when evaluating techs.

For example, if any enemy ship is encountered with Subspace Teleporters, then the perceived value of Subspace Interdictors would go through the roof. Similarly, an enemy with spores would greatly increase the value of antidotes. Large planetary shields would increase the value of spores, as would being in a position to not care about diplomatic repercussions (i.e. just 2 races left).

[RefSteel]

That makes a lot of sense. My concern about AIs being "too predictable" came from thinking about bioweapons though. If I can tell in advance that I won't get spored by an AI just by looking at the bargraphs ("in their position, they're not going to risk the diplo hit...") ... ... Actually, now that I write that out, it seems like a good thing for the most part. Even so, I wouldn't want to ever be 100% sure in advance that I can e.g. ignore antidotes and concentrate on shields in lieu of active defenses in spite of an enemy with spores, especially if they're erratic!

Also, if there are many situations where an AI will not use death spores, a smart AI will need to be able, in advance of deciding whether to start on them as its next tech project, to predict whether it would be willing to use them around or after the time research and a subsequent round of ship-building would complete. This seems like a really complex thing to program, and probably one of the reasons the original designers gave up and eliminated the diplo penalty for AIs who use the things.

More broadly, I agree that making AI reasonably competent is key; it's just that too-predictable behavior - especially for an AI that has to face humans who can remember previous games - is itself "an obvious mistake that I can exploit." Better predictably decent than randomly stupid, definitely, but the ideal would be for AIs to be able to adapt their strategies in multiple viable directions, and for their knowledge of the galactic status quo to weight but not strictly determine the viable adaptation they ultimately choose.

That last bit's not really applicable to the next alpha though, I'd imagine. What brought this up was just thinking about exploits and imbalances in the game, and how removing the AI's immunity to various diplo penalties will affect one of the scariest things in original MoO: The threat of an incoming spore fleet.

[Ray F]

That last bit's not really applicable to the next alpha though, I'd imagine. What brought this up was just thinking about exploits and imbalances in the game, and how removing the AI's immunity to various diplo penalties will affect one of the scariest things in original MoO: The threat of an incoming spore fleet.

I agree. That threat always needs to be present, and it doesn't make sense that some diplomatic penalty governs the behavior of every race.

It seems to me that Erratic leaders should not be concerned about the diplomatic penalty for using bioweapons -- after all, they will declare war at the drop of a hat. In addition, Ruthless leaders will not care, nor will they hold it against other races. To them, all is fair in love and war, but especially war. And finally, rebelling against the New Republic should probably mark you for termination by any means necessary.

I think the net result of this could be a partitioning of the races into rogue states and those aligned against them.

[Psillycyber]

About tech trades --
In Java MOO, each race (including the player) retains a list of enemy ship designs that he has scanned. Although this is useful for the player, the real intent is to give the AI some sort of data to better assess the value of technologies.

Ermagerd, are you for real? This is genius!

By the way, in case you haven't implemented it yet, I have one more thing on my JavaMoO wishlist to make all of my dreams come true: additional stats screens that show, in line graph form, stats for one's own empire across time (with the x and y axes clearly labelled), and maybe even have the ability to show the same info on rivals, at least for years where one had spy access on them...maybe with the values in the years between interpolated where the spy reports were old.

In Java MOO, the AI ShipDesigner currently tracks the best shield level of his enemies when determining what the "best" weapon is when designing a new fighter ship. However, I am hoping to eventually go beyond that very simple example when evaluating techs.

For example, if any enemy ship is encountered with Subspace Teleporters, then the perceived value of Subspace Interdictors would go through the roof. Similarly, an enemy with spores would greatly increase the value of antidotes. Large planetary shields would increase the value of spores, as would being in a position to not care about diplomatic repercussions (i.e. just 2 races left).

Cowabunga! No more useless AI laser fighters when all of my ships have class IV deflector shields! No wonder the AI won't need the free colony ship cheats! This AI on impossible is going to be a beast compared to the old one!

[Sarissofoi]

Hello
I just want to tell that it this Java MoO looks really promising.
No, its not good word. It looks great.
So many improvements over original. The only thing I dislike is art but that is minor concern. Prefer original pixel art.
I honestly thank you for doing this. I bet God reserved you a place in Heaven.
Anyway I want to ask if you gonna increase max AI races on the maps?
The larger sized maps with more factions could be a very interesting scenarios.