Investigation of AI reassessment upon reload

[Old Celt]

I've seen the Protectorate bug in action myself, and seen the test results in graphic detail posted by others. I think that's the worst one yet as it can totally unbalance the game as the AI creates whacky Protectorate arrangements willy nilly as it reassesses after loads. I don't know why you can't duplicate results RB, but our purpose here is to analyze demonstrated, reproducible results. Until you give me a test sequence to reproduce your results, I can make no sensible comments about your data.

I would like to know if others here support my conclusion that the reassessment process is performing illogically? Specifically, when the AI has the means and time to execute a capture of a city, it will do so when no loading of game interferes, but deliberately choose not to capture the city ASAP when a load game happens.

[roguebolo]

I can send you the save games in question. Since they are midgame saves, I obviously can't describe each and every move from the start of the game in order to duplicate the situation. But I can let you try to establish a protectorate yourself from my save position. It won't work. And that is duplicatable.

I originally encountered the Protectorate exploit because I was bound and determined to have Gaul submit to Protectorate status after laying siege to their last remaining territory in Celtiberia, so that I wouldn't have to worry about having to keep it garrisoned. They would under no conditions yield to my demands. So I went online to see if anyone had any advice about Protectorates. That's when I encountered the exploit; the person described it as working every single time, without failure, so I figured I'd just do it the quick and easy way. After several tries, I decided it just would not work.

One of the developers had also posted in that forum (I wish they'd do that more), and described a process that was somewhat longer and more arduous. I followed his advice, and a few turns later they finally agreed to a Protectorate for about 3-4,000 Florins.

I had the exact same results with Spain in Lusitania a few turns later.

I have a close friend who's also an avid RTW player, and I asked him to try it as well, in his game. He experienced the same results.

The exploit seems to work well near periods of extrema, in the absence of threats or alliances.

However, I HAVE observed an AI faction submitting to Protectorate status to another AI faction, when it did not seem reasonable. Specifically, in the same game Germany is a Protectorate of Brittania -- yet Germany is still strong, with a number of territories. Although this might seem illogical, it certainly doesn't work in my favor. It creates a situation in which they function almost as a single faction -- their alliance and military access and trade agreements give them the combined strength of a very large faction, which seems to be what people are saying they would like to confront later in the game.

As the first respondent, I will state that I agree with you -- but only partially. My agreement is restricted specifically to rebel territories, in which the AI will illogically abandon the territory for no gain. I think this is an oversight in the game design because in general the rebels seem to be treated as "just another faction".

[Pode]

Blasted work and school are keeping me from being able to do this myself, but here's a test I'd like to see. RB has advanced the hypothesis (if I understood him correctly) that most of the AI flakiness we have observed after loads settles out as you approach midgame, especially with the player being active. However, this is tough to test on other machines since being active produces a unique campaign situation and so we don't have a controlled test. I'd like to see someone post a savegame file where we can all download it easily, one from a ways into a campaign. (I'd like to volunteer RB for this since he somehow seems to not have the bu . . fea. . . issue, and I'd like to dig into that a bit. However, anyone's game will do, so long as everyone is working from the same game file). Run the standard player-passive tests from that mid-game point, 20 turns w/o loading, followed by 20 turns loading every turn, and compare AI province conquests. 20 turns loading at different frequencies would be interesting too, and if we coordinated beforehand who would run what frequency test it wouldn't be massively time consuming for any one of us to do. Thoughts?

[roguebolo]

I think it's a very good idea, and I'm sure I'm going to get the response that people just don't have enough time, but I think multiple people should upload save games so that we also have tests that eliminate other factors. For instance, my tendency to create a huge number of alliances could affect the way that my games are behaving. In general, I'm finding that a large database of save games is allowing me to test different theories, and to challenge them in situations that are contrary.

At the very least, if I'm going to volunteer, I would like to volunteer the most enigmatic of the situations I've encountered. Specifically, this is the Turn 12 and Turn 14 scenario. Both are very close to the period of extrema, which means they should exhibit at least SOME of the AI behavior exhibited by the sanitized 20-turn test. However, turn 12 exhibits it in the purest fashion -- straight lines across the board under territorial rankings with nothing but save/loads and "End Turn". Turn 14 ends up the exact same way as when I was really playing, with a number of territorial acquisitions. The only difference is that I made three alliances in between the two turns. Those are the ones I would like to submit, because they are the ones that I understand the least. They are the reason for some of the theories that I've promoted here.

If you just want to see a save game where the AI alternately relieves and reinstates sieges, I could provide that too. It does give some information, but I don't think it's quite as useful, nor as enigmatic. You're basically going to come to the same conclusion I did -- you gotta kill 'em, whether you do it outside your castle walls or a few miles away.

One thing that would be very useful is the ability to turn OFF fog of war in mid-campaign. Do any modders out there offer this capability?

[*edit*] Actually, in retrospect, I might be able to hack it in the save game file. But I'd prefer to avoid this extra work if someone can just point me to a mod solution.

[*edit*] Actually, heh, now that I think about it even more, it might be possible to change the load game behavior by hacking the save game files. If "state" information is stored as a numerical value, then changing this "state" information could change the way in which the AI moves after a load game...

[Pode]

toggle_fow in the romeshell window works, although when I loaded your turn 19 game you emailed me I had to toggle it, load and toggle it again before it finally took and began truly toggling. Maybe it gets borked by loading too?

[Bhruic]

I think it's a very good idea, and I'm sure I'm going to get the response that people just don't have enough time, but I think multiple people should upload save games so that we also have tests that eliminate other factors. For instance, my tendency to create a huge number of alliances could affect the way that my games are behaving. In general, I'm finding that a large database of save games is allowing me to test different theories, and to challenge them in situations that are contrary.

It may have some utility, but I'm not sure how useful it would be overall. A fair amount of a game's understanding comes from playing the turns to get there. Starting from a set point in would mean that you may have missed a great deal of the give-and-take that goes on. For example, let's say on Siciliy, that Carthage takes the Greek city, but then loses it to the Scipii. If I saw a game after that point, I'd have no idea that Carthage took the city (and fought the Greeks). That could taint the results.

One thing that would be very useful is the ability to turn OFF fog of war in mid-campaign. Do any modders out there offer this capability?

Hit ~
type "toggle_fow"
(type "toggle_fow" again to turn it back on)

Bh

[therother]

The LM has webspace available to host files. PM me and I can arrange for the file to be uploaded.

On the issue of this thread, please look through your previous posts and remove anything that is not directly related to the investigation of this issue. That means discussion of patches, CA, Activision, bugs and so on. This thread was intended to be clean of such discussions, and it is our intention that it will be. One way or the other.

[roguebolo]

Actually, on my way to the store just a few moments ago, I realized that one of my statements about hacking the save files was incorrect. Although it might be possible to modify a save file so that the behavior of the game changes after a load, it would be almost impossible to do so without the benefit of the source code. The reason for this is that typically you attempt to save a game in two identical situations with only one factor changing...then you analyze what data changed in the save file.

Unfortunately, the one thing that I want to try to change is consistently the same after each load game. Hence, no changes in the save file.

I'm going to PM you in a few seconds. I'm going to send all three of these save games -- the ones called "Turn 12" and "Turn 14" are the enigmatic ones, and the one called "DefendThapsus" illustrates the behavior of the AI during midgame.

We can address the issue of Protectorates after you've taken a look at these.

[Kraxis]

Ahhh but rouge can also be meant in the context of you being 'a rouge', a sort of romanticised bandit. And since your name is in fact two names (first- and surname) then rouge is fair enough, especially since you have chosen to write it small.
But yeah it is also a colour. Anyway, I will refer to you as RB from now on, ok?

I agree with therother that what we call this issue is of little point here. To some it acts too strangely and thus to them it must be a bug, to others it is merely a weak code of some sort, others yet an unfinished feature. But what we can agree on is that loading after a save has a grave impact on the game if done fairly often. We don't agree on how grave it is.

[Bromley]

FIRST: Apologies, but I have missed out Rhodes (Greeks) in all of my tests. It's immaterial to the conclusions though.

Summary

Thanks to RB for mailing me his files. I've looked at them and have concluded that they in no way contradict the statement that loading the game damages the AI expansion.

Any other conclusions/theories about the effect of alliances on the AI are for others to advance. My only objective in this test was to determine whether or not save/load produced results that I might expect from earlier tests.


Details

RB supplied 3 files which I refer to as "12", "14" and "19". Those numbers don't directly translate to turn numbers. Assuming that we call the first turn (270s AD) Turn 1, then they refer to turns 14 (264w BC),16 (263w BC) & 20 (261w BC) respectively.

The tests use "12" and "14" and run until 261w BC, which is when "19" was saved. The first thing to note is that these tests cover relatively short periods (6 & 4 turns), so I've also performed an couple of extended runs on "14" that go for 6 turns to 260w.

Thanks to Bhruic for telling us how to turn FOW off on someone else's save. That made the comparison a lot easier and allowed me to see a lot of interesting stuff that I would have otherwise missed. BTW, anybody trying it may find that they have to type it into the console twice to get it working properly.

As well as recording the territory data, I've also listed a bunch of locations that I found under siege in some of the runs. This is not necessarily comprehensive, but they do account for many of the sieges. x means that it wasn't besieged, s means that it was (followed by who is doing it) and lifted means that it was no longer sieged (note that this may have occurred in the previous turn). Likewise, taken means it was taken (duh!) but not necessarily in this turn. If there's no mark then I didn't check on the status.

"12" Test

A simple one to conclude. Look at the increase in total territories from 74 to 81 when no loads are made. As we've come to expect, zero growth with save/loads every turn. Also note that the no-loads column (interval=NIL) has a bunch of cities under siege, whereas the save/load every turn (interval=1) column has none. This means that the difference between the two runs could be expected to increase if we continued for a few more turns.

"14" Test

You'll see that the total territory only increases by 1 up to 261w BC whether you load or not. However, this is not too surprising as there have only been 4 turns. As we're already 14 turns into the game, it's reasonable to expect that the AI has already taken many of the easy Rebels. Indeed, we know that there have been 6 Rebels taken by the start because I recorded to total territory in 270s in my earlier tests as 71.

When I extend the test by 2 turns to 260w BC, you see a huge change. The save/load run shows no change at 78. However, not saving or loading for those two turns allows the AI to expand by 8 territories! That's huge when you think that in the preceding 20 turns it had only expanded 7 territories.

The reason can be seen in the Sieges box. There are a huge number of outstanding sieges at the end of the first 4 turns without saving. A large number of these succeed, along with a bunch of others that I didn't record like Noricum.

So again, this is evidence of the loadgame feature being pants.

"19"

Included for completion, this is RB's save at 261w from when he was playing the game. One possible conclusion is that he reloaded approximately every 3 turns when he was playing, as the total territory increase is similar to my interval=3 test that I ran previously and that covers a similar length period. In fact this conclusion probably applies to "12" and "14" because they similarily exhibit relatively little expansion.

Strangely enough, I did not replicate RB's results where he saw territorial expansion compared to "19" using "14" up to 261w BC. However, I did note that Pontus managed to expand by 1 territory despite save/load every turn. Also, there were a few sieges in place when I finished that run. Maybe sometimes with save/load an army will siege a Rebel and, as the Rebels have their turn at the end, the Rebels may sally before the turn ends? That way, the AI would be able to win a siege in one turn even with walls being present.

[Kraxis]

http://dictionary.reference.com/search?q=rouge
http://dictionary.reference.com/search?q=rogue

The content of this post will be removed fairly soon. Sorry, I couldn't resist butting in.

-Simetrical

Hahaha.... I see I have made a fool of myself. I guess I use the term 'a rogue' too little. I wonder at how I missed the point. Maybe because rouge is pronounced in a french style (roosh).

Anyway, thanks Bromley.
Clearly the AI had built up its forces by the end of 14, thus enabling it to expand mightily.
But I think this is obvious proof of something not being right.

[Kraxis]

Rouge, while the harassment might be better against you, it is only better because your are a much better commander than the AI. You can outsmart it pretty much every time. But since other AI factions fight auto-calc that is not likely to be a good strategy against them.

Result is that the AI does not expand but gains something of a better chance at getting a draw with you. I would rather that it lost its assault on me and gained other cities so it wouldn't be worn down too fast. A war with Dacia as a two province state is no fun when we talk about 40 turns into the game. True that could happen in a no save/load game but at least the Dacians have had the possiblity of expanding correctly but have gotten beaten at it.

And I wouldn't say that harassment is effective. It only gives me more time to bring in reinforcements if my garrison is very small or weak, and if it is strong it means that I will most certainly not suffer much damage to my economy.
I have been half-beaten enough times in sallies for me to believe that a sustained siege is better than a run-around. And it isn't always possible to win a sally. If you sit with 2 Peasants, 3 Militia Hoplites and a unit of Peltasts there isn't much you can do against a large army consisting of archers and/or horse archers plus a bit of other troops.
In such a case a broken siege is very bad for the AI as it will result in it never gaining a relatively weakly defended city.