at what point does the AI decide you are having it too easy?

**tallstark** · July 27, 2008, 12:38 PM

Sorry for all the messages, but in the middle of a fun campaign since I just downloaded SS6.1 and I am noticing some things that are odd.

Now that I have finally gotten to most powerful faction 5 of the last 8 turns, the AI has really decided to stick it to me! Of my last 8 auto-resolve items (spy, assassination, princess with 7 stars trying to steal a general) that were over 80% probability, only 1 turned out favorable. Thats just since I started counting, but its been bad for awhile now. Sent 3 spies to a city and all 3 had 90%+ to get in and all 3 died...

My question is does the AI setting (vh/vh in this instance) change as the game progresses or am I just having bad luck? Mathematically speaking its too strong to be random...

**Empress Meg** · July 27, 2008, 12:40 PM

Since this is dealing with a mod I'll move it to that mod's forum. You may want to ask there since it pertains to a specified mod rather then un-modded M2TW.

Devoirs The Empress

**tallstark** · July 27, 2008, 12:44 PM

Originally Posted by Empress Meg

Since this is dealing with a mod I'll move it to that mod's forum. You may want to ask there since it pertains to a specified mod rather then un-modded M2TW.

Devoirs The Empress

OK thanks! I was wondering in general if the AI changes as the game goes on or its just a SS thing or I am imagining the whole thing LOL but I am sure I'll get feedback either spot...

**Barser** · July 27, 2008, 01:03 PM

I dont know but what I have noticed is that when your assasin/spy die and you then reload the same thing usually happens again. I dont know how this system works but it feels like I need to do another battle or something before I can get another try to assasinate/spy etc. and have a chance of succes.

**obbob** · July 27, 2008, 01:27 PM

it preserves a random seed. So if you want to reload cheat, just move another character then re attempt the mission

**Soul Forge** · July 27, 2008, 01:43 PM

i actually experimented with this a little a while ago. I had 4 spies with varying subterfuge skill near a city. they had like 60-80% chance of infiltrating said city. So i sent them in. The first two and last one i sent were killed, the third one sneaks in. I reload. I send them in again in a different order. Same result (the third one makes it in). Reload. Repeat. Same thing. It's interesting that the results are the same even though all the spies had different skill levels.

**Wolfcp11** · July 27, 2008, 04:42 PM

Originally Posted by Soul Forge

i actually experimented with this a little a while ago. I had 4 spies with varying subterfuge skill near a city. they had like 60-80% chance of infiltrating said city. So i sent them in. The first two and last one i sent were killed, the third one sneaks in. I reload. I send them in again in a different order. Same result (the third one makes it in). Reload. Repeat. Same thing. It's interesting that the results are the same even though all the spies had different skill levels.

That is due to the seed. Everytime a game is started, it loads a new "random seed" that basically determines the outcomes of variables.

I think the was an option for a new seed to be generated after reloading... or was that Civ4? :hmmm:

Originally Posted by Barser

In my experience it is not enough to just move a charecter.

try to move, save again, then reload

**LionQ** · July 27, 2008, 05:18 PM

Originally Posted by Wolfcp11

That is due to the seed. Everytime a game is started, it loads a new "random seed" that basically determines the outcomes of variables.

I think the was an option for a new seed to be generated after reloading... or was that Civ4? :hmmm:

try to move, save again, then reload

It's Civilization IV indeed which has this great option.

**Barser** · July 27, 2008, 01:45 PM

Soul Forge: but try and do a battle or skip turn instead of reload and you get a different outcome.

**Soul Forge** · July 27, 2008, 02:16 PM

yea i know about that. I was just pointing out that spy skills have less effect than previously thought.

**Musthavename** · July 27, 2008, 03:11 PM

In a recent Byzantine campaign. I had a merchant with a 60% chance of taking someones asset's. He failed, but I wanted to try something out. I kept reloading until I finally suceeded. The result, I didn't succeed. Ten goes and failure every single time.

Similarly in one english campaign in SS I wanted to get Istvan as a general (and I mean really wanted him, just to see what the fuss was about). When I got there my Princess had a 45% success chance. It took about 8 reloads to get it to work.

Now, taking note of this same merchant who periodically failed at the 60% chance, right next to him was someone with a 15% chance of suceeding. Out of boredom I clicked, and it worked. Just to test something out, I then reloaded and tried him again until i'd tested him 10 times. The result? Half of them ended in success.

I do remember having these weird probability problems in vanilla where it wouldn't succeed (as I would periodically spam reload sometimes back then), so I don't think its a problem with SS, but it must be the RNG itself. As you said, theres been too many instances of rare results thats its more likely bugged than random.

This probably does back up this:

it preserves a random seed. So if you want to reload cheat, just move another character then re attempt the mission

EDIT (adding a bit): This basically means that it roughly predertimes an outcome by this seed. So although things are possible, they may be very unlikely while they appear quite likely (or vice-versa).

**Barser** · July 27, 2008, 03:14 PM

In my experience it is not enough to just move a charecter.

**The Noble Lord** · July 27, 2008, 04:21 PM

Thats right, just moving characters is not going to accomplish anything.

**moosey** · July 28, 2008, 07:55 AM

if you want to try again and have a diferent outcome, i usually get a diplomat to just talk to anyone around them even if its just to close the menu immediately, after that it seems new outcomes are possible.

Other things also work such as battle, burning heretic, aquiring merchants, etc...

**Mune** · July 28, 2008, 10:08 AM

I haven't looked at the hard-coded math, but here's what I've found and/or what seems to be happening in the background:

As you guys have pointed out, causing any event to occur which has a percent chance of something happening (battles count, as well as diplomacy) will make a new random chance seed to generate.

However, there seems to be times when reloading after a failed chance-related event either will or wont make a difference. Here's how I picture the math in my head, which may or may not be accurate. We'll use spying as an example.

Say you have a spy who you want to spy on a city, with a 60% chance. Sometimes if you try, and fail, then reload a number of times, you will finally recieve the success you were seeking. Other times, it is hopeless and any number of attempts will end in failure (personally tested up to 6 or 7 loads). Assuming the existance of one random seed generated, why would this occur? My guess: There are 2 or more randomly generated seeds, one which is re-generated upon loading, of which the results are added together to check for success. Here's a hypothetical mathematic example, all numbers made up, again using our spying at 60% chance example.

Random Seed #1 - Generates numbers between 0 and 80 - Randomly generated a 50, calculated after the last chance-related event which occurred (or the beginning of your turn).

Random Seed #2 - Genrates numbers between 0 and 20 - Randomly generated a 5, calculated after the last chance-related event, beginning of turn, OR loading a game (whichever occurred most recently).

Add the two results together, you get a 55. Spy dies. Load game. This time, Random Seed #2 generates a 15, adding up to 65, spy succeeds.

As a separate example, say Random Seed #1 generated a 30. Any attempts to re-load will be futile, as the results could never add up to over 60. Time to give up, or find a way to re-generate Random Seed #1.

Again, the above example is just a made-up hypothetical, but it seems to make sense of the seemingly odd way the game handles chance-related events. From the perspective of a game-designer, I could certainly see reasons for implementing such a system. It pleases the hardcore players who don't like the idea of re-loading infinitely to allow a 100% success rate on all chance-type events, but also allows some control via re-loading over chance related events. The end desired result was likely that they wanted people to be able to re-load and try egain when chances were high (60% or more) but have little to no chance when it was low.

Here's some anomalies I have noticed:

It seems more likely that loading will make a difference if the chance is at least 60%, or depending on the random-chance event you are attempting (I seem to be able to get success via loading pretty easily with the burning of heretics, even at 20% chance).

Most chance-related events with low odd's have no chance of success. This would seem to indicate that Random Seed #1 is modified by a 3rd factor, which can cause it to result in a negative number. Or, Random Seed #2 can be a negative number. However, this does not seem true in all cases (like the burning of heretics example above).