Weird Observation

**Atalus** · 04-01-2007, 05:12

Just had a thought about re-trying failed agent missions using the save feature. Do you think that it's possible that when you turn starts or ends he's given a % that he needs to get otherwise he will always fail the mission barring stroke of genius? I am not sure how to describe this any better due to i just can't find better words atm.

**sapi** · 04-01-2007, 06:58

I believe that at the end of each turn, a base random number is generated.

This is then modified according to other factors to determine whether the agent succeeds.

**_Tristan_** · 04-01-2007, 07:44

Each time you save/load to retry an agent mission (mostly due to failure... I presume), you'll have to reload the seed (that is to say the way that the AI generates the chances of success)...

In order to do that, you have to make the AI do another calculation, either by trying another agent mission, or fighting a battle...

That way, you should get a different result from the one you had earlier...

**Whacker** · 04-01-2007, 11:09

Originally Posted by Tristan de Castelreng

Each time you save/load to retry an agent mission (mostly due to failure... I presume), you'll have to reload the seed (that is to say the way that the AI generates the chances of success)...

In order to do that, you have to make the AI do another calculation, either by trying another agent mission, or fighting a battle...

That way, you should get a different result from the one you had earlier...

Not true, as far as I've tested so far. Reloading the game and doing nothing else (this is tested on assassination attempts) but the action in question will yield different results. To be absolutely sure I parked my assassins directly next to and facing the targets, so literally there was nothing done but the attempt itself. There was some interesting discussion in a few other threads on this.

**Nebuchadnezzar** · 04-01-2007, 11:20

Originally Posted by Whacker

Not true, as far as I've tested so far. Reloading the game and doing nothing else (this is tested on assassination attempts) but the action in question will yield different results. To be absolutely sure I parked my assassins directly next to and facing the targets, so literally there was nothing done but the attempt itself. There was some interesting discussion in a few other threads on this.

But did you do the attempt in exact same order each time after reload or not? I believe if the order is altered then the results also are altered.

**Atalus** · 04-01-2007, 19:20

Ok well lets say that a base number is generated and it doesn't reseed when you load. Very, and i mean Very rarely you can pass a mission off of a save/load. But I trying to do it again and again but I not seeing this happen more than 1:100.

For the record no i didn't test this 100 times. I have a life outside of this game

but maybe for those who play warhammer or games similar there is that chance that a second roll is made out of 100 and by chance should the computer decide that the character gets a 100 you pass. but on the reload when it rolls again the chances of getting that 100 .... well we know the odds

**vonsch** · 04-01-2007, 19:44

My testing suggests the saving will force a new seed. So when I did save-try-load-save-load-try-load-save-load-try-etc I got varied results. When I kept reloading the same save I always got the same result. My test pool wasn't huge. I think I did 10. The save-load-try sequence varied results considerably. The try-load-try sequence varied it not at all.

**_Tristan_** · 04-02-2007, 08:11

Originally Posted by vonsch

My testing suggests the saving will force a new seed. So when I did save-try-load-save-load-try-load-save-load-try-etc I got varied results. When I kept reloading the same save I always got the same result. My test pool wasn't huge. I think I did 10. The save-load-try sequence varied results considerably. The try-load-try sequence varied it not at all.

My own testing suggests the same and whatever the agent (spy, assassin, merchant, priest...)

@Whacker : I even did it at the death of my king to prolong his lifetime...

Anyway, it always worked for me... I'm surprised you didn't get the same results...

**PutCashIn** · 04-02-2007, 08:51

The Seed for the Pseudo Random Number Generator may be based on the compluters internal clock, thus reloading within the same minute, say, may result in the same 'random number.'

...?

**Whacker** · 04-02-2007, 09:12

Originally Posted by Tristan de Castelreng

My own testing suggests the same and whatever the agent (spy, assassin, merchant, priest...)

@Whacker : I even did it at the death of my king to prolong his lifetime...

Anyway, it always worked for me... I'm surprised you didn't get the same results...

@Nebuchadnezzar:

No sir, I literally didn't do a single thing except the assassination attempts during my testing. Quickload followed immediately by clicking on my guy, then the target, and lettin' 'er rip.

@Vonsch and Tristan:

My testing did not include saving prior to reloading, so it's entirely possible that Vonsch is correct. Without CA actually telling us how they implemented their code, the best we can all do is just guess. It's entirely possible that they've set up their code (assuming it's C++, could be C for all we know) to use the basic srand and rand routines, and they only employ the srand function once when the game loads up. Or perhaps they do it each turn. Or perhaps they do it before each action that requires a random number generated. Or perhaps they don't even use these functions and wrote their own, kinda like Bioware supposedly did. /shrug

As you pointed out Tristan, you saved your King for another turn by reloading. I'm guessing that there's some kind of function that checks how old the codger is and there's an increasing chance he'll die of old age as he gets up there in years, presumably ending in a 100% chance after a certain age to prevent Methusalahs running around. Saving/Reloading or a combination of the two may not have any affect on this either, as I'm guessing this calculation is done during the end of turn stuff after you hit that button.

I'm going to go out on a limb here at this point and suggest that the whole "random seed" part is probably pretty moot. CA's been doing this for awhile, I don't seem to remember having complaints about agent mechanics in terms of success rates in MTW or RTW... As I've said in a hojillion other posts, I still think something is busted with M2TW agent mechanics, at least with the assassins. At this stage of the game based on my own experiences and reading what others have to say, my best guess is that the probability of success displayed is simply not accurate, because it doesn't take into account a number of factors that mainly lower one's chances. In short, my guess is the game is actually working how it's supposed to, it's just that the player isn't given an accurate picture of what the situation is, hence the complaints about useless assassins, tests that don't match up with what's displayed in game, etc...

**_Tristan_** · 04-02-2007, 10:10

Don't think so...

I've reloaded some games one or two days later and got the same results...

I think the seed is based on the actions taken during game...thus moving an army or agent, accomplishing any mission, will reset the seed and gie you completely different results on reload...

**_Tristan_** · 04-02-2007, 10:25

Originally Posted by Whacker

I'm going to go out on a limb here at this point and suggest that the whole "random seed" part is probably pretty moot. CA's been doing this for awhile, I don't seem to remember having complaints about agent mechanics in terms of success rates in MTW or RTW... As I've said in a hojillion other posts, I still think something is busted with M2TW agent mechanics, at least with the assassins. At this stage of the game based on my own experiences and reading what others have to say, my best guess is that the probability of success displayed is simply not accurate, because it doesn't take into account a number of factors that mainly lower one's chances. In short, my guess is the game is actually working how it's supposed to, it's just that the player isn't given an accurate picture of what the situation is, hence the complaints about useless assassins, tests that don't match up with what's displayed in game, etc...

I agree on that last point you've made...

There is certainly a behind-the-scene calculation that is not figured in the probability of success displayed thus the discrepancies between what your agent should accomplish and what he really does accomplish...

Nevertheless, on the save/reload question, I'm sure that whether it is a end turn calculation (such as character demise) or in turn calculation (assassination attempts, buying assets or spying), if the results don't match the success you should have expected, a save/reload should obtain new results...

I did this a lot at the start of my campaign... I couldn't afford to lose 500$ assassin with 70% chances of success to the unpredictability of X-unknown factor not figured in the chances of success displayed...

Thus I saved before the attempt and if failed, I reloaded (moreover I don't have the anti-trait fix so you know what I mean hero to zero in a second...)

I would then go on with my game moving at least one other character or agent and saving anew...
Then I reloaded and got back to my first assassination attempt and voila, this time, it was a success...

I've done it quite a few times, and speaking from experience, it is not a guaranteed success but you can get the feeling that something different has been happening...

Hope you see my point...