Accuracy of agent success percentages

[Tusk]

I'm not sure if anyone else also has questions about this - if it has already been posted I apologize, but a fairly thorough reading of the forum does not make it apparent...

I have had some surprisingly low success rates with agents, assassins + merchants in particular. I'm not a save/reload player, but I decided to try something. I saved before an assassination mission, failed, then reloaded and tried again. Even with a 75% chance for success, my assassin failed to take out his target in 10 reloads! Are the actual results of the mission pre-calculated before you undertake them?

I tried something a little different, and used a spy on the target immediately before the same assassination mission, and succeeded first time. I now pair spies with all agents, as it seems to increase their chances of success since I have already lost far too many agents on what should be relatively easy missions.

Has anybody else encountered this?

[FactionHeir]

There are 2 outcomes to an action before/after load. You can reset that by performing a battle, entering diplomacy or using an agent to perform a special action. This will set it back to 0.

I.e.:
You save. Send an assassin, fail. Reload, send again. Those are 2 seperate calculations.
Now if you reload and do that again, you will always get the second outcome.
However, if you reload and use another agent, that agent will only have 1 outcome as well as the generator is set on only 1 outcome now for any agent (be it success or failure).
Therefore, the best way to reset the generator is by using an assassin/spy/priest/merchant that has a 95%-100% chance of suceeding and performing a mock mission (note that at 95% you can still fail).
After that mission, save and you get 2 tries again.
et cetera of course.
Best is to use a spy in a settlement to spy on a nearby foreign army/agent.

[Rothe]

So, do not try to do statistical tests like that, since it either results in 100% success or 100% failure...

What this means is that the random number generator has a seed number that does not change if you save-load and repeat the same action. You have to do certain things (mentioned in the above post) to get a new "fresh" value, otherwise you are getting to see the effect of the already calculated value over and over again.

[Tusk]

Thx - I noticed this with some other games also...wasn't sure if it was the case with MTW2 because I did notice at least 2 different results. This also explains why using a spy tends to increase success rates for agents when there has been a reload involved somewhere...

What about autoresolved battles? I NEVER autoresolve land battles, but I wonder if the same applies.

[Murf the Less than Average]

I'm having a hard time with assassins. In my most recent game, I didn't use them much until I go the assasins guild. Then I produced a few and "trained" them on low level diplomats/princesses/captains/preists. I finally got one to a full rating and went hunting for big dogs. I killed a couple generals, good. I notice that my enemy the Malinese have 3 cardinals with a perfect peity rating and have the top preferati cardinal. I go after him and kill him with a 24% chance. Awesome I think. Another Malinese cardinal takes his place and he's in the same general area. Again 24% chance and a take him out. I try another time at 24% and miss, but my rating stays up and my assassin doesn't die. I move on to practice on a lower level agent to keep all my stats up. I find a 0 piety priest (I think it was a priest, but it might have been some other agent with a 0 rating) and get a 95% chance of success. I go for it and my assassin dies. Was it just the odds catching up with him, or is this more random than the percentages would have you to believe?

Don't roast me too hard on this, first post and all.

[FactionHeir]

Its random. I had skill 10 assassins fail on a 95% mission before on the first try. Just reload and try again or use my above method till you get your kill.

Tusk: It also applies to auto resolve, yes, as it is a generated event. It even applies to events and generals you get at the start of a new turn (you can influence those by using a spy before ending turn for example)

[katank]

Percentages can be funky. I've had 16% assasinations work the very first time before while I've had 93% infiltrations for spies on the first turn fail before. Losing your starting spy on turn 1 can suck big time.

[Dunc]

I've had a lot of 95% assassinations fail and I've never had a 5% assassination succeed.

[Tuidjy]

There is something fishy about assassins. In one of my games, I had an
assassin fail every single mission during a play session. Yes, he failed
18 times in a row, without dying. The chances for success were anywhere
between 35% and 5%, he never succeeded, and he never died. I play
without reloading, so all this happened in one uninterrupted session. By the
end, of course, he was a level 0 assassin.

The next day, he succeeded in taking out the King of Portugal at 5%, and
died on the next turn. So, whatever was going on, the reload fixed it.

[Oaty]

///there are also retinnues that give + to asassination/sabotage but don't increase the chance visibly. Get a level 10 asassin with lots of retinues and watch his killing rampage steamroll anything that he has 50 percent chance or better on, He'll eventually die but he knew the consequences of his actions.

[Whacker]

My experience with agent success percentages is that they are completely and utterly worthless. I never have anything close to the success rate "as advertised". Some time ago I did come across some information that would lead me to believe that there are undocumented and undisplayed factors at work, specifically such as spies in the vicinity. The theory is that the percentage given when you open the screen is correct, but there are certain modifiers that are taken into account when the actual result is calculated depending on who's within a certain range and how far away they are. While still aggravating to no end, it does help explain the discrepancy. Personally, I think (if this is the case) that it stinks, and that a spy several miles away is not going to affect the outcome of one of my weasels trying to off some poor merchant squatting in the woods... /shrug

Cheers!