Agent Success Rates

[Marauder]

I've been messing around with agents for a while, specifically assassins and spies. Both of these have a certain percentage of success ranging from 5% to 95%. I've found that these numbers can be misleading. Using the magic of time travel (ie, reloading my saved game a bunch of times) I find that assassins/spies seem to hit a "barrier" in their missions, where they're almost guaranteed to fail on that turn. I think CA has a tally of how many successes they've had in a row and reduces the chance of the next mission. Has anyone else delved in this or noticed any patterns?

I can reload my saved game 20 times and the assassin will still fail a mission that the is listed as a 30% chance of success.

[PBI]

I believe the problem is that when you save the game, it saves the random seed as well, so the next "random" number generated will in fact always be the same number as before. Just by reloading the game and repeating the action (takeover/assassination), you will always get the same outcome whatever you try.

In order to get a different outcome, you need to do something else first that relies on generating a random number. So another agent action, or I think also diplomatic offers and autocalc battles use random numbers.

One thing I did notice though, I think mission outcomes may work as you describe. I once got a reward for a "you will be rewarded with some units" mission which I didn't like, so repeatedly reloaded to get a better outcome (yes, yes, I know, naughty me.) Anyway, the units it gave me became progressively worse each time until eventually it was just giving me 3 peasant archers every time.

[Marauder]

lol @ 3 peasant archers. Totally worth starting a war with that faction by blocking their port?

I don't think the random number is saved in the saved game, because I can get it to eventually kill someone on the 30%+ range. The percentages just don't match the number of reloads it takes to complete the mission.

Along the same lines as your mission rewards finding, I autocalced a battle to see if I could do better than the computer seiging a town with cannon towers (I couldn't unless I brought culverins to take them out first), I reloaded and autocalced again with a worse result than the first time. After 5 reloads and getting the same (bad) result, I decided to wait a turn for my seige equipment to arrive.

[Marauder]

Something I didn't think of trying is quitting the program and restarting. That would refresh any temporary variables stored. That would take forever though since M2TW isn't exactly speedy on loading.

[FactionHeir]

Nope, that wouldn't do it. You can change files though (like replace the AI file) which will reset the seed, as the AI file is loaded each time and recalculated with chance based events.

[Sentinel]

I think that PBI is correct in saying the game saves the seed for the random number. The game has more than one random number depending on which action you are trying to do. Some of these actions require only one of these seeds being used, whereas some require a series of different seeds.

If you end a turn and save the game a set of seeds are saved. These saved seeds are not the same as the seeds that the game is actually using at that time, but a different set.
For example,
End turn, save game, then use an assassin and you will get an outcome, result 1.
Reload the saved game, and use your assassin and you will then get a different outcome. result 2.
You can reload the game 100 times, but you then always get result 2. Because the seed for that calculation was reloaded with your save.

The same works for all the decisions made by the computer.

There are different seeds saved for different types of calculations.

Reload a saved game (A), then auto calculate a battle and you will get a results (X), a set of numbers giving your losses, their losses, prisoners taken etc.
Reloading save A will always give you exactly the same result (X), the same set of numbers.
If you reload save A, but this time take a unit out of a settlement before auto calculating the same battle, you will get result Y.
This is because one of the battle seeds is used to choose the captains name for the unit you just removed from the settlement and the computer generates a new seed for the next action.
You can repeat this 10 times and if you remove only 1 unit from the settlement you will always get result Y.
If however you remove 2 units (1 at a time) you will then get result Z.

However, If you replaced the battle with an assassination, then removing units from a settlement this will have no effect on the outcome of the assassination.

This suggests to me that the game saves more than 1 seed.
At any given time it has more than one seed for use in different types of calculation.

Actions like assignation, merchant takeovers, priests converting heretics, spies entering settlements all use the same type of seed.

Larger events such as ending a turn, navel battles, auto-calculating sieges etc require a lot of random numbers and therefore seem to use (therefore change) all the seeds.

This doesn’t mean that the % success is meaningless, it’s just that the effect that the random number has on the calculation is so large that it make it look that way.

[Uunts - a proud Estonian rebel]

Hi guys, a good tip for you. If assassination or spying fails, then you should load the game and before new attemt fight a battle. Then try again, almost every time they succeed. So reading your debate here makes sense on how the succeedcalculator works.

[Marauder]

Thanks for the details Sentinel, its interesting to see how they coded the game. There could be multiple seeds involved with an assassination since it isn't just pass/fail. There are different movies shown and the agent can either die or not die on a failure attempt.