On begging

[Donald Duck1]

Hi, so far I hardly ever beg for gold, so I read oedali's description of the begging mechanism at https://forums.civfanatics.com/threads/pushing-and-begging.253208/#post-6224468

If each turn there is a 5 % chance the AI will have forgotten a previous request, then begging every single turn will result in a successful beg once every 20 turns, on average.

Similarly, if every 10 turns there is a 40% chance the AI will have forgotten, then begging every 10 turns will result in a successful beg once every 25 turns, on average.

So having more turns inbetween begs increases the chance the beg is succesful, but decreases the average number of successful begs per X turns for all values of X. The more successful begs / turn, the sooner you have some of the gold, and gold now is better than gold later.

Now, obviously it would be very impractical to do a round of begging every single turn, while keeping track of the amount you can ask from each AI. But it can be an option to do this for a couple of turns in desperate situations, as there is no downside: A failed request resets the counter, but every turn beyond the first one after the request increases the success chance by a diminishing number. It's just more work.

I am aware that all of the above does not affect the total amount you can get from begging, as whatever you could have gotten in a failed request will still be available for the next request. Fewer turns between begging just gets you the gold a little bit earlier, on average.

Several people recommend about 20-30 turns between requests. I'd say it is the lower the better if you want the gold early, but in practice it just depends on the number of failed requests you can stomach. Using oedali's table:
1 turn...... 5% chance they forget
5 turns..... 23%
10 turns ... 40%
15 turns ... 54%
20 turns ... 64%
25 turns ... 72%
30 turns ... 79%
35 turns ... 83%
40 turns ... 87%
45 turns ... 90%
50 turns ... 92%
60 turns ... 95%

..then with 5 turns, it's about 3 failures out of 4, with 10 turns it's 3 failures out of 5, with 20 turns it's about 1 failure out of 3.

Any thoughts? Or am I talking nonsense overlooking something stupid?

 

[Fippy]

I cannot proof it, but my begs seem much more succesful when i have high power.

 

[sampsa]

Any thoughts? Or am I talking nonsense overlooking something stupid?

To me what you say makes perfect sense. It's a lot better to beg more often than every 20T.

 

[krikav]

It does sounds reasonable.
I haven't thought about it that way, but it makes sense to beg more frequently to get the goodies sooner.

When I play though, I don't really take care and keep track of how much I have begged and how much I can.

One thing that this strategy doesn't account for though, is that you play out a very important trump card.
The #1 utility in begs imho is the peace treaty, so I am hesitant to use it before I need that.
It's so useful to beg from onlookers before declaring on their friend, or begging preemptivly on that warmongerer close to you etc etc... If you beg for economical reasons you don't have that chance, you will just have very high odds of getting rejected.


*EDIT*
Oh, and also... It can be better to have those favours stored up at the AIs, than to make a large stack of cash that you have, even larger. If you do, that in turn increases the chance of Alexander comming and demanding all of your gold.
Better to beg when you are running deficit research and are getting close to emptying your coffers.

*/EDIT*

 

[Donald Duck1]

That's true, there are the diplomatic downsides of sacrificing credits, and counter-demands. Thanks for pointing that out.

 

[coanda]

When you make a request of the AI, it calls CvPlayer::handleDiploEvent (in CvPlayer.cpp) with a DIPLOEVENT_MADE_DEMAND message. That in turn leads the game to increment MEMORY_MADE_DEMAND_RECENT - telling the AI "you've made one more demand recently than before." I'm pretty sure (but not totally certain) that this function is called regardless of whether or not it was a successful beg (unsuccessful begs still increment the counter).

At the start of each turn, CvPlayerAI::AI_doTurnPre() is called. This is a general function that handles a lot of the early-turn AI tasks, like updating research and production priorities. Part of this is that it in turn calls CvPlayerAI::AI_doCounter(). AI_doCounter(), among other things, checks the memory counters for the AI. For each non-zero memory counter, it checks against the individual leader's memory decay setting for that type of memory (these are defined in CIV4LeaderHeadInfos.xml) with a 1-in-Value chance of decreasing the memory by one.

For every single leader in the base, unmodded game, the recent demand memory decay is defined as 20. This means there is a 1-in-20 chance, at the start of each turn, of their MEMORY_DEMAND_MADE_RECENT counter being decreased by one.

This does not mean that begging every turn should get one success every 20 turns though. If you beg on turn 1, 2, and 3, odds are high that you get one successful beg, two failed begs, and the counter is now at 3. Which means you'll probably need another 60 turns before you can get another success. Continuing to beg every turn is a good way to ensure you never succeed again at it, as you are pushing the counter up 1/turn and it is dropping on average 0.05/turn.

Every time you successfully beg, you know the counter is exactly one afterwards. Every time you unsuccessfully beg (assuming you know the amount requested was below the threshold), you know the counter is at least two afterwards.

So offhand I'd guess the "best" strategy is probably to wait about 25 turns after a successful beg, about 50 turns after a failed beg, before trying again.

Edit: Also, relative power does impact how much AIs will be willing to give assuming they aren't still remembering a recent demand. Weaker civs can beg less gold; stronger civs can beg more.

 

[sampsa]

Continuing to beg every turn is a good way to ensure you never succeed again at it, as you are pushing the counter up 1/turn

If this is the case it doesn't work like I thought it did and pretty sure @Donald Duck1 has the same misunderstanding as I do. I didn't know there is a counter that goes up, or if I did know that, I had already forgotten.

 

[coanda]

It's possible that the handleDiploEvent() function is only called by successful begs, in which case the "counter" would be limited to just 1 or 0 and essentially a boolean. But I'm fairly sure (based on how the code seems to be written, and a quick test where I begged 5 times in a row, then blitzed through 50 turns, then had two begs in a row fail) that it's acting as a counter which can go higher than 1.

 

[Donald Duck1]

Ok, then I misunderstood. So there is a 5 % chance to forget only applies to the most recent request. Then indeed it is not a good idea to beg often.
Thanks.

 

[6K Man]

I had forgotten about this table, thanks DD for reposting:

Spoiler :

1 turn...... 5% chance they forget
5 turns..... 23%
10 turns ... 40%
15 turns ... 54%
20 turns ... 64%
25 turns ... 72%
30 turns ... 79%
35 turns ... 83%
40 turns ... 87%
45 turns ... 90%
50 turns ... 92%
60 turns ... 95%

I tend to beg every 30 turns, and if they say no, wait another 50 turns. This is on Epic speed usually. Are the chance-to-forget percentages the same across all speeds?

 

[Gumbolt]

I wonder if this link explains that Mylene.

Land target beging. I think Kaitzilla also posted the begging formula last game.
https://forums.civfanatics.com/threads/land-target-and-manipulating-the-ai.561148/

'Begging_Ceiling = [((Turns_Known + 50) * 2 * ((Our_Power + 100)/(Their_Power + 100)) * 3 if Land Target]
Our_Power = Our Soldier count from demographics divided by 1000
Their_Power = Our_Power / 0.9
Divide the Begging_Ceiling by 2 if not in Financial Trouble to find the Gold_Ceiling.

We met Peter on T78 according to page 1.
It is now T138?
60 turns

60+50 * 2 * (0.9 roughly) / 2 = 99 ceiling

Ya, I'm 99% sure we can beg 40 safely'
Likely taken from the Homies SGOTM 26 post. I forget whch one.

 

[coanda]

Yes, forgiveness chances are the same regardless of game speed. Of course, the amount of gold you can beg also doesn't scale with game speed so it's not really a particular advantage (if you could successfully beg 50 gold on epic, you could also beg 50 gold on normal).

 

[Seraiel]

Hi, so far I hardly ever beg for gold, so I read oedali's description of the begging mechanism at https://forums.civfanatics.com/threads/pushing-and-begging.253208/#post-6224468

If each turn there is a 5 % chance the AI will have forgotten a previous request, then begging every single turn will result in a successful beg once every 20 turns, on average.

Similarly, if every 10 turns there is a 40% chance the AI will have forgotten, then begging every 10 turns will result in a successful beg once every 25 turns, on average.

So having more turns inbetween begs increases the chance the beg is succesful, but decreases the average number of successful begs per X turns for all values of X. The more successful begs / turn, the sooner you have some of the gold, and gold now is better than gold later.

Now, obviously it would be very impractical to do a round of begging every single turn, while keeping track of the amount you can ask from each AI. But it can be an option to do this for a couple of turns in desperate situations, as there is no downside: A failed request resets the counter, but every turn beyond the first one after the request increases the success chance by a diminishing number. It's just more work.

I am aware that all of the above does not affect the total amount you can get from begging, as whatever you could have gotten in a failed request will still be available for the next request. Fewer turns between begging just gets you the gold a little bit earlier, on average.

Several people recommend about 20-30 turns between requests. I'd say it is the lower the better if you want the gold early, but in practice it just depends on the number of failed requests you can stomach. Using oedali's table:
1 turn...... 5% chance they forget
5 turns..... 23%
10 turns ... 40%
15 turns ... 54%
20 turns ... 64%
25 turns ... 72%
30 turns ... 79%
35 turns ... 83%
40 turns ... 87%
45 turns ... 90%
50 turns ... 92%
60 turns ... 95%

..then with 5 turns, it's about 3 failures out of 4, with 10 turns it's 3 failures out of 5, with 20 turns it's about 1 failure out of 3.

Any thoughts? Or am I talking nonsense overlooking something stupid?

When failing a beg, the timer restarts at 0% chance.

Kind regards, Seraiel.

 

[Major Tom]

I find begging humiliating and shameful.

 

[Donald Duck1]

I find begging humiliating and shameful.

It's not begging, it's taking what's rightfully yours.

 

[ArchGhost]

If each turn there is a 5 % chance the AI will have forgotten a previous request, then begging every single turn will result in a successful beg once every 20 turns, on average.

Similarly, if every 10 turns there is a 40% chance the AI will have forgotten, then begging every 10 turns will result in a successful beg once every 25 turns, on average.

A lesson from proc effects in Vermintide:

pragmatism

If you have a 10% chance to do something, running the experiment 10 times should result in an average result of 1 proc per ten experiments, right?

No.

Average is a mathematical (calculated) median (as opposed to the ACTUAL median, which is ordinal) of the data set, so it pulls the low results higher and the higher results lower from each extreme.

Expecting an average result means right in the middle of possibilities....or half that. Proccing once in 10 trials at a 10% chance is the high result, at that rate. 0 is the low.

Much more practical then to assume a 1/10 gives a 50% chance of proc in 10 experiments, not 100%. If you were to extend that 1/10 chance to more trials, like 20 tries...you could be more expectant of a proc. Still not guaranteed, but more likely. probably 75% chance to proc at least once in at a 1/10 chance in 20 trials. 87.5% in 40 trials. More and more likely approaching infinity as the trials are extended. Etc Etc.

I am running on little sleep, haven't been in school in years so can't back any of it with math, and am most likely talking out my ass, but it seems to fit a whole lot better after a lot of frustration with the concept with proc effects in the Vermintide games.

 

[Donald Duck1]

Ok I when I wrote 'average' I meant 'mean'. I guess you may be right that the median may be a more practical metric than the mean.