|
kipley
|
 |
« Reply #2 on: April 28, 2010, 15:53 » |
|
I decided to do a bit of a "Mythbusters-style" test to see if the AI was winning more money gambling than it should (after performing many actions, so having less time left on the clock). If you have less time on the clock, you should generally win less money from the pub. From the FAQ:
"GAMBLING AT THE PUB
Once you enter the Pub, you're guaranteed to win some money, but you also end your turn. However, the more time you have left when you go to the Pub, the more money you're likely to win."
So anyways, I created a training game with me and 3 AIs. On my turns, I ran directly into the pub, doing nothing else. The AIs, of course, did their normal turns of developing their plots, hunting the wampus, and assaying land. So in theory they should have won less money from the pub, on average, than me, because they would have less time remaining on their clocks. Here's how it went in my first test game:
Round 1 - Me: 143, AIs: 83, 142, 78
Round 2 - Me: 229, AIs: 193, 171, 150
...then we had a fire in the store, which created a food shortage, invalidating the test (the test is only meaningful if everyone has full food. I wasn't producing food, but I was buying food until full each turn). So I created a second test game, which went as follows:
Round 1 - Me: 178, AIs: 224, 191, 70
Round 2 - Me: 177, AIs: 154, 235, 213
Round 3 - Me: 71, AIs: 118, 174, 71
Round 4 - Me: 250, AIs: 250, 100, 192
Round 5 - Me: 136, AIs: 248, 242, 229
Adding up the totals for all seven test rounds, we get:
Totals - Me: 1184, AIs (averaged): 3528 / 3 = 1176
So, it would seem from this (admittedly small sample-sized) test that a human who does nothing on his turn (and has maximum clock time when entering the pub) will receive from gambling just about the same amount of money as an AI that does a regular "busy" turn (and should have a lot less clock time when entering the pub).
In my mind this test lends credence to my original theory that the AI is winning more money than it should. On some of the computers turns, it does a LOT (developing, catching the wampus, assaying 3 plots), and STILL gets a whole lot of money gambling.
On the other hand, the test did surprise me in how much variance to how much you'll win from pub gambling. I was under the impression it was much more consistent than that. I guess after so many games the pub amounts just become "noise" and you don't really notice how much the differ. It actually explains some things... I've often been confused after the first round why one bot is in the lead instead of another, given their production, etc.... and I now know it probably was just that one got much luckier in the pub than the other.
Both tests were done with "observe computer development" turned off.
|
