Hubert Cater -> RE: AI cheating (4/1/2021 8:59:50 PM)
|
quote:
ORIGINAL: nosfer121 I don't think that the AI cheats on the battle field, but in the economical part of the game. My concerns started with my last game as the Axies against the AI. I played with the default difficulties, and all went quite normal for me. In the June 1940 turn I have captured Nanning, and in the May turn Changsa and the towns between them. I looked into the statistics and I was astonished to see, that the Chinese have 44 units left. They start with 30, and I know that with the fall of Changsa and Nanning a human chinese player gets 1 army each. It is the event "Destroyed forces of Changsa rebuild there army at ...". With the units I have destroyed so far, the chinese have had to build more than 2 units a turn and replenish there troops in oder to achieve 44 units. I know that this is impossible to a human player. When I played the chinese the MPPs were barely enough to replace the losses, but not to build up my forces by 50%. The AI as mentioned above does receive free units to it at no cost and these are listed in the OPTIONS->ADVANCED->SCRIPTS screen under the UNIT event list. The Chinese specific ones are listed starting on page 18 and listed under ALLIED AI: China - etc. Based on your description and the advances you achieved, and how things are scripted for the Chinese AI here, they would receive about 10 units. For example, had you not taken Nanning and Changsha and not pushed towards Sian they would have received less. Replacing losses for the AI is the same as a human player, e.g. 60% cost, plus the Chinese have 4 units in the Production Queue that would have arrived by the end of June 1940, so having around 44 units would make sense within this entire context. Again, these units are added in to help the AI accordingly and can be disabled at any time in the OPTIONS->ADVANCED->SCRIPTS screen, however, if the goal is to get a better feel of what it is like to play a human player I almost hesitate to suggest turning these off. I just suggest this as the entire reason they have been added is to make up for some of the AI shortcomings versus how a human player would play. Typically the AI has a harder time making up for losses and planning accordingly for this, this can snowball quickly in a bad way for the AI, which is why a few extra units here and there based on game play (in this case to help it defend a bit better while it recovers from losses) was found to help it the most in this regard. But of course the option is indeed there to disable it which is why we've added it in and made it accessible to players to do so as they see fit. quote:
I start a new game, and wrote down the start situation of every Allies nation in an excel file. One sheet per nation. I made screenshots all over the world to cover all units and to see which units will be replenished. Then I played two turns as the Axies, saved the game, load again and take the control of the Allies. In this way I could see, what the AI Allies player did in his first turn. In this manner I played the first 4 turns, taking fresh screenshots, fill in the lines in the excel file etc. Here some examples: England starts with 125 MPP. In the first turn there overall research costs raises from 1200 to 1350. How? In the same turn they bought a Canadian light cruiser (225 MPP) and replenished troops in Africa for 45 MPP. How with 125 MPP at the start? Thanks for these very specific examples as this is helpful and I was able to dig down and figure out what was happening exactly. It looks like there is a bug within the AI research chit assignment routine, e.g. if you watch the turn the unit reinforcements are acting as they should, reinforce and MPPs are being deducted accordingly, then at the very end of the turn the AI performs its research checks and this is where it is assigning a chit without paying for it. I run through a series of loops here and the error doesn't always happen, as there is some randomness in what the AI will invest in based on the scripts and in my case I had to run through a few times before I saw it. For me it was a 125 MPP investment into Production Technology that put it at 1325 research MPP spent when it only had 2 MPP in the bank at that point. Regarding the Canadian Light Cruiser, I didn't see this on my end, I checked the Production Queue and at the start the Canadians are already set to receive 2 Destroyers and there is a New Zealand Cruiser also already in the queue, perhaps this is the one you saw? If not please do feel free to send me a saved turn to support@furysoftware.com and I'd be happy to take a closer look. quote:
In the second turn, they again invest 175 MPP in the research, but also 169 MPP in replenishments. They start turn 2 with 277 MPP. So they spend 67 MPP more than they start the round with. In every 4 turns England uses more MPP than they start the round with. France starts the third round with 128 MPP. But they can achive a research option in that turn for 175 MPP. China starts the 4th round with 206 MPP, but is able to raise the research budget from 550 to 800. My impression is, that the AI orders units, research options and replenishments as is needed, regardless of the MPP costs. That is a bit more cheating than a unit or two more once in a while. As I stated above, I play the AI mainly to test strategies for the next PBeM.But it is really difficult to evaluate the results of submarine strategies or a war of attriction, when MPPs doensn't matter for the AI. No the AI as mentioned should act within the same rules as a human player and if it doesn't then it is either an oversight or a bug. I only say this as there are just too many ways for players to verify that the AI is acting accordingly for us to get away with any of these kinds of cheats while at the same time claiming the AI does not cheat. For example, this is why anyone is able to see with Fog of War off what units the AI reinforces and how much it has spent to see it is above board. Comparing what the AI starts with in the Production Queue when viewing in the Editor and then comparing against what it has purchased in the first turn, or in any subsequent turns is easy enough to see that it is spending within its limits. Comparing the results of a turn in Fog of War versus not in Fog of War can also verify that everything is above board as well. Looks like there was however a small bug in the Research routine, and to be honest I'm surprised it hasn't been noticed until now, so thank you for this and consider it fixed for the next update.
|
|
|
|