NEW Poll: AI needs Improvement

[nejxis]

So I've been playing this game for some time now, and I am seeing that once you learn how to play it efficiently you always end up the richest in the world in a matter of years and being able to buy most of the other competitors. The reason I believe this happens is due to the AI not being good enough.

1) For example I use like 700% difficulty settings with very low capital for me and massive for them, but I use debt to leverage my assets and build apartments and commercial buildings. In 20 years I will have dominated the real estate market and can easily expand in other areas. They don't use debt as leverage to expand, so they are destined to a very slow growth.

the fix would be to make a ratio of assets to debt that would fluctuate based on the Charimen / CEOs Risk Appetite, also it should be based on the industry they're in. For example real estate is widely known as one of the most, if not the most industry that uses debt to leverage their assets and expand.

2) I've also noticed that they keep too much cash reserves and not expand as they should. I mean a company worth 1bil should not have 500mil cash in the bank doing nothing.

I would suggest making a capping of max cash reserves as a percentage of the total net worth, and making it in relation to current economy. For example huge companies like Apple and Berkshire Hathaway are currently keeping a lot of cash waiting for the next recession to be able to buy out other companies / assets at cheaper prices.

3) They are not competing enough. For example in the operating systems I rarely, if ever see two operating systems, unless I created them. For sake of competition I sometimes create like 3 OS & software companies and sell all their shares early on to make it more fun and challenging, and let them expand independently.

I see two simple fixes here. Either increase the number of competitors so that if the game mechanism is based on probability of say 5% that it makes an OS, then the higher the number the greater the chance (Law of Large Numbers). Or straight up just make a minimum direct competitor. You can either add a mix of both.

4) expanding in random industries. I've seen companies sell software and then create furniture. Who does that? If they want to buy out a furniture company and become a conglomerate then yeah, but Adobe will not sell Photoshop and a cupboard. Never.

Simple. Just do not allow companies to expand more than their Class, and if they merge with another company who has products in another class, they should create a subsidiary and transfer those firms/research to them.

5) I see a lot of products that are not being developed by anyone, and leave too much "open space" for any competitor to just enter and dominate 100% the product easily. It should definitely not be like this.

6) They will not try to maintain their dominance of their product. For example in the telecom industry there is a company that starts off with 7 telecoms around 7 cities and dominates most of them. If I decide to enter the market and reduce my prices, they will not even try to reduce it a bit to keep prices competitive. In a matter of years, I would have gained at least 50% market share from them due to lower prices without competition

[buells]

I find the AI tends to become extremely bloated later in the game. They diversify beyond their product category expertise and beyond what their technology assets and high expertise farms can justify. While diversification is fine and some AI competitors build successful franchises across several different products, they should not choose to continue losing massive amounts of money on products that simply don't move off shelves. AI can have literally hundreds of unprofitable retail stores and dozens of useless factories / R&D facilities. They need more discipline about both making new investments and realizing when it is time to throw in the towel in a given product or product category.

Having debt financing play a bigger role for the AI may help with this because then they would be more constrained by the ROI on their late game investments.

[David]

Different AI CEOs have different tendencies and preferences for business expansions. Did you observe common behaviors from various AI companies that expanded overly aggressively resulting in bloated and inefficient operations? If so, please provide us with your save game and the details. We will then look into it.

[David]

So I've been playing this game for some time now, and I am seeing that once you learn how to play it efficiently you always end up the richest in the world in a matter of years and being able to buy most of the other competitors. The reason I believe this happens is due to the AI not being good enough.

1) For example I use like 700% difficulty settings with very low capital for me and massive for them, but I use debt to leverage my assets and build apartments and commercial buildings. In 20 years I will have dominated the real estate market and can easily expand in other areas. They don't use debt as leverage to expand, so they are destined to a very slow growth.

the fix would be to make a ratio of assets to debt that would fluctuate based on the Charimen / CEOs Risk Appetite, also it should be based on the industry they're in. For example real estate is widely known as one of the most, if not the most industry that uses debt to leverage their assets and expand.

2) I've also noticed that they keep too much cash reserves and not expand as they should. I mean a company worth 1bil should not have 500mil cash in the bank doing nothing.

I would suggest making a capping of max cash reserves as a percentage of the total net worth, and making it in relation to current economy. For example huge companies like Apple and Berkshire Hathaway are currently keeping a lot of cash waiting for the next recession to be able to buy out other companies / assets at cheaper prices.

3) They are not competing enough. For example in the operating systems I rarely, if ever see two operating systems, unless I created them. For sake of competition I sometimes create like 3 OS & software companies and sell all their shares early on to make it more fun and challenging, and let them expand independently.

I see two simple fixes here. Either increase the number of competitors so that if the game mechanism is based on probability of say 5% that it makes an OS, then the higher the number the greater the chance (Law of Large Numbers). Or straight up just make a minimum direct competitor. You can either add a mix of both.

4) expanding in random industries. I've seen companies sell software and then create furniture. Who does that? If they want to buy out a furniture company and become a conglomerate then yeah, but Adobe will not sell Photoshop and a cupboard. Never.

Simple. Just do not allow companies to expand more than their Class, and if they merge with another company who has products in another class, they should create a subsidiary and transfer those firms/research to them.

5) I see a lot of products that are not being developed by anyone, and leave too much "open space" for any competitor to just enter and dominate 100% the product easily. It should definitely not be like this.

6) They will not try to maintain their dominance of their product. For example in the telecom industry there is a company that starts off with 7 telecoms around 7 cities and dominates most of them. If I decide to enter the market and reduce my prices, they will not even try to reduce it a bit to keep prices competitive. In a matter of years, I would have gained at least 50% market share from them due to lower prices without competition

From time to time, we received emails from users complaining that the game runs too slowly at late games. Here is our standard reply to them:

In Capitalism Lab, whenever a new firm is set up and in operation, it requires additional CPU processing time. At the late stage of the game, where there are thousands of firms, the game will inevitably run more slowly.

We would like to draw a parallel to other strategy games like Civilization. When you play in a late game of Civilization, the AI also takes a long time to think and execute orders in a single game turn. In fact, it is inevitable that the more complex the game has become, the more CPU resources are required to process it. Please understand that this is not really a bug or a problem with the game, but merely a fact that more data require more CPU time to process.

Therefore, we would recommend that you start a game with fewer AI competitors (15 or fewer competitors) and fewer cities (4 or fewer cities) and see if your computer can handle it well.

Another tips is that if you just want the game's time to pass faster without any interaction with the UI, then you may open the Manager's Guide or the Farmer's Guide which requires less graphical processing and the game as a whole runs faster.

The thing is that coding a more intelligent AI will inevitably increase the CPU usage. And the more firms the AI sets up, the slowly the game will run, and in turn this will lead to more users complaining about game slowness. There is no such thing as a free lunch - having increasingly complex game logic with introductions of new features from various DLC and demanding more intelligent AI and yet expecting the game run fast as before. I know that the later is not what you expect, but some users may expect that or get upset with the slower performance. So perhaps the current state of the AI is at a decent point of balance considering all this?

What are your views on this?

[nejxis]

Having a better AI would probably result in better performance of the game on the CPU as well in my opinion, and here's why.

If the current AI for example expands in retail and opens 5 stores in a city. Currently the locations are not always the best, sometimes they open a shop in a zone of 12 customers rather than 60+.

If the AI was better he would either open in the good location (60+), or just not open at all, which means the same amount of firms or actually less.

Same thing with opening extra factories without maximising the efficacy of the previous one first and so on and so forth.

Improving the AI can also make things more organized and care about their brand and make it more realistic. Just because a company is expanding in software and furniture and opened 30 firms, doesn't mean that more firms will be opened if the AI decided to make the furniture section in a subsidiary company.

To conclude I believe better AI will actually be better in itself for the performance of the game as it will close down extra firms, not open firms in bad locations in the first place and so on.


- To answer your question directly, it is true that it slows down year on year, however in my opinion and some people will agree and others will disagree, I would much rather have a slower passing game with better AI.

This is because with better AI I would be able to play on 1-4 cities with less competitors, but they will be harder to dominate - Currently if I start low budget and they massive with 4 cities and 20 competitors I would probably dominate all industries in 30 years.




- Just a couple of small questions pls:

1) regarding performance, would a bigger Map, but less cities make it better in terms of speed? That is increase the size of each city and lower the number of cities?

2) Would creating bigger factories that can handle more production than the current big ones be better for performance as for example 1 Factory could handle 5 times as much as the biggest one currently, but only 1 firm was built?

3) With the addition of say shopping malls, there will be for example 10 shops per mall, in terms of performance would it make it better than say opening 10 shops? Because if that were the case we should start thinking about implementing the same strategies for different things.

[David]

I will communicate with the dev team about your suggestions.

I think the sensible way will be to improve an aspect of the AI one at a time. As it is going to be very time consuming to improve all aspects of the AI and the dev team is currently busy working on other tasks. It will be wise to pick an AI improvement task that is easiest to implement and yet has the most impact to start with. Do you have any suggestion on which one this might be?

[lagrelax]

I will communicate with the dev team about your suggestions.

I think the sensible way will be to improve an aspect of the AI one at a time. As it is going to be very time consuming to improve all aspects of the AI and the dev team is currently busy working on other tasks. It will be wise to pick an AI improvement task that is easiest to implement and yet has the most impact to start with. Do you have any suggestion on which one this might be?

One thing I can suggest is the profit-cost awareness of AI, as I observed the following:

-- In the expertise area of the CEO, AI actually randomly pick one or two product to produce. This usually leads to producing less-needed goods and leaving more profitable ones open, assuming all input components are available. My strategy is always starting with the one of the most revenue in the product details tab. This should be easy to implement

-- AI also do not care much about the operation overhead of factories, 99% of time they establish factories of way more capacities than actual needed. Combined with the point above, they can't make a decent profit on the product they dominate. Here I also share my rule of thumb to decide the proper size of factory -- based on the total revenue of the product details, build the cheapest large factory when >2M, then middle for goods between 750k to 1M, and small for anything below.

[chengtsai]

1. Each unprofitable firm needs to do a break even point analysis. If failed to pass this test then close it down immediately. Of course, they can listed on sales months before shut them down.

2. AI needs to be better job on finding raw material/parts that is cheaper, higher quality and cost less in freight charges.

3. Stop AI from entering a industry that is not related to it's core business.

4. If AI found a way to be profitable in certain product, then they need to go all in and expand as quickly as possible.

5. Technology movements between companies are too slow or even happening. AI lost millions to open R&D, and it is way cheaper to purchase a tech than develop one.

6. Any firms have low customer traffic should consider to relocate.

[David]

Could someone please set up a poll listing all suggestions for improving AI and see which one is most agreed by the community as the most important one for the dev team to work on first?

[nejxis]

I added a poll with the suggestions above from the different posts.

Please vote so as we can start getting improvements.