Global Stock Market

[JasonLJJ]

Guys

For starters, thank you for all the hard work on the global stock market - I think the layout is pretty perfect and absent a few minor tweaks, this should be a viable game mode.

Immediate Reaction on gameplay:
Very hard to decide when to invest and what to invest in given the lack of information
Dividend yield goes from 3-5 % in the initial phases of the game to <1% in later stages of the game
Generally, stocks trend lower overtime (more than half the stocks in the market over a 20+ timeframe have lost value in my game)

Proposal:
Stocks be grouped into two categories - Growth stocks and value stocks
Growth stocks - price more volatile, with volatility in-line with nation economic growth cycle
Each upward movement from lowest point to highest point should be 50-100%
In short, P/E ratio should swing from lowest of 15 to highest of 30 times
should have dividend yield between 1 - 3% with dividends increasing between 5-10% every year
Each stock market boom-bust cycle should be around 5 years
Price action wise (over a 20 year horizon), the price chart should look like 1) Steep, increasing slope; 2) M-shape with the end point of 5 years 20-30% higher than beginning of the year

Value stocks - price stable; limited volatility (-10 - +15% swing in P/E ratio over 5 years)
In short, PE ratio should go from 9 to 11.5 times
Dividend yield is around 5-6% with yearly increase off 1-3% every year


Growth stocks - TMT, Capital Goods, Energy, Banking, Consumer Goods, Retail

Value stocks - Insurance, Utilities

[buells]

I don't feel that it needs to move with the in-game economy so much, but that's fine if that's the preference.

In my view, there should be some correlation of expected return to the variables shown (P/E, P/B, div yield, momentum too perhaps) when you buy. Then some meaningful random/pseudo-random variation in the return. My suggestion is conditioned on the assumption that that doesn't require a ton of computing power. If it was set up this way, you could also correlate each stock to a certain degree depending on industry exposure, but that's probably too much work.

Perhaps there can be a set of return time series that are randomly assigned among the different stocks at the outset of each game? They could be broken up into 10 year periods or something so the sequence is not predictable (i.e., you wouldn't be able to easily recognize a pattern going on with one stock and know that's the best one to buy).

To me the key thing with investing in stocks is the portfolio return can be pretty lousy relative to the mean historical return for an unpredictably long time (say 10 years), but mean reversion eventually takes over. Individual stocks each exhibit much greater variance in lifetime returns, but holding a basket should cancel these out. How you arrive at a global stock market that simulates that doesn't matter much to me.

[JasonLJJ]

I could not agree more with you - it would be great if the devs could implement this in the game to fix global stock market.

Then I think this would be perfect

[David]

What do you guys think about the ideas in this post: http://www.capitalism2.com/forum/viewto ... =52&t=6957

[David]

I recall that the dev team originally didn't include the global stock market as part of the Banking and Finance DLC because they told me that it wasn't easy to find a way to simulate the global stock market.

Then I got this user suggestion: http://www.capitalism2.com/forum/viewto ... 580#p23580 and forwarded it to the dev team.

If I understand correctly the current implementation is based on the ideas of this suggestion with some modifications.

I would be happy to forward your suggestions to the dev team as well. But could you provide some algorithms on how your suggestions could be implemented? The aforementioned post is a good example, which outlines an algorithm that could be easily understood and implemented.

[JasonLJJ]

Which post do you mean? Happy to work with you guys!

[David]

Which post do you mean? Happy to work with you guys!

Your original post states the desired outcomes of the implementation but it will be important to provide an algorithm about how programmers could actually implement it. The aforementioned post is a good example, which outlines an algorithm that could be easily understood and implemented.

ie. the algorithm suggested by that post:

the total revenue growth for 50 combined companies are correlated to 50 virtual companies, as well as their profitability, P/E, P/B ratios ...etc.

[JasonLJJ]

Also, I just found another major bug where the default rate in one of the year spikes for AAA loans - where do I access saved game files so I can send you it?

I can also send you the file where the bond purchasing led to a crash

[David]

You can find save game files of Capitalism Lab from:
C:\Users\<your user name>\Documents\My Games\Capitalism Lab Post-Release Beta\SAVE

You may provide us with your save game file by posting it here on the forum as a ZIP file attachment or emailing the save game file to us at info@enlight.com.

If your save game file is too large for attachment, you may check out this post about how to send us large files: https://www.enlight.com/forum/viewtopic.php?f=13&t=2473

[buells]

I mean, my suggestion is simply taking historical data for a bunch of companies (price, P/E, P/B, div yield) for the last 30 years and randomly assigning them to different companies at the outset of each game. For example, let's say Company A tracks the data from Home Depot during the first 30 years of the game. Next, you have the changes in each variable (price, P/E, P/B, div yield) follow those in another randomly selected company's data set for the next 30 years and so on. The only issue is that timing of economic cycle impacts shouldn't be so predictable. Accordingly, it may be better to use 10-year periods and put them out of order. I.e., instead of using 30 years of HD returns every time starting from 1989, maybe you use 30 years starting from 1969, followed by 10 years of another company's returns starting from 1989, followed by 10 years of a different company's returns starting from 2005. All of the GSM companies should be based on data from the same time period at any given time, however, so the returns are somewhat correlated and the whole market is simultaneously impacted by real world historical events like the global financial crisis or the dotcom boom. The GSM company data could be based on stocks in the same industry as the company selected (e.g., Goldman Sachs would trace the returns of different financial stocks in each period).

The reason I suggest this method is that all the potential sequences can be predetermined and it will give the feeling of a realistic stock market without any additional burden of simulation. The only downside I see is that the GSM will not be correlated to the local stock market.