Managing the Economy

[steve39]

Methodology

All my trials involved 20 tests so if a result states "50%" incidence of a fall in the economy, that means 10/20.  The "raw data" is as simple as that - and there is no "interpretation" except to present these figures as a percentage.  I'm sorry, I thought I'd made that clear. 

The error, put simply, is 1/20 on the raw data, which is +/- 1, and hence +/- 5 on the percentage figure.  For example, 50% is accurate within 45-55%.  That's why in one case I stated that 50% and 55% were not significantly different.  In fact, say, 40(+/-5)% and 30(+/-5)% are not significantly different, or at their limit of significance.  For greater precision the sample size would have to be scaled up. 

It's a while since I did advanced statistical calculations but there is no need to employ the whole shooting match here; this is good enough for our purposes at the moment.  For a clearer resolution it would need a game situation in which every possible combination can be tried, and at a sample size of, say, 50 where the accuracy would be 1/50, ie. +/- 2 on the percentage, or a sample size of 100, etc.  Followed perhaps by a regression analysis. 

But that is quite a task and before starting it's necessary to identify fruitful trends first - and that's where we are at the moment, with some success I would say.  It's easy to play Devil's Advocate - but muddying the waters with unquantified opinion; no reference points; and errors of logic is not helpful.  This problem will only be solved by knuckling down to some hard graft and the application of science (bit of a clich

[JSS]

Well. I managed to get some time and quantify my un-quantified opinion. Skeptics will most likely not be satisfied with my test methodology but nevertheless here it is.

The test map:

The map used to supply the data is that of a game I eluded to in an earlier post. It is a POS2 game that was started with a basic connection as well as train setups in place. All cash was spent and the company has issued one lot of shares. This map was saved before actually starting any train traffic. A copy of the game can be obtained here.

Test 1:

The game was loaded, un-paused and usually run until Oct. of the second year. When there was a change in economy the game was paused and the change was recorded. The economy was listed in the form of a number from 0 to 4. (Panic being 0 and Boom being 4)

If there was no change at the appropriate time slot, the game was stopped the following month and the status was also recorded.

There were neither changes made in the train schedule nor any changes in consist. This was done forty times. The results of that test can be seen here.

Test 2:

This test was identical to Test 1 except in February of the second year the game was paused and as many shares bought as personal cash would allow. Based on train income and when it happened, made it impossible at times to purchase any shares. You will see an added column that shows the number of shares obtained. The results of Test 2 can be seen here.

Conclusions:

First group of tests:

At the column bottom of each period

[Gwizz]

Very impressive JSS.

[djf01]

I want to second that.  Very impressive array of data.

I will pick you up on one or two of your conclusions though.

The result of this test seems to indicate an overwhelming positive change in the frequency of economic improvements in period 3 by a count of 6 to 14. That is in Test 1 (no shares purchased) there were 6 increases of the economic climate between September and March and I Test 2 (purchasing shares most of the time) there were 14.

I don't think it's valid to directly compare to results from any one specific period in isolation (other than the first one).  That's because you are not necessarily starting from a uniform base.

To summarise your data:

[table][tr][td][/td][td]Q1[/td][td]Q2[/td][td]Q3[/td][td]Q4[/td][/tr]

[tr][td]Test 1 (Control)[/td][td]10[/td][td]7[/td][td]6[/td][td]12[/td]  [/tr]

[tr][td]Test 2 (Buy Shares)[/td][td]8[/td][td]4[/td][td]14[/td][td]13[/td]  [/tr]

[/table]

After Q2, in test 1 there have been 17 economy boosts, but only 12 in test 2.  So comparing Q3's results on there own makes no sense, because the starting sample is different.  But what you can do is track the total number of changes from the start at the end of each period.  In which case the data looks like this:

[table][tr][td][/td][td]Q1[/td][td]Q2[/td][td]Q3[/td][td]Q4[/td]  [/tr]

[tr][td]Test 1 (Control)[/td][td]10[/td][td]17[/td][td]23[/td][td]35[/td]  [/tr]

[tr][td]Test 2 (Buy Shares)[/td][td]8[/td][td]12[/td][td]26[/td][td]39[/td]  [/tr]

[/table]

It means the buying shares *may* be helpful, but it delays the growth.

[loco_motive]

It would be quite a job to wade through all the horse pucky in this thread and figure out what might be worth testing and what might not be. From what I understand, the underlying question is what variables are included in the formula that determines the economic condition to be set at the end of Feb and Aug of each year.

I guess PopTop tells us that it is random and independent of player activity but does include a momentum factor for the direction in which it is headed. It wouldn't be the first thing they said that was inaccurate. Nor does it necessarily reflect hard-coded differences built into some scenarios, or other differences that do or could exist related to starting year, MAP vs MP2, bad seeding of random numbers, missed remnants of changed programming, etc.

(I'm thinking in particular of the fact that when I play an MAP file, I get 1-2 engine explosions in the first 2-3 years far too often if I am running a game with no saves, but that if I save and reload before starting to play, these explosions tend to go away. Anyhow its only an impression I'm sure.)

There seems to be strong opinion and appears to be some support for the idea that some actions taken by the players, including undoubtedly the AI players, are reflected in the formula.

I have taken a quick look at JSS's data to see if anything else can be inferred from it. I agree that comparing year 2 in two independent runs is not great, and prefer steve39's method of rerunning an immediate "change/no change" over and over from the same base with minimal time lapse.

Nevertheless some interesting data on frequency and direction of changes can be examined with care since apparently actions can affect that too. Of course djf01's numbers need to reflect JSS's -1's as positive counts, but in any case he could consider restating his frequency of economic changes from 10% per year number given in an earlier post to something like 50-60%/yr based on this data. Also of course his conclusion is pretty much in line with what I was mentioning in my first paragraph.

I extracted the following from JSS's data to try and normalize both sets of data. This may be subject to the same criticism as the whole set of numbers, but still gives interesting results.

Test 1 econ going 2 2 1: 1 case

Test 1 econ going 2 2 2: 15 cases

Test 1 econ going 2 2 3: 3 cases

Test 2 econ going 2 2 1: 0 cases

Test 2 econ going 2 2 2: 19 cases, 12 one share and 7 two share purchases

Test 2 econ going 2 2 3: 11 cases, 8 one share and 3 two share purchases

From this scant data, the econ increased 16% of the time in test 1 and 37% of the time in test 2, enough for me to think there might be something here for this factor and likely several other factors as well. Strangely, there was a higher percentage of 2 share purchases in the 2 2 2 group than in the 2 2 3 group, which also begs retesting on account of that variable or some other variable that caused a different number of shares to be purchasable.

-------------------------------------

PS: There is a function called "insert code" indicated by a # sign among the editing icons that can be used with data, and some editing, to preserve its format as well as making the data capturable by us. It has its pros and cons for the user to decide upon, the worst one for me is the small type that I don't think can be enlarged.

[steve39]

I'm not happy about comparing apples with oranges either, but can't get too involved for a week because of a heavy work load.  I should however underline one aspect of my tests - I deliberately chose an established  position which I knew was going to trip the economy, in fact with a 50% chance. 

Under normal conditions when a company is being run profitably and share values and earnings are rising, the chance of the economy turning down is close to 0%. 

I can now get an economy into Boom inside 2-3 years and with care keep it there for another 15yrs.  Please, none of this "if it works for you

[djf01]

There seems to be strong opinion and appears to be some support for the idea that some actions taken by the players [effect changed in the economy].

.

I think it's more than strong opinion.  Two players have presented credible evidence of it.

Of course djf01's numbers need to reflect JSS's -1's as positive counts, but in any case he could consider restating his frequency of economic changes from 10% per year number given in an earlier post

The 10% number was inferred from Steve's data.  His data suggests that the chances of a purely random downturn in the economy is 10%, and the rest was triggered by player actions.

I don't think JSS's experiment is able to isolate all random and triggered economic movements.    It's a very good, very well defined experiment, however the control and test sequence have many other things going on in them.  This experiment was specifically designed to show what impact, if any, purchasing shares had on the movement in the economy.  It does this very very well. 

The problem we're all having is interpreting the data because it seems somewhat inconclusive.

What I think we can say from JSS's experiment is:

1) Player action definitely *does* impact movement in the economy.

2) Buying Shares (the thing JSS was testing) does not specifically trigger increases in the economy in and of itself.

I'm hypothesising here (ie adding to the horse mucky), but what I think is going on is the movements in the economy are linked in some way to the player's personal financial position.

[JSS]

Well, what I set out to do was to try to see what will happen if one does opposite of what Steve39 said needed to be done for a better economic game and the way it turned out showed that one can play a different game and still have the same overall economic state. Perhaps I was just lucky.

Over those 80 runs tested it again reconfirmed the effect of randomness on an overall outcome. Perhaps randomness is not the correct term because some seemingly random behavior can also be affected by fixed rules.

For instance, in all test games the two trains playing the primary role were being loaded in cities opposite from each other. The train in Hammond had to wait for one additional load of Passengers and the train in Cleveland had to wait for two loads of Passengers, both being six car trains. The arrival of the first train would vary between the end of November to sometime very early in the next year. The two main reasons determining the arrival were the seemingly random generation of loads and the fixed rule that whatever train occupies a single track square first has the right of way. That at times meant that the train having the longer distance to travel interfered with the train that would give you an earlier arrival.

This all then had a cascading effect. When a train came in very early the share price would go up at the end of the year which in turn attracted the A.I. player to purchase shares which then left you out in the cold when waiting for February to come around.

On the other hand, if a train came in late, at the beginning of the second year, the share cost was lower and you only competed with perhaps one A.I. payer owning one lot of shares.

What those tests also showed was that waiting for February for your stock purchases could do your early PNW a lot of harm from which you might not be able to recover.

In any case, just to try to weed out all those intangibles here is another test.

The Test map.

The map was generated in the editor and left void of any cities or features. There was only one company (

[Gwizz]

Steve 39,

RRT3 must have about the same affects on the economy.  Hawk's site has been talking how RRT3 is affected in different ways.  I don't think many players understood these little affects.

I know I didn't.