Low percentile, high hits > winww2 and mbt

[Grumbler]

Remember NOTHING in a computer happens by random. Nothing. It is impossible. Computers work by creating patterens. The closest to random you get is a very small part of a very large pattern.
The Programmer will use a pattern with 1.2 x 10 E <23> or somthing along that order in the theory that you will not ever see the entire pattern. That works fine for some things, but I think in SP, where you have thousands of 'random' factors included in the calculations every move, then standard patterns are not big enough.
Every time your uber panzer grinds forward a hex the program checks to see if it's spotted. I''ll bet for every unit. Then It calculates it it's spotted any units. FOR EVERY UNIT. Throw in breakdown calculations and that is a lot of 'random' events that have to be figured out by a non-random system. To move one hex.

[Narwan]

This has come up in the past and extensive testing then has shown that there is no systematic divergence of the expected spread of hits.

The results posted in the first message say very little. It's not a 200-times test either. For that to be the case each category should have been tested 200 times or more. Now it's a test varying between 2-times and 58-times. These sort of numbers are far too few to show any meaningful results as even one extreme result (which WILL happen once in a while) will throw off the curve. The total of 200 fire results is large enough to have some extreme results in them affecting some categories. But as the total per category is far too low to compensate. For a test to show reliable results you'll have to have at least hundreds of results per category. If memory serves me right, the amount used to check these claims even ran into the thousands.

It is also expected that in a test with 10 categories like this at least one wil be off to some degree. Again due to the law of probability. Repeats of the same extensive test will show whether the same category is consistently off by a significant margin. That has not been the case to my knowledge.

I have also kept track of results during games a while back when this came up. Conclusion: no significant divergence of expected results over the long run. What did happen once a while was a 'string' of several 'unexpected' results (2 to 4 in a row). In the long run these didn't skew the results but in a smaller sample they may very well seem to do that. Whether these short strings are an artifact of the pseudo-random generator the system uses or again an expected probability over the long run I'm not sure.

As to the categories claimed to give better results than they should, my very recent gaming experience showed exactly the opposite. My soviet armor in a '63 game scored far less hits in the lower odds margins (up the 30% score) than I expected. A lot. It happens.

Remco