BigDork Posted August 2, 2004 Share Posted August 2, 2004 Hi all... I just got done playing a game for a tournament over at BoB and ran across a little difficulty. When I got the AAR screen my score showed 60% to 40% while my opponent saw 59% to 41%. What's going on with the different results and who has the right score? Thanks! 0 Quote Link to comment Share on other sites More sharing options...
Sergei Posted August 2, 2004 Share Posted August 2, 2004 Wasn't there the same bug in CMBB at some point but got squashed in a patch? 0 Quote Link to comment Share on other sites More sharing options...
BigDork Posted August 2, 2004 Author Share Posted August 2, 2004 So I did a little experiment and loaded the last file ten times. Five times I got 59% and five times I got 60%. 0 Quote Link to comment Share on other sites More sharing options...
WWB Posted August 2, 2004 Share Posted August 2, 2004 What were the point totals? Do you have a screenshot or a savegame? WWB 0 Quote Link to comment Share on other sites More sharing options...
Brent Pollock Posted August 2, 2004 Share Posted August 2, 2004 Dunno...is it one of those floating decimal point things...maybe your right on the edge of rounding up or down, and it depends on some random number generator? I really have no clue as to how this black box operates. 0 Quote Link to comment Share on other sites More sharing options...
xerxes Posted August 2, 2004 Share Posted August 2, 2004 It doesn't sound like the same problem that existing in CMBB where the scores could end up wildly different. This sounds like rounding error. Split the difference and call the score 59.5% to 40.5%. 0 Quote Link to comment Share on other sites More sharing options...
Redwolf Posted August 3, 2004 Share Posted August 3, 2004 Is one of the machines an AMD one and the other one with a Intel CPU? 0 Quote Link to comment Share on other sites More sharing options...
BigDork Posted August 3, 2004 Author Share Posted August 3, 2004 For a screenshot click on the 60% to 40% in my first post. As for different processors, I don't know. I've got an Intel. I can e-mail my opponent and find out what he has. 0 Quote Link to comment Share on other sites More sharing options...
Sergei Posted August 3, 2004 Share Posted August 3, 2004 Since you experienced the same variation on your own 'puter too, it shouldn't depend on processor. Unless yours is the only one which does this randomness. You could ask your opponent to load the file several times too and see if results vary. 0 Quote Link to comment Share on other sites More sharing options...
BigDork Posted August 3, 2004 Author Share Posted August 3, 2004 Okay I just got - this - screenshot from my opponent. 40% to 59%??? So now the scores don't even add up to 100%. 0 Quote Link to comment Share on other sites More sharing options...
BigDork Posted August 3, 2004 Author Share Posted August 3, 2004 Wow that could be the problem... If you look at both screenshots the KIA numbers are different. In the first it's Axis-31 Allies-21 and in the second screenshot it's Axis-28 Allies-23. So how's that happening? 0 Quote Link to comment Share on other sites More sharing options...
GreenAsJade Posted August 3, 2004 Share Posted August 3, 2004 Different KIAs ... now that's wierd. 1% due to rounding I wouldn't have cared about! 0 Quote Link to comment Share on other sites More sharing options...
Joachim Posted August 3, 2004 Share Posted August 3, 2004 I id'd 3 problems in this thread: a) Men killed differ Points for prisoners differ (but not htier number) c) Rounding errors ad a) Look at the "Men killed". It explains part of the problem. While the "total casualties" are always the same, the killed differ. IMHO they are calculated after the battle as a random number from the total causalties. IIRC the same problem is in ops - you can have more killed in the last but one battle than in the AAR screen after the op is done. BigDork - please do your test again but check not just the percentage but the killed and the captured, too. Depending on the killed (of both sides) you should get 59% or 60%. ad A bit strange is the prisoners - you get different points for the same number of prisoners? Do they calculate it depending on rank or wpn in the sqd? What were the prisoners? Which unit types and how many? Differing points for the captured is another problem. Please check if the points are always the same for each side. ad c) The 3rd problem is the "40+59=99". A classic in statistics if you have rounding errors. Calculating the points from your friends AAR you get 708 : 1040 or 40.5034325 : 59.4965675 Both rounded down? No, then you would always get 40:59 Or do they calculate internally using just one decimal (cutting the rest)- not exact but works in most calcs. Example: Points total: 1748. 100/1748= 5.7% 608/1748= 34.7% Sum is 40.4% rounded to 40% Same for Allies: 200/1748= 11.4% 799/1748= 45.7% 41/1748= 2.3% Sum is 59.4% or 59% Reminds me of what my old Algebra teacher said during my first term: "Computers make errors when calculating. But they do this amazingly fast." Gruß Joachim 0 Quote Link to comment Share on other sites More sharing options...
BigDork Posted August 3, 2004 Author Share Posted August 3, 2004 Yep, I'll definatly load up the last turn a few more times when I get home from work today. The prisoners are 6 fallschrimjagers, all from the same squad. 0 Quote Link to comment Share on other sites More sharing options...
Redwolf Posted August 3, 2004 Share Posted August 3, 2004 Now that the mencount is different it can't be that anymore, but AMD and Intel do slightly different rouding on floating point arithmetic. Always makes for nice surprises for anyone who think FP is deterministic. But this is a bug. Can you manually count the men taken prisoner or were some moved off the map? 0 Quote Link to comment Share on other sites More sharing options...
BigDork Posted August 3, 2004 Author Share Posted August 3, 2004 It was 6 total taken prisoner. 0 Quote Link to comment Share on other sites More sharing options...
BigDork Posted August 3, 2004 Author Share Posted August 3, 2004 And what's with the big blocky letters? Definatly not nice. 0 Quote Link to comment Share on other sites More sharing options...
GJK Posted August 3, 2004 Share Posted August 3, 2004 Not that it's 100% accurate, and not that it will <u>prove</u> anything, but I thought it might be interesting to plug in those figures from the screenshots in to the VL Planner at the Proving Grounds. 1st screenshot: </font><blockquote>code:</font><hr /><pre style="font-size:x-small; font-family: monospace;">VL Points: A:100 / D:200 Attacker Totals: 708 Ratio: 0.676 40.32% Defender Totals: 1048 Ratio: 1.480 59.68% Casualty Ratio: 1.32 Attacker to 1 Defender Total Pts: 1756 Result: Allied Minor Victory</pre> 0 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.