Combat Replays Are SO SCREWED UP!

[jwilkerson]

Uh yup ... as reads "replay" should read "report" !!!
[:)]

[Nikademus]

personally i like the out of sync....given that the replay makes me look like i'm doing well whilst the actual combat report and situation in the game's i'm playing paints me as Bad General(tm)

[:D]

[joliverlay]

I supposed I do not communicate as well as I could.  I understand there is a replay bug.  I understand there is a random variation from replay to replay.   What I was trying to point out is the accumulating effect of the different random number on cumulative rounds of combat.
 
This single difference in a random number did the following things, each in different rounds of combat.  If any changed some should have favored me.  None did.  The replay bug:
 
1.  It decreased damage caused by US forces in surface combat.  The US BB did not get any significant hits.
2.  It increased the damage caused by Japanease forces.  The US BB sunk rather than not. 16/20 shops sunk!
3.  It decreased the number of allied aircraft that flew by a very large number.
4.  It decreased the number of bomb and torpedo hits by a very large number.
 
My opponent has informed me that rather than 3BBs sunk his ships have no critical damage. 
 
Now, for a random event I would expect either 1 or 2 or 3 or 4 to be different, but not all four events.  Difference should favor either player equally.  If each one was truely random, the probability of all four being significantly different AND FAVORING THE SAME PLAYER can be calculated as much less than (.5)(.5)(.5)(.5) or less than 6.25%.  (Less likely than that really because the most likely outcome should be that the combat rounds are about the same, rather than 50% to favor one player or the other.)
 
I do scientific calculations for a living, and publish alot in the scientific literature.  I'm not stupid, (although i may not be an very good player) but I do think I can recognize odd behavior that seems rather different than random.
 
I still like the game.  The only consolation is the example at MIDWAY given by one of the posters.  I will have to assume that something done by the skipper of the BB in the 3ed or 4th round of combat sealed my fate, and ignore the differences in the number of planes that flew and counter attacked.  Maybe by CVs were so shaken by the encounter they refused to fly against the BBs attacking them.

[panda124c]

ORIGINAL: Nikademus

personally i like the out of sync....given that the replay makes me look like i'm doing well whilst the actual combat report and situation in the game's i'm playing paints me as Bad General(tm)

[:D]

Sort of like having a bunch of computer generated "Yes" men. [:D]

[pompack]

ORIGINAL: joliverlay

I supposed I do not communicate as well as I could.  I understand there is a replay bug.  I understand there is a random variation from replay to replay.   What I was trying to point out is the accumulating effect of the different random number on cumulative rounds of combat.

This single difference in a random number did the following things, each in different rounds of combat.  If any changed some should have favored me.  None did.  The replay bug:

1.  It decreased damage caused by US forces in surface combat.  The US BB did not get any significant hits.
2.  It increased the damage caused by Japanease forces.  The US BB sunk rather than not. 16/20 shops sunk!
3.  It decreased the number of allied aircraft that flew by a very large number.
4.  It decreased the number of bomb and torpedo hits by a very large number.

My opponent has informed me that rather than 3BBs sunk his ships have no critical damage. 

Now, for a random event I would expect either 1 or 2 or 3 or 4 to be different, but not all four events.  Difference should favor either player equally.  If each one was truely random, the probability of all four being significantly different AND FAVORING THE SAME PLAYER can be calculated as much less than (.5)(.5)(.5)(.5) or less than 6.25%.  (Less likely than that really because the most likely outcome should be that the combat rounds are about the same, rather than 50% to favor one player or the other.)

I do scientific calculations for a living, and publish alot in the scientific literature.  I'm not stupid, (although i may not be an very good player) but I do think I can recognize odd behavior that seems rather different than random.

I still like the game.  The only consolation is the example at MIDWAY given by one of the posters.  I will have to assume that something done by the skipper of the BB in the 3ed or 4th round of combat sealed my fate, and ignore the differences in the number of planes that flew and counter attacked.  Maybe by CVs were so shaken by the encounter they refused to fly against the BBs attacking them.

Just to step in one more time: the results you see are not the result of one random change, they are the results of tens of thousands of individual random calculations. "I would expect either 1 or 2 or 3 or 4 to be different, but not all four events": actually I would be shocked if any were the same.

[treespider]

The differences are explained by Detection Level which is affected by damaged ships

So when comparing the two combat reports after the 1st round of combat in the Japanese replay we see 5 damaged Japanese ships (2BB, 2CA and 1 CL) when compared to Round 1 in the Allied version there are 7 damaged Japanese ships (4BB and 3CA)

After the second round the Japanese version now sees 6 damaged ships (+2BB) and in the Allied version we see 8 damaged ships (+1CA)

After the third round the Japanese version sees 9 damaged ships (+3CA) and the Allied Version sees 13 damaged ships (+4CA and 1DD) and so forth and so on...


So by the time Daylight arrives the DL of the Japanese TF is greater in the Allied replay version as compared to the Japanese replay version so more planes fly and more Japanese ships are sunk.


It all went back to the 1st round of combat and the number of ships that were damaged affecting the DLs in later rounds.

ORIGINAL: joliverlay

I supposed I do not communicate as well as I could.  I understand there is a replay bug.  I understand there is a random variation from replay to replay.   What I was trying to point out is the accumulating effect of the different random number on cumulative rounds of combat.

This single difference in a random number did the following things, each in different rounds of combat.  If any changed some should have favored me.  None did.  The replay bug:

1.  It decreased damage caused by US forces in surface combat.  The US BB did not get any significant hits.
2.  It increased the damage caused by Japanease forces.  The US BB sunk rather than not. 16/20 shops sunk!
3.  It decreased the number of allied aircraft that flew by a very large number.
4.  It decreased the number of bomb and torpedo hits by a very large number.

My opponent has informed me that rather than 3BBs sunk his ships have no critical damage. 

Now, for a random event I would expect either 1 or 2 or 3 or 4 to be different, but not all four events.  Difference should favor either player equally.  If each one was truely random, the probability of all four being significantly different AND FAVORING THE SAME PLAYER can be calculated as much less than (.5)(.5)(.5)(.5) or less than 6.25%.  (Less likely than that really because the most likely outcome should be that the combat rounds are about the same, rather than 50% to favor one player or the other.)

I do scientific calculations for a living, and publish alot in the scientific literature.  I'm not stupid, (although i may not be an very good player) but I do think I can recognize odd behavior that seems rather different than random.

I still like the game.  The only consolation is the example at MIDWAY given by one of the posters.  I will have to assume that something done by the skipper of the BB in the 3ed or 4th round of combat sealed my fate, and ignore the differences in the number of planes that flew and counter attacked.  Maybe by CVs were so shaken by the encounter they refused to fly against the BBs attacking them.

[treespider]

-

[qgaliana]

joliverlay, I think you're assuming these events are statistically independent?

From Mike's explanation that doesn't seem to be the case. One or two bad rolls could lead to a tactical disadvantage that just leads to a cascade of disasters.

[treespider]

One other consideration as to the differences in the number of aircraft flights between the replays is the amount of running done by the Allied CV's.

In the Japanese replay version all of the screen is either damged and/or sunk which implies the CV's did a bit more running and hence would have fewer OPs available for Flight ops come daylight.

In the Allied replay version two DD's are untouched and the BB survives implying the CV's did a bit less running and likely had more OPs avaible for Flight Ops.

[joliverlay]

qgaliana, you are quite correct.  I assumed the air combat rounds would be more or less indepent of the surface combat rounds.  I can see that this might not be true.   I do think I detect a problem with that explanation.  After an engagement involving the prior two friendly surface combat forces, the companion air combat task force should not have been suprised, problaby not even engaged (the Transport TF should have been perhaps) nor had problems locating the fleeing surface TF for air strikes.  These subsequent rounds should have been independent of the first round with respect to sighting because of radios and the buring ships from the prior battles.  The surface combat rounds with the ensuing TFs should not have been effected so much by an initial sighting and thus been independent.  Surely the first two TFs had radios.
 
Someone commented they would be suprised if the results were the same.  I agree they should not.  I'm only arguing that a different repy should differ in a random way, not having every the subsequent rounds with different TFs and types of units ALL favor the same player.  Thats not random.
 
I also think the problem was magnified by a 25 ship enemy surface TF engaging my 10 ship surface TF.   In this combat I see no disadvantage to putting most of your capital ships into a few large surface TFs.  It does not seem to slow them or effect their resistance to air attacks.   

[ADM Halsey]

I was shocked in this battle to catch the allied carriers in a surface action. My Bombardment TF was not expecting any allied ships to be in the area save the PT boats. To my surprise the third round revieled the carriers and the escorts did a good job of screening the carriers from my surface fleet. I am sure the carriers hauled butt to put distance between them and the 18.1 inch guns raining down on the TF.

I was sue I would lose a lot of ships in the airstrikes to come during the day. When I checked ship damage no ship was above 20% combined damage. I am glad that the carriers of joliverlay escaped the surface action to fight another day.

Jack if you want to try the sync bug fix I will start it with the next turn. If you wish we can trash this game and start a new one with me being the allied and you as japan.

[qgaliana]

ORIGINAL: joliverlay
I do think I detect a problem with that explanation.

Just an observation - I couldn't possibly begin to explain this. [:D]

ORIGINAL: joliverlay
I'm only arguing that a different repy should differ in a random way, not having every the subsequent rounds with different TFs and types of units ALL favor the same player.  Thats not random.

Given that it's a pseudo-random generator, it's not impossible the results could get streaky, even if over time the distribution is fairly correct. I've noticed for example it is common that air missions to fail to find targets in clusters of 4 or 5 scattered all over the map.

[rtrapasso]

One thing i noticed from playing AI only games: the results DO differ despite the same gamesave... why?

i think the reason is that the game does not clear out the memory completely when it loads... there is a problem (theorized) for "sticky slots" that can be seen under some different circumstances that i won't go into right now.

However - a couple of years ago i demonstrated the "lack of consistency" in replays by doing the following:

1. run the game for a while, make orders, etc.
2. save the game.
3. run the turn - note the results.

Then - 4. turn off your machine - do a cold boot- load the game, load the save game file from above.

5. Run the turn - note the results.
6. Check the results from part 3 vs. part 5. They will NOT be the same in most cases. i spent many, many hours doing this and posted the results to the forum.

if you do a cold boot again and run the turn again - the results will be the same as what you got in part 5.

Now, these were done about 2 years ago - i don't know if any patches eliminated these flaws, but i hadn't read of anything in the patch notes that would lead me to believe that things were fixed.

Note that according to what was just posted by the devs - the results SHOULD BE THE SAME NO MATTER WHAT - it was the same save game file. But somehow, stuff is being held in the game memory area that messes things up.

i am not sure how the PBEM games are handled to be sure something similar is happening to trigger sync issues, but i suspect there might be.