Blizzard Combat Bug

[PMCN]

I just went thru the fourth week of blizzard and the soviets finally went on the offensive but when viewing the results there is obviously a bug in the routine.

1. Units in cities were not getting their full numbers of combat forces showing up. I had a 99% readiness unit in a city attacked and what showed up to fight was maybe 20% of the defending force (a panzer korp with a CV of 180). I had this happen 2 times in different cities (although 1 unit only had readiness of 70% at miniumium 200 squads of infantry should have showed up ... 85 did).

2. Units are being severly hit by readiness losses. I had units with 400 squads 70% readiness normal german experience (80%+) in the mountains, in a scale 5 entrenchment shatter...again 85 squads showed up to battle. Somehow "I don't think so" is hard to avoid thinking. I have difficulty imagining that you can force a retreat in that situation.

If someone official wants a saved game file (which I have replayed 5 times now and seen all sorts of truely unbelievable things occur) let me know at pmcneely@t-online.de and I will gladly send it to you to look at.

[Culiacan Mexico]

Originally posted by Paul McNeely:

2. Units are being severly hit by readiness losses. I had units with 400 squads 70% readiness normal german experience (80%+) in the mountains, in a scale 5 entrenchment shatter...

I have the same problem.

[RickyB]

I have just reported this as a major issue to Arnaud, based on everyone's comments here and a test I just ran where in the first blizzard turn, I launched practically every Soviet unit into an attack and almost every one of them succeeded, even 1 army against a Panzer Korps that was at a CV of 64 before the battle and entrenched to 4. Hopefully, he will be able to get a fix prepared for this, as it makes the Axis job impossible, I think.

Thanks for the reports, and sorry we missed it.

------------------
Rick Bancroft
Semper Fi

[RickyB]

Thanks again for the comments on the blizzard, and a save game from Paul. In looking further at this, I managed to do the following: I played both sides from the first blizzard turn on 12/7 to the start of the Axis turn on 1/4/42. I basically just plotted straight ahead attacks for the Soviets, without any reinforcements or air attacks, and only doing special supply when the units were below 15 for strength, and not always then, taking no more than 5 minutes per turn. For the Axis, I worked as hard as I could trying to stop the Soviets. Quite historical, I would think.

By January, the Axis had two large 4-5 hex holes in their front lines, and had lost the game equivalent of 600,000 men, in just 4 weeks. Overall, more than 40% of the total men and equipment in the east were lost during this time. This was based on my game againg another tester, and the Axis were pretty well entrenched in almost every area.

Anyway, I sent this, with more details, off to Arnaud along with a debugger trace file so that he can take a look at what is happening. He is on it, and hopefully can make some changes quickly. Hang in there.

Thanks for the files, Paul.
------------------
Rick Bancroft
Semper Fi

[This message has been edited by RickyB (edited February 26, 2001).]

[Vasquez]

hello together,

last time I played WIR as the russians and crushed in the Blizzard turnes 1941 (8 or 9 weeks) over 1.000.000 mens by encirceling 2 or 3 big pockets...?!?!?!
this could not be right..

(sorry for my english)

Greets from München, Germany
Vasquez

[silkworm]

It isn't just during blizzard turns, either. Playing the Soviet side I often don't see the right number of units show up for combat, as would result from multiplying experience and readiness. I think what I actually get is one half of what I expect. Is this right? I'm using good leaders, rated at 8 or 9.

[RickyB]

Originally posted by silkworm:
It isn't just during blizzard turns, either. Playing the Soviet side I often don't see the right number of units show up for combat, as would result from multiplying experience and readiness. I think what I actually get is one half of what I expect. Is this right? I'm using good leaders, rated at 8 or 9.

I don't have the manual in front of me, but I believe there is a drop in Soviet readiness before combat in 1941 (and smaller in later years), compared to what you you would expect from before combat. Also, the artillery and aircraft bombardments result in lower readiness and losses, which I think would account for fewer squads than you expect. Don't know if one or the other or both apply, but I think this explains what you are seeing.


------------------
Rick Bancroft
Semper Fi

[Svar]

Originally posted by silkworm:
It isn't just during blizzard turns, either. Playing the Soviet side I often don't see the right number of units show up for combat, as would result from multiplying experience and readiness. I think what I actually get is one half of what I expect. Is this right? I'm using good leaders, rated at 8 or 9.

In looking into the Blizzard 41/42 effects, I have seen that if the Soviet player doesn't have enough operation points in the immediate headquarters that the game engine divides the effective readiness by 2. Make sure that your HQ has plenty of operation points before executing combat. I suspect that there are a lot of undocumented effects for that Blizzard that need to be found before any more adjustments are made and will try to run a comparative analysis of the versions 1.1 and 3.0 this weekend. Right now I'm having too much fun just playing version 3.0, it plays much more realistic than version 1.1.

Svar

[Heinrici]

The blizzard bug is alive and well. While playing an email game I entrenched my units up to level 5 had 50-70% readiness(using special supply) My opponent shattered 5 panzer korps and 2 infantry korps in just 8 attacks (87.5% shatter rate). If he had wanted to he could have taken out the entire eastern front within 4 weeks. We finally agreed on an armistice line where I could pull back and he wouldn't attack me during blizzard turns. This is a game breaker as far as play balance. I was just wondering if the programmer is still looking into this issue as I have not seen a post on this recently.

[RickyB]

Originally posted by Heinrici:
The blizzard bug is alive and well. While playing an email game I entrenched my units up to level 5 had 50-70% readiness(using special supply) My opponent shattered 5 panzer korps and 2 infantry korps in just 8 attacks (87.5% shatter rate). If he had wanted to he could have taken out the entire eastern front within 4 weeks. We finally agreed on an armistice line where I could pull back and he wouldn't attack me during blizzard turns. This is a game breaker as far as play balance. I was just wondering if the programmer is still looking into this issue as I have not seen a post on this recently.

Arnaud is planning on having something ready in about 3 weeks, if all goes well. He is ready to get it out for testing in a few days, and then a couple of weeks to test it out and then put it out there for everyone.

He discovered the key to the shatter problems, which was exposed by the weakening of air power but it explains a lot of other behavior also. The old blizzard routine for 1941 quarters the actual readiness. However, there is code that even Arnaud hadn't found before that basically put a minimum readiness of 25% anyway, so high readiness and low readiness in these blizzard turns had no effect - the readiness would be 25% no matter what :rolleyes: . Losses are based on the actual readiness (comes out of total ready) while the effective readiness is the same no matter what, so the higher the readiness the higher the losses and the greater the chance of shattering. This would also explain why Soviet units that receive special supply in 1941 have always been very susceptible to shattering, as I would bet they have some kind of special readiness reduction also.

[PMCN]

Great Work!!!

I had noticed the Soviet shattering myself but I thought it might be due to low operations points. But even in situations were the op points were good I have seen soviet formations shatter when I could never figure out why. When playing the germans the last time around (with the latest version) I noticed that I was getting a large number of soviet shatter results. I think perhaps the most I have ever seen.

Can Arnaud confirm if the game uses the length of the blizzard as a balancing factor? I noticed when playing the computer that the blizzards are long and hard (my record is 16 straight weeks) but when playing the soviets or during PBEM games it is usually very short generally historical. These were under balance even BTW.

[Svar]

Originally posted by Paul McNeely:

Can Arnaud confirm if the game uses the length of the blizzard as a balancing factor? I noticed when playing the computer that the blizzards are long and hard (my record is 16 straight weeks) but when playing the soviets or during PBEM games it is usually very short generally historical. These were under balance even BTW.[/QB]

Paul,

I am in the middle of a weather test for WIR v 3.0. After 10 tests starting in 2 Nov 41 and going through 4 Apr 42, the most blizzard turns total have been 14 and the most consecutive blizzard turns has been 12. I suspect that while the probabability for blizzard weather in Dec, Jan, and Feb will be over 90%, the probability for blizzard weather in Nov and Mar will be below 50%. I also suspect that the possibility of blizzard weather in the first two turns of Nov and the last two turns of Mar are zero so the maximum number of turns of consecutive blizzard weather would be 16 but the probability would be very low. You were just very unlucky, the average number of consecutive blizzard turns is about 8 with my limited data set. I'll run 40 more data sets for a total of 50 for the whole time period then work out the statictics and repoert them in this thread.

Svar

[Svar]

The data is in on 5 sets of 10 runs each with slightly different starting conditions. All tests were run from just before 11/2/41 through 4/5/42 with balance set to even. The first test set was with no combat playing human versus human at a test location approximately historical. Test 2 was with no combat playing human versus human at the 6/22/41 starting position. Test 3 was a repeat of test 1 playing the Soviet versus computer so computer directed combat started when the test began. Test 4 was from a saved game where I player the Germans versus the computer set to max help for the Soviets so there had been heavy combat and the location was near historical. For test 4 the Soviets were playing against the computer. Test 5 was a repeat of test 4 with the Germans playing against the computer.

The good news is, it was statistically insignificant what the starting conditions were, the results were remarkably consistent. The average number of blizzard turns for the 50 tests was 11.66 turns while the maximum was 15 turns and the minimum was 8. Interestingly enough those occurred in the same test set (test 4). The maximum consecutive string of blizzard turns was 13 and the minimum consecutive string of blizzard turns was 3. Again those occurred in the same test set (test 5). Test sets 4 and 5 were from the same starting conditions the only difference being the Soviet versus the computer (test 4) or the German versus the computer (test 5). These two test sets also gave the greatest average number of blizzard turns (12.1 test 4) and the least average number of blizzard turns (11.4 test 5). If there is a bias, the game favors the human player over the computer with respect to the blizzard weather of the winter of 41/42. I don't think that is the case; it's just a statistical anomaly.

All the data points generated the probability of any given weather by month. The first two weeks of November have the probability of rain 46% of the time and snow 54% of the time. The last three weeks of November have the probability of rain 4% of the time, snow 81% of the time and blizzard 15% of the time. December has the probability of rain 6.5% of the time and blizzard 93.5% of the time. January has the probability of rain 7.5% of the time, snow 6.5% of the time and blizzard 86% of the time. February has the probability of rain 3% of the time, snow 11% of the time and blizzard 86% of the time. The first three weeks of March have the probability of rain 3% of the time, snow 78% of the time and blizzard 19% of the time. The last two weeks of March have the probability of rain 60% of the time and snow 40% of the time. Finally the first week of April has the probability of rain 90% of the time and clear 10% of the time. The likelihood of having blizzard weather for all 16 turns in which it is possible is extremely unlikely. In test set 5 the two biggest surprises for the German player occurred when once the blizzard started in 11/23/41 and lasted unabated until 1/8/42 for a brief respite. The other surprise was another time when the blizzard started in 12/7/41 and went through 3/15/42 with only one snow turn in the middle.

I didn't run any tests for weather as PBEM games but would be surprised if it is any different. Also, I doubt that the game program uses probabilities any more accurate that 5% so those numbers in the previous paragraph can be rounded to the nearest 5%.

Svar

[Mist]

Svar! The great job has been done by you! Thank you, thank you, thank you!!!! These data are very interesting and important. I wouldn't dare to do that. Thank you again!

[PMCN]

Thanks Svar,
Did you notice if the weather is handled in a logical fashion? Is it completely random or does it depend on the previous turn?
I noticed that you could tell what sort of a winter you were up for by the weather in october and november. If you skip the mud-rain phase and go from clear to snow (or have only a short rainy season) you are generally in for a long hard blizzard, oops, I mean winter. If you have a more normal one than the blizzard period is likely to be mild-short.
I am have difficulty believing that my luck has been so extreme but it is possible. It is just that when playing against a live opponent or as the soviets I have never seen a blizzard turn beyond the historical 3 blizzard turns (baring the odd one) but my data set is rather modest.
If this disparity is gone I will be glad, since it may mean it is possible when playing the soviets to regain some lost ground. Well currently it is more than possible but after that bug is squashed...
Thanks again, much appreciated!

[Svar]

Originally posted by Paul McNeely:
Thanks Svar,
Did you notice if the weather is handled in a logical fashion? Is it completely random or does it depend on the previous turn?

I noticed that you could tell what sort of a winter you were up for by the weather in october and november. If you skip the mud-rain phase and go from clear to snow (or have only a short rainy season) you are generally in for a long hard blizzard, oops, I mean winter. If you have a more normal one than the blizzard period is likely to be mild-short.

I am have difficulty believing that my luck has been so extreme but it is possible. It is just that when playing against a live opponent or as the soviets I have never seen a blizzard turn beyond the historical 3 blizzard turns (baring the odd one) but my data set is rather modest.

If this disparity is gone I will be glad, since it may mean it is possible when playing the soviets to regain some lost ground. Well currently it is more than possible but after that bug is squashed...


Thanks again, much
appreciated!

The weather appears completely random depending only on the probabilities for a give turn. A rain turn can appear in the middle of many blizzard turns. Also, for all 50 data sets the, November week with the most blizzard turns was 11/16/41 and the March week with the most blizzard turns was 3/8/42 but those were just statistical oddities. A different set of data should give slightly different results but the overall numbers should be very close.

I didn't run the October weather but didn't see this in the November weather. Once there was a snow turn the first week of November then 4 rain turns before the first blizzard turn in December. For that run, the total number of blizzard turns was 12 and the longest string of consecutive blizzard turns was 7, pretty much along average lines for the total data set.

A total of 20 runs were made as the Soviet versus the computer and one of these sets of 10 were the highest total average number of blizzard turns for the total data set (average of 12.1 turns per winter). But one of the runs did have only 8 total blizzard turns with the longest string being 4 turns.

I hope the disparity is gone but don't know about PBEM. I'm currently playing a PBEM as the Germans so should at least check, I hope it isn't different. BTW the old version 1.1 didn't have as long a severe winter as the Matrix versions. When I first started playing this, I noticed it and posted the differences in September 2000 I think, I'll try to find it and repost in this thread.

Svar

[ April 23, 2001: Message edited by: Svar ]

[Svar]

Here is the post from September 2000. This data was generated by month. There were only 10 runs for each month and I rounded the percentages to the nearest 5%. You can see why we get long strings of blizzard weather in version 3.0 (the weather wasn't changed from version 2.0 to 3.0) compared to the 1.1 version.:


The results of the weather check after running both versions through January and February 1942 (I assumed December 1941 was unchanged since previous versions of the game had blizzard weather almost all the time):

WIR TMP v 2.0
Month Blizzard Snow Rain
Jan 42 90% 5% 5%
Feb 42 90% 5% 5%

WIR v 1.1
Jan 42 70% 20% 10%
Feb 42 25% 65% 10%

As can be seen there is a big difference between the two. If this was intended it is not documented. All that severe weather (blizzard) makes life more difficult for the Axis side.

Svar

[PMCN]

Thanks Svar,
Most of my experience has been with v1.1 and certainly under those rules when playing the russians I didn't notice the same weather as I did as the germans. The 16 straight weeks of blizzard was during a v1.1 game. But realistically my sample is too small to extract the probability distribution function.

I must say the new weather seems a bit extreme to me. 90% chance of blizzard is far too high. The winter of 41 was more or less a typical Canadian winter, and I think we both know that many weeks (12 or whatever the average is) of straight of blizzard is simply impossible. But I suspect that this is not something Arnaud can fix easily

It is too bad that the weather is not linked, I thought it was based on what I had seen in the game.

The effect of such extended blizzard periods is going to strongly depend on what fixes are made to the blizzard bug. I would be very curious to see if there is a difference in PBEM or human vrs human games.