Question, how often does the IJN fire torpedoes with no response from the USN?
This was one of the advantages of the LL torpedoes, being able to launch at longer ranges thereby decreasing the chance of being spotted and IIRC the launch system used by the IJN did not produce a 'flash' when torpedoes were launched.
The USN gained an advantage from their radar only when it spotted enemy ships and the TF CO ordered an immediate radical change of course and speed, because it was assumed that the torpedoes were already on the way. When this happen there was fewer torpedo hits. The gun fire control was generally better for the USN so the guns got more hits 'when they had a target'. Visually the IJN had much better night optics than the USN, this allow the spotting of enemy ships at longer ranges, this is what was offset by the USN's radar. The radar on the USN ships in the early part of the war was cranky, and unreliable, the operators were new and the commanders not very trusting of it.
I don't know if UV take into account where one side spots the other and is itself not spotted. The IJN's tactics (I know it's an operational game) was to spot the enemy launch a full spread of torpedo's withdraw out of visual range reload and go in again for a second launch and gun battle thus allowing the first spread to reek havoc and confusion. These tactic did not work later in the war because the USN's radar and radar use improved giving the USN first sighting capability. This would allow the USN to fire first but due to the powder used in USN ships they would be spotted at the same time, until even later in the war.
I assume the purpose of these test is to determine if game engine generates the appropriate results for the tactics used by the two navies. This should give the IJN first shot advantage and the possibility of not being spotted at all. Unless the USN radar gave warning of their presence at which point both sides would start firing, but the first shot would still go to the IJN. Until the USN night experience increased to a point where radar would start to make a difference as to which side is spotted first.
And for us non-statistic types do your result bear out the advantage of one side or the other being spotted first or does it merely show that both sides do almost equal damage to each other with a slight edge going to the IJN?
As for the LL torpedoes not being a ship killer, keep in mind the it was developed to be used against Battleships in a battle line.
Falling off my soap box, I yield the floor.
Reports from Area 51
Moderators: Joel Billings, Tankerace, siRkid
Version 2.0
Has anyone seen or heard of any upcoming patch 2.0 changes that may invalidate any of the testing we are currently doing? Things like changes to weapon capabilities, spotting, radar, etc.
Just wondering whether there was any reason to shelve the testing until the new patch.
Just wondering whether there was any reason to shelve the testing until the new patch.
Have no fear,
drink more beer.
drink more beer.
-
CapAndGown
- Posts: 3078
- Joined: Tue Mar 06, 2001 10:00 am
- Location: Virginia, USA
=========================================
Test set 7: Test high USN experience, no radar
=========================================
USN experience: 70
IJN experience: 20
No radar
Damage
IJN USN Diff
0 0 0
52 0 52
24 0 24
0 0 0
0 9 -9
122 106 16 USN crosses T
54 0 54
57 0 57
32 100 -68
35 35 0
173 80 93
29 104 -75 USN crosses T
0 0 0
0 19 -19 USN crosses T
19 0 19
133 147 -14
139 123 16
30 19 11 USN crosses T
103 0 103 Both surprised
111 25 86
-----------------------------------------------------------------------
Remarks: Even though the USN has more than three times the experience level of the IJN, it is unable to translate this into a decisive advantage. It may win a majority of the battles, but nowhere near as decisively as one might expect. Again, it appears that the Long Lance is a big equalizer.
Test set 7: Test high USN experience, no radar
=========================================
USN experience: 70
IJN experience: 20
No radar
Damage
IJN USN Diff
0 0 0
52 0 52
24 0 24
0 0 0
0 9 -9
122 106 16 USN crosses T
54 0 54
57 0 57
32 100 -68
35 35 0
173 80 93
29 104 -75 USN crosses T
0 0 0
0 19 -19 USN crosses T
19 0 19
133 147 -14
139 123 16
30 19 11 USN crosses T
103 0 103 Both surprised
111 25 86
-----------------------------------------------------------------------
Remarks: Even though the USN has more than three times the experience level of the IJN, it is unable to translate this into a decisive advantage. It may win a majority of the battles, but nowhere near as decisively as one might expect. Again, it appears that the Long Lance is a big equalizer.
-
CapAndGown
- Posts: 3078
- Joined: Tue Mar 06, 2001 10:00 am
- Location: Virginia, USA
I have run a few regression analyses on the data I have collected so far. These regressions were run over a data set that included 120 battles (refer to my previous posts).
I am a little hesitant to do this at this point because I am not anywhere near an expert on statistical analysis and I wish to check on whether my interpretation of the results is correct. So bear in mind that the following interpretation is subject to correction/revision. But I felt that some people might like to see some of the results, and if there are any statistics wizards out there they can weigh in with their own opinions/corrections.
The most important relation I have been trying to get at is the relationship between experience level and chance of victory. This is modified, of course, by several other factors, such as TF commander rating, ship captain rating, (suposedly) crossing the T, etc. For purposes of this regression, only two variables are considered: IJN experience - USN experience, and SG Radar
The dependent variable is IJN damage - USN damage
R=.556 (the regression explains 55% of the cases)
................................................Unstandardized Coefficients
.............................................................B........Std. Error........Sig.
Constant..........................................-32.9..........12.7........ .011
SG Radar...........................................81.773......18............ .000
IJN experience - USN experience............-1.212...... .2........... .000
SG Radar ranges between 0 (no radar) and 1 (radar present)
From my reading of this relationship, SG Radar is worth a differential of about 67 experience points (call it 70 for simplicity's sake). Thus an IJN TF with 90 experience can expect to beat a USN TF with 20 experience and SG radar. If, however, the USN TF has a higher experience rating (which is always the case) then, on average, the IJN can be expected to lose.
One other interesting thing can be seen from this relationship. If SG radar is set to 0 (i.e. no SG radar) and the difference between USN experience and IJN experience is 0, then the Constant is the amount of damage that will occur, on average. Since the Constant is -32.9 (+/-12.7) this correlates with visual inspection of the data that shows the significant impact of the Long Lance. In other words, if the USN does not have SG radar and experience levels are equal, then the IJN can be expected to dish out 32.9 (+/-12.7) more damage than it recieves. This extra damage must be due to the Long Lance.
Other (in)signifcant coorelations:
The number of cases where one side or the other crossed the T is rather small. The USN crossed the T in 15 cases, the IJN crossed the T in only 7 cases. Nevertheless, it is very bothersome that there is no significant correlation between crossing the T and victory. The signifigance for the coorelation between the USN crossing the T and IJN - USN damage is 12%. For the IJN, the signfigance for IJN - USN damage is a somewhat higher at 5.9% (though most social scientists choose the 5% level for a cut off). At least, however, the variables point in the right direction, even if they are significant only at the 12% and 6% level: USN crosses T== .142; IJN crosses the T== -.173 Then again, these correlations are so small that even if we allow for a 6% cut off, then in only 17% of the cases will the IJN benefit from crossing the T. Over all, it appears that crossing the T has little to no beneficial impact on the outcome of a battle.
Now that the test set has grown from 60 cases to 120, the coorelations look somewhat different from what was last reported.
IJN experience is significantly correlated with USN damage.
Correlation: .343; Sig. .000
USN experience is significantly correlated with IJN damage.
Correlation: .298; Sig. .001
On the other hand, SG radar has a much greater correlation with IJN damage
Correlation: .525; Sig. .000
This, of course should be expected from the above regression
What has changed from the previous run is the correlation between experience and damage suffered. Thus, damage suffered is generally related to the enemy's experience, and not one's own, though in the case of the USN, there is a very slight correlation between experience and the damage it recieves:
Correlation: -.252; Sig. .005 I would not put much emphasis on this however, because of the variability of the radar variable (hence the purpose of the regression analysis).
The next task is to run a set of tests using SC radar rather than SG radar. It will be most interesting to see what the magnitude of the impact of SC radar is versus differences in experience.
I am a little hesitant to do this at this point because I am not anywhere near an expert on statistical analysis and I wish to check on whether my interpretation of the results is correct. So bear in mind that the following interpretation is subject to correction/revision. But I felt that some people might like to see some of the results, and if there are any statistics wizards out there they can weigh in with their own opinions/corrections.
The most important relation I have been trying to get at is the relationship between experience level and chance of victory. This is modified, of course, by several other factors, such as TF commander rating, ship captain rating, (suposedly) crossing the T, etc. For purposes of this regression, only two variables are considered: IJN experience - USN experience, and SG Radar
The dependent variable is IJN damage - USN damage
R=.556 (the regression explains 55% of the cases)
................................................Unstandardized Coefficients
.............................................................B........Std. Error........Sig.
Constant..........................................-32.9..........12.7........ .011
SG Radar...........................................81.773......18............ .000
IJN experience - USN experience............-1.212...... .2........... .000
SG Radar ranges between 0 (no radar) and 1 (radar present)
From my reading of this relationship, SG Radar is worth a differential of about 67 experience points (call it 70 for simplicity's sake). Thus an IJN TF with 90 experience can expect to beat a USN TF with 20 experience and SG radar. If, however, the USN TF has a higher experience rating (which is always the case) then, on average, the IJN can be expected to lose.
One other interesting thing can be seen from this relationship. If SG radar is set to 0 (i.e. no SG radar) and the difference between USN experience and IJN experience is 0, then the Constant is the amount of damage that will occur, on average. Since the Constant is -32.9 (+/-12.7) this correlates with visual inspection of the data that shows the significant impact of the Long Lance. In other words, if the USN does not have SG radar and experience levels are equal, then the IJN can be expected to dish out 32.9 (+/-12.7) more damage than it recieves. This extra damage must be due to the Long Lance.
Other (in)signifcant coorelations:
The number of cases where one side or the other crossed the T is rather small. The USN crossed the T in 15 cases, the IJN crossed the T in only 7 cases. Nevertheless, it is very bothersome that there is no significant correlation between crossing the T and victory. The signifigance for the coorelation between the USN crossing the T and IJN - USN damage is 12%. For the IJN, the signfigance for IJN - USN damage is a somewhat higher at 5.9% (though most social scientists choose the 5% level for a cut off). At least, however, the variables point in the right direction, even if they are significant only at the 12% and 6% level: USN crosses T== .142; IJN crosses the T== -.173 Then again, these correlations are so small that even if we allow for a 6% cut off, then in only 17% of the cases will the IJN benefit from crossing the T. Over all, it appears that crossing the T has little to no beneficial impact on the outcome of a battle.
Now that the test set has grown from 60 cases to 120, the coorelations look somewhat different from what was last reported.
IJN experience is significantly correlated with USN damage.
Correlation: .343; Sig. .000
USN experience is significantly correlated with IJN damage.
Correlation: .298; Sig. .001
On the other hand, SG radar has a much greater correlation with IJN damage
Correlation: .525; Sig. .000
This, of course should be expected from the above regression
What has changed from the previous run is the correlation between experience and damage suffered. Thus, damage suffered is generally related to the enemy's experience, and not one's own, though in the case of the USN, there is a very slight correlation between experience and the damage it recieves:
Correlation: -.252; Sig. .005 I would not put much emphasis on this however, because of the variability of the radar variable (hence the purpose of the regression analysis).
The next task is to run a set of tests using SC radar rather than SG radar. It will be most interesting to see what the magnitude of the impact of SC radar is versus differences in experience.
-
CapAndGown
- Posts: 3078
- Joined: Tue Mar 06, 2001 10:00 am
- Location: Virginia, USA
============================================
Test set #7: Test high IJN experience, USN SC radar
============================================
IJN experience: 95
USN experience: 35
USN ships have SC radar
Damage...................................Ships Hit
IJN...........USN..........Diff.........IJN..........USN
151..........127..........24............7..............5 IJN crosses T
0..............137..........-137.........4..............5 Both surprised
199..........127..........72............5..............5 Both surprised
13............69............-56...........2..............3
16............0..............16............3..............1
58............124..........-66...........4..............3
0..............0..............0..............1..............0
127..........331..........-204.........5..............7 Both surprised
28............12............16............1..............3 IJN crosses T
103..........394..........-291.........6..............6 Both surprised
42............108..........-66...........3..............4
0..............0..............0..............2..............1 USN crosses T
176..........121..........55............6..............7
69............149..........-80...........3..............6 USN crosses T
59............91............-32...........4..............2 Both surprised
89............289..........-200.........6..............4 Both surprised
17............14............3..............1..............1
36............77............-41...........2..............3 Both surprised
74............141..........-67...........5..............4 Both surprised
52............127..........-75...........5..............3 Both surprised
------------------------------------------------------------------------
Remarks: I have included the number of ships hit here because I find it interesting that in many cases the USN hit more ships than the IJN even though it came off worse in the damage department. I think I shall have to do a regression analysis that compares number of ships hit with experience and radar as independent variables. This may be even more revealing of the impact of radar than using the difference in damage as the dependent variable.
Again, we see that crossing the T yeilds no advantage.
Test set #7: Test high IJN experience, USN SC radar
============================================
IJN experience: 95
USN experience: 35
USN ships have SC radar
Damage...................................Ships Hit
IJN...........USN..........Diff.........IJN..........USN
151..........127..........24............7..............5 IJN crosses T
0..............137..........-137.........4..............5 Both surprised
199..........127..........72............5..............5 Both surprised
13............69............-56...........2..............3
16............0..............16............3..............1
58............124..........-66...........4..............3
0..............0..............0..............1..............0
127..........331..........-204.........5..............7 Both surprised
28............12............16............1..............3 IJN crosses T
103..........394..........-291.........6..............6 Both surprised
42............108..........-66...........3..............4
0..............0..............0..............2..............1 USN crosses T
176..........121..........55............6..............7
69............149..........-80...........3..............6 USN crosses T
59............91............-32...........4..............2 Both surprised
89............289..........-200.........6..............4 Both surprised
17............14............3..............1..............1
36............77............-41...........2..............3 Both surprised
74............141..........-67...........5..............4 Both surprised
52............127..........-75...........5..............3 Both surprised
------------------------------------------------------------------------
Remarks: I have included the number of ships hit here because I find it interesting that in many cases the USN hit more ships than the IJN even though it came off worse in the damage department. I think I shall have to do a regression analysis that compares number of ships hit with experience and radar as independent variables. This may be even more revealing of the impact of radar than using the difference in damage as the dependent variable.
Again, we see that crossing the T yeilds no advantage.
Posted by cap_and_gown
Mulder, you're alive!:p
I know you're just testing to establish the impact of experience but bear in mind that the 95/35 exp mix will not be the average for early naval encounters in UV. Keep going with the figures that give you the best correlation to track experience but also keep in mind the early averages for the game would be along the lines of 80/45 for IJN/USN experience.
An explanation for more USN hits on the IJN but less IJN damage overall could be assumed to be from the fast firing secondary batteries of the USN cruisers. From watching my test battles, I noticed about 2/3 of the hits that USN CA/CL's got on IJN ships were from the 5" batteries (which normally wont penetrate IJN CA armour). The IJN CA's were a 50/50 mix (or better) of 8" and 5" hits on USN ships.
I have a pet theory about my test battles involving IJN CA's vs USN CA's. It goes something like this (clears throat).
Battle begins.
Better experienced IJN more likely (initially) to hit with 8" gun than lesser experienced USN (+SC radar).
USN is less accurate but if the range is short to medium (under 9000 yds), their rapid firing 5" guns will have many more chances to hit. About every second or third 5" hit creates a fire on an IJN cruiser. Any USN cruiser firing at the burning IJN ship has its 8" accuracy markedly improved (to the extent it now hits as often as the IJN 8" guns or better).
Conclusion :
a) that the USN CA's will tend to get more hits on their IJN counterparts overall (8" and 5" combined).
b) the 5" hits will do little in the way of damage.
c) battles tend to be heavily influenced by how many ships are set on fire and how quickly.
Recommenation : modify IJN and USN cruisers to eliminate their secondary batteries so as to allow better correlation between experience and damage. 8" guns and torps are really the only weapons that cause damage to an opposing cruiser. Therefore only those hits need to be considered in this excercise as 5" guns cause no real damage to cruisers but do introduce the "fire" wildcard which can make correlation difficult. I am assuming you are fighting 2 identically composed TF's (same no of cruisers and destroyers). In most battles like that, cruisers tend to fire at cruisers and destroyers at destroyers (at least for the majority of the battle).
When I restart my testing in 2 weeks (if needed), I was planning to test using CA's without secondaries (I hope the editor accepts the mod).
Remarks: I have included the number of ships hit here because I find it interesting that in many cases the USN hit more ships than the IJN even though it came off worse in the damage department. I think I shall have to do a regression analysis that compares number of ships hit with experience and radar as independent variables. This may be even more revealing of the impact of radar than using the difference in damage as the dependent variable.
Mulder, you're alive!:p
I know you're just testing to establish the impact of experience but bear in mind that the 95/35 exp mix will not be the average for early naval encounters in UV. Keep going with the figures that give you the best correlation to track experience but also keep in mind the early averages for the game would be along the lines of 80/45 for IJN/USN experience.
An explanation for more USN hits on the IJN but less IJN damage overall could be assumed to be from the fast firing secondary batteries of the USN cruisers. From watching my test battles, I noticed about 2/3 of the hits that USN CA/CL's got on IJN ships were from the 5" batteries (which normally wont penetrate IJN CA armour). The IJN CA's were a 50/50 mix (or better) of 8" and 5" hits on USN ships.
I have a pet theory about my test battles involving IJN CA's vs USN CA's. It goes something like this (clears throat).
Battle begins.
Better experienced IJN more likely (initially) to hit with 8" gun than lesser experienced USN (+SC radar).
USN is less accurate but if the range is short to medium (under 9000 yds), their rapid firing 5" guns will have many more chances to hit. About every second or third 5" hit creates a fire on an IJN cruiser. Any USN cruiser firing at the burning IJN ship has its 8" accuracy markedly improved (to the extent it now hits as often as the IJN 8" guns or better).
Conclusion :
a) that the USN CA's will tend to get more hits on their IJN counterparts overall (8" and 5" combined).
b) the 5" hits will do little in the way of damage.
c) battles tend to be heavily influenced by how many ships are set on fire and how quickly.
Recommenation : modify IJN and USN cruisers to eliminate their secondary batteries so as to allow better correlation between experience and damage. 8" guns and torps are really the only weapons that cause damage to an opposing cruiser. Therefore only those hits need to be considered in this excercise as 5" guns cause no real damage to cruisers but do introduce the "fire" wildcard which can make correlation difficult. I am assuming you are fighting 2 identically composed TF's (same no of cruisers and destroyers). In most battles like that, cruisers tend to fire at cruisers and destroyers at destroyers (at least for the majority of the battle).
When I restart my testing in 2 weeks (if needed), I was planning to test using CA's without secondaries (I hope the editor accepts the mod).
Have no fear,
drink more beer.
drink more beer.
-
CapAndGown
- Posts: 3078
- Joined: Tue Mar 06, 2001 10:00 am
- Location: Virginia, USA
======================================
Test set #8: Test equal experience, SC radar
======================================
USN experience: 65
IJN experience: 65
USN has SC radar
Damage
IJN USN Diff
60 100 -40 Both surprised
279 143 136 USN crosses T
171 369 -198
75 93 -18
85 173 -88
133 180 -47
62 15 47
30 102 -72
105 87 18 Both surprised
16 0 16
26 46 -20
34 143 -109
52 26 26
135 60 75 USN crosses T
197 292 -95 USN crosses T
84 123 -39
47 0 47
77 96 -19 Both surprised
0 0 0
238 299 -61
--------------------------------------------------------------------
Remarks: Well the IJN seemed to come out OK here. However, it should be kept in mind that right along the IJN has been able to inflict more damage due to at least two factors: the Long Lance, and the fact that the USN CA's used in these test runs have half the belt armor. In this case, the IJN inflicted on average 22 more damage on the USN per engagement. When the USN had no radar and equal experience, the IJN inflicted on average 35 more damage on the USN per engagement. According the the regression analysis (see above) the IJN could be expected to deliver 33 more damage (+/-13) on the USN if experience was equal and the USN had no radar. With only 20 test cases here, it is hard to say what this result will mean for the regression constant, but it does appear that SC radar is having an impact on the outcome. What the magnitude of that impact is will have to wait until after the next set of tests.
After the next set of tests I will include the spreadsheet I am using to store this data so that interested parties can play with the numbers on their own.
Interestingly, the USN actually won two battles where it crossed the T.
Of course, then it turned right around and lost one where it crossed the T. :rolleyes: Moral of the story: if you are watching the replay and it says "Lee crosses the T", don't get your hopes up if you are the US. (And don't despair if you are play the Japs.)
Test set #8: Test equal experience, SC radar
======================================
USN experience: 65
IJN experience: 65
USN has SC radar
Damage
IJN USN Diff
60 100 -40 Both surprised
279 143 136 USN crosses T
171 369 -198
75 93 -18
85 173 -88
133 180 -47
62 15 47
30 102 -72
105 87 18 Both surprised
16 0 16
26 46 -20
34 143 -109
52 26 26
135 60 75 USN crosses T
197 292 -95 USN crosses T
84 123 -39
47 0 47
77 96 -19 Both surprised
0 0 0
238 299 -61
--------------------------------------------------------------------
Remarks: Well the IJN seemed to come out OK here. However, it should be kept in mind that right along the IJN has been able to inflict more damage due to at least two factors: the Long Lance, and the fact that the USN CA's used in these test runs have half the belt armor. In this case, the IJN inflicted on average 22 more damage on the USN per engagement. When the USN had no radar and equal experience, the IJN inflicted on average 35 more damage on the USN per engagement. According the the regression analysis (see above) the IJN could be expected to deliver 33 more damage (+/-13) on the USN if experience was equal and the USN had no radar. With only 20 test cases here, it is hard to say what this result will mean for the regression constant, but it does appear that SC radar is having an impact on the outcome. What the magnitude of that impact is will have to wait until after the next set of tests.
After the next set of tests I will include the spreadsheet I am using to store this data so that interested parties can play with the numbers on their own.
Interestingly, the USN actually won two battles where it crossed the T.
Of course, then it turned right around and lost one where it crossed the T. :rolleyes: Moral of the story: if you are watching the replay and it says "Lee crosses the T", don't get your hopes up if you are the US. (And don't despair if you are play the Japs.)-
CapAndGown
- Posts: 3078
- Joined: Tue Mar 06, 2001 10:00 am
- Location: Virginia, USA
==========================================
Test set #9: Test high USN experience; SC radar
==========================================
USN experience: 85
IJN experience: 45
all USN ships have SC radar
Damage
IJN USN Diff
71 60 11
120 38 82
184 68 116
58 0 58
76 101 -25
98 64 34
244 135 109 Both surprised
197 164 33 Both surprised
34 0 34
214 74 140
0 20 -20
62 102 -40 Both surprised
84 108 -24 Both surprised
141 207 -66 Both surprised
0 0 0
51 108 -57 Both surprised
0 71 -71
161 77 84 Both surprised
71 44 27 Both surprised
81 33 48 Both surprised
-------------------------------------------------------------------
Remarks: In case anybody was wondering, I am using different experience numbers in each test run in order to generate more data points for the regression and correlation analyses. I have included the spread sheet I have been storing this data in here for anyone who may be interested. But because of the way I have been varying the data, any simple comparison of cases with out using statistical analysis may be problematic. But, should you wish to run your own analysis, or if you would just like to look at the numbers, you have the table.
The next test run will varying the surface rating of the different TF comanders. It would be nice to know whether Tanaka and Lee really make that much of a difference.
Test set #9: Test high USN experience; SC radar
==========================================
USN experience: 85
IJN experience: 45
all USN ships have SC radar
Damage
IJN USN Diff
71 60 11
120 38 82
184 68 116
58 0 58
76 101 -25
98 64 34
244 135 109 Both surprised
197 164 33 Both surprised
34 0 34
214 74 140
0 20 -20
62 102 -40 Both surprised
84 108 -24 Both surprised
141 207 -66 Both surprised
0 0 0
51 108 -57 Both surprised
0 71 -71
161 77 84 Both surprised
71 44 27 Both surprised
81 33 48 Both surprised
-------------------------------------------------------------------
Remarks: In case anybody was wondering, I am using different experience numbers in each test run in order to generate more data points for the regression and correlation analyses. I have included the spread sheet I have been storing this data in here for anyone who may be interested. But because of the way I have been varying the data, any simple comparison of cases with out using statistical analysis may be problematic. But, should you wish to run your own analysis, or if you would just like to look at the numbers, you have the table.
The next test run will varying the surface rating of the different TF comanders. It would be nice to know whether Tanaka and Lee really make that much of a difference.
- Attachments
-
- uv stats.zip
- (13.98 KiB) Downloaded 13 times
-
CapAndGown
- Posts: 3078
- Joined: Tue Mar 06, 2001 10:00 am
- Location: Virginia, USA
And now, after 180 test battles, it is possible to start coming to some conclusions. Of course, Scully has promised us some more tests that would eliminate the 5" guns to see what kind of difference that would make. Her theories about the course of the battle and why a lower experienced USN is able to do as well as it appears to be doing seem to be highly probable.
Well first off, somewhat to my surprise, a regression analysis of the effects of SC radar show that it is having no statistically significant impact on the outcome of the battle.
dependent variable: IJN damage - USN damage
R=.433
...................................B..............Std. Error...............Sig.
Constant..................-31.8.............9.3...................... .001
SC radar..................19.23............13.218................. .148
IJN experience -
USN experience .......-.863........... .168.................... .000
According to this, SC radar is worth a differential of about 20 experience points. But since the calculation only shows a significance at the 15% level, (and the Std. error is almost equal to the size of this impact!) one can say that for all practical purposes, the impact of SC radar is very small indeed. Perhaps if more test cases were run (and I will be doing them) then the impact of SC radar could be more accurately determined, even if it were very small. But for now, I think it safe to say that SC radar is not playing much of a role in these battles.
This is in sharp contrast to SG radar. In that case we saw that SG radar was worth a difference of almost 70 experience points and that result was statistically significant.
Another thing to note is the constant. In my first regression, the constant was -33 (+/-13). Here it is almost the same: -32 (+/-13) What this says is that the number of test cases used (60 for SG radar and 60 for SC radar and 60 for no radar) is large enough to even out the impact of the random number generator. Basically, the constant is saying that for the ships I chose to use for my tests, the IJN ships have a raw ability (unmodified by experience, US radar, etc.) of delivering 33 (+/-13) more damage on the USN ships chosen.
In the previous regression, the impact of the difference in experience levels was -1.2 (+/-.2) per point of experience differential. Here it is -.86 (+/-.17) per point of experience differential. Since these numbers overlap they seem to point strongly to a conclusion that each 1 point you have in more experience than your opponent, you can, on average, expect to inflict 1 point more damage on him than you recieve yourself.
Indeed, every regression I have run points to this same conclusion: 1 point in experience gain means 1 point more damage inflicted on the enemy.
What this might mean is that the experience variable is effecting accuracy more than it is effecting probability of firing, or at least it is also effecting accuracy as well as probability of firing. Anyway, I think what Drongo and I (and some others) are seeing is that experience is not given as much weight as we expected and so we are somewhat surprised that the early IJN advantage is not as visible as we would have thought.
This is confirmed by some correlations:
IJN experience - USN experience :: IJN damage - USN damage
-.427
This correlation is after 180 battles. It is interesting that this number is not much different from the previous two times I ran this analysis.
after 60 battles: -.48
after 120 battles: -.434
(note that all these correlations are significant at the 1% level)
Now it would be silly to think that there would or should be a 1 to 1 correlation (i.e. 100%) between experience advantage and victory. As they say, Sh#t Happens. The japs, despite all their training, did have their heads handed to them on occasion. But because the correlation is only 43% we are not seeing as many victories as we had expected. Were the correlation on the order of 60%, then I think there would be less questioning of the combat model. This, at base, is where I think the source of doubt arises.
So much for the numbers. I will be running some more tests. These will first be geared towards gaugin the impact of the TF commander. Then I want to look at the impact of TF size differences. More test runs should also make it possible to firm up these numbers and make those that are now not significant (i.e. SC radar) significant (whatever there level of impact might be, large or small). I think that Scully's test, when she runs them, should also be of interest. But there does now seem to be a large enough data base to work with in discusions of the combat model. Both Drongo and Nikademus have put forth very interesting ideas that should be explored further.
Well first off, somewhat to my surprise, a regression analysis of the effects of SC radar show that it is having no statistically significant impact on the outcome of the battle.
dependent variable: IJN damage - USN damage
R=.433
...................................B..............Std. Error...............Sig.
Constant..................-31.8.............9.3...................... .001
SC radar..................19.23............13.218................. .148
IJN experience -
USN experience .......-.863........... .168.................... .000
According to this, SC radar is worth a differential of about 20 experience points. But since the calculation only shows a significance at the 15% level, (and the Std. error is almost equal to the size of this impact!) one can say that for all practical purposes, the impact of SC radar is very small indeed. Perhaps if more test cases were run (and I will be doing them) then the impact of SC radar could be more accurately determined, even if it were very small. But for now, I think it safe to say that SC radar is not playing much of a role in these battles.
This is in sharp contrast to SG radar. In that case we saw that SG radar was worth a difference of almost 70 experience points and that result was statistically significant.
Another thing to note is the constant. In my first regression, the constant was -33 (+/-13). Here it is almost the same: -32 (+/-13) What this says is that the number of test cases used (60 for SG radar and 60 for SC radar and 60 for no radar) is large enough to even out the impact of the random number generator. Basically, the constant is saying that for the ships I chose to use for my tests, the IJN ships have a raw ability (unmodified by experience, US radar, etc.) of delivering 33 (+/-13) more damage on the USN ships chosen.
In the previous regression, the impact of the difference in experience levels was -1.2 (+/-.2) per point of experience differential. Here it is -.86 (+/-.17) per point of experience differential. Since these numbers overlap they seem to point strongly to a conclusion that each 1 point you have in more experience than your opponent, you can, on average, expect to inflict 1 point more damage on him than you recieve yourself.
Indeed, every regression I have run points to this same conclusion: 1 point in experience gain means 1 point more damage inflicted on the enemy.
What this might mean is that the experience variable is effecting accuracy more than it is effecting probability of firing, or at least it is also effecting accuracy as well as probability of firing. Anyway, I think what Drongo and I (and some others) are seeing is that experience is not given as much weight as we expected and so we are somewhat surprised that the early IJN advantage is not as visible as we would have thought.
This is confirmed by some correlations:
IJN experience - USN experience :: IJN damage - USN damage
-.427
This correlation is after 180 battles. It is interesting that this number is not much different from the previous two times I ran this analysis.
after 60 battles: -.48
after 120 battles: -.434
(note that all these correlations are significant at the 1% level)
Now it would be silly to think that there would or should be a 1 to 1 correlation (i.e. 100%) between experience advantage and victory. As they say, Sh#t Happens. The japs, despite all their training, did have their heads handed to them on occasion. But because the correlation is only 43% we are not seeing as many victories as we had expected. Were the correlation on the order of 60%, then I think there would be less questioning of the combat model. This, at base, is where I think the source of doubt arises.
So much for the numbers. I will be running some more tests. These will first be geared towards gaugin the impact of the TF commander. Then I want to look at the impact of TF size differences. More test runs should also make it possible to firm up these numbers and make those that are now not significant (i.e. SC radar) significant (whatever there level of impact might be, large or small). I think that Scully's test, when she runs them, should also be of interest. But there does now seem to be a large enough data base to work with in discusions of the combat model. Both Drongo and Nikademus have put forth very interesting ideas that should be explored further.
-
LargeSlowTarget
- Posts: 5053
- Joined: Sat Sep 23, 2000 8:00 am
- Location: Hessen, Germany - now living in France
A minor remark: As luck would have it I have just read about the Battle of the Java Sea in S.E. Morison. He writes that during the engagement Captain Waller of HMAS Perth observed 'a row of flashes' along a Japanese column and - assuming this meant torpedo launches - ordered his ship to turn away. The other ships in the Striking Force followed and no torpedo struck home - on this occasion. But this may be an isolated example of luck, instinct or 'battle awareness', so it does not necessarily negate your claim of a flashless launch system.Originally posted by pbear
This was one of the advantages of the LL torpedoes, being able to launch at longer ranges thereby decreasing the chance of being spotted and IIRC the launch system used by the IJN did not produce a 'flash' when torpedoes were launched.
The 'flash' was from the torpedo's hitting the water and causing the 'florencence' effect (I'm not sure but I believe that this is caused by plankton in the water at these latatudes). The 'flashless' launch system developed by the Japanese was due to the use of pure oxygen as a torpedo propelent. While oxygen will not in itself burn it promotes burning in any other material, this means that for saftey reasons you don't want a spark of any kind around pure oxygen.Originally posted by LargeSlowTarget
A minor remark: As luck would have it I have just read about the Battle of the Java Sea in S.E. Morison. He writes that during the engagement Captain Waller of HMAS Perth observed 'a row of flashes' along a Japanese column and - assuming this meant torpedo launches - ordered his ship to turn away. The other ships in the Striking Force followed and no torpedo struck home - on this occasion. But this may be an isolated example of luck, instinct or 'battle awareness', so it does not necessarily negate your claim of a flashless launch system.
The American torpedos were launched with an explosive charge thus a 'flash' was sometimes seen. This is not a problem when using compressed air to drive a torpedo but with an oxygen torpedo you could set off all the torpedo warheads in the launcher which could totally ruin your day. Please note that this applies only to surface launched torpedos, submarines used a compressed air launch.
From Japanese Destroyer Captain by Capt. Tameichi Hara.(paraphrased)
At 1927 the Japanese fired 72 torpedos at 6,000 meters Capt. Hara estimated less than 5% of the torpedos would hit, about 3. The enemy fleet made a rapid turn to the west and about half of the torpedos passed ahead of the enemy the rest of the 'fish' reached the target area all seemed to miss then there was a tremendous flash. The Dutch Destroyer Kortenaer was hit and sank instantly. One hit. "What poor aiming and how excellent were the enemy's evasion tactics!"
So much for the great Japanese torpedo tactics, of course this was open sea with lots of manoeuvre room. And the launch was spotted.
Mulder,
Just letting you know I'm going to bail out from Area 51 testing for a while. I've just upgraded to 2.0 and the save/load delays makes testing too time consuming (even though its all head to head) on games started in 1.4.
If I get a chance to set up test scenario's from scratch under 2.0, I might give it another go later.
Until then, I'll spend my time constructively by trying to raise our alien baby as a normal human child.
Keep those child maintenance cheques coming otherwise I'll see you in court.
Scully
Just letting you know I'm going to bail out from Area 51 testing for a while. I've just upgraded to 2.0 and the save/load delays makes testing too time consuming (even though its all head to head) on games started in 1.4.
If I get a chance to set up test scenario's from scratch under 2.0, I might give it another go later.
Until then, I'll spend my time constructively by trying to raise our alien baby as a normal human child.
Keep those child maintenance cheques coming otherwise I'll see you in court.
Scully
Have no fear,
drink more beer.
drink more beer.
-
CapAndGown
- Posts: 3078
- Joined: Tue Mar 06, 2001 10:00 am
- Location: Virginia, USA
You go sleeping around with some strange alien, whose name you probably don't even remember, and which you couldn't pronounce anyway, and you want me to pay child support?!
How long have we been together now Scully? Six seasons, at least. And have you ever offered to do the wild thing with me? No! But I have to pay for this stupid baby who will probably mature within a week (meaning no time to save up for college). I say, you want the child support, you got's to put out wench!
How long have we been together now Scully? Six seasons, at least. And have you ever offered to do the wild thing with me? No! But I have to pay for this stupid baby who will probably mature within a week (meaning no time to save up for college). I say, you want the child support, you got's to put out wench!
Posted by Mulder
You mean it wasn't you? F*ck, either your lying, I was pissed or that alien was a very good shapechanger (he even got your little Mr. Happy right).
Do DNA paternity tests work on alien babies? Is Tau Ceti IV outside our court jurisdictions? This could justify an extra series. $$$$.
You go sleeping around with some strange alien, whose name you probably don't even remember, and which you couldn't pronounce anyway, and you want me to pay child support?!
You mean it wasn't you? F*ck, either your lying, I was pissed or that alien was a very good shapechanger (he even got your little Mr. Happy right).
Do DNA paternity tests work on alien babies? Is Tau Ceti IV outside our court jurisdictions? This could justify an extra series. $$$$.
Have no fear,
drink more beer.
drink more beer.