Matrix Games Forums

Interesting, well done; sure looks like research to me!

Thank you for your effort. I would add some considerations. Particular pilots/crews are predestined in the game engine, somehow, to advance quickly and that is not easy to test, nor do I know if this can be changed in the editor. This applies to officers of whatever sort as well. You did not use the same pilots in each test or as part of a control group. Generally those would be minor deviations in any case. With that caveat, I accept your results in whole. I have not advocated having zero aircraft, or one aircraft; I do not have any personal experience with that. I do firmly (until somebody rubs my nose in my fallibility) believe that 1/3 of the allowed complement is more than sufficient for training purposes and would encourage all to proceed confidently with the understanding that that is so.

ORIGINAL: geofflambert

Thank you for your effort. I would add some considerations. Particular pilots/crews are predestined in the game engine, somehow, to advance quickly and that is not easy to test, nor do I know if this can be changed in the editor. This applies to officers of whatever sort as well. You did not use the same pilots in each test or as part of a control group. Generally those would be minor deviations in any case. With that caveat, I accept your results in whole. I have not advocated having zero aircraft, or one aircraft; I do not have any personal experience with that. I do firmly (until somebody rubs my nose in my fallibility) believe that 1/3 of the allowed complement is more than sufficient for training purposes and would encourage all to proceed confidently with the understanding that that is so.

It's not. The relationship is linear with respect to aircraft. Your most critical resource when it comes to training as Japan is air groups. Every one you dedicate to training cannot be used for combat. Maximizing them is more important than saving aircraft. And yes, the officers can make a difference, but I know that and corrected for it (as much as I can) before I ran the test. Moreover, the results would likely have been more skewed towards the 1 plane squadrons had I not corrected for it. One leader affects the advancement ALL pilots in the air group.

I've done a lot more testing on this than what you see here. I literally have over 10 pages of notes on research I've done, testing various aspects of the Pilot training issue and reading posts on the forums. See below:

That looks like damn fine research in my book. I stand corrected, and will definitely change how I have been training. I feel like I have wasted almost 2 years of the game now. The amount of info in this game is just staggering at times.

I am not mistaken.

This is a screenshot of VS-7D14 taken on 16 Dec 1942. The unit has only 2 aircraft and is at 100% training.



A check through my USA, Brit and NZ fighter fighter/medium bomber units in 100% training with only 1 or 2 aircraft on 16 Dec 1942 shows the following.

USA - fighter units

1 a/c - 34 pilots - 5 green
2 a/c - 33 pilots - 6 green
1 a/c - 33 pilots - 3 green
2 a/c - 33 pilots - 0 green
1 a/c - 33 pilots - 3 green
2 a/c - 34 pilots - 6 green

USA - medium bomber units

0 a/c - 21 pilots - 0 green (but previous turn it did have green pilots and currently has 11 brown pilots)
2 a/c - 21 pilots - 3 green
1 a/c - 21 pilots - 3 green
1 a/c - 21 pilots - 4 green

Brit - fighter units

2 a/c - 22 pilots - 4 green
2 a/c - 21 pilots - 6 green

Brit - medium bomber units

2 a/c - 16 pilots - 2 green

NZ - fighter units

2 a/c - 16 pilots - 8 green

This is hard data from an actual game, not some vanity test which has not taken into account all the relevant variables.

Alfred

ORIGINAL: Alfred
This is hard data from an actual game, not some vanity test which has not taken into account all the relevant variables.

And thus is more suspect, not less. Your test is invalid BECAUSE you do not control variables. The question is "What is the effect of the number of aircraft on training?"

To test that, you control the OTHER variables and then vary the number of aircraft:

1) Type of Aircraft
2) Number of pilots
3) location of training (weather affects whether training flights occur)
4) leadership rating of commanders
5) experience levels of pilots
6) morale of pilots
7) etc. etc.

ALL of the above are controlled for in my test. I am testing the game engine, not some random sampling of groups. And by the way, your percentage of pilots of pilots receiving skill increases vs mine:

For bombers 23/200 = 0.115 = 0.115 11.5%
For medium bombers: 10/84 = 0.119 = 11.9%
My 15 plane squadrons: 28/100 = 0.28 = 0.28 28%

Vanity exercise? Grow up and show some scientific method. Do not try to hurl insults at someone that has the audacity to challenge the views of the almighty Alfred. Refute my facts and avoid the personal attacks, k?

The one thing I will say from what I can see is that it is no wonder your #4 test group only got one pilot advanced one point in air skill. The unit commander is 'poor' to say the least. I wouldn't let him train my dog.[:D]

ORIGINAL: geofflambert

Alfred seems to have knowledge of the source code so that would make him one of those things he is "many" of.

I believe Alfred wrote the original manual and had to ask a lot of questions about the game engine to frame his information. He also needed to ask the developers what things should not be described in detail to avoid revealing too much of the code. He referenced meetings with developers several times, but he was not a developer per se.

He has a rigour of thought akin to a University Prof or Scientist - one that expects students to put in honest efforts and gives them their come-uppance when they do not. I have been guilty of lack of research too often and got my ears boxed accordingly. I'm thinking of having them bronzed! [8D]

EDIT: wrote this before I saw Alfred had joined the thread. I will concede the floor to him rather than speculate further on how he came to be so knowledgeable about the game.

ORIGINAL: rustysi

The one thing I will say from what I can see is that it is no wonder your #4 test group only got one pilot advanced one point in air skill. The unit commander is 'poor' to say the least. I wouldn't let him train my dog.[:D]

Per tm.asp?m=2350193 the leader's Leadership skill influences experience gain. Most of the leaders in the original test had Leadership in the high 30's, a few were in the low 40's and both groups had roughly equal leaders. I did not get 10 identical leaders, but I tried to get ones close in Leadership skill and other stats. I wanted lowish leader skill so that I could try to look at effects of plane numbers only. In other words, crappy leadership affected both groups roughly equally.

According to the Pilot Management Addendum (pdf in your /manuals directory), Leader skill kicks in on skill gain:

"- if the pilot’s experience is less 50 (plus pilot’s missions and kills) and less than the leader’s skill "

It is not stated if the leader's contribution to experience gain is affected by the amount by which Leadership exceeds the skill. Therefore I wanted leaders which exceeded the average exp of the pilots training, but not by much in case the effect was magnified for leaders with big advantages in skill. The actual results did not correlate much between the Leader's Leadership and the number of skill advances, but the number of trials was limited. Higher skill leaders had pilots both exceed and fall short of the average.

What is clear from the tests is that there is a strong relationship between number of planes and advances.

Initial test: all leaders in high 30's/low 40's:
1 plane groups got 3 advances in 100 trials: 3%
15 plane groups got 28 advances in 100 trials: 28%
Pilots in the 15 plane groups advanced skill 9.33 to 1 times more often than those in 1 plane groups.

I've run the test 5 times tonight and got similar results each time. In no case did a 1 plane group ever get more than one skill increase. In ALL cases, the 15 plan groups got AT LEAST 3 (max 9).

Most recent test (re-selected all leaders in 42-44 range for Leadership):
15 plane groups: +6, 3, 5, 4, 5 = +23 (23% of pilots got a skill increase)
1 plane groups: +1, 1, 1, 0, 0 = +3 (3% of pilots got a skill increase)

The 15 plane groups skill rate actually declined slightly with higher leadership leaders from the initial trial, but pilots in the 15 plane groups still advance 7.67 to 1 times more often than those in 1 plane groups.

ORIGINAL: InfiniteMonkey

ORIGINAL: Alfred
This is hard data from an actual game, not some vanity test which has not taken into account all the relevant variables.

And thus is more suspect, not less. Your test is invalid BECAUSE you do not control variables. The question is "What is the effect of the number of aircraft on training?"

To test that, you control the OTHER variables and then vary the number of aircraft:

1) Type of Aircraft
2) Number of pilots
3) location of training (weather affects whether training flights occur)
4) leadership rating of commanders
5) experience levels of pilots
6) morale of pilots
7) etc. etc.

ALL of the above are controlled for in my test. I am testing the game engine, not some random sampling of groups. And by the way, your percentage of pilots of pilots receiving skill increases vs mine:

For bombers 23/200 = 0.115 = 0.115 11.5%
For medium bombers: 10/84 = 0.119 = 11.9%
My 15 plane squadrons: 28/100 = 0.28 = 0.28 28%

Vanity exercise? Grow up and show some scientific method. Do not try to hurl insults at someone that has the audacity to challenge the views of the almighty Alfred. Refute my facts and avoid the personal attacks, k?

What rubbish you write about what I posted and then as a free kick insults are thrown in. Typical of someone with the ego of a small planet who simply cannot accept his tests are fundamentally flawed.

1. You cannot properly control because you do not know nor properly understand the variables.

2. You have done no such thing as a valid "scientific test". To do so requires only a single variable input to be tested. Over many iterations. And the entire set of test conditions are fully published so that any one else can replicate exactly the same test. You have not done any of this so don't insult me or any other reader with your self perceived "superior" scientific test.

3. Because your ego is so huge and the need to invalidate me is consuming you misrepresent my post. I presented the results of the game engine. Those results completely and utterly contradict your claims. Which I remind you is the claim that air units with fewer aircraft will train fewer pilots than fully TOE equipped air units.

4. Anyone with a scintilla of understanding knows that in a contest between game engine results and a mickey mouse test bed, the game engine results always trump what the mickey mouse test bed throws up.

5. A good tester, when confronted with what the diametrically opposed outcomes, would go back to his mickey mouse test bed and try to see where they failed.

The simple fact of life is that you are one of the numerous reverse engineers who frequent the forum who is incapable of accepting the truth when it is uttered by a dev or by someone with a better mind than yourself. This then makes you envious.

Practical players play the game as it is. They are not in a position to play the game in a vacuum divorced from all the externalities that exist in every single situation. I showed what the game engine does. If there was any validity to your mickey mouse test bed which you employed to purportedly show what the game engine does, there would not be such a discrepancy in outcomes.

There is no fact for me to further disprove because you have not produced a single fact about how this game engine operates. I show game reality, you show your make believe world.

I will repeat it again. Players do not need aircraft inside air units in order to generate pilot training. Air units with 1 or 2 aircraft in them can advance more pilots than air units with 100% TOE. Just as some air units with 100% TOE can also advance more pilots than other air units with a very low TOE%.

Show me a single dev comment that shows I am wrong and you are correct. You will not find one; not in the manual, not in the pilot addendum, not in the patch notes, not in any post in any thread in the forum. So on what basis can you sustain your superiority.

Constantly my comments get challenged by "reverse engineers" but they always fail, although they never apologise. In one particular 2014 (IIRC) thread after several pages of vehement antagonism against me Symon, in one of his last posts pointed out that I was completely correct. The very next post in that thread came from Symon who wryly noted how quickly silence descended when a dev spoke.

Just like the Greek Gods have left Mt Olympus, so have the devs left AE. I am as close as AE players can get to the knowledge of the devs precisely because I closely research what they have posted, and have often provided their source comments for independent verification. You provide ... give me a moment I'll think of something ... really there must be something you do ...

Alfred

ORIGINAL: rustysi

The one thing I will say from what I can see is that it is no wonder your #4 test group only got one pilot advanced one point in air skill. The unit commander is 'poor' to say the least. I wouldn't let him train my dog.[:D]

And while I am at it, let me re-emphasize what the /manuals/Pilot Management Addendum.pdf says:

Leader skill increases skill gain IF

"- if the pilot’s experience is less 50 (plus pilot’s missions and kills) and less than the leader’s skill"

If that is an accurate description (and it is in the manual, so I am going with it until testing tells me different), then as long as the Leadership is higher than the experience/skill, you are wasting better leaders on those trainee pilots. It doesn't say if the number is 10 higher, or 20 higher. It just says higher. Haven't tested this, but do not assume that medium or high skill leaders do any better than low skill leaders at training your greenest trainees.

ORIGINAL: Alfred
What rubbish you write about what I posted and then as a free kick insults are thrown in. Typical of someone with the ego of a small planet who simply cannot accept his tests are fundamentally flawed.

I said you were mistaken. That is hardly throwing stones.. You, however wrote "not some vanity test which has not taken into account all the relevant variables". And I'm the one throwing insults? So you are pure hearted and always correct, but I'm posting as a "vanity test"? What did you expect in my response?

ORIGINAL: Alfred
1. You cannot properly control because you do not know nor properly understand the variables.

2. You have done no such thing as a valid "scientific test". To do so requires only a single variable input to be tested. Over many iterations. And the entire set of test conditions are fully published so that any one else can replicate exactly the same test. You have not done any of this so don't insult me or any other reader with your self perceived "superior" scientific test.

I invite anyone who doubts me to replicate the test. Create identical air groups in as close to identical conditions as the game allows using the scenarios editor, then vary ONLY the number of aircraft in the group and see what happens to pilots that train in those respective groups. I have attached a screenshot for anyone that wants to duplicate my results.
1. Go into the scenario editor and open Scen 1.
2. Make TEST-1 through TEST-10 as you see TEST-1 in the screenshot.
3. On TEST-2/4/6/8/10, make the 'Ready' number of planes 1 isntead of 15.
4. File->Save As Scenario to another scenario slot.
5. Once in game, start a new game as Japan Go to Oita and pull pilots from the reserve to each air group until you have 20 pilots in each air group.
6. Pull leaders for each air group from the bottom of the leader list ranked by Leadership skill. Try for a number in the 40ish range.
7. Run the turn.
8. Look at pilot list for each group.
9. Run as many turns as you like to get confidence that number of planes has an impact on skill advancement.

ORIGINAL: Alfred
3. Because your ego is so huge and the need to invalidate me is consuming you misrepresent my post. I presented the results of the game engine. Those results completely and utterly contradict your claims. Which I remind you is the claim that air units with fewer aircraft will train fewer pilots than fully TOE equipped air units.

I'm not interested in invalidating you. In a recent post, I recommended a newbie read your back posts. That is hardly representative of invalidating you. I did note (correctly as my experience here confirms and posts confirm) that you are a crotchety guy, but that your information is usually good. My tests support my assertion, but your sample of data does not. You have not controlled other variables. By your own admission, the cases you cite are just a sampling of the pilot experience gains in a handful of squadrons that happen to have 1 or 2 aircraft in them. I've told you I created duplicate squadrons - the only difference in them was the number of planes and the leaders, though the leaders were as close as leader selection would allow me to make them.

ORIGINAL: Alfred
4. Anyone with a scintilla of understanding knows that in a contest between game engine results and a mickey mouse test bed, the game engine results always trump what the mickey mouse test bed throws up.

I used the game engine to perform the test, therefore my test is a test of the game engine. Not sure how you think I got the results without using the game to test. I am measuring dx/dy where dx is the change in experience and dy is the change in number of planes. All other variables are held constant.

ORIGINAL: Alfred
5. A good tester, when confronted with what the diametrically opposed outcomes, would go back to his mickey mouse test bed and try to see where they failed.

You might apply this to your own "test". Say, put a full complement of planes in an air group with the same starting experience, the same air craft, the same base, ample support, the same morale, attempt to match up leader stats, and pull fresh pilots form your reserve. Then compare the two groups where one has 1 or 2 aircraft, and the other has 20 something. You did not do that. I did.

ORIGINAL: Alfred
The simple fact of life is that you are one of the numerous reverse engineers who frequent the forum who is incapable of accepting the truth when it is uttered by a dev or by someone with a better mind than yourself. This then makes you envious.

Lol. MIT trained software engineer bud, I make my living doing this kind of stuff. I am very good at it and I am very well paid to do it.

ORIGINAL: Alfred
Practical players play the game as it is. They are not in a position to play the game in a vacuum divorced from all the externalities that exist in every single situation. I showed what the game engine does. If there was any validity to your mickey mouse test bed which you employed to purportedly show what the game engine does, there would not be such a discrepancy in outcomes.

Odd statement when you have no control groups and have not kept any of the variables that I cited constant in your sample. To disprove my position, you would have to present air groups with similar variables and larger numbers of aircraft training that did not advance at a greater rate. You haven't. You have effectively proven nothing.

ORIGINAL: Alfred
There is no fact for me to further disprove because you have not produced a single fact about how this game engine operates. I show game reality, you show your make believe world.

I will repeat it again. Players do not need aircraft inside air units in order to generate pilot training. Air units with 1 or 2 aircraft in them can advance more pilots than air units with 100% TOE. Just as some air units with 100% TOE can also advance more pilots than other air units with a very low TOE%.

I've shown you results that contradict your statement here. Not sure I can say anything that will ever convince you.

ORIGINAL: Alfred
Show me a single dev comment that shows I am wrong and you are correct. You will not find one; not in the manual, not in the pilot addendum, not in the patch notes, not in any post in any thread in the forum. So on what basis can you sustain your superiority.

Observed results. You develop a theory, you design a test for that theory, you measure the results. You draw conclusions from the results. Frankly, if a dev came on here and said you were right, my first action would be to send him the above scenario file and say, "Okay, then why do I get these results?" Devs are not infallible any more than you or I are.

ORIGINAL: Alfred
Constantly my comments get challenged by "reverse engineers" but they always fail, although they never apologise. In one particular 2014 (IIRC) thread after several pages of vehement antagonism against me Symon, in one of his last posts pointed out that I was completely correct. The very next post in that thread came from Symon who wryly noted how quickly silence descended when a dev spoke.

Just like the Greek Gods have left Mt Olympus, so have the devs left AE. I am as close as AE players can get to the knowledge of the devs precisely because I closely research what they have posted, and have often provided their source comments for independent verification. You provide ... give me a moment I'll think of something ... really there must be something you do ...

Alfred

I said you were wrong. I based it upon tests of the game engine using air groups that attempted to eliminate as many variables as might be involved in the test. The results were not just sorta clear, they were clear by an order of magnitude. Not once in your rant have you even considered that you, the devs, or anyone else might be wrong. If it walks like a duck, if it quacks like a duck, it's probably a duck. If I vary ONE thing, and that thing is the number of planes, and I get 7-9 times as many skill advancements, then I feel pretty comfortable thinking that the number of planes has a significant impact on the rate of training advancement.

I am as close as AE players can get to the knowledge of the devs precisely because I closely research what they have posted, and have often provided their source comments for independent verification.

From where do you draw this vast index of dev comments?

I for one would think it a massive boon to the community if it was to be shared publically.

Having an AE oracle is all well and good, but sometimes you want a quick question resolved without having to queue or sacrifice a goat. Or a gorn.

ORIGINAL: bushpsu

Only an observation but I am pretty sure this was covered (no surprise) by Alfred and there is supposedly NO difference in training based upon amount of planes - including 0 - or pilot/plane ratio. I remember when I read this I stopped worrying about it and have trained this way ever since, so I have never taken the time to compare how quickly the training occurs.

No, number of planes matters without doubt (or you play a different game or some mod?), as well good air support and bigger airfields matter. The huge airfields ofc will provide much better training infrastructur then a some muddy field in the middle of the jungle. So it makes perfect sense, as well perhaps good supply + HQ present(last 2 points not confirmed but would make sense too)... [;)]

Comment on CR Corsair example: those are very low in exp this may be the reason they train decent even with only 2 planes As most know the lower skill/exp the faster they learn... there is a threshold some point when it slows down... When I get a turn back perhaps I can post examples too if this issue is still open for debate [:)]

ORIGINAL: Alfred

ORIGINAL: InfiniteMonkey

ORIGINAL: Alfred
This is hard data from an actual game, not some vanity test which has not taken into account all the relevant variables.

And thus is more suspect, not less. Your test is invalid BECAUSE you do not control variables. The question is "What is the effect of the number of aircraft on training?"

To test that, you control the OTHER variables and then vary the number of aircraft:

1) Type of Aircraft
2) Number of pilots
3) location of training (weather affects whether training flights occur)
4) leadership rating of commanders
5) experience levels of pilots
6) morale of pilots
7) etc. etc.

ALL of the above are controlled for in my test. I am testing the game engine, not some random sampling of groups. And by the way, your percentage of pilots of pilots receiving skill increases vs mine:

For bombers 23/200 = 0.115 = 0.115 11.5%
For medium bombers: 10/84 = 0.119 = 11.9%
My 15 plane squadrons: 28/100 = 0.28 = 0.28 28%

Vanity exercise? Grow up and show some scientific method. Do not try to hurl insults at someone that has the audacity to challenge the views of the almighty Alfred. Refute my facts and avoid the personal attacks, k?

What rubbish you write about what I posted and then as a free kick insults are thrown in. Typical of someone with the ego of a small planet who simply cannot accept his tests are fundamentally flawed.

1. You cannot properly control because you do not know nor properly understand the variables.

2. You have done no such thing as a valid "scientific test". To do so requires only a single variable input to be tested. Over many iterations. And the entire set of test conditions are fully published so that any one else can replicate exactly the same test. You have not done any of this so don't insult me or any other reader with your self perceived "superior" scientific test.

3. Because your ego is so huge and the need to invalidate me is consuming you misrepresent my post. I presented the results of the game engine. Those results completely and utterly contradict your claims. Which I remind you is the claim that air units with fewer aircraft will train fewer pilots than fully TOE equipped air units.

4. Anyone with a scintilla of understanding knows that in a contest between game engine results and a mickey mouse test bed, the game engine results always trump what the mickey mouse test bed throws up.

5. A good tester, when confronted with what the diametrically opposed outcomes, would go back to his mickey mouse test bed and try to see where they failed.

The simple fact of life is that you are one of the numerous reverse engineers who frequent the forum who is incapable of accepting the truth when it is uttered by a dev or by someone with a better mind than yourself. This then makes you envious.

Practical players play the game as it is. They are not in a position to play the game in a vacuum divorced from all the externalities that exist in every single situation. I showed what the game engine does. If there was any validity to your mickey mouse test bed which you employed to purportedly show what the game engine does, there would not be such a discrepancy in outcomes.

There is no fact for me to further disprove because you have not produced a single fact about how this game engine operates. I show game reality, you show your make believe world.

I will repeat it again. Players do not need aircraft inside air units in order to generate pilot training. Air units with 1 or 2 aircraft in them can advance more pilots than air units with 100% TOE. Just as some air units with 100% TOE can also advance more pilots than other air units with a very low TOE%.

Show me a single dev comment that shows I am wrong and you are correct. You will not find one; not in the manual, not in the pilot addendum, not in the patch notes, not in any post in any thread in the forum. So on what basis can you sustain your superiority.

Constantly my comments get challenged by "reverse engineers" but they always fail, although they never apologise. In one particular 2014 (IIRC) thread after several pages of vehement antagonism against me Symon, in one of his last posts pointed out that I was completely correct. The very next post in that thread came from Symon who wryly noted how quickly silence descended when a dev spoke.

Just like the Greek Gods have left Mt Olympus, so have the devs left AE. I am as close as AE players can get to the knowledge of the devs precisely because I closely research what they have posted, and have often provided their source comments for independent verification. You provide ... give me a moment I'll think of something ... really there must be something you do ...

Alfred

Thank you, sir. May I have another? [:D]

ORIGINAL: Canoerebel

My experience is that aircraft squadrons with no aircraft will not train pilots;

They will but slowly

They also appear to train while being transported. This however, might be a bug where the info is not being cleaned when the group is loaded.

Training while in transport is the same mode as training without aircraft and is intentional. A number of fighter units trained their pilots in aerial gunnery while on board ships by shooting skeet off the stern of the ship. Some resourceful units that didn't have enough aircraft did formation flight training with bicycles (though not on ships). Other ground training classes could be done while the unit is in transit.

Bill

ORIGINAL: InfiniteMonkey
I invite anyone who doubts me to replicate the test. Create identical air groups in as close to identical conditions as the game allows using the scenarios editor, then vary ONLY the number of aircraft in the group and see what happens to pilots that train in those respective groups

Did not doubt you since training effectiveness depending on planes coincides with my observations. Yet did my own test in an editor just to confirm:

Set of identical conditions: 30 airgroups of 36-size IJA army Ki-27b fighter sentais with 38 pilots each, Train/Escort at 100% at 0 range at 5k, all groups tied to the same HQ which is present in the hex, 10 level airfield with excessive supply/support, 99 morale, 30 base xp (ranging from 21 to 37 for individual pilots), all commanders set at all skills 30, 1 turn run, scenario restarted 2 times (turns out no need for more for this particular test)

Difference: 10 sentais with 36 ready planes, 10 - with 2 ready planes, 10 - with zero planes

Results for average Air skill increases per sentai in 1 turn: 36 planes - 12.3, 2 planes - 0.15, 0 planes - zero.

Hypothesis of training not depending on # of planes can be refuted with 99+%. Alfred has a lot of theoretical knowledge but too big a mouth for his relevant empirical knowledge

Ok, folks, YES no plane units train but slow. See this Mavis unit (transport - I do not have any IJN transport for a while so could not give them planes, they arrived a week ago, bought back, unit was overrun in the south). Note it is a 10 airfield with plenty of AV at Tokyo: