Pilot skill (from Merchant Shipping Thread)

[vettim89]

ORIGINAL: herwin

OK, the RAF OR studies seemed to show that combat improved defensive skills. Offensive skills were better trained up or at least were not improved as quickly in combat. I speculate it was due to the greater availability of feedback in training.

I moved this out of the Merchant Shipping thread because I felt it was worth discussing but didn't want to hijack the OP.

Harry, don't you think this is somewhat of a self-fulfilling prophecy? By that I mean the neophyte pilots who survives to fly again probably did so by improving his defensive skills. Over time, those pilots that improved in that skill the most would be preselected to be overrepresented in the test group. I cannot recall what movie it was, but I recall a scene where the new guy asks the grizzled veteran if he had any advice for him. The reply was, "Try not to get yourself killed out there".

I realize that this is hard coded and cannot be altered except by Matrix, but it does seem a bit off that high skills in certain areas seem to equal such fantastic results

[herwin]

ORIGINAL: vettim89

ORIGINAL: herwin

OK, the RAF OR studies seemed to show that combat improved defensive skills. Offensive skills were better trained up or at least were not improved as quickly in combat. I speculate it was due to the greater availability of feedback in training.

I moved this out of the Merchant Shipping thread because I felt it was worth discussing but didn't want to hijack the OP.

Harry, don't you think this is somewhat of a self-fulfilling prophecy? By that I mean the neophyte pilots who survives to fly again probably did so by improving his defensive skills. Over time, those pilots that improved in that skill the most would be preselected to be overrepresented in the test group. I cannot recall what movie it was, but I recall a scene where the new guy asks the grizzled veteran if he had any advice for him. The reply was, "Try not to get yourself killed out there".

I realize that this is hard coded and cannot be altered except by Matrix, but it does seem a bit off that high skills in certain areas seem to equal such fantastic results

The RAF data seem to indicate that defensive skills for fighter pilots improved dramatically over their first six engagements, and it wasn't simply due to survival of the most skilled. Pilots learned. In game terms, each engagement increased their defensive skill by about 5 points. There was no such performance increase for offensive skills.

The data also show that RAF patrol bomber crews didn't get better over time in operations, but a few days of deliberate training improved their performance markedly.

Trained prewar Japanese torpedo bomber crews had double the hit probability of American crews. That's the difference between a 50 and a 90.

Early in the war, the VS squadron on American carriers was made up of the newbies. Their aircraft carried half the bomb-load that the veterans in the VB squadron carried.

[witpqs]

Summarizing my remarks on this subject from the merchant shipping thread and other threads:

---

I suspect that pilot skill above roughly 70 yields too much benefit to certain attack missions, specifically:

[b]
• Airfield Attack
• Ground Attack
• Naval Attack
• ASW[/b]

The concern is that at a certain point the benefit from skill level provides results that exceed what the technology (or technology plus doctrine) were capable of.

At certain points in the game the naval attack issue favors the IJ player, at other times it favors the Allied player, although I think it favors the Allied player more by the end of the game. The ground attack and airfield attack issues overwhelmingly favor the Allied player for the great majority of the game. The ASW issue could favor either depending upon player style but most likely favors the IJ player overall.

I feel that in those air missions listed above, the contribution from (the relevant) pilot skill greater than about 70 should be adjusted to be one of fiercely diminishing returns. By diminishing returns I mean that a pilot skill change of 65 to 70 would yield a significantly greater benefit than a pilot skill change of 70 to 75; and a pilot skill change of 75 to 80 would yield still far, far less benefit, and so on. I mean this on a point for point bases and only use 5-point divisions for clarity of illustration.

From what I have seen, Port Attack does not seem overly powered. I do not have as much experience with this attack mode, so this view could easily be mistaken.

Air to air combat is different from the surface attack modes as it has been shown that ever higher levels of pilot skill do IRL yield better and better air to air results. In other words, do not subject air to air to the same level of diminishing returns - leave it alone.

I would like to point out that this is different than proposals to, for example, prohibit the players from using army bombers on ASW missions. I disagree with those proposals and believe that players should retain the flexibility to train and use their forces as they see fit. For example, the IJ player should be able to use all the air units he sees fit on ASW duty, but each of those planes/pilots/crews should only be able to achieve what was possible (regarding chances to spot, hit, and/or kill a sub).

[herwin]

Let {x_i} be a collection of variables describing an air attack. Some of them might be usually considered inputs (skills, aircraft numbers, etc.) and some outputs (damage done, losses, etc). Some may even be intermediate variables. Ideally, you use the same variables used by the game. You need to know the conditional distribution of each variable with respect to the other variables. That might be null if there's no causal connection, or the variable might be a parameter and not vary. Otherwise use the game. Now wander through the variables that are actually variable and resample a few hundred times using the conditional distributions (or the game). The overall distribution should converge. Take about 100 or 1000 samples from the final distribution and compare to history. If they are not significantly different, you have a good model of the air attack. If you have to go back and fix things get history, you know something is wrong with the game (or your conditional distributions).