Need advice: serious problem with naval target selection

[Christof]

Hi all,
this has recently happened in a PBEM game (version 2.3):
My IJN air combat TF (2CVL, 2CV, 1CS - commanded by Yamaguchi) stumbles across an USN TF with 3CV, 1CVE.

All TF's are being discovered and some ship names detected during the air ops phase. I set my Zeros to "Escort" 50% CAP and all my bombers to Naval attack / Rest

The enemy carriers are 6 hexes away and react to shorten the distance to 5 hexes. My TF is set to "Do not react / do not retire".

Weather is clear.

The USN looses three morning strikes that find my carriers but do not do a lot of damage. My strike with some 90 bombers is directed against a TF purely consisting of DD's, so far away (9-10 hexes) that all Vals attack with reduced payload and my Kates with bombs.

Afternoon, same procedure. USN finds my carriers, my planes go after the DD's.

Next day same procedure in both morning and afternoon.

Only once 12 Vals arrive over his air combat TF but are ineffective.

End of story: all four carriers on my side sunk, aircrews gone. I damaged three DD's and the CVE. Gave up the game.

What did I do wrong?
Has something like this happened to anybody else before?

Sorry guys, but from my point of view this looks like a bug.


:rolleyes:

[Toro]

Hi Christof,

No, you didn't do anything wrong, and it appears to be (sorry for the term) just bad luck.

UV attack decisions are based strongly on sighting reports, and each turn there are two (I believe) "patrol/scout" periods which bring in targeting information on enemy TFs. It appears that, unfortunately for you, the sightings were much more current on the DD TF than the CV TF. As in real life, a commander wouldn't send aircraft after a "chance" target when a strong, valid one existed. Thus, the info on the CVs was just plain sketch, while the DDs were a known quantity, hence the attack went there.

Toro

[wobbly]

HI, I do wonder about that Toro...

I think it may have something to do with CAP levels. When the game was patched the designers changed the threshold at which bomber groups would contemplate an attack on the target depending on the amount of CAP as opposed to amount of escort you have. It was done to stop unescorted bomber raids attacking naval targets at bases that had massive CAP above the airfields.

I thought it only influenced LBA, but it also appears to influence Carrier - Carrier battles some. In my game with Herbie I have managed to survive his attacks but he has not in return. The main differences I have noted have been in approach to CAP - he put in large escorts, low CAP; I put in large CAP low escort. (Herbie right me if I am wrong here)

So my bombers don't have a large CAP to overcome - their reduced escort suffices; his attack has a large CAP to overcome (here's where the theory fails a little though as he is allocating a large escort) and doesn't attack. He also used many small isolated groups where I put my carriers into a large TFs for instance:

Herbie - JAP: 1 CV 2 CVL
Me - Allied: 4CV

So he has, with 20% CAP on all carriers (and assuming all planes fly and full carriers), 63 zeros of which about 12 will CAP and 50 should escort.
I have 144 F4Fs with CAP levels of 80 - escort 20 which means I have about 115 or so on CAP and 30 or so escorting.

What I am hinting at here is in this case the 50 zero escort may not be enough to overcome the 115 CAP numbers...
In christof's case he had better numbers but an alternative target - with little or no CAP I imagine. They chose the lesser of two evils.
I think this is just another call for "give me target priorities"!!!

[Christof]

Toro:
I understand the point you make. I would have no problem accepting a targeting decision by the AI like that if only the days first strike was concerned.

The case was that my planes kept attacking that dreaded DD's over and over again, and my CV's were at the same time under servere attack by an enemy CV-TF only five hexes away. Then my commander suddenly sends in 12 Vals with plenty of escorts that burn through the allied CAP (I think some 40 F4F-4's) but get hammered by flak.

Next day same procedure, but this time not even a single plane selects to attack the origin of all the trouble.

I'm not really looking for a setting like target priorities here. I like NOT having complete control over everything and making operational decisions.
But the decision making process of the AI engine is flawed, because the commander will make the decision to "react to enemy" if an enemy air combat TF is encountered (CV's were high priority targets in WW2), but refuse to attack and go after a DD force instead.

Complete madness and a serious problem whether a bug or a cheat...

[Joel Billings]

Toro:
I understand the point you make. I would have no problem accepting a targeting decision by the AI like that if only the days first strike was concerned.

The case was that my planes kept attacking that dreaded DD's over and over again, and my CV's were at the same time under servere attack by an enemy CV-TF only five hexes away. Then my commander suddenly sends in 12 Vals with plenty of escorts that burn through the allied CAP (I think some 40 F4F-4's) but get hammered by flak.

Next day same procedure, but this time not even a single plane selects to attack the origin of all the trouble.

I'm not really looking for a setting like target priorities here. I like NOT having complete control over everything and making operational decisions.
But the decision making process of the AI engine is flawed, because the commander will make the decision to "react to enemy" if an enemy air combat TF is encountered (CV's were high priority targets in WW2), but refuse to attack and go after a DD force instead.

Complete madness and a serious problem whether a bug or a cheat...

More than likely, Toro is right. In Gary's tests on the last set of bugs, the biggest answer to bad attacks was bad intel in the spotting phase which does not always become clear to the player. The checks on CAP vs Escort are set to allow even poorly escorted strikes to fly against known enemy carriers. Based on your info, this should not have been a factor (if I remember right, if you have at least 25 escorts, the computer is not going to say you don't have enough escorts versus CAP). However, if you can send me a save, just prior to the execution phase, I will check into why the attacks don't go off. The only caveat here is that since we have the bug in PBEM not consistently resolving the same on all computers, I may get a totally different result. Please send the saves to 2by3@2by3games.com. I'm sorry for the frustration. Gary devoted a lot of time and effort on 2.3 to try to resolve this issue, and the logic that I know of in the code made sense. All his tests during this time showed that bad attacks went off based on random factors having to do with the spotting reports. This did happen often during the real battles. If you didn't know the sequence of events during Coral Sea, and the quality of the sighting reports, you would think the carrier forces were often idiots. If your main strike went off against a DD and a tanker while enemy carriers were nearby, you'd be screaming at the computer.

Joel

[Thrashman]

More than likely, Toro is right. In Gary's tests on the last set of bugs, the biggest answer to bad attacks was bad intel in the spotting phase which does not always become clear to the player. Joel

In this type of situation, if it is intel that has caused the "bad attack", then why not present the intel to the player as it is presented in the code. What I mean is, and I don't know the order in which air search is done within the code, but if lets say the search routines are done in two phases, then present them to the player in two phases.

If the morning search spots two task forces, one containing DD and CL, and one containing DD and CA. Then the afternoon search discovers that the one TF with DD and CA also has a CV, then don't present this to the player until after the afternoon search phase. It appears that if the player see's the CV first thing in the morning, the player assumes his TF comanders know this intel when within the code they do not know that there is a CV out there until later in the day.

Is this a possibility?