$15.00
$5.00
$20.00
$30.00
Recommend
1 
 Thumb up
 Hide
8 Posts

Fire in the Lake» Forums » Rules

Subject: NVA Bot Attack Space Selection rss

Your Tags: Add tags
Popular Tags: [View All]
Bruce Wigdor
United States
Lake Hiawatha
New Jersey
flag msg tools
Hi Oerjan--I'm still trying to write the code for my program's NVA Attack Bot based on this new literal interpretation, and I wanted to confirm that I've got something right.

I believe that you have said that when prioritizing spaces for selection using the first four bullets, you don't consider how many losses will be eliminated from a space using the 5th through 8th bullets, you consider how many losses could be eliminated using the game rules and trying to maximize the losses from that one space. Do I have that right?

If so, is it correct that any space with at least one NVA Guerilla will count as a space where at least two enemy pieces could be be removed? Because they could be, with a successful attack by that Guerilla.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Oerjan Ariander
Sweden
HUDDINGE
flag msg tools
mbmbmbmb
brucewig wrote:
Hi Oerjan--I'm still trying to write the code for my program's NVA Attack Bot based on this new literal interpretation, and I wanted to confirm that I've got something right.

I believe that you have said that when prioritizing spaces for selection using the first four bullets, you don't consider how many losses will be eliminated from a space using the 5th through 8th bullets, you consider how many losses could be eliminated using the game rules and trying to maximize the losses from that one space. Do I have that right?

Yes.

Quote:
If so, is it correct that any space with at least one NVA Guerilla will count as a space where at least two enemy pieces could be be removed?

Almost. NVAbot never Attacks (as opposed to Ambushes) with groups of 3 or fewer Guerrillas, but if you replace "one NVA Guerrilla" above with "four NVA Guerrillas" then it is correct.

Regards,
Oerjan
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bruce Wigdor
United States
Lake Hiawatha
New Jersey
flag msg tools
Quote:
Almost. NVAbot never Attacks (as opposed to Ambushes) with groups of 3 or fewer Guerrillas, but if you replace "one NVA Guerrilla" above with "four NVA Guerrillas" then it is correct.


But that information is in the 7th bullet, and I thought we didn't consider how the bot actually would do things when selecting target spaces; we only considered what was possible within the rules. For example, you have stated that you count potential ambushers when prioritizing Target spaces even if there is no chance the Ambush will ever happen.

So Ambushes that could happen (but will not because of the 5th through 7th bullets) are considered. Why aren't single Guerilla attacks that could happen (but will not because of the 5th through 7th bullets) considered in the same way?

One more thing on the whole NVA attack bot procedure. When you told us to forget everything you said, you outlined the following:

Quote:
The NVAbot Attack process consists of three steps:

1) Check if the bot Attacks at all (the first flowchart diamond): it Attacks if its Troops will take Control somewhere, or remove a COIN Base, or remove the last COIN piece from a space, or remove 4+ COIN pieces from a single space. Guerrillas don't count at all for this decision, regardless of what their Attacks or Ambushes might achieve.

2) If the bot does execute an Attack Op, select all potential Attack spaces before resolving any of them (the first 4 bullets of 8.6.2), remembering that the bot can't select more spaces than it has Resources (3.1). In this selection process, LoC Ambushes removing pieces in South Vietnam are dead last in the priority order due to a number of quirks in the phrasing (that I'm fairly certain we didn't intend, but they're in the rules nevertheless).

3) Resolve the selected Attack spaces in the order established in step 2, only slightly modified by additional instructions if step 2 left any Attack spaces tied. During this step the bot can end up aborting Ambushes (due to an additional quirk); if it does, it does not select new Attack spaces to replace them (but might occasionally convert them into Attacks with 4+ Guerrillas or 2-3 Troops, if there are such groups of units in those spaces along with suitable victims).


This appears to me to be a consecutive step procedure. But then in your last post to me, you gave an example of the bot aborting an attack all the way in step 3.

Quote:
At this point the bot realizes that the Attack Op doesn't meet any of the conditions that would allow the bot to choose that Op in the first place! Note that the first diamond on the NVA flowchart says "Troop Attack will ...", not "...can...", so unless Attacking NVA Troops (not Guerrillas) are guaranteed to achieve at least one of the four tasks listed in the diamond the bot will skip to the next diamond (the one where it checks if it will use Terror). As we've seen above an Attack Op in this situation would only remove a single ARVN cube, which means that it doesn't achieve any of the four Attack tasks - so the bot won't Attack.


Which is correct? Is checking to see if an attack Op is possible indeed Step 1 of the procedure, or is it something that there is no reason to even bother checking until after Steps 2 and 3?

Thanks for the help as always!

 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Oerjan Ariander
Sweden
HUDDINGE
flag msg tools
mbmbmbmb
Yes, checking if an Attack Op is possible at all is the first part of the procedure. That check is made in the decision diamond where the NVAbot flowchart starts; NVAbot can only enter the Attack instruction box if the answer to that question is "yes". If the answer is "no", the bot continues to the second decision diamond.

What confuses the issue is that in order to answer the question in step 1 (the decision diamond), the bot needs to look ahead at the Attack instructions in the box and evaluate what effects an Attack Op would have at least until a positive answer is reached. If no positive answer is reached (which was the case in your example), that means tracing out the entire Attack process in advance while the bot is still technically in the decision diamond (i.e., step 1).

What you interpreted as "aborting in step 3" was the end of such a step-1 decision diamond evaluation that turned out to give the answer "no". I'm sorry for not emphasizing that fact more; that's what the initial "As we will soon see, NVAbot won't choose the Attack Operation at all in this situation" was meant to explain.

Regards,
Oerjan
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bruce Wigdor
United States
Lake Hiawatha
New Jersey
flag msg tools
OK, I see your intent on the 3 step process, so I won't question it further.

I am still confused about the Target selection. Do we consider the 5th through 8th bullets when selecting Target spaces, or do we not?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bruce Wigdor
United States
Lake Hiawatha
New Jersey
flag msg tools
Please forgive me, Oerjan--I said I wouldn't question the 3 step process any further, but upon programming this and reading the rulebook (as opposed to the flowchart), I need to ask again.

Quote:
(Oerjan) What confuses the issue is that in order to answer the question in step 1 (the decision diamond), the bot needs to look ahead at the Attack instructions in the box and evaluate what effects an Attack Op would have at least until a positive answer is reached. If no positive answer is reached (which was the case in your example), that means tracing out the entire Attack process in advance while the bot is still technically in the decision diamond (i.e., step 1).


Do we really need to look ahead? The flowchart seems to suggest that we do, because the first decision diamond says (emphasis mine):

Quote:
Troop Attack will add Control or remove Base, last COIN, or remove 4+ enemies from a space


But the rulebook says:

Quote:
they Attack if NVA Troops Attacking alone (without any Ambush or Bombard) would either add any NVA Control, or remove any US or ARVN Base, or remove the last or at least 4 enemy pieces from a space.


This wording ("would" instead of "will") seems to suggest that looking ahead is not necessary. What's more, since there is further detail in the rules than in the decision diamond, I would think the rulebook would take precedence.

And requiring the decision diamond to look ahead is so cumbersome! You need to do everything, including rolling dice to randomly select spaces when you prioritize them, and then, if Ambush is not possible, roll more dice to randomly select the Bombardment sites before you can check to see if any NVA troops will attack alone and fulfill the requirement. Do we really need players to go through this entire attack process every time the NVA flowchart is consulted, even when this first decision diamond will result in a "no?"

There is not another decision diamond on the entire flowchart that has a requirement that needs to check an end result. I don't think the NVA Attack decision diamond should either.

I hate doing this, but this is a very unusual circumstance where you have admitted that the wording of the rule and your original intent don't match up. So, given the fact that the flowchart and rules seem (to me, at least) to be at odds, I've got to ask--are you sure about your ruling that the decision diamond needs to look ahead?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Oerjan Ariander
Sweden
HUDDINGE
flag msg tools
mbmbmbmb
No, "would" doesn't suggest that. "Could" might suggest it, but that's not the word used.

"Would" is the correct future-looking form of the verb at the point in time when you haven't yet determined whether or not to execute the Operation. It asks what would actually happen in case the bot selects the Attack Operation, instead of asking what could possibly happen during an Attack.

As for being cumbersome, I believe I've mentioned that I consider the current NVAbot rule to be badly messed up and in need of replacement. To be more explicit, it is a train wreck between three different and mutually contradicting sets of good intentions. That being the case, are you sure you want to code instructions you already know will be replaced eventually? If you're willing to do that, why not use your project to try out other alternatives and see if the resulting bot actions make sense? You'll have to re-code anyway when the bot instruction is revised, after all.

Regards,
Oerjan
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bruce Wigdor
United States
Lake Hiawatha
New Jersey
flag msg tools
Quote:
"Would" is the correct future-looking form of the verb at the point in time when you haven't yet determined whether or not to execute the Operation. It asks what would actually happen in case the bot selects the Attack Operation, instead of asking what could possibly happen during an Attack.


Not necessarily. You can point to a space where an attack would occur and say "here, if Troops Attack alone, without an Ambush or a Bombardment, the attack would eliminate the last COIN piece in this space." That works semantically too, without actually executing the whole procedure.

Quote:
I believe I've mentioned that I consider the current NVAbot rule to be badly messed up and in need of replacement. To be more explicit, it is a train wreck between three different and mutually contradicting sets of good intentions.


Actually, it was all pretty clear when you were interpreting the rules as they appeared to read. I still struggle to read the rules this new way, because there are other sentences and other structures that to me seem at odds with these new interpretations.

Quote:
That being the case, are you sure you want to code instructions you already know will be replaced eventually?


I don't, but I have to do something. But the more I look at this current interpretation, the more cumbersome it gets (for both player and computer), and the more it seems like the bot will be making suboptimal choices.

If you're saying it's broken and needs to be replaced, though I believe that I will try to program the AI so that it matches what I believe your original intent was for the time being. I actually think that was pretty good.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.