About RBC

[alexzhz]

When attacker is PO, defender is more likely to RBC, but played the same scenario , I barely found out that defender RBCed when I strongly attacked, so what's the problem here? It seems PO is programmed to cheat [X(]

[Panama]

ORIGINAL: Spider

When attacker is PO, defender is more likely to RBC, but played the same scenario , I barely found out that defender RBCed when I strongly attacked, so what's the problem here? It seems PO is programmed to cheat [X(]

I wouldn't be surprised since the PO is not capable of thought nor planning. Something has to be done to make a single player game challenging. One of the common things is to give the PO in a game advantages that the human player doesn't receive. Otherwise it's a cakewalk and boredom sets in.

HOI3 is a good example. It's better than it was before the update but still, you have to be exceptionally dim witted not to win as the German player. Not tried the USSR nor US player yet but I'm sure it's not that hard to win with either of those.

[sPzAbt653]

It seems PO is programmed to cheat

AHAH ! I KNEW it !!

[ralphtricky]

ORIGINAL: Spider

When attacker is PO, defender is more likely to RBC, but played the same scenario , I barely found out that defender RBCed when I strongly attacked, so what's the problem here? It seems PO is programmed to cheat [X(]

Can you please post or email directions on how to reproduce that, possibly with a save game.

Thanks,
Ralph.

[ralphtricky]

ORIGINAL: sPzAbt653

It seems PO is programmed to cheat

AHAH ! I KNEW it !!

I'm pretty sure we tested RBCs thoroughly, but we may have missed a case.

Elmer cheats in the following ways only, unless I made a mistake.

He gets multiple tries at RBCs, but they are at 4:1. humans get one try, but it's at 4:1 * random (0.7-1.3). That may sometimes give him a slight edge, sometimes the human, but not significant. It would require a major rewrite to make him follow the same rules.

He understands the exact strengths of even unknown but visible units. He doesn't know about visible units. I might be able to do something down the line to have him make probing attacks or understand the makeup of the enemies forces, or something.

He uses the strengths of surrounding units, I think that includes invisible ones, to set his stance only, also the overall strengths of the forces for the same purpose. That may make him hold instead of advancing, or retreat instead of holding but that's probably the extent of the effect.

That's it in 3.4. 3.2 ran at a higher handicap level, but I've removed that for 3.4.

If you've got test scenarios that show differently, I'll be glad to look into it again.

I'm pretty proud of him tactically, strategically I'm depending on the scenario designer.

RBCs do take supporting artillery into account, it's possible that Elmer just manages them better, he tries to keep them in reserve mode at all times.

Ralph

[Panama]

With the emerging technologies (memristors) there will come a time when the AI will have to be handicapped to give the human player a fighting chance. [;)]

http://cosmiclog.msnbc.msn.com/_news/20 ... moores-law

[sPzAbt653]

I'm pretty proud of him tactically, strategically I'm depending on the scenario designer.

Since I am mostly interested in solitaire play, I think Elmer does very well. Playtesting by different players is a huge help. I can't think of anything that's come up that can't be overcome by changing some settings or jiggling the objectives. I didn't know that valued objectives made a difference to Elmer until recently. I have been biased against placing objectives all over the maps, I might have to change that once I get a grip on what difference it makes.

As far as him cheating, I swear he knows when he will get a 'double turn/initiative change'. No proof, just a paranoid suspicion.

Elmer shouldn't bombard or air attack the same hex each round when he is causing 0% casualties. I can't see what he is doing other than wasting his supply, and in the larger scenarios, its a waste of time. Seems like he should realise after the second round that he is having no effect.

He also tends to send units thru breaks in the line putting his units at terrible risk. In some scenarios this is desired, but I would like to think that the 'Advance' order would cover that. Most of the time he has no regard for where his neighboring friendly flank units are. This happens even under cautious bias.

I have no knowledge or training in these matters, just making some freshman observations.

[ralphtricky]

ORIGINAL: sPzAbt653

I'm pretty proud of him tactically, strategically I'm depending on the scenario designer.

Since I am mostly interested in solitaire play, I think Elmer does very well. Playtesting by different players is a huge help. I can't think of anything that's come up that can't be overcome by changing some settings or jiggling the objectives. I didn't know that valued objectives made a difference to Elmer until recently. I have been biased against placing objectives all over the maps, I might have to change that once I get a grip on what difference it makes.

I just changed that behavior actually. Prior to 3.4 he didn't care very much about objective values. Now he will try to take any objective that has a value, and try to take the area for the rest of the objectives. There's more I can do there, obviously.

As far as him cheating, I swear he knows when he will get a 'double turn/initiative change'. No proof, just a paranoid suspicion.

Elmer shouldn't bombard or air attack the same hex each round when he is causing 0% casualties. I can't see what he is doing other than wasting his supply, and in the larger scenarios, its a waste of time. Seems like he should realise after the second round that he is having no effect.

That requires what's called an influence map. I need to track casualties in an area to help him avoid trying to take the same spot time after time. I can do it, it's just not a priority yet. I hadn't thought about tracking the bombarments separately, though.

He also tends to send units thru breaks in the line putting his units at terrible risk. In some scenarios this is desired, but I would like to think that the 'Advance' order would cover that. Most of the time he has no regard for where his neighboring friendly flank units are. This happens even under cautious bias.

I need to document how he works better, He does pay some attention to flanking in some circumstances, but it's hard to balance making an advance and not getting flanked. Right now, he's all out aggressive.

I have no knowledge or training in these matters, just making some freshman observations.

I don't either, I'm afraid. All the advice seems to be pretty fuzzy too. Eventually, I want him to recognize killing grounds and weak points, be able to do terrain analysis for chokepoints, and set goals to work for cooperatively. I'm working through the armchair general tactics articles now to see if they help. It's going to be the next game before I can start putting some of those ideas into place, but I am working on them in my mind.

I also know I need to add more flexibility to the AI routines, I think I can do part of that for the next patch to give the designers a little more control.

Ralph

[ralphtricky]

ORIGINAL: Panama

With the emerging technologies (memristors) there will come a time when the AI will have to be handicapped to give the human player a fighting chance. [;)]

http://cosmiclog.msnbc.msn.com/_news/20 ... moores-law

Not without massive advances in software[X(] The thing that caught my attention was that it takes a neuron something like 1 Millisecond to respond to a change, and they timed monkeys watching an object. It took 20 milliseconds for them to see the object (excluding the time it took for them to look at it directly.) That means that the shortest path for neuron firing to recognize an object was only 20 neurons 'deep'. I know a bit about neural networks, but I can't even begin to conceive how something like that works on a large scale like the brain.

Computer transistors work at 3 Ghz, roughly 3 million times faster... That should help you sleep better tonight <evil grin>

Ralph

[alexzhz]

ORIGINAL: ralphtrick

ORIGINAL: Spider

When attacker is PO, defender is more likely to RBC, but played the same scenario , I barely found out that defender RBCed when I strongly attacked, so what's the problem here? It seems PO is programmed to cheat [X(]

Can you please post or email directions on how to reproduce that, possibly with a save game.

Thanks,
Ralph.

I just finished my test on Korea 50-51 as UN, even there's no evidence I still have to say Elmer really had enjoyed his advantage at RBC especially at the beginning of the scenario, he almost beat back all my front units without using a single combat round, and I'm so sure his advantage is significant, because when I did the same thing as Communist, there's no RBC at all, how terribly unfair it is!

It's not the complaints about the difficult level, for it's much easier to win this scenario as UN force.

I agree that Elmer should have some advantage while against human, but not in this way, we already have the cheat options to give Elmer some extra advantages, if we are not satisfied with the current difficult level we'll set computer+1 or computer+2.

[sPzAbt653]

I let the PO run the Communist side in Korea 50-51 and as you said, he rbc'd most of the front line units and by the end of his turn the entire line was gone. I restarted and played the Communist side myself, and while I didn't get as many rbc's that Elmer did, I got several. At the end of the turn I was in about the same situation that the PO had, with the entire UN line gone. I played it with no handicap to the PO. So there must be something different between your test and mine, because you said that you got no rbc's when you played the Communist side.

[alexzhz]

ORIGINAL: sPzAbt653

I let the PO run the Communist side in Korea 50-51 and as you said, he rbc'd most of the front line units and by the end of his turn the entire line was gone. I restarted and played the Communist side myself, and while I didn't get as many rbc's that Elmer did, I got several. At the end of the turn I was in about the same situation that the PO had, with the entire UN line gone. I played it with no handicap to the PO. So there must be something different between your test and mine, because you said that you got no rbc's when you played the Communist side.

To simulate the way how PO attack, I didn't plan any attack at the beginning, only by right clicking to try a RBC. Make sure you tested it at the same way.

There is a simple way to distinguish RBC attack from planned attack, a battle result window would come out after a planned attack, while no imformation about combat result after RBC attack.

[alexzhz]

BTW, I've turned on combat report.

[sPzAbt653]

Understood - we are talking about the same thing.

[ralphtricky]

It looks like there is a bug. With 178, Elmer has a better chance to RBC than a human does. It's fixed with the latest internal build. Sorry about the mistake.

I did some tests with Korea, and it does slow him down a little, but he still makes pretty good progress.

Thanks Spider and sPzAbt653,
Ralph

[jmlima]

ORIGINAL: sPzAbt653

I let the PO run the Communist side in Korea 50-51 and as you said, he rbc'd most of the front line units and by the end of his turn the entire line was gone. ...

Just yesterday, on Frozen Steppes, I saw the Axis PO RBC the entire Soviet Uranus offensive... fair enough the scenario unit values are pro-German (proficiencies of 85 and 95!), but even then...

[sPzAbt653]

ORIGINAL: jmlima

ORIGINAL: sPzAbt653

I let the PO run the Communist side in Korea 50-51 and as you said, he rbc'd most of the front line units and by the end of his turn the entire line was gone. ...

Just yesterday, on Frozen Steppes, I saw the Axis PO RBC the entire Soviet Uranus offensive... fair enough the scenario unit values are pro-German (proficiencies of 85 and 95!), but even then...

I love rbc's ! I'm not a fan of only getting one shot at a stack, but that's life.

I experienced my first reverse rbc. Attacking one enemy unit (a defense factor of 4) that was surrounded by five similar units and one with a df of 2. When I attacked with three of the similar strength units, the defender rbc'd the non-attacking weaker unit and excaped! An excellent function, but as with the one shot at a stack, I can also noodle thru why it maybe isn't realistic. Not complaining, just whining.

[Panama]

I feel it's equivalent to a unit in a pocket breaking out through a weakly held line. Might force people to forget about dividing recon battalions into companies to 'surround' divisions.

Much of what Ralph has done will make people rethink strategies of the past.

[sPzAbt653]

ORIGINAL: Panama

I feel it's equivalent to a unit in a pocket breaking out through a weakly held line. Might force people to forget about dividing recon battalions into companies to 'surround' divisions.

Much of what Ralph has done will make people rethink strategies of the past.

I agree 100%. But I wonder if the entire unit is breaking out, or if part of it is rear-guarding the other 5 hexsides ?

[Panama]

There would have to be a rear guard and it would engage only the units pushing and how heavily engaged would depend on how hard they are being pushed. I guess that would be decided by the orders emphasis: minimize they would receive probing attacks. limited they may be in contact the entire way but not heavily engaged. ignore losses they would be heavily engaged, maybe to the point where they get creamed.

Is this getting into the tactical end of things or are we still talking about operational? It's getting kind of foggy. [;)]