RE: Historical AI?

Matrix Games Forums

Forums Register Login Photo Gallery Member List Search Calendars FAQ

My Profile Inbox Address Book My Subscription My Forums Log Out

RE: Historical AI?

View related threads: (in this forum | in all forums)

Logged in as: Guest

Users viewing this topic: none

Printable Version

All Forums >> [Current Games From Matrix.] >> [World War I] >> Guns of August 1914 - 1918 >> RE: Historical AI?

Page: << < prev 1 [2]

Message

<< Older Topic Newer Topic >>

RE: Historical AI? - 10/2/2007 12:56:26 AM

SMK-at-work

Posts: 3396
Joined: 8/28/2000
From: New Zealand
Status: offline

i've only seen it do this when it thinks it is attacking an empty hex - usually with the Russians - at het same time I move in 2-3 corps and the Ruskies get crushed. I don't think I've ever seen it with the French.

(in reply to EUBanana)

Post #: 31

RE: Historical AI? - 10/2/2007 1:04:05 AM

hjaco

Posts: 872
Joined: 3/23/2007
Status: offline

I have the same experience as SMK.

(in reply to SMK-at-work)

Post #: 32

RE: Historical AI? - 10/2/2007 1:52:13 AM

EUBanana

Posts: 4552
Joined: 9/30/2003
From: Little England
Status: offline

Really? I've seen this constantly. Its probably the biggest reason why I now thrash the AI every time. It throws away troops in offensives like theres no tomorrow.

_____________________________

(in reply to hjaco)

Post #: 33

RE: Historical AI? - 10/2/2007 2:21:32 AM

SMK-at-work

Posts: 3396
Joined: 8/28/2000
From: New Zealand
Status: offline

Get 1.2!!

(in reply to EUBanana)

Post #: 34

RE: Historical AI? - 10/3/2007 5:33:35 PM

Joel Rauber

Posts: 195
Joined: 10/4/2000
From: Brookings, SD, USA
Status: offline

quote:

ORIGINAL: hjaco

Perhaps the solution is a new Militia/Garrison unit instead to all powers ? They should be next to useless in ordinary field battle but get a bonus if attacked in a home city ?

After all the British had Territorial Guards of substantial size in both Egypt and Kuwait. Although i have no knowledge of this it would make sense that a garrison would be in place at all cities with major harbor facilities in Europe ?

Something like this seems like a good idea and may help with some of the invasion problems (Wilhelmshaven??) that people notice. Perhaps something like all cities have a 4-point garrison (or other value) It should be large enough to throw some doubt on an amphibious invasion but small enough that a reasonably sized land attack would usually be successful; and/or the quality rating adjusted so that the above would happen. What would work best depends on the defence bonus that defenders from amphibious invasions receive vs. the city bonus for defence against land attacks.

_____________________________

Any relationship between what I say and reality is purely coincidental.

Joel Rauber

(in reply to hjaco)

Post #: 35

RE: Historical AI? - 10/7/2007 11:32:19 PM

Graf Leinsdorf

Posts: 23
Joined: 9/27/2007
Status: offline

quote:

ORIGINAL: Graf Leinsdorf

As former boardgamer I was taught that playing the role of the force historically defeated was a greater challenge, like for instance winning at Waterloo as Napoleon or at Gettysburg as Lee. I would hope that Gun of August be developed in a way that AI is hard to be beaten AT LEAST when it plays as EP and the human player is the historical underdog.

Playing the beta 1.2 update version as CP against the Entente AI, I was just appreciating a more challenging behavior of Entente Allied on the Western Front, when, in the November-December 1916 game turn, Britain all of sudden surrendered. A quite mysterious as well unhistorical oucome, considering the successful conduct of operations in France at the time. Thus a critical situation for Germany is strangely reversed and, in the end, I�m not able to remark any apparent improvement of AI resilience in comparison to previous versions, as already complained in this thread.
Moreover, together with some notable improvements in game mechanics and interface, 1.2 update displays perhaps some new flaws:
- inability to strategic move various units along the game, in spite of plenty of rail points and selection of eligible hexes;
- inability, twice in the game, to activate a Bulgarian HQ, in spite of its 2 offensive points, against Rumanians and Russians.

(in reply to Graf Leinsdorf)

Post #: 36

RE: Historical AI? - 10/8/2007 12:49:41 AM

FrankHunter

Posts: 2111
Joined: 3/26/2004
Status: offline

quote:

- inability to strategic move various units along the game, in spite of plenty of rail points and selection of eligible hexes;

This is happening? I haven't received any saved games about this.

quote:

- inability, twice in the game, to activate a Bulgarian HQ, in spite of its 2 offensive points, against Rumanians and Russians.

If you have a save where I can run this I'd be happy to fix it.

(in reply to Graf Leinsdorf)

Post #: 37

RE: Historical AI? - 10/8/2007 9:58:53 PM

Graf Leinsdorf

Posts: 23
Joined: 9/27/2007
Status: offline

quote:

ORIGINAL: FrankHunter

quote:

- inability to strategic move various units along the game, in spite of plenty of rail points and selection of eligible hexes;

This is happening?� I haven't received any saved games about this.

quote:

- inability, twice in the game, to activate a Bulgarian HQ, in spite of its 2 offensive points, against Rumanians and Russians.

If you have a save where I can run this I'd be happy to fix it.

Unfortunately I haven't a save of my game.

However I�m afraid that the first �flaw� was rather a misperception of the new, very fine Iron X counters loaded, properly a (small, they're wonderful) flaw of the counters themselves, as Bulgarian HQ and Infantry Corps counters look rather similar in attire and beret and thus probably I clicked twice on the false counter.

As to the inability of strategic moving units, it happened 4 or 5 times in the entire game: I could move the units in the Orders Phase, but in the Playback Phase they were mysteriously repositioned in the pristine hex. I suppose it�s the same flaw complained in the other thread "Strat movement bugs?".

(in reply to FrankHunter)

Post #: 38

RE: Historical AI? - 10/13/2007 3:06:18 PM

Graf Leinsdorf

Posts: 23
Joined: 9/27/2007
Status: offline

Playing in the role of CP, after a couple of games with the Pro-Entente bias switched on (as suggested by SMK-at-work in another thread), AI has been a tougher nut to crack indeed (playing update 1.2 beta 2), in spite of a good execution of Schlieffen plan (1 win and 1 defeat) which in all previous games did assure me an easy win.

However, contrary to the enhancement of difficulty levels in chess computer programs, I suppose that in GOA increasing the bias has only effect on odds calculation of combat. I could not remark indeed any great difference in the conduct of operations by the Entente AI, as Russians in the east still were slow in exploiting the weakness of CP (stemming from their huge effort in the West), Serbians were still too easy to dispose of (historically they were a very hard nut to crack for Austrians) and Ottoman Empire had no troubles in funneling all its resources to Germany and Austria, due to the non-existent pressure on its fronts in the Caucasus and (especially) Middle East and Mesopotamia. At least with bias on I could appreciate that Britain didn�t regularly surrender in 1916, as instead regularly happened in my games when the pro-Entente bias was switched off, having a more resilient fleet in North Atlantic.

I wonder whether, through some diffrent programming, AI could be made to somewhat replicate in East Prusia-Galicia, Serbia and Turkey (as well as in the Salonika area too!) the aggressive conduct of operations it regularly displays on the Italian Front: in every game I did, just after entering war, Italy�s onslaught alongthe Alps, in spite of the difficult terrain, has paid huge dividends and CP could not afford � as in East Prussia and Galicia � to leave the defense only to scanty and scattered troops, as this brought about a deep penetration of Italian and Allied corps in Austrian territory.

(in reply to SMK-at-work)

Post #: 39

RE: Historical AI? - 10/14/2007 3:12:52 AM

SMK-at-work

Posts: 3396
Joined: 8/28/2000
From: New Zealand
Status: offline

I'm not sure exactly what increasing the bias does....but with beta v2 winning is pretty easy due to the excessive casualties casued by artillery!! :)

(in reply to Graf Leinsdorf)

Post #: 40

RE: Historical AI? - 10/15/2007 9:11:49 AM

FrankHunter

Posts: 2111
Joined: 3/26/2004
Status: offline

Since the AI never used artillery in the way it was being used by players, it would be good to give the AI a look with the current artillery rules.

quote:

I suppose that in GOA increasing the bias has only effect on odds calculation of combat.

I'm trying to recall if the AI gains a "cheat" benefit in combat when the handicap is on. It may, I'd have to go back and look. However, off the top of my head I believe the only AI advantage is 50% more economic points. Since the AI will never spend as wisely as the human player I don't see this as being much of an advantage myself, more of an "evening up".

quote:

as Russians in the east still were slow in exploiting the weakness of CP

The Russians have a very long front and as the AI thinks defence first they will move slowly when on the offensive because they just don't do as well as a human at allocating troops and HQs so as to free up a reserve for offensive purposes.

quote:

I wonder whether, through some diffrent programming, AI could be made to somewhat replicate in East Prusia-Galicia, Serbia and Turkey (as well as in the Salonika area too!) the aggressive conduct of operations it regularly displays on the Italian Front

It is probably possible to increase aggressiveness.

(in reply to SMK-at-work)

Post #: 41

RE: Historical AI? - 10/17/2007 12:58:05 AM

Graf Leinsdorf

Posts: 23
Joined: 9/27/2007
Status: offline

quote:

ORIGINAL: FrankHunter

It is probably possible to increase aggressiveness.

Re: programming Entente AI against CP human player, I would suggest, in order to simulate Russian offensive efforts historically put in the Eastern front to relieve CP pressure against Western Allies, that a more aggressive conduct by AI Russians (and Serbians) be devised in a way that it works only for a couple of turns after each lowering of the moral level of either France or Britain (e.g. strong to good, or weak to wavering).

Moreover, for the sake of some historical flavour, I would recommend, if possible, an increase of Entente amphibious capability or other ways which encourage a kind of Gallipoli operation (or disembarkment on Turkish Mediterranean coast) by the British whenever a Russian predicament (e.g. loss of Batum, Tbilisi or both) occurs in the Caucasian region. Conversely Western Allies aggressiveness could be enhanced whenever moral level of the Russian decreases. And more links could be thought of (e.g. affecting aggressiveness of Saloniki expedition or of Mesopotamian- Middles East operations by the British). This kind of linked developments is rather easily programmed with the Events engine of Matrix game "The Operational Art War", I suppose they could are far more easily programmable with the GOA engine.

I would add here that the list of news appearing in the Strategic phase and detailing exotic events occurring on minor fronts, just after the two first games played, becomes a real annoyance. It would be better to devise random variable events each game which actually may affect the conduct of operations. As samples, there are plenty of such what-ifs events to draw on, well sketched in history books or in boardgames like XTR The Great War 1914-1918 or GMT Paths of Glory (e.g. Armenian uprising, Polish restoration, Albert King of Belgium caving in, etc.).

(in reply to FrankHunter)

Post #: 42

RE: Historical AI? - 10/26/2007 1:14:24 AM

Disintegration

Posts: 27
Joined: 10/17/2007
Status: offline

quote:

ORIGINAL: FrankHunter

The current AI, which doesn't use scripting, actually plays better than the previous one although it has a tendency towards cautiousness.

That's the biggest problem I've seen by far. Granted, I've only played 5 games, but in the three I played as the CP, the AI just defended on both major fronts - I was still attacking France in late 1915 and the Russians had yet to move a single unit into German territory or even stack many units on the front line. Likewise, when I attacked Russia first, the sole TE response was a few tepid bombardments from the French border. The Brits didn't even show up anywhere except Basra until the Italian DoW.

Playing the TE it was even worse. I've done that twice and both times the Germans stacked a huge army on the Western Front and then did absolutely nothing with it, again save for a few bombardments - and also did nothing as the Russians rolled to Berlin completely unopposed.

Honestly, I'd rather the AI cheat a little than be so ineffective. Would it be possible to let it look at the human player's setup just for the purpose of deciding on a simple strategy - i.e., whether to take a generally offensive or defensive posture on any given front - and then forget those dispositions and go back to FOW when giving orders? Then the AI could afford to be a lot more aggressive without reaching Plan XVII levels of bad judgment.

The game itself is fantastic - thanks so much for your work on it!

(Would be nice to have a more detailed manual, though...)

(in reply to FrankHunter)

Post #: 43

RE: Historical AI? - 10/26/2007 5:24:00 AM

Joel Rauber

Posts: 195
Joined: 10/4/2000
From: Brookings, SD, USA
Status: offline

I'm currently fighting a CP AI, and the Germans are being relatively agressive in the east. While trying to hold a line in the west. I.e. its pursuing an East first strategy. Its not doing to badly either.

_____________________________

Any relationship between what I say and reality is purely coincidental.

Joel Rauber

(in reply to Disintegration)

Post #: 44

RE: Historical AI? - 10/27/2007 4:10:51 PM

Graf Leinsdorf

Posts: 23
Joined: 9/27/2007
Status: offline

quote:

That's the biggest problem I've seen by far. Granted, I've only played 5 games, but in the three I played as the CP, the AI just defended on both major fronts - I was still attacking France in late 1915 and the Russians had yet to move a single unit into German territory or even stack many units on the front line. Likewise, when I attacked Russia first, the sole TE response was a few tepid bombardments from the French border. The Brits didn't even show up anywhere except Basra until the Italian DoW.

Finally someone has got to the heart of the matter.
I don't see any reason to play a historical wargame which is neither really historical nor really challenging for the human player, whatever merits a clever design like this may have (and provided of course that one has no time to play PBEM).
Thus at the time I've placed Guns of August on the shelf and I will not take and play it again before the first update is released which seriously addresses the AI and makes it some more challenging, at least for average experienced players.

(in reply to Disintegration)

Post #: 45

RE: Historical AI? - 10/27/2007 9:46:35 PM

Shawkhan

Posts: 125
Joined: 3/1/2006
Status: offline

...I would love to have an option in the game where the human player could choose which historical war plan the AI would use. There was an old wargame called '1914' where a player had the choice of several different OOBs based on the historical war plans of both sides in WWI. This enhanced replayability enormously, making the game so enjoyable I still pull out one of my copies of it now and again to try different combinations.
...For those of us wanting AI aggression the chance to fight the allies using plan17 and the Tannenburg opening for the Russians should fit the bill.

(in reply to Graf Leinsdorf)

Post #: 46