AI testers

Post by **IainMcNeil** » Mon Oct 27, 2008 12:00 pm

We're looking for a few people who had strong views on the AI to help out with Beta testing of the patch. Anyone interested?

If so, please post here.

Post by **IainMcNeil** » Tue Oct 28, 2008 9:47 am

Come on - you guys were complaining about the AI but you don't want to try out a new improved version and help us make it better?

If you think the AI is poor - we need you to help make it better!

joerock22 · Post by **joerock22** » Tue Oct 28, 2008 2:53 pm

I'm willing to beta-test, but I've never done it before so you would need to tell me exactly what to do and how to report my findings.

Post by **IainMcNeil** » Tue Oct 28, 2008 3:45 pm

All we're looking for people to do is grab the beta patch, play the game and tell us if the AI is better than before and if not, what it's doing wrong. JOhan thinks he's sorted out the worst issues but its hard to be sure as different people play different ways it shows up different issues in the AI.

firepowerjohan · Post by **firepowerjohan** » Tue Oct 28, 2008 4:58 pm

We should add that you need the Slitherine version of the game since the patch is only Slitherine beta.

joerock22 · Post by **joerock22** » Tue Oct 28, 2008 5:32 pm

Sorry if this is a dumb question, but where do I get the beta patch?

Post by **IainMcNeil** » Tue Oct 28, 2008 6:05 pm

Once we've got a few people signed up we'll get it out to you. We just need a few more volunteers

Davedj · Post by **Davedj** » Wed Oct 29, 2008 12:16 am

I like the game and would love to help make it better. I tried to get in the original beta test and wasn't picked, but I still want to help. thanks Dave

Davedj · Post by **Davedj** » Wed Oct 29, 2008 12:16 am

I like the game and would love to help make it better. I tried to get in the original beta test and wasn't picked, but I still want to help. thanks Dave

Post by **IainMcNeil** » Wed Oct 29, 2008 9:58 am

You can grab the beta patch here - http://www.slitherine.co.uk/Files/cnaw_patch1.01.exe

Please send us your feedback. We're especially interested in the 1805 scenario and the French AI.

joerock22 · Post by **joerock22** » Thu Oct 30, 2008 3:14 am

OK, here it is: my first report w/ the beta patch.
Scenario: 1805, playing as the Coalition
- Large French Advantage
- Fog of War off (so I report on what the AI is up to)

The French immediately built many new units, all infantry (line & light) and light cavalry. Wiped out half the Austrian army before I could move. The French move sluggishly toward Vienna, squandering the early advantage and allowing me to bring in reinforcements. Vienna doesn't fall until June 18, 1806, despite my willingness to give it up. I spent my PP on Russian tech. for later in the war, and pulled the Russian units away from Vienna when the French brought their numerical superiority to bear.
I occupied the Eastern half of Austria after its surrender, but the French turned their attention northwards. While this is a perfectly reasonable course of action, they putzed around at the border until September, 1806, when Prussia and Saxony automatically joined the war. The French blitzed their massive army toward Berlin, and I evacuated, setting up a defensive line across the river with Russian and British reinforcements.
After a short counterattack at the beginning of 1807, things have settled into a stalement in Prussia. My Russian tech. finally kicked in. I stopped at March 1807, deciding to report my findings before going any further.

The Good:
1. The AI knows how to pick off weak or exposed units, and can coordinate attacks well.
2. AI researched technology well. By 1807, French infantry tech. was 3 and light cavalry tech. was 7. French troops are consistently superior to anything I can throw at them, and there's not much I can do about it.
3. AI continually used wagon moves to bring reinforcements to the front, halting my counterattack in Prussia before I could reach Berlin.

The Bad:
1. AI kept far too many units at home. In 1807, 13 infantry units were stationed immediately around Paris, and 6 more were stationed just East of Amsterdam. These two put together outnumber the French troops on the Prussian front. The armies were there, but smaller, from the start, but the AI occasionally adds troops to each group as the game progresses.
2. AI built no heavy cavalry and no new artillery in two years of combat. 'Nuff said.
3. AI didn't attempt to build a navy, even though they constantly had a surplus of PP (Large French Advantage). While this may be historical, the British are only getting 50 PP a turn when their ships are in port, so if the French were more aggressive at sea they may have been able to overwhelm me. This would also have prevented me from shipping Wellington and his Brits off to rescue the Prussians.
4. [Related to 1-3] AI did not manage manpower well. In 1807, manpower was at 266, well into the yellow. This is largely a product of the huge reserve they have back in France. The AI needs to do a better job with this, because as I cause casualties at the front, they're being replaced by crappy troops. This shouldn't be happening on large French advantage.
5. AI seemed reluctant to invade Saxony; it sat at the border for a few turns doing nothing. The Saxons were cowering in fear, but the French refused to declare war...such gentlemen!

The Ugly:
1. The AI doesn't seem to understand that all it has to do to finish off Austria is capture Vienna. I found myself shaking my head; the AI seemed bound and determined to capture that city only after every other Austrian and Russian unit in the area was destroyed. The AI could have saved itself months of combat if it had focused its offensive on Vienna instead of the units on the flanks.

Is this an improvement? Yes. The first time I played this scenario (sans patch), I was able to stop the AI from even capturing Vienna. But this patch clearly needs more work. Thanks for taking the time to work on the AI for us, and if there's anything else I can do let me know. I'll report in again the next time I play.

Post by **IainMcNeil** » Thu Oct 30, 2008 8:37 am

Thanks for the feedback. I'm sure this will help Johan improve thigns further.

firepowerjohan · Post by **firepowerjohan** » Thu Oct 30, 2008 12:28 pm

i) Saxony thing, good point and this is one of small tweaks that eventually could make it into the new tree based formulas

ii) The AI try to keep 5 times as much infantry as cavalry. With the 20 extra units, if some of them were cavalry then it is unlikely that the AI decided to buy many more. If you kill of a lot of enemy cavalry they should buy new ones, can you confirm this? AI tries to keep 150 horses in store for repairs before buying new cav also, so their existing units do not run short of replacements.

iii) France focus on privateers while GB on Frigates and Ships. You might succeed grabbing naval superiority from GB in 1.00 but the GB naval AI is improved now. You can try playing as France instead and see if it works. It could work, but it could also cost more than what it is worth. Therefore, the French AI will use Privateers trying to loot instead of battling for Naval superiority. Please proof me wrong and I will adjust the AI

iv) One thing need to be noted though, is that the passivity problems are fixed but this new "Paris Defence" one instead tied to the difficulty level of France 20 extra units to start with (so is NOT the same problem as in the 1.00 version). AI uses better formulas now for invading Prussia and Russia at the right time. The new tree based variations added to the general thinking based formulas are in its infancy and are expected to be expanded and improved for future patches also.

But at "Big Advantage" the French start with 20 extra ground units in 1805 so the passivity is conerned how the AI uses its extra units. The unit density is not intended to go that high this early in the game, but eventually once Prussia join those units come into play since AI then has a good objective.

There are thresholds for when the AI decide to build up a 2nd group because sending 100% of your forces into one attack group will mean sneaky counters are expected. For example, if you invade Russia and Austria+Prussia decide to revolt you need to have more than one group to prevent these revolters to eat up all your land behind your back. This threshold is reached much sooner now if France start with 20 EXTRA units, so temporarily (and since the AI is not built from having France with twice as many units in 1805-1807 as normal) some of these units are then temporarily placed near Paris until more offensive objectives are created (=when Prussia joins).

Have you tried the 1805 on small or moderate advantage also?

Thanks for the feedback

joerock22 · Post by **joerock22** » Thu Oct 30, 2008 2:55 pm

ii) yes, when I killed a few cavalry units in Austria the French AI built a lot more. So much, in fact, that they ran out of horses. I would still really like to see them recruit heavy cavalry and especially artillery.

iii) I understand why the French didn't go for naval superiority, but they didn't buy any ships, even privateers. The only privateer on the map was mine, and he had free reign to loot French merchants at leisure.

iv) so you're saying I should keep playing and see if the AI deploys these reserve troops?

No, I haven't played on small or moderate yet, though it's on my 'to do' list. Not sure when I'll get to it. Wish I had more time to play.

firepowerjohan · Post by **firepowerjohan** » Thu Oct 30, 2008 3:06 pm

I think I got it. When you play does the AI target Prussia and keep a force near Austria as well? It could be that the threshold is not met so that AI instead target Prussia only and keep rest at Paris, it should really be targetting both Prussia and Austria (despite that Austria is conquered).

Especially due to the massive amount of extra units, that affect the influence over areas and affect the final result of targetting countries. I tested the max difficulty level myself and the French AI had 7 extra units (i.e units that are not stationed on a resource) around Paris in 1806.

Note that in this test I did not move my Coalition forces I just let the AI invade the land

Does this screenshots differ from your test result much when it comes to French offence/defence distribution of forces ?

joerock22 · Post by **joerock22** » Thu Oct 30, 2008 3:41 pm

No, that's pretty much exactly what I had. Right down to that French heavy cavalry unit occupying that Austrian horse farm. I left only garrison units in Eastern Austria and moved all my spare troops to Prussia. I think once I stopped them in Prussia they really started building more troops, then their manpower got into the yellow. The problem was that most of those new recruits sat at home and didn't move to the front.
And you see how untenable the Prussian position around Berlin is. The AI can definately attack en masse fairly well. And they did sneak a cavalry unit behind my lines and grab the Prussian horse farm; they just didn't follow up so I was able to retake it.

firepowerjohan · Post by **firepowerjohan** » Thu Oct 30, 2008 3:54 pm

But in your game, did the French have more units than this? Because from the current AI, once the unit number reaches a certain level it starts to distribute to its 2nd and 3rd objectives.

So if threshold is at 18 units, primary objective is Berlin and secondary is Paris then that means

Units on 1st objective, units on 2nd objective
18:0
18:1
19:1
19:2
20:2

so it starts to build up the 2nd objective but the units will always be more on the primary objective of course. Units on defensive duty (like in Cities or other resources) do not count towards the objective numbers. A high threshold means the AI will be more aggressive since it will attack with larger % of forces into their primary (offensive) objective but also more vulnerable to counters. In the screenshot I posted in my last reply can be seen that AI has roughly 25 units aimed at Berlin and rougly 7 free units (not in cities) around Paris so it works according to the 18 unit threshold seemingly. 18+7=25 (Berlin) 0+7=7 (Paris)

The changes available are of course

1) to assign the objectives differently, Austria as secondary objective so that Paris becomes 3rd objective means the units will be around Vienna instead of Paris close for reinforcing losses or defending counters.

2) Vary the thresholds (according to any factors in game. Currently for the threshold for when the secondary and tertiary objectives start to kick in varies with if AI is behind in VP or if the game has few turns left. When game is near end or in short scenarios AI is more aggressive and direct. If it is currently winning or if it is a long time left it is more defensive to avoid counters)

The AI primary threshold is currently set at 18 units (when behind in VP and in a long scenario) so it means once 18 units are assigned for Berlin (in this case) the secondary and primary starts to get one unit each (=50%) of all additional units. Sometimes there are even 3 objectives and another threshold for when the tertiary objective starts to get used but that is not too relevant so do not need to post all the maths on it here

A too high threshold will mean counters and exploitive moves are easier to pull off (for example GB landing with units near Paris and grabbing it while all French units are busy invading Russia) or if set too low then the AI will be too defensive.

What would you like to see changed in the AI from your last testing?

Bern · Post by **Bern** » Thu Oct 30, 2008 5:43 pm

Didn't realise the beta patch was out. The real world is cruel just now. I don't have much time for gaming at present, but I did d/l the beta and have played the first ten turns of 1805 as Coalition with mod. French advantage.

Not much to go on I know, but some thoughts. I was disappointed that we still have the same naval scenario. French fleet is destroyed in two turns. The Batavian fleet did do some damage in the channel with some v. fortunate die rolls but then, for some reason sailed into the Bay of Biscay where it ran into units of the Brit fleet returning for refit and was also promptly sunk. I suppose the idea of keeping the navies apart would require too much programming, but it is a pity IMO. As it is the naval war is effectively over.

The land campaign has gone very much the way of version 1. The initial Austrian forces, with the support of newly created units and two initial Russian units held off the French assaults with little difficulty until Russian reinforcements were able to arrive. It seems that the AI gets easily distracted by any perceived threat on its flanks, rather than concentrating on Vienna and its immediate defence.

What could have made all the difference is the use of the units remaining in France and Belgium. I cant see why there is a need for six units around Paris at this stage of the game. The only threat would be from a GB invasion. I suppose it could be possible, although very 'gamey', but I wouldn't try that so early. A French naval threat would stop it altogether

. I did notice that there was an effort by the AI to bring additional units to the front but this is far too late to make a difference. If this movement had started at turn 1, the situation in Austria could have been very different.

As things stand, the French army is bleeding to death around Vienna and Nappy himself is in serious danger.

I probably wont have time for any more, but it does seem clear that additional French units supporting those already in Austria at an early stage would provide a stern test for even the best Coalition General.

Bern

joerock22 · Post by **joerock22** » Thu Oct 30, 2008 6:07 pm

firepowerjohan wrote:But in your game, did the French have more units than this? Because from the current AI, once the unit number reaches a certain level it starts to distribute to its 2nd and 3rd objectives.

What would you like to see changed in the AI from your last testing?

No, the AI had between 18 and 22 units attacking Berlin. I'm wondering what the AI does after it achieves it's primary objective (Berlin). Does Paris become the primary objective? I would like to see Moscow become the next primary objective after capturing Berlin; that would cause the AI to aggressively attack my armies in Prussia. The AI became complacent after it captured Berlin
I think it's smart that the AI wants to keep some units at home, but six or so should be the maximum non-defense units. Thirteen is too many. I don't know how you would fix this, but I think it might be a good idea to make Paris the 3rd objective. Maybe that would make those extra units attack me in Austria or invade Spain, or something like that.

firepowerjohan · Post by **firepowerjohan** » Thu Oct 30, 2008 6:09 pm

@Bern:
Thanks, then it is probably the thresholds that need to increase in certain conditions (certain years) so that the primary objective (in this last case Vienna) can collect more units without distributing forces to the rear. When game is into 1807+ year then that threshold can be brought back to "normal" again since by then France would have several fronts (Austria, Prussia, Russia) to take care of and being sure that a silly "all-in" against one of them is not done on the cost of losing all the rest.

Thanks for the feedback

@joerock22
As mentioned above, varying the threshold correctly can make a dramatic impact so early in the war the threshold might be set much higher. Once Berlin is captured, the primary will be Konigsberg since that is the second capital of Prussia and once both Austria and Prussia is conquered France will go for the big bear Russia. Sound fair?

Another thing which comes to mind is, even with more units it is very difficult to use this in a narrow front so it is understandable and not too foreign to belive that a clever human player can stop France in Austria already. Another tweak to the game would be to increase the technology benefits so that the French in 1805 has even more superior technology and more punching power.

Corps d Arme' for example (excuse my spelling) give the French infantry +8% in 1805 but perhaps if that tech gave 13% could make some difference. Would it be better than just ramping up the number of units?

AI testers

AI testers

First Report