$18.00
GeekGold Bonus for All Supporters: 127.79

7,568 Supporters

$15 min for supporter badge & GeekGold bonus
47.7% of Goal | left

Support:

Recommend
 
 Thumb up
 Hide
8 Posts

Memoir '44» Forums » General

Subject: The choice between M44 Online's AI bots rss

Your Tags: Add tags
Popular Tags: [View All]
Ragnar K
msg tools
mb
Initially I posted this in the video game forum, but as it appears to be rather inactive place, I repost the question here as well.

There are two AI "robots" which can be chosen for an opponent while playing solo game. Memoir '44 Online manual v1.1 says that "Hermann is an experimental robot we use to code alternative strategies". And then there's added that "Note: as of version 1.0, the two robots have identical strategies. This might change at anytime thereafter".

I have played Pegasus Bridge scenario as Allies more than 30 times now. I started with Johnny as an AI opponent and then switched to Hermann in some point. For some reason Hermann seemed to be more difficult, causing more problems even when I had good cards. (And this after I had gained already some experience while playing against Johnny, as mentioned.) Curiously, when I looked at my stats, there was no apparent difference between the two (I have lost three times to both and the win ratio is almost even for both AI versions). So it's hard to give a verdict.

I know that it's recommended to play with human opponents instead (as the AI makes indeed some obvious mistakes occasionally), but still...

What do you think - is Hermann currently slightly better bot than Johnny? Thanks.

Poll
Which of the two available AI robots is a tougher opponent?
Johnny
Hermann
They are practically equal
      12 answers
Poll created by ragnarkuusik


Note: Of course, from the fully practical point of view, one could say, that if someone (like I) finds that Hermann is at least not worse than Johnny and Hermann is declared to be under development, then there's no reason not to choose always the former AI bot, forget the latter - and the dilemma would disappear.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dave C.
United States
Maine
flag msg tools
mbmbmbmbmb
I almost always play against Herman (200+ plays) and find two things to be relatively consistent.

1. He ALWAYS gets great starting cards that I NEVER get to start: ie; ALL units in Center - 2,2,2 - Direct From HQ - Infantry Assault, etc.

2. He rarely knows how to defend. He is all about attack, attack, attack, which makes for predictable and somewhat boring gameplay eventually.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ragnar K
msg tools
mb
By the way, on the Days of Wonder site, both bots have their own user information page like any other player. Johnny has slightly higher score than Hermann, but I think it can be attributed to the fact that Johnny is the default AI and his performance gets regularly boosted by games with newcomers who are still learning.

I hope and rather believe that dice rolls and card deals are both fair.

Both bots are compulsive attackers, at least that is sure.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Jeroen Nieuwenhuis
Netherlands
Epe
flag msg tools
mbmbmbmbmb
When they programmed Memoir'44 online Days of Wonder wanted to make two different AI-players. But it is my understanding they never got round to it. So Hermann and Johnny are not practically equal they are equal Johnny = Hermann.

In reaction to Dave.
1. In my experience I have gotten those cards at the start of games. But I rarely play them at the start because it is almost always better to position your troops before playing those big cards.
Johnny/Hermann is programmed (roughly) to get the maximum of dice every turn, so he will consistently play the best cards first. This is why it is such a bother when you lose connection for a free seconds. Johnny/Hermann will have spent your best card by then.

2. As stated above, yes, Johnny/Hermann is all about attack. Because he maximizes the dice he will leave cover to get that extra die. And also he usually plays with his front troops only until you have killed them and then he has to start moving fresh troops from the baseline. (Which a human player probably would have done earlier to support the front troops.)


So if you keep this in mind the next you play Johnny/Hermann. You can let him come to you and spend his best cards while you are in cover. And then retaliate. Mind you, the short-sighted strategy of Johnny/Hermann can at times be very effective. In desert and Japanese scenarios (with the AI playing the Japanese) it is very daunting.

Good luck everybody in defeating Johnny/Hermaan and human opponents!

And don't forget to have fun!
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ragnar K
msg tools
mb
Thanks, Dave and Jeroen. It's a very cool game!

If the two AI-players are indeed still the same, I wonder why they didn't design Hermann just to a simple defensive version of Johnny.

Keep the base algorithm, but make the AI appreciate/value the advantages of Terrain and Obstacles - at least of those positions which are held by the AI-player in the scenario set-up. Take into account unit and firepower concentrations. Protect on-the-map objectives (Victory Medal hexes). Almost by the definition, a defensive AI should show more static behavior (and therefore, as a bonus, can afford to be less creative). A few adjusting rules of thumb might make already a difference in results I think.

I'm a beginner in this game, but many scenarios appear to have one side which is expected to move forward, be more aggressive, take more risks. (Of course, both sides can use offensive tactics - especially if the cards are good and lucky dice rolls have given an initiative in some sector(s).) If there were available two types of AI - attacking and defensive version -, a player who desires more challenging solo game could choose deliberately a robot whose mindset would suit better particular side (would offer more promising opposition).

Based on those two AI versions, there could be finally third (called, say, Albert) which would use some basic heuristics in order to switch between attacking and defensive mindset depending on developments on board and cards held.

It just seems like a missed opportunity (although it probably has some good reasons for being missed).

All the best!
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Timothy Young
United Kingdom
Keighley
West Yorkshire
flag msg tools
mbmbmbmbmb
I imagine it is harder to program defence into a bot than attack. I have not played the M44 bots, but when playing strategic computer games, the opponents are often readily able to attack, but poor at defence.

Attack requires a friendly unit to find the nearest enemy unit, move close to it and open fire. With a bonus for choosing enemy units that have less troops and are not protected by terrain.

Defence requires a friendly unit to stay put much of the time, seek defensive cover and watch the line of assault of the enemy forces, to ensure you have a good line of fire when they get close enough. Much harder to define in an AI program.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Jeroen Nieuwenhuis
Netherlands
Epe
flag msg tools
mbmbmbmbmb
Since DoW stopped developing the online game, our speculations might be useless. (But you never know wha the future brings! )

I feel Tim has it right. Programming defense is very hard. Try some games where there are some key objective medals and laugh at the antics of the AI.

It's why I play the AI only to get to know the lay of the land of a new scenario. Knowing the human opponent will almost always be the greater challenge.

Keep fighting gentlemen and hopefully we will meet on a battlefield somewhere!
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ragnar K
msg tools
mb
Right. I have somehow missed that the M44 is from 2004 and the M44OL from 2011 - old games according to nowadays standards. And I stumbled on Yann's sad announcement in the DoW forum too. What a pity.
 
 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.