Recommend
13 
 Thumb up
 Hide
4 Posts

Race for the Galaxy» Forums » General

Subject: Never Give Up!!! rss

Your Tags: Add tags
Popular Tags: [View All]
rain
Canada
flag msg tools
badge
Avatar
mbmbmbmbmb
Hi all,

Recently my favourite thing when playing Keldon's AI is losing till I win. I have been taking advantage of the handy Undo Round feature in borgemik's great patch. When I lose a game I thought I could have won, I undo a round and see if I could win with different choices. I do this recursively until I win. This may sound a little like cheating, but the point here is not really the victory. Instead, its about learning what moves would really have been better. Sometimes I also do this with games I won, but just to see if I could have achieved a better final score.

So far I've learned a couple of things:

1. Sometimes I worry too much about tempo. I play crappy cards that end up forcing the end of the game when instead I should worry more about card quality. This is especially true in BoW, where individual cards can be worth so many points it doesn't matter if you're up three cards in tempo.

2. There's a lot of randomness in Race. Veteran players lose track of this because play skill does have a very high correlation to performance. So when considering a whole game, or a series of games, a skillful player will do much better than a novice. No surprises there. However, when it comes down to individual choices, sometimes what seems like the objectively best move may end up being less effective than another choice. The way I have seen this is in that a better play may change which cards you draw off the top of the deck, which may end up with you getting a worse selection and/or your opponent(s) getting a better one. Conclusion? C'est la vie, I guess -- not to worry too much about small differences in plays, except in the situations where all information is static (like in the last turn usually).

I would suggest trying out this game if you'd like a challenge:
Seed 1307405721
3p BoW, TO, G
Starting Hand: (UMF, GS) (NS, SC, RCM, TW, RO, RL)
That one took me many undos before getting it (61-48-37)
13 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Runcible Spoon
United States
flag msg tools
badge
Avatar
mbmbmbmbmb
I don't think it's cheating...it's more like a training scenario. Think of it like baseball hitters in a batting cage, it's just a way to practice fundamentals.

But I like your philosophy behind it. I have never actually played Keldons AI, all of my games are face to face, but now I just might try it out and do exactly what you suggest. It really does seem like a good way to learn to appreciate the game even more deeply is to explore each branch in the decision tree and see where it leads.

...yes I think that is very clever idea indeed...thanks for the suggestion... have a thumbsup
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
J P
Germany
flag msg tools
mbmb
Don't forget that when you undo to make a choice differently, the result will not necessarily reflect the strategic quality of that choice.

There are many reasons for this:
It might be a bad idea for you to call settle, but you try anyway and draw a card that fits really well into your tableau off the settle bonus. The strategically superior choice (inferior by design of this example) to call produce would not net you this card. The fact that you don't have this information when playing regularly (with no undo) still makes calling produce the right choice, even though calling settle would have a better outcome (due to information you didn't have access too).

A simple analogy in RL:
Suppose you faced with the incredibly stupid choice to take a gun and shoot your foot. Why would you? It's stupid given any regular circumstance.
Now then you are given a rewind button for your life and you say "what the heck, I'm gonna give it a shot (literally)". You shoot and miss and break the floor instead of your foot. Behold, you discover a secret cache with a lot of gold. You are glad you tried to shoot your foot because you found the gold and learned that shooting feet is good.
Obviously this is a false lesson as based on the information you had (without the rewind button) shooting your foot was an incredibly bad idea.

The whole thing becomes even more subtle if you alter the draws of the AI as you don't see what's going on. The right choice for you might net the AI a good card it really needed to take off which could make it seems like a bad choice.


I'm not saying don't use the rewind button to learn, I'm saying use it with caution.

9 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
rain
Canada
flag msg tools
badge
Avatar
mbmbmbmbmb
Good points J P.

Perhaps there is another lesson in here: the best play does not guarantee the best outcome. I'm not suggesting not trying to make the best play you can. But I think it is reassuring to know that its okay to take a chance and do something unusual if it seems good, instead of going with the standard strong decision, since you can't ever really be sure of the outcome. I've noticed playing a little offbeat to be very effective in 2pa, for example, although this is less effective in multiplayer.

I also appreciate the implication that you don't need to beat yourself up too much for making a suboptimal choice, since the best decision may actually not have had a great result. I've noticed this using undo that sometimes better (local) decisions result in worse games.
 
 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.