GeekGold Bonus for All Supporters at year's end: 1000!
8,066 Supporters
$15 min for supporter badge & GeekGold bonus
21 Days Left

Support:

Recommend
1 
 Thumb up
 Hide
24 Posts

Race for the Galaxy: Rebel vs Imperium» Forums » Strategy

Subject: Develop Wins Rebel versus Imperium in my group rss

Your Tags: Add tags
Popular Tags: [View All]
Dave J McWeasely
United States
Louisville
Kentucky
flag msg tools
designer
Avatar
mbmbmb
I'm turning the page to Brink of War, and looking back on my games on Flex and in person, one trend is enormously clear:

At our skill level, the Develop strategy is dominating. This is my go-to strategy for bludgoning my regular opponents. It is also the strategy I most fear.

The pool of players on flex is certainly inbred - there's not really a regular player there who has my number, but there were two tiers above me on Genie.

The Gathering Storm added zero cards which were useful to the develop strategy. RvI added a ton:

Banksters
Crash R&D x2
Galactic Developers
Pan Galactic Research

With honorable mentions to Advertisers and Salon, both of which make excellent development targets, and draft the Develop strategy's main weakness: the consume phase.

Now there's a full ecosystem of cards allowing a self-contained develop strategy to thrive. Early Bank/Works/GalDevs give you the cardflow you need. Mid-game Crash R&D lets you build the first 6-dev without blowing your hand. And late game there are decent numbers of 6s you can build which continue to reinforce the strategy.

There are 3 weaknesses I'm aware of:
1) Your opponent will also execute a Develop strategy, and do it better.
2) Your opponent will get lucky with Terraforming Robots and a Small Military, and win the game with some combination of Goals, Terraforming Guild, and maybe Improved Logistics. You need the Terraforming Robots for cardflow - calling trade costs you a settle action, and it often helps the develop player.
3) inadequate cardflow leads to stalls

Not on that list: produce-consume strategies. That's because they need to amass cash, call settle 2-3 times to get a decent crank arm length, then turn the crank 2-4 times. There just isn't enough time for that to happen against a develop spammer.

Also, I've noticed a propensity for my opponents to call Explore to combat this strategy. That's like calling Settle against a Produce-Consume player: a good way to lose! That's because the Develop strategy fails mainly due to inadequate cardflow. Showing me just 2 cards more seems to give me a decent development over half the time. Each explore call is like handing the develop player about 4vp, once you factor in 6-devs that will score of the resulting dev's power, and cardflow generated by the build, and so forth.

The new "Search" action is a good way to not help the Develop player, but sadly it isn't in RvI.

We're kind of stumped.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rob Neuhaus
United States
New York
NY
flag msg tools
Avatar
mbmbmbmbmb
A few thoughts. For what it's worth, I think I am probably significantly better at TGS than I am at RVi.

The chance of getting the 4+ dev goal or the first 6 dev are reduced. Both of these work against the dev strategy. The first to 8 size tableau offsets this a little bit, but I think the overall, goals tended to shift away from dev heavy strats a bit. You have to expect that there would be some new dev bonus cards when the deck expands by ~1/5 , or else it would get weaker. But I could buy the argument that the dev strategy has gotten better relative to TGS, mostly from the galactic bankers and galactic developers. Certainly blue prod gets weaker from TGS -> RVI which gives it some slack to pick up. On the other hand, the addition of of Rebel Alliance, Imperium Seat, Hidden Fortress, and the 8 alien and 9 rebel, as well as the most Rebels goal probably gave military another boost from TGS. Further, explore and combine helps military strategies disproportionately.

I'd say if you want to tilt the meta game away from dev heavy strategies, bias yourself more towards military strategies. Be more willing to hoard the big military worlds. Be more willing to make somewhat speculative settle calls, where you need an opponents dev call to place say, a big alien mil world, instead of a small rare military world if the dev doesn't occur.

It's possible that you've simply gotten better. It's also possible that the lack of ratings means you don't see that there is a tier above you on flex.
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dave J McWeasely
United States
Louisville
Kentucky
flag msg tools
designer
Avatar
mbmbmb

Quote:
It's possible that you've simply gotten better.

In the base game, was the dev strategy actually viable, and I missed it the whole time? I remember concluding that it was not workable, but I also remember my initial move being E+1+1 / Develop to get out GalFed. That didn't work at all. But knowing what I know now, (e.g. little devs lead the way, keep a big hand, cardflow, cardflow, cardflow) I wonder if it would be workable. My gut tells me that there are only 4 cardflow cards (two duplicates no less), so it would be pretty rare.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rob Neuhaus
United States
New York
NY
flag msg tools
Avatar
mbmbmbmbmb
The dev strategy is viable in TGS. From base to TGS, all that happened to it was that it got quite a bit worse via dilution (i mean, which cards help the dev strat that are new in TGS?), and possibly a bit better via goals. So yeah, in base game, the dev strategy is very viable. In fact, I'd say that my general approach to the base game is wish for GalFed, and content myself with produce consume if I don't get it.

My group used call GalFed "gg" until TGS came out, when TG become the "new gg" and GalFed the "old gg."
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Eric Jome
United States
Franklin
Wisconsin
flag msg tools
designer
badge
Avatar
mb
One of the things that keeps the game still interesting to me is that none of the strategies in the game seem dominant in repeated play - I guess I'd call that "balance" and the metastrategy of the game is adapting to what you're given in your hand.

Are Developments good? Sure. The main weakness of Development as a strategy is that it includes no card drawing mechanism. Very few developments will get you any number of meaningful new cards in hand, certainly not enough to pay for expensive developments over and over... do you find yourself really grinning when you draw Interstellar Bank in your opening hand? That's small potatoes compared to a Trade...

Which makes me think you are winning with Develop a lot because others are playing Settle for you. One of the first hard lessons I think lots of players have to learn is that Settle and Develop are "hardly ever" cards, not "whenever I think I need it" cards. If people are playing Settle freely, it's easy to exploit them to your advantage.

So, if Developments are so dominant, how are you getting the cards to do it?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dave J McWeasely
United States
Louisville
Kentucky
flag msg tools
designer
Avatar
mbmbmb
cosine wrote:
So, if Developments are so dominant, how are you getting the cards to do it?


Have 2 of: Bank,GalDevs,Banksters. Call Develop for cards. Like twice a turn, even if I have nothing to develop. Then get draw one of the other meta-develop cards. The advantage given to other players by calling a build phase like develop diminishes under the pile of spam. Meanwhile, R&D becomes much more powerful if you're getting compounded spam bonuses, and the other guy isn't really going to get to develop anything extra because you're putting it down. It increases your abilty to roll out those 6s prior to actually having those 6s.


2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Eric Jome
United States
Franklin
Wisconsin
flag msg tools
designer
badge
Avatar
mb
MrWeasely wrote:
Have 2 of: Bank,GalDevs,Banksters.


That's a pretty small window of opportunity. You can't be maintaining a very high win percentage with this strategy.

Quote:
Call Develop for cards. Like twice a turn,


Ah... here's yer problem. You are playing 2 player advanced. 2 player advanced is extremely easy to play... you've got more than double the amount of control over how the game will go.

Play more 3 player. I don't think I can think of anything more likely to challenge more players than playing lots of 3 player games.

And frankly, I don't know how 2 player advanced became the dominant form of this game...
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rob Neuhaus
United States
New York
NY
flag msg tools
Avatar
mbmbmbmbmb
cosine wrote:
MrWeasely wrote:
Have 2 of: Bank,GalDevs,Banksters.


That's a pretty small window of opportunity. You can't be maintaining a very high win percentage with this strategy.

Quote:
Call Develop for cards. Like twice a turn,


Ah... here's yer problem. You are playing 2 player advanced. 2 player advanced is extremely easy to play... you've got more than double the amount of control over how the game will go.

Play more 3 player. I don't think I can think of anything more likely to challenge more players than playing lots of 3 player games.

And frankly, I don't know how 2 player advanced became the dominant form of this game...


I think 2pa is so popular (among people who post here), because the chance of a >2 player game online going super slow is very high.

On the other hand, I am not sure how you are measuring ease to play. I am not sure that the game is less strategically deep in 2 player than in 3+. Certainly somethings are more important in multiplayer, like accurate phase prediction, but others are less important, like really stabbing at an opponents weakness.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
United States
SoCal
flag msg tools
badge
Avatar
mbmbmbmbmb
cosine wrote:
MrWeasely wrote:
Have 2 of: Bank,GalDevs,Banksters.



And frankly, I don't know how 2 player advanced became the dominant form of this game...
AFAIK, one ain't more dominant than the other. At least on BGG. With my groups, 3p+ games see about 4 plays for every 1 play of 2pa. I know some here on BGG can only find one other person to play RftG with, so go figure with those situations.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
JW
Australia
Doncaster
Victoria
flag msg tools
Avatar
mbmbmbmbmb
I agree 100% with MrWeasely. There is nothing he is saying here that I disagree with. Of course, I mainly play 2padv. I would say that the develop strategy is less dominant in 3p or more, but still quite strong nevertheless.

cosine wrote:
Are Developments good? Sure. The main weakness of Development as a strategy is that it includes no card drawing mechanism. Very few developments will get you any number of meaningful new cards in hand, certainly not enough to pay for expensive developments over and over... do you find yourself really grinning when you draw Interstellar Bank in your opening hand? That's small potatoes compared to a Trade...

So, if Developments are so dominant, how are you getting the cards to do it?

I'm speaking for 2padv only, and I understand your comment was meant to be for 3p or more. However, I want to address this by saying that the develop strategy doesn't happen in isolation. You can Settle trade a windfall and get a full hand before double developing your develop discounts. You don't even need to start doing your develop until the midgame, where you just switch from whatever you are doing and start developing.

UlyZed wrote:
My hunch is that it isn't so much that a heavy develop-strategy is dominant, but that inexperienced players have the most trouble identifying and defending against it.

I disagree. I don't think there is anything you can do against someone who settle trades first turn then plonks down Interstellar Bank and Public Works then turn after. Or Galactic Developers plonking down Interstellar Bank on an Explore/Dev first turn. This might happen once every four to five games or so (anecdotal evidence only), and it is definitely way stronger and much more likely to happen than ELC/ATS, which got its own thread.

And you don't even have to spam double develop after plonking down 2 of those cheap developments. You can just do other things and block the develop phase out for the opponent.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dave J McWeasely
United States
Louisville
Kentucky
flag msg tools
designer
Avatar
mbmbmb
Here's an example of a game with GalDevs, where I opened with a classic "develop" strategy, calling D/D with no development-discounting-developments in my hand. Hey, maybe I'll draw some on Develop powers?

No such luck, but with the Most Developments goal out, I feel forced to drop two devs this turn. I go with Contact Specialist, and $$$ expensive-o $$$ Terraforming Robots. Starting T2 with 3 cards to Ancient Race's 7+R&D, I felt I had no choice but to E+1+1/E+5. He called E+1+1/Settle, inexplicably discarded Universal Symbionts, and settled a totally random Deserted Alien Grey World (a game-losing mistake, as it turns out). Meanwhile, on my side of the table, my super-explore netted me three excellent cards: Galactic Federation to further my Developers, Primitive Rebel Race for the pending settle, and Terraforming Guild, which would let me go berserk in the Settle phase. After some hemming and hawing, I abandoned my attempts at the Develop strategy, instead to go with the TERRAFORMING strategy. Either would have won the game and the Most Developments prize, but I had already two settle powers to only one Develop power, so the settling seemed more robust.

From the high ground of an advantage of developments, I could run downhill to whatever strategy suited me.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Eric Jome
United States
Franklin
Wisconsin
flag msg tools
designer
badge
Avatar
mb
rrenaud wrote:
On the other hand, I am not sure how you are measuring ease to play. I am not sure that the game is less strategically deep in 2 player than in 3+.


You are controlling a great deal more of what happens in the game. And you have a lot less difficulty in predicting what the other players will do.

I dunno... we all have different opinions on Keldon's AI, but I'll tell you this - I blow it away on 2 player constantly and I struggle to come second in 3 player... and win more than my share at 5 player and up. I don't play online against other human beings - perhaps I should. I get lots of game time in front of multiplayer groups live.

But my experience, limited as it is, indicates to me that 2 player advanced is much, much easier to play. You control the game a lot more, instead of trying to guess what others will do.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Rob Neuhaus
United States
New York
NY
flag msg tools
Avatar
mbmbmbmbmb
The problem about trying to make generalizations about the game from the AI is that the AI has an unfair advantage when playing in multiplayer games. Since it is playing in nearly the same homogeneous (all opponents are runinng the same logic as itself). If you take a bunch of players who are all about equal vs the AI in some setup, and force the AI to play against a bunch of those human players with only a single instance of itself, I think the AI will to substantially worse than equal.
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mark Delano
United States
Cary
North Carolina
flag msg tools
mbmbmbmbmb
rrenaud wrote:
The problem about trying to make generalizations about the game from the AI is that the AI has an unfair advantage when playing in multiplayer games. Since it is playing in nearly the same homogeneous (all opponents are runinng the same logic as itself). If you take a bunch of players who are all about equal vs the AI in some setup, and force the AI to play against a bunch of those human players with only a single instance of itself, I think the AI will to substantially worse than equal.


Probably true, except in this situation the AI does not take advantage in 4-6 player games. It is very easy to beat in 6 player, probably easier than in 2 player advanced. 3 player is by far the toughest opponent.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Once the Geek has you, there is no escape...
United States
Binghamton
New York
flag msg tools
badge
Avatar
mbmbmbmbmb
I haven't figured exactly to what degree "collusion" occurs between the AIs, but I'm certain it does. I've seen one AI clearly throw games to another AI.

3 player against the AI is definitely more difficult, but each instance of the AI also plays better individually than it does in 2PA.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Guy Srinivasan
United States
Kirkland
Washington
flag msg tools
Avatar
mbmbmbmbmb
TheMadVulcan wrote:
I haven't figured exactly to what degree "collusion" occurs between the AIs, but I'm certain it does. I've seen one AI clearly throw games to another AI.

3 player against the AI is definitely more difficult, but each instance of the AI also plays better individually than it does in 2PA.

They don't throw games to each other. What would be clearly throwing a game if you could infer any intent at all is in the case of these AIs simply bad behavior that you can also exploit.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Once the Geek has you, there is no escape...
United States
Binghamton
New York
flag msg tools
badge
Avatar
mbmbmbmbmb
AI 1 has matched AI 2 for the Most Develop Goal. I've long since stepped out of that race. One of AI 1's developments is R&D Crash. In the final round, the only build action called is Develop (by AI 2). In placing its development (Export Duties), AI 1 decides not to spend one of its 8 cards in hand to pay for the development and instead discards R&D, losing the Goal to AI 2. Doing so does not have a chance of preventing the end game because the game is guaranteed to end on VPs. I lose by a point or on tie-breakers (I don't recall) to AI 2.

I don't believe that this is necessarily explicit collusion (thus my use of quotes above), and it is clearly 'bad behavior', but this sure seems to be more than that. I've seen several other instances of AIs behaving in such a way that it makes no significant difference for me or itself, but it does allow the other AI a significant advantage.

As another example of "collusion", I'm fairly certain that under specific circumstances one AI will choose Develop and the other Settle, fully knowing what the other picks. Training the AIs against each other could reasonably produce this effect. You wrote somewhere (in a conversation with David DesJardins(sp?), I believe) about a hypothetical convention in which a higher level player will always choose one phase (Dev or Set) and the lower level player will always choose the other. I think the AIs might have developed something like this convention using some fairly ambiguous (to human eyes) variable.

This may simply be the effect that Rob mentions above. Two human players against 1 AI would very likely produce fewer than 1/2 the AI wins than a 2 AIs against 1 human setup.

Please don't take my comments as sour grapes. I'm immensely grateful to Keldon for all the time and energy he's spent creating his AI. And I'm content with my record against the AI in both 2PA and 3P. However, 1) I think that the training of an AI against itself could reasonably produce some "collusion"-like results and 2) I think it is unarguable that the AI's endgame performance needs work.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Eric Brosius
United States
Needham Heights
Massachusetts
flag msg tools
badge
My favorite 18xx game for six players is two games of 1846 with three players each.
Avatar
mbmbmbmbmb
GreedyAlgorithm wrote:
They don't throw games to each other. What would be clearly throwing a game if you could infer any intent at all is in the case of these AIs simply bad behavior that you can also exploit.


I agree. It may be less noticeable when an AI is throwing the game to you, but it happens. Just today I had a game where I was able to play Consume x2 every turn because one of the AIs was playing Produce every time (though it was obviously losing ground every turn.)
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Guy Srinivasan
United States
Kirkland
Washington
flag msg tools
Avatar
mbmbmbmbmb
TheMadVulcan wrote:
AI 1 has matched AI 2 for the Most Develop Goal. I've long since stepped out of that race. One of AI 1's developments is R&D Crash. In the final round, the only build action called is Develop (by AI 2). In placing its development (Export Duties), AI 1 decides not to spend one of its 8 cards in hand to pay for the development and instead discards R&D, losing the Goal to AI 2. Doing so does not have a chance of preventing the end game because the game is guaranteed to end on VPs. I lose by a point or on tie-breakers (I don't recall) to AI 2.

This is the sort of thing that has nothing to do with collusion and everything to do with artificial neural networks' well-known poor performance in highly tactical situations. Or as you say, "the AI's endgame performance needs work".

TheMadVulcan wrote:
As another example of "collusion", I'm fairly certain that under specific circumstances one AI will choose Develop and the other Settle, fully knowing what the other picks. Training the AIs against each other could reasonably produce this effect. You wrote somewhere (in a conversation with David DesJardins(sp?), I believe) about a hypothetical convention in which a higher level player will always choose one phase (Dev or Set) and the lower level player will always choose the other. I think the AIs might have developed something like this convention using some fairly ambiguous (to human eyes) variable.

Yes, this is the sort of thing that could come about via groupthink training effects. My guess is that when they "need" both they naturally pick the one that would be best (most discounts, whatever) for themselves in other similar-except-for-needing-both-phases situations, since that's how neural networks end up training. When it happens that they have asymmetric discounts (or whatever "best" means to the NN), they'll pick distinct roles.

Do they manage it more often than humans? Dunno. Do they do it explicitly? Definitely not, read the code. Does it come up often enough in training, and are the right features present in the input to the NN, that it can actually implicitly collude in this circumstance? Not sure, but I doubt it - I've learned to put a very high prior probability on confirmation bias.
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Mark Delano
United States
Cary
North Carolina
flag msg tools
mbmbmbmbmb
The AI is not well trained for the last turn. It tends to keep barrelling ahead when it should really be doing some endgame optimizing.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
United States
SoCal
flag msg tools
badge
Avatar
mbmbmbmbmb
I also see everynow and then... I'll Consume x2, but all the other 3 will explore, expecting that someone would've choosen II or III, but to no avail.
 
 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.