|
Brian Lee
United States
Northridge
California
|
Any thoughts regarding this card?
|
Chris Linneman
Canada
Vancouver
BC
|
It fits in very nicely alongside Galactic Genome Project, gene windfall world and any production world for a strong produce-consume engine. That's pretty specific, but at cost 3 with a $4 good it is not a bad Settle-Trade option. Usually, though, this ends up as cash.
|
ackmondual
United States
Virginia
|
QBert80 wrote: It fits in very nicely alongside Galactic Genome Project, gene windfall world and any production world for a strong produce-consume engine. That's pretty specific, but at cost 3 with a $4 good it is not a bad Settle-Trade option. Usually, though, this ends up as cash. It's consume one good for 1 VP IV power can be equally a curse and a blessing. The only way to trade the good without having to call produce to replenish its good is a naked trade. Otherwise, you settle it, and if someone else choose IV, then it's going away as a VP instead. Be nice to have something like a blue good to use as cannon fodder VP fodder so you still have the green good to trade for next round.
|
Chris Linneman
Canada
Vancouver
BC
|
Of course the Settle/Trade usage works best in 2p Advanced. Or in the case that you have another good to consume instead. Or if no one calls Consume. Or if you call naked Trade. Or if you have Genetics Lab/Galactic Engineers/Terraforming Guild and someone calls Produce. So yeah, there are a lot of times you can Settle it and be reasonably confident you can trade the good.
|
Serge
Canada
Vancouver
British Columbia
|
ackmondual wrote: It's consume one good for 1 VP IV power can be equally a curse and a blessing. I see this general idea on here a lot, and i really think it's vastly overstated. It's nearly never a problem for me. All it requires is a little forethought or a backup plan. This might deserve a strategy thread of its own... Forethought: if you can see that someone has a juicy good they are going to trade, or the produce-consumer is on an x2 turn, then don't call settle this turn to avoid sabotaging your windfall good. Or backup plan: call settle with the intent to settle your windfall, but if someone calls consume then play a different world you held onto as a backup for just that situation.
|
Chris Linneman
Canada
Vancouver
BC
|
Serge you are absolutely right. As long as you are not dumping your hand and saying, "oh shit" when someone Consumes (and are you really ever shocked when a player with an Alien good plays trade or the guy with Free Trade Association and 3 blue goods consumes double?) the VP consumption is rarely much of a problem.
In fact, it can be a nice boost if you know Player X is going to Develop, Player Y is going to Consume and Player Z is going to Produce. Settle your windfall, drop your Genetics Lab on Develop and BAM! you are ready to Trade next turn--with an extra VP chip to boot.
|
Dave J McWeasely
United States
Louisville
Kentucky
|
I find it coming into play pretty often. Galactic Genome Project hungers for green goods, and this is the cheapest way to get 2 green production going. Except for Plague World which is toxic deprecated because of the 0 vp thing. ackmondual wrote: It's consume one good for 1 VP IV power can be equally a curse and a blessing. The only way to trade the good without having to call produce to replenish its good is a naked trade. Otherwise, you settle it, and if someone else choose IV, then it's going away as a VP instead. Be nice to have something like a blue good to use as cannon fodder VP fodder so you still have the green good to trade for next round. Either you call produce, in which case you by definition have at least two green cards, so you have an armored trade, or someone else calls produce for you, in which case it can't be chipped before you have a chance to call trade, or its the turn it comes into play and you're calling Settle yourself, in which case you have a point. Most of the time this case reduces to the case where you have two green windfalls to settle. You call Settle with the intention of calling trade next turn. If someone calls Trade this turn, simply settle the non-symbiote world, and save the symbiotants for the next settle.
Last edited on 2009-09-26 22:55:20 CST (Total Number of Edits: 2)
|
BT Carpenter
United States
Reston
Virginia
|
MrWeasely wrote: I find it coming into play pretty often. Galactic Genome Project hungers for green goods, and this is the cheapest way to get 2 green production going. Except for Plague World which is toxic because of the 0 vp thing. Toxic? If I see this in the first half of a tableau and I'm already working a Produce/Consume engine, this thing can easily spit out 6-8 VPs over the course of the game. How is that toxic?
|
ackmondual
United States
Virginia
|
QBert80 wrote: Serge you are absolutely right. As long as you are not dumping your hand and saying, "oh shit" when someone Consumes (and are you really ever shocked when a player with an Alien good plays trade or the guy with Free Trade Association and 3 blue goods consumes double?) the VP consumption is rarely much of a problem.
In fact, it can be a nice boost if you know Player X is going to Develop, Player Y is going to Consume and Player Z is going to Produce. Settle your windfall, drop your Genetics Lab on Develop and BAM! you are ready to Trade next turn--with an extra VP chip to boot. Until you get that kind of setup, you either have to settle something else instaed, decline a settle, or settle US and find another way to trade a good. It's inherit with many cases where you settle a windfall, but the good gets forced consumed b/c you have a consume power for it. byronczimmer wrote: MrWeasely wrote: I find it coming into play pretty often. Galactic Genome Project hungers for green goods, and this is the cheapest way to get 2 green production going. Except for Plague World which is toxic because of the 0 vp thing. Toxic? If I see this in the first half of a tableau and I'm already working a Produce/Consume engine, this thing can easily spit out 6-8 VPs over the course of the game. How is that toxic? The card draw is another nice thing. It gives you more options to leech off your opponents' II/III
|
Rob Renaud
United States
New York
New York
|
MrWeasely wrote: I find it coming into play pretty often. Galactic Genome Project hungers for green goods, and this is the cheapest way to get 2 green production going. Except for Plague World which is toxic deprecated because of the 0 vp thing. If you say it's one of the weakest production worlds in the game, I'd totally agree with you. And so would 22000 games of Race 
|
Serge
Canada
Vancouver
British Columbia
|
rrenaud wrote: If you say it's one of the weakest production worlds in the game, I'd totally agree with you. And so would 22000 games of Race ;) The hatred of Plague World is another thing i think may be overstated. 0vp instead of 1-2 isn't all that big of a difference, and the card draw makes up for it. I feel like it may be some kind of cognitive bias against the number zero, when what's vastly more important are the powers (the long term). To me the Genie data says that it is likely underplayed/ignored/misplayed, not necessarily that it's bad.
|
Tom Lehmann
United States
Palo Alto
California
|
It's one thing to assemble a large data set; it's quite another thing to interpret it correctly. We had between 10-15K playtest games of RFTG before it was published, across six different test groups. Since then, we've roughly doubled this number of games. It was interesting watching the different groups. On one hand, there was a fairly common trajectory of strategies that groups developed over time (which I've since seen mirrored in many comments by players since the game has been published). On the other hand, there are some fairly significant group-think differences that have persisted over time (around frequency of exploration, naked trades, second produces, and willingness to switch strategies when behind), even when we had some top players from different groups play against each other. In some areas, it appears that RFTG has cyclical over time, rather than dominate, meta-strategies (though I don't have enough data to be very confident about this conclusion). However, seeing this makes me leery of relying on stats from a single group, no matter how strong some of those players may be. It was quite useful having Wei-Hwa and I be in separate playtest groups for most of the game's development. When Rob first posted the Genie data regarding start worlds, I took a glance at it. My first reaction was to note that the range from top to bottom in winning % was fairly small (roughly 55% to 45%), which I found reassuring from a balance point of view. His first reaction was to proclaim that Alpha Centauri rocks... My second reaction was to look at the top start worlds and note that they were all worlds that give players early card advantage. The led me to hypothesize that the play on Genie *taken as a whole* was what I would call "intermediate" level, based on comparisons with the playtest groups' experiences with start worlds. (This is not to say that some Genie players aren't extremely strong or that this won't change over time.) Now, to give Rob credit, he did go back and take a harder look at the Alpha Centauri data and found that it fared less well in winning % among the stronger players. To me, that was expected. Card advantage is something that is fairly easy for beginning and intermediate players to exploit (as opposed to leeching or explore powers, etc.). But this experience certainly means that I take "data-based" RFTG claims with a boulder or two of salt. It's very easy to generate data and statistics and make claims. It's very hard to generate meaningful metrics and sound conclusions from data sets. And only now are some of these data sets approaching the number of games that we, as developers, have been able to observe.
|
Rob Renaud
United States
New York
New York
|
The draw on consume seems nice, but I don't think it is really all that great in Plague World's case, since it is limited to green only. Usually when you add a draw on consume power to an engine early in the game, it adds to the number of cards drawn on a produce/trade cycle. However, this is often not the case when adding plague world to an engine early, since it is likely to be traded away. Once you have started consuming, you get the bonus on consume, but by this time, it's blue/brown competitors are much more likely drawing one or two on produce (consumer markets are mining conglomerate are cheaper and more frequent than PGL), and the blues are even somewhat likely to be drawing on consume anyway.
|
Serge
Canada
Vancouver
British Columbia
|
Tom Lehmann wrote: second produces What is a "second produce"? It's probly not this, but the following needs a name as well: when produce helps your opponents more than you but you call it to get a good on your windfall, because you know someone else is calling it this turn.
|
Dave J McWeasely
United States
Louisville
Kentucky
|
While I appreciate the discussion that this has produced, I find it silly that this has kicked off a minor umbrage war over Plague World. Look, sometimes you just don't play certain cards. That's not to say that it is a bad design or un-fun. It just often doesn't scratch an itch in your tableau. No shame in that.
Last edited on 2009-09-27 22:31:35 CST (Total Number of Edits: 1)
|
Rob Renaud
United States
New York
New York
|
I like arguing about this stuff , and I thoroughly enjoy picking Tom's brain. I hope no one takes offense to my posts, certainly I wouldn't have played thousands of games of race if I didn't think it was brilliant and I don't mean to come of as nasty or hostile. What is the right metric for card strength? Where is the ideal skill level that you think the game is balanced? Do you need to play 1000 games to get there? 5000? If the winning percentages were 55 to 45 in the long for highly skilled players, after many many games, would you be happy with that? What kind of data did the playtesters keep? I've recorded homeworlds/goals/scores/names/a few special cards played for ~250 games manually, but it got to be exhausting, and genie is a much cleaner source for the stuff anyway. I think the argument from the data is stronger here than the Alpha vs other homeworld argument. This isn't 55% vs 50%. This is more like being a solid 40% under the mean number of wins among comparable worlds. I think the bottom leftness on the win-rate/playrate shows that Plague World is quite weak. Here is the raw data among non-homeworld prod worlds for total wins, win rate, and probability of being played. This just gathering storm, 2p advanced. I do believe that green prod is more viable in RvI, so likely the green worlds here that are bringing up the bottom will be stronger.
Mining World 2375 1.099 0.109
Comet Zone 2355 1.091 0.109
Gem World 2270 1.073 0.107
Lost Species Ark World 2032 1.087 0.094
Spice World 2004 1.020 0.099
Artist Colony 1997 1.016 0.099
New Survivalists 1920 1.035 0.093
Galactic Studios 1918 1.137 0.085
Rebel Miners 1899 1.045 0.092
Secluded World 1868 1.025 0.092
New Vinland 1814 1.046 0.087
New Earth 1808 1.050 0.087
Space Port 1707 0.989 0.087
Prosperous World 1661 1.007 0.083
Bio-hazard Mining World 1651 1.012 0.082
Lost Alien Battle Fleet 1630 1.130 0.073
Volcanic World 1504 0.964 0.079
Imperium Armaments World 1279 0.935 0.069
Hive World 1232 0.994 0.062
Alien Robotic Factory 1112 1.006 0.056
Malevolent Lifeforms 1099 0.998 0.055
Plague World 995 0.878 0.057
Distant World 939 0.849 0.056
Last edited on 2009-09-27 23:57:38 CST (Total Number of Edits: 1)
|
Tom Lehmann
United States
Palo Alto
California
|
This data is mostly meaningless to me. What does a "win rate" of 1.09 mean? It's not a probability. It's some "normalized" something that's not defined.
When you were talking about 2-player win %s for the various start worlds, that meant something in real terms. This doesn't.
|
Rob Renaud
United States
New York
New York
|
C'mon Tom, you say I have the wrong metrics, and you don't even know my metrics?  From http://racestats.mastersofspace.netQuote:
A brief discussion about Winning Rates
An n player game is worth n points. The wining rate is the number of points accumulated divided by the number of games played. Thus, if you win a 4 player game, lose a 3 player game, and lose a 2 player game, your winning rate would (4 + 0 + 0) / 3 = 1.33. Thus, a totally average and optimally balanced homeworld will have a winning rate of near 1 after many games. Likewise, a player whose skill is totally representative of the distribution of the player population will have a winning rate of 1.
But these are all 2p advanced games, so win probability (if ties count as half a win) is exactly win rate / 2. Edit: It's not actually wins that were listed on the left, it was winning points, which is just wins * 2 + ties for this case, but it's just a linear shift, so the order is preserved and the story is still the same.
Last edited on 2009-09-28 10:24:45 CST (Total Number of Edits: 1)
|
Randall Bart
United States
Seattle
Washington
|
Tom Lehmann wrote: This data is mostly meaningless to me. What does a "win rate" of 1.09 mean? It's not a probability. It's some "normalized" something that's not defined. Tom, I'm surprised you missed this. It's normalized and the normalization is obvious. It's win rate times number of players, which is the most usable metric for a game with variable player count.
|
Tom Lehmann
United States
Palo Alto
California
|
No, it's not obvious.
I tried to make sense of this statement: "This is more like being a solid 40% under the mean number of wins among comparable worlds" with the data presented.
If you think this is so goddamn obvious, why don't you explain how these two things relate?
I tried to think it through and couldn't see any way to relate these two items. So, I gave up.
|
Rob Renaud
United States
New York
New York
|
Tom Lehmann wrote: No, it's not obvious.
I tried to make sense of this statement: "This is more like being a solid 40% under the mean number of wins among comparable worlds" with the data presented.
If you think this is so goddamn obvious, why don't you explain how these two things relate?
I tried to think it through and couldn't see any way to relate these two items. So, I gave up.
Certainly, if you don't understand it, I am just not writing clearly enough. I thought you had read my stats page and had enough context. Take the average of the number of win points for each of those worlds, (the second column) get 1698.65. Divide the winning points for Plague World by this number, get 0.585759. From here, see that plague worlds occurs on a winning tableau about 40% less than the average production world. That is as simple of an argument for Plague World being weak as I can make from the data. The graph on http://rftgstats.com shows it better, but there the point is more subtle, it's not a raw number of wins. I totally buy the argument that raw wins isn't quite the right metric. If two worlds have the same number of wins, but one is played half as much as the other (and consequently, it's twice as often to win given it's played), I would believe that the infrequently played but highly correlated with winning card is better. However, plague world fairs poorly on both the play rate and the winning rate front. Also, correlation is not causation, surely sometimes a non-producer will throw down a random world, and it's almost certainly not going to be plague world because of the poor cost/point ratio, and some of the other production worlds get a bonus from this effect. Still, 40% below the mean quite a margin to overcome. The Spike in me is rarely going to be pleased by having plague world in hand.
Last edited on 2009-12-03 21:04:50 CST (Total Number of Edits: 2)
|
Guy Srinivasan
United States
Seattle
Washington
|
Yeah, um, Rob, unless you start factoring in something else, the whole win-rate thing definitely can't notice things like "for some reason everyone hates Plague World so they only play it when they're behind since sometimes then they have no other choice even though it's fine". entranced wrote: I feel like it may be some kind of cognitive bias against the number zero, when what's vastly more important are the powers (the long term). To me the Genie data says that it is likely underplayed/ignored/misplayed, not necessarily that it's bad.
This is IMO almost certainly correct. Remember when you first produced the statistics, and I started a thread about how to actually measure a card's strength? The same criticisms still apply, applying your win rate metric to anything but homeworlds which are forced into the tableau (or players if they don't cherry-pick their opponents) is subject to wildly huge confounders such as "what if people are irrationally biased against playing Plague World?", since it assumes (essentially) a uniform distribution over game types in which cards are played. (ETA: anything far out on the +win/played axis I'd say is legitimately a "good card", that sort of thing will survive more detailed accounting, I'm sure. Except the ones played very rarely, since it could be that you're almost always sad to see them in your hand rather than another card, but you won't play them unless they're winning. Galactic Renaissance, Imperium Lords, Rebel Homeworld, and New Economy, specifically.) But all that's overshadowed by not taking player ratings into account. As it stands, the more tricky a card is to use well, the lower it will be ranked with your metric - poorer players will play it in the wrong circumstances, use it poorly, and lose. Here's a simple change that should help (this only helps fix the skill level bias): whenever computing exp_wins for 2p, instead of adding 1/2, add EloProbability(player_result['name']'s ELO, other ELO). For 4p it's not as clear since your ELO computation for 4p ends up being "probability A wins and B does not", which in any specific game is not independent of "probability A wins and C does not", so it's not at all obvious to me how to combine them to form an expected score.
Last edited on 2009-09-28 12:58:43 CST (Total Number of Edits: 2)
|
Tom Lehmann
United States
Palo Alto
California
|
But, this is where I don't understand the point of your "normalization".
If you are normalizing things such that the win rate of an average card is defined as 1.00 and the win rate of Plague World is .88, then it would seem to me that the correct statement is to say that Plague World, on average, when played wins 12% less. That is, if an average player sits down against an average opponent in a 2-player game, then, of all the tableaus that this player plays which involve Plague World, 44% are winning tableaus and 56% of them are losing tableaus (ignoring ties). Is this understanding correct?
By ignoring your own normalization and dividing raw win points by average raw win points for a class of related cards, you're conflating the *frequency of a card being played* with its *win correlation*. I don't see how this measures the "strength" of a card in any meaningful way.
I will certainly agree you can construct a metric by doing this. Obviously, such a metric is biased against specialized cards. However, I would call this metric "how generally useful" a card is, not a proxy for a card's "strength". A card that is not generally useful may still be overpowered or underpowered.
To then make statements such as Plague World is "one of the weakest production worlds in the game" (as you did above), as opposed to, say, Plague World is one of the less generally useful production worlds in the game, seems to me to be miss-stating what your data is actually telling us.
I will certainly agree that RFTG includes some very specialized cards. This is by design.
|
Jason Woolever
United States
Mountain View
California
|
I am suspect of any rating that is conditioned on the times cards are played. It would seem better to include all instances where cards are drawn. Consider the follow made up card for 2-player games:
Desperate Longshot
World [cost: 1]
When Desperate Longshot comes into play, roll a die. If it is a 1 or 2, you win immediately. Otherwise you lose immediately.
This card would by definition have a 1/3 win rate and thus would have a lousy score if you only consider games the card is played. However, if you counted all instances where the card is drawn, it would probably appear quite good.
The same effect could happen in reverse. If there are cards that tend to be played in running-up-the-score situations, it'll appear better than it really is. Granted, there aren't real cards as egregious as my made up example, but I worry the effect could also skew values of real cards.
|
Rob Renaud
United States
New York
New York
|
GreedyAlgorithm wrote: Yeah, um, Rob, unless you start factoring in something else, the whole win-rate thing definitely can't notice things like "for some reason everyone hates Plague World so they only play it when they're behind since sometimes then they have no other choice even though it's fine". I guess I tend to think more of the players on genie ;P. Although, at least one reasonably highly rated player has said he has a distaste for 0 value cards (in that case, it was mercenary forces). Quote: This is IMO almost certainly correct. Remember when you first produced the statistics, and I started a thread about how to actually measure a card's strength? The same criticisms still apply, applying your win rate metric to anything but homeworlds which are forced into the tableau (or players if they don't cherry-pick their opponents) is subject to wildly huge confounders such as "what if people are irrationally biased against playing Plague World?", since it assumes (essentially) a uniform distribution over game types in which cards are played. If you assume players are rational, it's not a big problem. You actually get a lot more information than just homeworld/goals/win from a tableau. I think you can make interesting observations from that information. I agree that you can also imagine an adversarial universe in which I make false deductions based on this data, but I don't think the universe is actually so mean to me ;P. Large samples should shake out biases. Do you believe that stronger players are more rational than weaker ones? If I bucket the stats by win level, and then animate the graph by increasing skill level, would that address the rationality concern? Basically, are you willing to accept the premise that skill level correlates with increasing rationality? Quote:
(ETA: anything far out on the +win/played axis I'd say is legitimately a "good card", that sort of thing will survive more detailed accounting, I'm sure. Except the ones played very rarely, since it could be that you're almost always sad to see them in your hand rather than another card, but you won't play them unless they're winning. Galactic Renaissance, Imperium Lords, Rebel Homeworld, and New Economy, specifically.)
But all that's overshadowed by not taking player ratings into account. As it stands, the more tricky a card is to use well, the lower it will be ranked with your metric - poorer players will play it in the wrong circumstances, use it poorly, and lose. Here's a simple change that should help (this only helps fix the skill level bias): whenever computing exp_wins for 2p, instead of adding 1/2, add EloProbability(player_result['name']'s ELO, other ELO).
I'll do the Elo normalization eventually (shouldn't post stuff like this when I have homework due in a couple days :P), but I doubt things will move much. I've done something similar before, and it showed that good players were slightly more inclined to play some of the non-obvious cards (New Economy can score quite large on a prod table, even if its ability is possibly the weakest of all the 6s), but no drastic difference. Quote: For 4p it's not as clear since your ELO computation for 4p ends up being "probability A wins and B does not", which in any specific game is not independent of "probability A wins and C does not", so it's not at all obvious to me how to combine them to form an expected score. Why do you think this is a big problem?
|
|
|
Posted Sat Sep 26, 2009 8:05 am
