Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Donald X.]

What I would call a "fun" set is a set that is in the A or B categories.

This suggests a direct approach. Pick 10 cards at random. Count how many strategies you see. If too low, try again with a new 10. You have certain broad strategies that you identify, and then card-specific ones where certain things are needed to make them plausible enough. This is the method-of-producing-option-rich-sets that minimizes the bias towards specific cards (though it doesn't get rid of it).

[Donald X.]

It's possible that I might be wrong about that assumption (what casual players like).  Another thing that I am trying to get Goko to do is to have a way to get user feedback on kingdoms.  A simple way to start would be, after a game, let the player give a simple "thumbs up / thumbs down" on whether they liked the kingdom they just played.  I'd then adjust the generator to compensate.

If people are rating sets, then you could use those lists directly. Have all games provide the ranking option. If the players both thumbs up a set (or at least two do if it's multiplayer), store the list. Once you have enough, add a mode that's "play a set two people liked." It just picks a double-thumb'd up set and that's that. These too are ranked. After enough data piles up on those, have a mode that's "play a set that a bunch of people liked."

[Davio]

I know some people might disagree with me here, but I do think people should choose what sets get chosen for the randomizer. As it is, someone can buy Prosperity and have just Prosperity and base come up. In real life, many people only play with two sets at a time. So, I think that the option for full random or choosing which expansions should exist.

Absolutely.  It's why the expansion sets are at the top of my list in my ugly-looking HTML interface.  Almost certainly it's going to be the most requested feature in any set generator Goko uses.  The tricky bit with anything that has to face a casual player, though, is to not overwhelm them with options they don't understand yet.

Well, you could just make a collapsible section with a button "Advanced Options" or something.

[onigame]

What I would call a "fun" set is a set that is in the A or B categories.

This suggests a direct approach. Pick 10 cards at random. Count how many strategies you see. If too low, try again with a new 10. You have certain broad strategies that you identify, and then card-specific ones where certain things are needed to make them plausible enough. This is the method-of-producing-option-rich-sets that minimizes the bias towards specific cards (though it doesn't get rid of it).

This direct approach was my first approach.  I eventually abandoned it because (1) I couldn't be confident that I was comprehensively identifying all the broad strategies; (2) The generation would tend to choke for partial data -- say the user wanted to play with five specific cards, and the generator would take a long time to find five cards to add that satisfied its conditions; (3) There wasn't a great way to "reverse the polarity" and bias away from those decks.

I think it works well if you're trying to find a few good kingdoms that you're willing to play over-and-over again, but not so hot for a more general-purpose random filler.

[LastFootnote]

Well, a "new randomizer" has apparently been implemented on Goko. Apparently it's used for all non-Pro games, if I understand correctly. So, Wei-Hwa, I'm writing to try to summarize my feelings. As you suggested, I'm doing it here rather than on my bug report in Goko's GetSatisfaction.

I feel like, if you'd created this randomizer for your own use and put it on your website, or if you'd gotten them to implement it in a special, separate mode in Goko, or if it was just a non-default option for casual (or even pro) games on Goko, I might be helping and supporting you in this endeavor. Maybe not. But in theory it sounds nice to create a randomizer that magically only makes 'fun' sets, and I respect your chutzpah.

Now that it's been implemented on Goko as the only way to randomize non-Pro games, I'm livid. I'm just so, so angry. I just don't see why you've wrenched away the status quo in order to put in place your new, still highly experimental algorithm. Yeah, I could play only Pro-rated games, but a lot of times I just want a lower-stress experience or I want to include certain cards to practice with them. That's another effect of this change: randomly-generated casual games no longer 'match-up' with pro games. Practicing in casual games will no longer actually prepare me for what I may encounter in pro games (non-fun sets).

I could use a randomizer external to Goko and then painstakingly enter the cards into a new set each game, but that's just not a hoop I'm willing to jump through.

Why did you feel you had to do this? Why couldn't you at least wait until they were willing to make it a choice, rather than the only option? Why are your opinions so much more important than mine?

I guess it would be hyperbole to say that I feel betrayed or duped. I feel a little like the rug was pulled out from under me. I paid for the service, with the assumption that I knew more or less what I was getting and that functionality wouldn't be taken away. I feel…unhappy. A lot of my desire to play on Goko has just sort of leaked away. They've already got my money, so no skin off their teeth, I guess…

 

P.S. Good luck with your randomizer.

[Beyond Awesome]

The randomizer is already up? I thought it was supposed to let you choose from sets and whatnot. Maybe your right though. I clicked through the randomizer 10 times (not enough for a true sample size), but each time there was a village.

[WanderingWinder]

Well, a "new randomizer" has apparently been implemented on Goko. Apparently it's used for all non-Pro games, if I understand correctly. So, Wei-Hwa, I'm writing to try to summarize my feelings. As you suggested, I'm doing it here rather than on my bug report in Goko's GetSatisfaction.

I feel like, if you'd created this randomizer for your own use and put it on your website, or if you'd gotten them to implement it in a special, separate mode in Goko, or if it was just a non-default option for casual (or even pro) games on Goko, I might be helping and supporting you in this endeavor. Maybe not. But in theory it sounds nice to create a randomizer that magically only makes 'fun' sets, and I respect your chutzpah.

Now that it's been implemented on Goko as the only way to randomize non-Pro games, I'm livid. I'm just so, so angry. I just don't see why you've wrenched away the status quo in order to put in place your new, still highly experimental algorithm. Yeah, I could play only Pro-rated games, but a lot of times I just want a lower-stress experience or I want to include certain cards to practice with them. That's another effect of this change: randomly-generated casual games no longer 'match-up' with pro games. Practicing in casual games will no longer actually prepare me for what I may encounter in pro games (non-fun sets).

I could use a randomizer external to Goko and then painstakingly enter the cards into a new set each game, but that's just not a hoop I'm willing to jump through.

Why did you feel you had to do this? Why couldn't you at least wait until they were willing to make it a choice, rather than the only option? Why are your opinions so much more important than mine?

I guess it would be hyperbole to say that I feel betrayed or duped. I feel a little like the rug was pulled out from under me. I paid for the service, with the assumption that I knew more or less what I was getting and that functionality wouldn't be taken away. I feel…unhappy. A lot of my desire to play on Goko has just sort of leaked away. They've already got my money, so no skin off their teeth, I guess…

 

P.S. Good luck with your randomizer.

I don't think blaming the implementation of this on Goko on Wei-Hwa is fair. I mean, I assume he permitted them to use it, but it's not like he is the one who forced it to happen or made the final decision.

[LastFootnote]

I don't think blaming the implementation of this on Goko on Wei-Hwa is fair. I mean, I assume he permitted them to use it, but it's not like he is the one who forced it to happen or made the final decision.

Forgive me, but I'm struggling to come up with a scenario where he's not mostly to blame.

[WanderingWinder]

I don't think blaming the implementation of this on Goko on Wei-Hwa is fair. I mean, I assume he permitted them to use it, but it's not like he is the one who forced it to happen or made the final decision.

Forgive me, but I'm struggling to come up with a scenario where he's not mostly to blame.

Wei-Hwa wants to come up with a way to make kingdoms he thinks are fun. He designs this tool. He thinks, 'hey, maybe other people like this too, I should at least give them the chance to use it, because I think it's neat'. He tells other people (including this board) about it.
Goko thinks 'Hey there's this really cool tool that we think makes awesome kingdoms. Let's make this the default and incidentally give nobody the option to play pure random on casual. People like having FUN right?'
Goko asks Wei-Hwa 'Hey, can we use your set generator on our site?'
Wei-Hwa says yes.
Goko implements.

I'm really struggling to place blame on Wei-Hwa in this scenario, which is sorta my null assumption of what happened.

 I wouldn't at all be surprised that he pushed for this to be used on their site, but even so, I would still be very surprised, given his statements, if he pushed to make full random not be an option.

[ftl]

Having said all that, I do believe that my current implementation of the Set Generator weights engine games too high and I am trying to find ways to lower it.  There are many fun sets that don't have an engine and I don't think my generator is finding them frequently enough.

It seems like your generator is fundamentally always going to be engine-biased. You say on the page that the aim is to have

- sets that have bits of everything

and those are fundamentally engine sets. If you have a village, card draw, an attack, +buy, trashing, alt-vp in the set - you know, a little bit of everything - that'll be an engine set unless all the components suck.

[LastFootnote]

I don't think blaming the implementation of this on Goko on Wei-Hwa is fair. I mean, I assume he permitted them to use it, but it's not like he is the one who forced it to happen or made the final decision.

Forgive me, but I'm struggling to come up with a scenario where he's not mostly to blame.

Wei-Hwa wants to come up with a way to make kingdoms he thinks are fun. He designs this tool. He thinks, 'hey, maybe other people like this too, I should at least give them the chance to use it, because I think it's neat'. He tells other people (including this board) about it.
Goko thinks 'Hey there's this really cool tool that we think makes awesome kingdoms. Let's make this the default and incidentally give nobody the option to play pure random on casual. People like having FUN right?'
Goko asks Wei-Hwa 'Hey, can we use your set generator on our site?'
Wei-Hwa says yes.
Goko implements.

I'm really struggling to place blame on Wei-Hwa in this scenario, which is sorta my null assumption of what happened.

 I wouldn't at all be surprised that he pushed for this to be used on their site, but even so, I would still be very surprised, given his statements, if he pushed to make full random not be an option.

I don't think he pushed for full random not to be an option either. I just don't think he pushed very hard to make sure it was.

Here are some assumptions I'm making. If any are false, I'd very much appreciate being corrected.

• Wei-Hwa has a close relationship with Goko. If this weren't the case, we'd have seen the conversation that led to this implementation take place. It would have been a thread in the "Ideas Under Consideration" area of Goko's GetSatisfaction or something.
• He approached Goko about implementing his algorithm, not the other way around.
• He didn't tell Goko that some people had concerns about his algorithm.
• He didn't say, "You may use this algorithm, but make sure it's only an option, not mandatory."

[Toskk]

Goko definitely implemented something.. although I would agree with a few others here that the algorithm is significantly over-prioritizing certain cards. Using only Base, Intrigue, half of Seaside, and Governor, I got 6 games in a row today with Governor, Laboratory, Throne Room, Mine, Workshop, and Duke. They all had either Village or Fishing Village, as well.

[LastFootnote]

I'm currently generating sets at dominionsetgenerator and tallying up how often each card comes up. After 1000 runs on the default settings with only the base set selected, here's how often each card came up, along with the percentage compared to full random (50% means the card only appears half as often as it would in full random).

Card Name	#Occurrences	Proportion
Festival	954	238.50%
Market	943	235.75%
Laboratory	921	230.25%
Mine	799	199.75%
Village	730	182.50%
Remodel	693	173.25%
Moneylender	626	156.50%
Spy	601	150.25%
Feast	345	86.25%
Workshop	305	76.25%
Woodcutter	294	73.50%
Gardens	281	70.25%
Chapel	276	69.00%
Cellar	256	64.00%
Council Room	256	64.00%
Witch	229	57.25%
Thief	223	55.75%
Bureaucrat	218	54.50%
Library	197	49.25%
Throne Room	192	48.00%
Militia	181	45.25%
Adventurer	159	39.75%
Moat	137	34.25%
Smithy	137	34.25%
Chancellor	47	11.75%

Here's Intrigue:

Card Name	#Occurrences	Proportion
Harem	966	241.50%
Duke	963	240.75%
Mining Village	916	229.00%
Great Hall	906	226.50%
Conspirator	852	213.00%
Upgrade	724	181.00%
Trading Post	616	154.00%
Bridge	484	121.00%
Tribute	473	118.25%
Wishing Well	378	94.50%
Baron	376	94.00%
Shanty Town	366	91.50%
Scout	362	90.50%
Ironworks	201	50.25%
Pawn	195	48.75%
Torturer	190	47.50%
Nobles	149	37.25%
Saboteur	147	36.75%
Masquerade	145	36.25%
Steward	144	36.00%
Swindler	108	27.00%
Minion	99	24.75%
Secret Chamber	87	21.75%
Courtyard	84	21.00%
Coppersmith	69	17.25%

And Seaside:

Card Name	#Occurrences	Proportion
Merchant Ship	918	238.68%
Bazaar	834	216.84%
Treasury	770	200.20%
Explorer	556	144.56%
Native Village	555	144.30%
Haven	547	142.22%
Caravan	494	128.44%
Wharf	481	125.06%
Fishing Village	430	111.80%
Lighthouse	410	106.60%
Salvager	382	99.32%
Pearl Diver	380	98.80%
Ghost Ship	375	97.50%
Ambassador	356	92.56%
Island	337	87.62%
Outpost	330	85.80%
Embargo	300	78.00%
Sea Hag	299	77.74%
Treasure Map	279	72.54%
Smugglers	172	44.72%
Warehouse	165	42.90%
Cutpurse	155	40.30%
Pirate Ship	146	37.96%
Tactician	139	36.14%
Lookout	133	34.58%
Navigator	57	14.82%

[LastFootnote]

Prosperity:

Card Name	#Occurrences	Proportion
Grand Market	930	232.50%
Peddler	869	217.25%
Worker's Village	833	208.25%
Trade Route	766	191.50%
Forge	619	154.75%
Expand	567	141.75%
City	558	139.50%
Rabble	539	134.75%
Vault	432	108.00%
Bishop	426	106.50%
Watchtower	411	102.75%
Talisman	352	88.00%
Mountebank	313	78.25%
Royal Seal	287	71.75%
Quarry	278	69.50%
Monument	273	68.25%
Hoard	272	68.00%
Goons	249	62.25%
Contraband	222	55.50%
Bank	179	44.75%
Venture	170	42.50%
Mint	155	38.75%
Loan	141	35.25%
Counting House	86	21.50%
King's Court	73	18.25%

Hinterlands:

Card Name	#Occurrences	Proportion
Border Village	718	186.68%
Highway	704	183.04%
Cache	695	180.70%
Farmland	657	170.82%
Crossroads	645	167.70%
Nomad Camp	555	144.30%
Silk Road	543	141.18%
Ill-Gotten Gains	532	138.32%
Spice Merchant	498	129.48%
Stables	470	122.20%
Mandarin	464	120.64%
Cartographer	425	110.50%
Margrave	398	103.48%
Scheme	362	94.12%
Jack of all Trades	306	79.56%
Haggler	249	64.74%
Duchess	233	60.58%
Oasis	232	60.32%
Trader	215	55.90%
Tunnel	214	55.64%
Noble Brigand	189	49.14%
Oracle	176	45.76%
Fool's Gold	176	45.76%
Inn	139	36.14%
Develop	124	32.24%
Embassy	81	21.06%

Dark Ages:

Card Name	#Occurrences	Proportion
Bandit Camp	999	349.65%
Forager	908	317.80%
Beggar	901	315.35%
Vagrant	835	292.25%
Marauder	663	232.05%
Altar	477	166.95%
Armory	418	146.30%
Junk Dealer	413	144.55%
Count	359	125.65%
Mystic	304	106.40%
Counterfeit	292	102.20%
Market Square	263	92.05%
Catacombs	257	89.95%
Urchin	228	79.80%
Band of Misfits	192	67.20%
Fortress	184	64.40%
Hunting Grounds	175	61.25%
Rats	166	58.10%
Pillage	163	57.05%
Sage	162	56.70%
Cultist	159	55.65%
Death Cart	156	54.60%
Graverobber	151	52.85%
Ironmonger	143	50.05%
Knights	142	49.70%
Rebuild	136	47.60%
Hermit	134	46.90%
Feodum	110	38.50%
Wandering Minstrel	89	31.15%
Storeroom	86	30.10%
Squire	77	26.95%
Procession	72	25.20%
Rogue	72	25.20%
Scavenger	66	23.10%
Poor House	48	16.80%

Apparently, you essentially cannot have a fun Dark Ages-only set without Bandit Camp.

[Beyond Awesome]

There does seem to be some bias going on there. That's for sure. lol at KC.

[Donald X.]

I'm currently generating sets at dominionsetgenerator and tallying up how often each card comes up. After 1000 runs on the default settings with only the base set selected, here's how often each card came up, along with the percentage compared to full random (50% means the card only appears half as often as it would in full random).

Thanks for the data, it's interesting. I should say up front that I will tell Goko I think pure random is a mandatory option.

The biases you have charted are due to the algorithm weighting cards for filling particular roles. Chancellor fills no roles and so is low. Festival fills multiple roles - +2 Actions, +1 Buy, costs $5 - and so is high.

I think overall the difference between high and low is too extreme. If I were actually picking how to bias sets on a per-card basis, I could see wanting to see Chancellor less often than average (since I'd just replace it if I had the chance), and Festival more often than average (due to the roles it fills). The actual amounts are extreme though, and not all of the winners/losers are hits/duds. Spy for example is at 150% and yet is another card I'd replace, since it bogs down games without adding much; Squire meanwhile is a very flexible compelling card that only rates 27%. Some of the cards seem like they have to be mistakes, I mean really, how does Squire do so poorly via this algorithm? Is it because every game already has Bandit Camp? Most attacks are below average, and yet I have attacks in the game on purpose.

Anyway it doesn't seem ready for prime time, but regardless, for sure people should be able to play random too.

[Donald X.]

Update: Ted says they will go back to random in a day or two, and hold off on onigame's thing until they can have a full interface for it (understanding that random needs to still be supported).

[serakfalcon]

Yeah, poor Squire! I love squire, it is such a good card! Also KC, I don't see why its rated so bad either.
An easy way to deal with the generation vs. rating problem is simple:
Include the generation method used as information somewhere in the game, so people can read that before they join. It ideally shouldn't show exact cards, but rather how you generated the set. If you have it as a side button that expands or something, people can read that and decide for themselves if they want to join or not. After they join, they will be hit by the resign penalty if they don't like it.
If you don't make it obtrusive, novice players will ignore it (or be unaware that it exists), and skilled players can use it to get an idea if they want to play or not (letting them know beforehand what the biases are, so they can decide for themselves if they want to risk it or not). That's a more organic way, since, if you agree to play with certain preconditions, its your own fault if you lose. If you don't know enough about the game to care about the preconditions, it won't matter anyways. Everyone wins!

[onigame]

I'm currently generating sets at dominionsetgenerator and tallying up how often each card comes up. After 1000 runs on the default settings with only the base set selected, here's how often each card came up, along with the percentage compared to full random (50% means the card only appears half as often as it would in full random).

I want to say, thanks a lot for doing this!  It saves me a lot of the work of doing it myself.

[onigame]

• Wei-Hwa has a close relationship with Goko. If this weren't the case, we'd have seen the conversation that led to this implementation take place. It would have been a thread in the "Ideas Under Consideration" area of Goko's GetSatisfaction or something.

Confirmed.  I've been using my spare time in the last two months to go to their offices and work on things.  I spend about 2/3rds of my time there answering questions on rules interactions and creating test cases.  I'm not getting paid for any of the work, except for the fact that I own a stake in the company.

• He approached Goko about implementing his algorithm, not the other way around.

Confirmed, if you replace the word "implementing" with "integrating".  I've had this idea for months (if you dig around in the old Goko feedback site you can find the original post by me that was the genesis of the idea).  The implementation of the algorithm is all mine.  The source is readable for anyone who knows how Javascript works and is copyright by me and not Goko.  I provided the code and an API to Goko.  If you dig around Goko's visible client-side code you can see that my code is just loaded on the side (presumably invoked by something in theirs).

• He didn't tell Goko that some people had concerns about his algorithm.
• He didn't say, "You may use this algorithm, but make sure it's only an option, not mandatory."

This is a bit of a mix.  There were definitely people at Goko who liked it and wanted to put it in.  I said something to the effect of "Don't put it in everywhere, the people on the Professional Leaderboard will revolt."  I was asked, "What if we put it for casual play but not on the Professional Leaderboard then?"  I responded that that could probably work.

They put it on a test server and gave me a link to test it.  But (and this is the part I'm definitely not happy about) it went live on the main site before I got around to testing it.

Generally there's a pretty aggressive attitude towards implementing and pushing out new features there, which is not what I'm comfortable with in regards to game design.

The ironic thing is that I originally went to try to get more work on Race for the Galaxy done but it seems like all my time there is spent on Dominion now.  Sigh.

[onigame]

Yeah, poor Squire! I love squire, it is such a good card! Also KC, I don't see why its rated so bad either.

This is definitely a big surprise to me too, especially Squire.  Right now I suspect I probably just didn't type in the right flags for Squire.

[timchen]

Seems to me either the algorithm has some systematic error, or it thinks "fun" very differently to what I think.

Some highlight examples (just from the same set to be comparable):

Smithy is 1/5 as fun as Moneylender and less than half as fun as Wood Cutter. Throne Room is not much better.
Treasure Map is doubly as fun as Tactician.
Courtyard is 1/10 as fun as Harem. So is Minion (Probably fit thematically but... I don't think Duke should be almost as fun then.)
Border Village is the most interesting card among Hinterlands.

Ok, I understand that the algorithm is not ranking how fun each card is, but instead looking at their functionality on the board. But this method seems really biased upon those card who can fit multiple roles while those card may actually not be that interesting.

One improvement might be to normalize one card's appearing probability as a specific role by how many roles it can be? This way the board can still be interesting by having the needed roles, without overemphasizing those cards.

[onigame]

Now that it's been implemented on Goko as the only way to randomize non-Pro games, I'm livid. I'm just so, so angry. I just don't see why you've wrenched away the status quo in order to put in place your new, still highly experimental algorithm. Yeah, I could play only Pro-rated games, but a lot of times I just want a lower-stress experience or I want to include certain cards to practice with them. That's another effect of this change: randomly-generated casual games no longer 'match-up' with pro games. Practicing in casual games will no longer actually prepare me for what I may encounter in pro games (non-fun sets).

I'm very sorry that I did things that led to cause you emotional stress, and I'd like to try to make things better.

I really do apologize for this and I am willing to take some responsibility for allowing this to occur.  I'm sorry.  The scenario of wanting to play with the "Pro" generator as practice without it actually affecting your "Pro rating" is important, and it wasn't a use case that had occurred to me (or had ever been pointed out to me) in the short amount of time before it went live.  I honestly thought that the change would have just been on the beta site (you know, where tests are supposed to happen) for a while before it went to the live site.

I've been working furiously on a design revamp for game selection that would allow complete freedom in choice of randomizing options and been just typing and writing and putting off all my non-Dominion-related work for the last week.  I just posted a summary last night on the getsatisfaction forum in a new thread.  I know that saying that I've been working hard doesn't absolve me of the responsibility of making you feel bad.  So, again, I'm sorry.

Why are your opinions so much more important than mine?

I guess it would be hyperbole to say that I feel betrayed or duped. I feel a little like the rug was pulled out from under me. I paid for the service, with the assumption that I knew more or less what I was getting and that functionality wouldn't be taken away. I feel…unhappy. A lot of my desire to play on Goko has just sort of leaked away. They've already got my money, so no skin off their teeth, I guess…

I feel I have to address this even though you may not enjoy the answer.  My opinions are "more important" than yours because Goko has a lot more of my money than they have of yours (as far as I can tell; I don't actually know for sure that you also aren't one of the investors).  I invested in the company rather early, when it was more hope than actual implementation, and a lot of the work I'm putting in is motivated by making that investment be profitable.  I know that the company can't succeed without having customer loyalty, but also that it can't succeed without a wider, casual, audience.

We all have invisible biases towards our own experiences.  I enjoyed isotropic when it first came out, but soon got tired of it, because I felt I was just playing too many games where the kingdom was a one-strategy kingdom and that to do well I had to invest a lot of time to sense and execute tiny improvements in strategy and that having a good time seemed to be more about winning games rather than actually having fun with different game mechanics (which is the whole reason I enjoy playing Dominion).  I don't feel that a completely faithful clone of isotropic would ever be monetarily successful.

It's very hard treading the line between getting new users and satisfying old users.  In this case a confluence of events, some of which I was responsible for, tipped the ship too far in one direction.  I'm doing my best to course-correct.

It's never been about getting *my* code into Goko.  It's been about doing what I feel is important to make a niche game appeal to a wider audience.  I realize that it's not very honorable to come out and say "You accuse me of vanity, but the only sin I'm actually committing is greed," but I hope at least you understand.

To put some stuff in perspective -- the integration and push all happened in the space of a week.  It's actually not hard to roll that aspect back, and I'm in full support of doing this.  Again, I'm sorry.

[onigame]

Seems to me either the algorithm has some systematic error, or it thinks "fun" very differently to what I think.

I am leaning more towards the "systematic error" side of things.  Addressing your specific examples:

Smithy is 1/5 as fun as Moneylender and less than half as fun as Wood Cutter. Throne Room is not much better.

Smithy: I noticed last week that my weighting on "Carddrawers" (which Smithy gets a high score in) was not high enough -- the intent is that the generator (on default settings) favors Carddrawers until there's about +6 Cards total or so, then discourages them.  The problem is that the discouraging factor is too large.  I made a change yesterday afternoon to try to correct this.  I don't know if LastFootnote's stats were before or after this change.  I agree, Smithy needs to be better.

Throne Room: Gets a boost from a "Brancher" attribute (if you combine it with a +1 Action you get a splitter) but a threshhold-based penalty from a "Variance" attribute (if you draw it without any Actions it's useless).  I think right now the "Variance" penalty threshhold is too low, and that's almost certainly because of my own personal biases -- I really hate it when I draw a TR with no actions, and so I almost never buy them.  I will raise that threshhold to compensate.

Treasure Map is doubly as fun as Tactician.

Tactician was showing up waaay too much in early versions of the generator (almost every Seaside game).  I eventually realized that it was because it drew cards, gave you +Buys, and gave you an extra action -- so it does everything, why would you ever not want it in the game?  But the +Buys is kinda only half of the time since it's often more like you're pushing your Buy to the next turn.  Anyway I changed some of the Tactician attributes and I probably nerfed it too much.

Courtyard is 1/10 as fun as Harem. So is Minion (Probably fit thematically but... I don't think Duke should be almost as fun then.)

Courtyard is getting hit by the Carddrawer bug I mentioned above.  Minion I dunno, I'll look into that.  Harem and Duke got a bonus because Donald made a strong argument that including Victory cards in the set changes the focus of the game significantly and makes things more interesting.   While I don't doubt that for Gardens, I'm not so sure Harem has that property.

Border Village is the most interesting card among Hinterlands.

Huh, I actually don't see a problem with that (although the magnitude of the favoritism may be a bit high, I admit).  Sorry, we might just disagree on this statement. 

[dondon151]

I don't actually think that biases on any cards need to be strong in order to influence the creation of "interesting" kingdoms. You can probably get away with very slight biases such that dud cards (like Chancellor) appear just a little less often than normal and good cards (like BV) appear just a little more often than normal, and you'll find a middle ground that most people can probably agree on. I think that most Dominion sets are fairly interesting anyway, even the boring ones, because sometimes even little decisions matter, and it's those decisions that tend to draw out the difference between a L30 and a L40 player.