Print

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

[jsh357]

That hamlet is second best village.

Hamlet is the 4th ranked village on the list, and it's better than all the ones below it. Cheapness is a very valuable asset for villages.

I disagree.  Having to discard cards is a pretty big disadvantage a lot of the time, especially if your engine relies on cards that draw 2 cards.  Hamlet is a killer card for disappearing hand engines, and it of course comes through in a pinch, but generally I'd prefer a $4 reliable village in most decks.

[Awaclus]

especially if your engine relies on cards that draw 2 cards.

It probably shouldn't.

[jsh357]

especially if your engine relies on cards that draw 2 cards.

It probably shouldn't.

No, but plenty of times it does.  With trashing, 2 card draw is acceptable.

However, even with better draw options, discarding from Hamlet can hurt, even if it's just a Copper. 

[Beyond Awesome]

That hamlet is second best village.

Hamlet is the 4th ranked village on the list, and it's better than all the ones below it. Cheapness is a very valuable asset for villages.

I disagree.  Having to discard cards is a pretty big disadvantage a lot of the time, especially if your engine relies on cards that draw 2 cards.  Hamlet is a killer card for disappearing hand engines, and it of course comes through in a pinch, but generally I'd prefer a $4 reliable village in most decks.

This is so true.

[HiveMindEmulator]

I disagree.  Having to discard cards is a pretty big disadvantage a lot of the time, especially if your engine relies on cards that draw 2 cards.  Hamlet is a killer card for disappearing hand engines, and it of course comes through in a pinch, but generally I'd prefer a $4 reliable village in most decks.

Of course you'd rather *have* a $4 village. That's why they cost $4 instead of $2.  I'd also rather *have* Farming/Mining Village than Village, but Village is still better. The ease with which you can *get* Hamlets (especially since they give +buy) is a big part of what makes them good. Add that to the situations in which the discarding is a benefit, and you have an overall stronger card. That's why it's higher on the list. It did not gain 16 percentage points over vanilla Village just because of random $2 hands in non-engines.

[Holger]

Everybody loves the card rankings that Qvist puts out, but I've always secretly wondered how these lists compare to the way the top players really play the cards.

So I've collected over 90,000 game logs of the top 100 Dominion players on the Isotrophish rankings (2-player, pro rating, no bots) and analyzed them to determine how the best players in the world approach the cards. 

Thanks for the great work!

I'm quite surprised that Rebuild was only gained in 74% of games.
Did you use all the game logs of the current top 100 players, or only those that were played by a player while they were in the top 100 (which would probably be much more work...)? And did you use a time cut-off (e.g. only games played in the last 12 months) or did you go back till the beginning of Goko?
Either could explain Rebuild's low gain rate (since it was massively underestimated when new), and the first possibility might also explain the high gain rate of the very worst cards.

[liopoil]

74% feels right. 26% of boards have a good engine, that's all.

[ben_king]

Thanks for the great work!

I'm quite surprised that Rebuild was only gained in 74% of games.
Did you use all the game logs of the current top 100 players, or only those that were played by a player while they were in the top 100 (which would probably be much more work...)? And did you use a time cut-off (e.g. only games played in the last 12 months) or did you go back till the beginning of Goko?
Either could explain Rebuild's low gain rate (since it was massively underestimated when new), and the first possibility might also explain the high gain rate of the very worst cards.

Thanks.  The data set I used had all games of current top 100 players -- I don't think there's a way on gokosalvager to figure out what their rank was at the time the game was played.  And it goes as far back as the gokosalvager logs go.  I'm not sure if that's all the way back to the start of Goko or not (I would guess that it's not).

[Holger]

Thanks for the great work!

I'm quite surprised that Rebuild was only gained in 74% of games.
Did you use all the game logs of the current top 100 players, or only those that were played by a player while they were in the top 100 (which would probably be much more work...)? And did you use a time cut-off (e.g. only games played in the last 12 months) or did you go back till the beginning of Goko?
Either could explain Rebuild's low gain rate (since it was massively underestimated when new), and the first possibility might also explain the high gain rate of the very worst cards.

Thanks.  The data set I used had all games of current top 100 players -- I don't think there's a way on gokosalvager to figure out what their rank was at the time the game was played.  And it goes as far back as the gokosalvager logs go.  I'm not sure if that's all the way back to the start of Goko or not (I would guess that it's not).

I think Goko's first months were not covered, the logs start around May, 2013. Since the strength of Rebuild (and to a lesser degree, other DA cards) only became "common knowledge" in late 2013, it would be interesting to only look at logs from 2014 onwards. Yes, accurately accounting for the top 100's rating changes is probably impossible, but you might want to discard e.g. the older half of the games each top 100 player played, to make it likely they were similarly strong when they played the games as they are now. But I don't know how much work that would be...

[SCSN]

Since the strength of Rebuild (and to a lesser degree, other DA cards) only became "common knowledge" in late 2013

I don't think that's very relevant. The thing about good players (who should be overrepresented in the top 100) is that they realise that whatever manages to pass for "common knowledge" always lies somewhere between a disfigured caricature of the truth and misleading nonsense, so they just opt to use their own brain instead.

It took me personally about one game to figure out that Rebuild was powerful, and I left iso as a level 33, which is fairly mediocre. These are my Rebuild gainrates over different periods (I started out on Goko with Base+DA only, and gradually added the other expansions during the next two months):

03/15/2013 - 06/30/2013: 67 out of 76 pro games, or 88.2% of the time.
07/01/2013 - 12/31/2013: 108 out of 129 pro games, or 83.7% of the time.
01/01/2014 - 01/29/2015: 42 out of 57 pro games, or 73.3% of the time.

Before looking at my own data I thought that 74.4% was too low as well, but my own play convinces me enough to not have a strong opinion on the matter, and if a Dominion Oracle emerged who could resolve such issues and someone offered me to place a lot of money on "80% is closer to correct than 75%", I would politely decline.

The larger point, though, is that ~75% is still an insane amount for what is basically a single card strategy, to the point that there's little qualitative difference between "this card makes other cards irrelevant on 85% of boards" vs. "it does so only on 75% of them". If you look at the top 10 of this list--all cards gained more frequently than Rebuild--you'll see that none of these dictate a particular strategy on their own, not even Governor. In fact, many of them open up strategic possibilities, they are expansive rather than constrictive; whereas in the vast majority of Rebuild games the path is paved and your sole job consists in not diverting too much from the racing line.

[c4master]

I still think that 75% is a lot and maybe even too much. On games with Shelters, Rebuild is still a strategy, but not as good as usual. Thus, most of these board could offer other strategies. Then, there are knights who like to eat up Rebuilds and Duchies. Next on the list of cards disabling Rebuild could be strong junking attacks like Mountebank or Cultist.

[Mic Qsenoch]

On games with Shelters, Rebuild is still a strategy, but not as good as usual.

SCSN! They keep saying it, they just won't stop saying it!

Junkers don't really disable Rebuild, though I guess Cultist specifically might hurt it enough and be able to get Provinces fast enough to beat it. I would guess in most Cursing games you just want Rebuild after you get your junker.

[SCSN]

On games with Shelters, Rebuild is still a strategy, but not as good as usual.

SCSN! They keep saying it, they just won't stop saying it!

"common knowledge" always lies somewhere between a disfigured caricature of the truth and misleading nonsense

Well, that took exactly one post for someone to chime in and illustrate the point!

Junkers don't really disable Rebuild, though I guess Cultist specifically might hurt it enough and be able to get Provinces fast enough to beat it. I would guess in most Cursing games you just want Rebuild after you get your junker.

Pure Rebuild (only buys Rebuilds, Yellows and Greens) beats Cultist-BM by something retardedly small like 51.5-48.5 over 200k games. Add any kind of non-trivial support (a powerhouse like Chancellor would certainly do) and it stops being close.

[AdamH]

Pure Rebuild (only buys Rebuilds, Yellows and Greens) beats Cultist-BM by something small like 51.5-48.5 over 200k games. Add any kind of non-trivial support (a powerhouse like Chancellor would certainly do) and it stops being close.

What about adding Cultist itself? I feel like the relevant number here is how much Rebuild+Cultist beats Cultist-only and Rebuild-only by, and then how do you prioritize your buys?

I remember a lot of this discussion about Jack: "Jack is OP it beats Mountebank+BM!" but Jack+Mountebank is better and that's something you would actually play if this statistic is relevant. I keep holding out hope that Rebuild is much more interactive than we give it credit for (like Jack turned out to be), but if Rebuild-only beats Rebuild+Cultist (or it beats Rebuild+X for 100+ kingdom cards) then that's a dim hope.

And I feel like that's not a discussion that's ever been had. Everybody whines about how un-interactive Rebuild is but these numbers haven't been shown, and until they are I'm not convinced. I will hold on to this hope until it's properly dead and mutilated.

[SCSN]

What about adding Cultist itself? I feel like the relevant number here is how much Rebuild+Cultist beats Cultist-only and Rebuild-only by, and then how do you prioritize your buys?

Rebuild+Cultist loses to both.

or it beats Rebuild+X for 100+ kingdom cards

Almost every kingdom contains an X for which Rebuild+X beats pure Rebuild (all terminal silvers except possibly Fortune Teller, for one; things like Scheme and Warehouse; Horse Traders, Jack, Baron), but the same holds for Big Money, and as with Big Money, you're still only playing a slightly stronger variant of essentially the same strategy. The only cards that bring some sort of qualitative change to a Rebuild strategy are Rogue, Graverobber, Duke, Tunnel and to some extend Feodum, and even these play pretty straightforward once you've figured them out.

And I feel like that's not a discussion that's ever been had. Everybody whines about how un-interactive Rebuild is but these numbers haven't been shown, and until they are I'm not convinced. I will hold on to this hope until it's properly dead and mutilated.

I've had these discussions in the early days, mostly with AI in private during playtesting for our article, then more (some in public I think) during the early Dominiate simulations and then even more within myself while testing variously competing things in the amazingly performing Dominulator.

The mirror article has been published and was later updated (as recently as 4.5 months ago) to include some qualitative guidelines based on the simulations. I've referenced other important results in various posts from mid-late 2013. Crude scripts for a bunch of Rebuild+X strategies are available in Dominiate (linked on gokosalvager.com), some more polished ones are included in Dominulator (available on GitHub).

For some reason almost no one seemed to care for the fine details, and various brands of fashionable nonsense--often asserted with proud conviction--continued to prosper. To my personal lamentation (and Mic's laughter) all my attempts to counter this have been an utter failure. I can certainly appreciate the funny side of it, but not without simultaneously experiencing this sense of bitterness that has replaced some of my former enthusiasm.

[Titandrake]

Maybe back when Iso was up, I would have been interested in the simulations for Rebuild + X. When I had access to all the cards, I played a lot of solo games to mess with card combinations, and the simulator was a nice shortcut for that. However, I've now realized that I prefer playing and learning Dominion in a very loose way. Try dumb strategies, find out which ones are less dumb. And most importantly do this against somebody else so that I get properly punished for doing something especially dumb.

[TheExpressicist]

I'm not quite picking up on the context of this; would you mind clarifying, SCSN? What is it that you have been saying, and what is the fashionable nonsense?

[GeoLib]

I'm not quite picking up on the context of this; would you mind clarifying, SCSN? What is it that you have been saying, and what is the fashionable nonsense?

SCSN has done a lot of simulation to figure out how rebuild works with various modifications and people keep saying things like "rebuild is bad with shelters" despite the fact that he demonstrated that this was not true.

[liopoil]

Okay, so, question: Rebuild is still slightly worse with shelters than without shelters, right?

[eHalcyon]

Okay, so, question: Rebuild is still slightly worse with shelters than without shelters, right?

SCSN's simulations showed that there was pretty much no difference.

Here's the article.  Shelters is just a small note at the end though.

[AdamH]

What about adding Cultist itself? I feel like the relevant number here is how much Rebuild+Cultist beats Cultist-only and Rebuild-only by, and then how do you prioritize your buys?

Rebuild+Cultist loses to both.

Well that's just sad. Judging from the amount of simulation work it appears you've done I think I believe that you programmed a competent Rebuild+Cultist bot, which means Rebuild just isn't a good card. It's so rare to see it really interact with other cards in interesting ways 

I don't remember hearing anything about your simulation results or anything like that. Re-reading the article on mirrors it seems that some stuff was edited in. I was under the impression that Rebuild mirrors were pretty well studied and uncontested Rebuild was a little bit lacking, but maybe that's not the case anymore. I'd certainly like to know what's out there.

[c4master]

Okay, so, question: Rebuild is still slightly worse with shelters than without shelters, right?

SCSN's simulations showed that there was pretty much no difference.

Here's the article.  Shelters is just a small note at the end though.

Ok, thx. I didn'tre-read this article for quite some time.

I still wonder how shelters don't affect Rebuild. Or more likely: How they don't support other strategies as well. Rebuild then needs 2-3 extra buys which makes up for at least 2 extra turns, that's what I thought. Actually, trashing the green shelter will give you +1 Card maybe enabling a Duchy. And buying a Duchy enables trashing your hovel, so maybe that's why Rebuild is just as fast as usual. Still, I wonder how other strategies are not supported better than Rebuild.

I'd also like to know how effective trashing attacks (Swindler, Knights, Saboteur etc.) can be against Rebuild.

[Holger]

Okay, so, question: Rebuild is still slightly worse with shelters than without shelters, right?

SCSN's simulations showed that there was pretty much no difference.

Here's the article.  Shelters is just a small note at the end though.

Ok, thx. I didn'tre-read this article for quite some time.

I still wonder how shelters don't affect Rebuild. Or more likely: How they don't support other strategies as well. Rebuild then needs 2-3 extra buys which makes up for at least 2 extra turns, that's what I thought. Actually, trashing the green shelter will give you +1 Card maybe enabling a Duchy. And buying a Duchy enables trashing your hovel, so maybe that's why Rebuild is just as fast as usual. Still, I wonder how other strategies are not supported better than Rebuild.

Buying a single early Estate (or Duchy) is enough in shelter games, you can rebuild OE twice before it becomes a Duchy. If you should hit $2 during your first 4-5 turns, you don't have to waste a single good buy. The main disadvantage is having 1 (initially 2) useless dead cards, but that's not so bad.

[WanderingWinder]

The thing is that engines vs Rebuild still isn't terribly explored. They don't tend to simulate well, and they often need several different kingdom cards to reach full potential, so it's hard to necessarily translate from one board to the next. Make no mistake, it takes a fairly potent engine to overcome the VP-destruction powers of Rebuild, but it can happen.

Of note, the strongest things which help an engine vs Rebuild are probably Colony and Vineyard. Neither of them help other money strategies, though.