Print

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

[Elanchana]

Hey guys. Ela here with some more faux-interviews.

There are risks involved in every Dominion play. But how much risk should a player take? On the one hand, you want to ensure that you don't crash and burn, but on the other hand, high risk can lead to high reward. If you can go for a strategy that is guaranteed to give you a small reward every time, do you ignore it and try one that can make you win much faster but is more likely to fail?

Simplified example scenarios with Tortoise (low risk player) and Hare (high risk player)

Game 1 - opening turns:
Tortoise opens by buying two low-powered cards that can work with almost any combination of Coppers and Estates, to be prepared for most orders in the second shuffle.
Hare opens by buying two cards with the intent of them being drawn together on the second shuffle, to kick start a high-powered strategy.

Game 2 - Vill/Smith mirror:
Tortoise prioritizes buying Villages to ensure always having actions available for terminals, even though it gets through the deck slower.
Hare prioritizes buying Smithies to get more draw and cycle faster, even though the cards might be drawn dead.

Obvi, there are tons more situations than these and they get much more complicated. But which player would you side with? Are you usually a low or high risk player? What factors influence how much risk you take? Discuss.

[Asper]

It depends on the relation between payoff and risk for me. And on my mood. Generally, i try to apply what i think are statistic probabilities.

Let's take Coppersmith, which is not that awesome most of the time. Still i might buy a Coppersmith in the hopes of hitting 8$ very early to buy a Prince (to play on Coppersmith). In fact, i might buy a Copper with my second opening buy to make hitting 8$ more likely - so i guess that's "playing it save while playing risky"? If the payoff isn't that great even if i succeed (for example, there are no reasonable picks at 7+), i'd rather skip it entirely.

[sudgy]

I remember some article somewhere saying that the less likely you are to win, the more you should take risks.

Edit: here.

[Eevee]

Ignoring edge cases such than playing someone much lower ranked than yourself (you want to play safer) or playing Stef (play to get lucky), deviating from what you perceive to be the middle ground seems suboptimal by definition. Sure, some situations call for hoping to get extreme shuffle luck (being far behind is an obvious example), and some situations call for playing super safe, but I always do whatever I think helps me win, and going into a game that  almost always means playing my perceived middle ground, not any looser or safer.

[qmech]

One of the factors that affect this decision in the opening is how important it is to buy certain cards during your second shuffle.  Some of the most aggressive openings for me are Swindler/Swindler, Ambassador/Ambassador and Steward/Steward.  The Ambassadors and Stewards declare an intention to get very thin and accept that your initial build-up is going to be slow.  The Swindlers accept the chance of collision for the opportunity to wreak havoc with your opponents purchases (and turn some Coppers into Curses).  In all three cases you're trading off chances to buy useful cards against the opportunity to clean yourself and/or junk opponents.  If you want to contest Labs, for instance, you'd often rather have a Silver over the second terminal.

[Eevee]

The double terminal openings with steward and ambassador are so good because you weren't going to buy anything the turn your opening buys collided even if you got a silver anyways.

[theblankman]

How much risk do you take?  Depends on the reward, in a way that can be quantified, though not easily for Dominion.  People have already mentioned specific cases like, "Take more risks when you're far behind."  In general we're talking about a concept out of statistics called "expected value."  I'll illustrate it with a WAY oversimplified example:

Take a potentially high-reward opening like, say, turn 1 Nomad Camp.  IIRC the chance of hitting $5 on turn 2 is around 40%.  Suppose the reward is turn 2 Mountebank or Witch or some other power $5, and a significant lead compared to an opponent who takes a "safe" opening like Silver/Silver.  Then the next questions are: what's your chance of winning the game if your gamble pays off?  If it backfires?  Expected value is a weighted average of the outcome of both strategies. 

Of course that T2 Mountebank or whatever is more or less dominant depending on the rest of the board, so I'll pick some sets of numbers out of a hat:
1) If you hit $5 on T2 you win 90% of the time, if you miss it's not that bad and you still can win 40%.  E[Nomad Camp] = (.9 * .4) + (.4 * .6) = .6 = 60%.  Better than even, so go ahead and risk the Nomad Camp. 
2) If you hit $5 on T2 you win 90% of the time, but if you miss it's a disaster and you win only 10%.  E[NC] = (.9 * .4) + (.1 * .6) = .42 = 42%.  Don't take the Nomad Camp here. 
3) If you hit $5 on T2 you win 60% of the time, if you miss 50%.  E[NC] = (.6 * .4) + (.5 * .6) = .54 = 54%.  Nomad Camp is a little better than even here, worth considering. 
4) If you hit $5 on T2 you win 60% of the time, if you miss only 40%.  E[NC] = (.6 * .4) + (.4 * .6) = .48 = 48%.  Nomad Camp is barely worse than even here, still worth considering. 

So that's plenty of cases even ignoring things like whether you're first player, what your opponent did, and of course the rest of the board and the rest of the shuffles after your first.  But the general idea is to pick the level of risk that maximizes your overall chance to win.  If you think getting that $5-cost on turn 2 will give you a huge advantage, then maybe the risk of putting an extra terminal like Nomad Camp in your deck is worth it. 

Of course extending this math to full games of Dominion is prohibitively time-consuming, so we don't do it.  That's what sims are for, at least for relatively simple strategies.  But the upshot is that we don't have a right answer to "Should I play safe or risky?"  The real question is, "How much risk should I accept right now?"

[ben_king]

I remember some article somewhere saying that the less likely you are to win, the more you should take risks.

Edit: here.

Yes, it depends on the opponent and how you're doing in the game.  There's a great article about this topic here (it's about basketball, but it applies equally well to any zero-sum game like Dominion): https://gravityandlevity.wordpress.com/2010/12/21/risk-is-the-ally-of-the-underdog/

If you think you're better than your opponent or ahead in the game, then adopting a low-risk strategy is best.  If you're way overmatched by your opponent or way behind, then you might want to think about Treasure Map or Black Market even if there's something safer and less swingy on the board.  If you both play it safe, you'll probably lose.  If you roll the dice, there's a chance you hit the jackpot and win when you didn't really "deserve" it.

[Mic Qsenoch]

I think most estimations of "risk", "swinginess" etc. are based on superficial things like: is it an attack that makes me feel bad, is it "craaazy" like King's Court or whatever. I don't trust them. They usually emphasize highly visible effects, and tend to neglect more subtle ones. Everyone notices when their terminals collide, but for some reason aren't as worried about a truly terrible thing: playing fewer terminals on average because you didn't buy enough. Also people think of a card as risky or swingy just because a bunch of people say it is on the forums.

I think you're better off focusing on and getting a feel for expected values, as others have suggested.

I really hate the advice to adopt uncertain/"swingy" strategies against stronger opponents or as 2nd player. The shuffles alone already provide sufficient randomness for weaker players to beat stronger players (or for P2 to beat P1). Picking a "risky" option just to do something different from your opponent is a sure-fire way to lower your chance of winning (very rarely will it increase it). The next best choice is often significantly worse than the best one. I would always suggest to pick what you think the strongest play on the board is, do your best and try to learn how to play it better. Playing mirrors against stronger players is a great way to reveal what mistakes you might be making.

I think deviations from strong general plans or strategies should happen when you get favorable/unfavorable shuffles for a certain strategy, not because you psych yourself out before the game even starts.

[Titandrake]

I don't think Stef's article said anything about playing a riskier strategy against a stronger player. It says to play a riskier strategy if you're 2nd player, or if you're losing. Usually, I don't think it's worth taking the risk in the opening, you start playing towards getting good luck when you need it to win.

Semi off-topic, but if you're behind your opponent, you should figure out how you can catch up assuming decent luck, and then execute it and hope it happens. Often, this means doing something different from your opponent. For example, if they're hammering the VP, you should build your engine more, instead of dipping into VP yourself. Presumably their deck is stronger at this point, so if you both get VP, their deck will continue to stay stronger, because neither of you are buying actions to improve your deck. In fact, when good engine players have a lead, they specifically buy VP in a way such that if you try to build more, you'll run out of time to catch up, and if you contest VP, your deck will choke first.

More succinctly, if you have the same game plan as your opponent, and your opponent has the better deck for that game plan, then sticking to that plan will just make you lose, and you should only do so if no other plan has remotely as good a chance to succeed.

[WanderingWinder]

I think most estimations of "risk", "swinginess" etc. are based on superficial things like: is it an attack that makes me feel bad, is it "craaazy" like King's Court or whatever. I don't trust them. They usually emphasize highly visible effects, and tend to neglect more subtle ones. Everyone notices when their terminals collide, but for some reason aren't as worried about a truly terrible thing: playing fewer terminals on average because you didn't buy enough. Also people think of a card as risky or swingy just because a bunch of people say it is on the forums.

I think you're better off focusing on and getting a feel for expected values, as others have suggested.

I really hate the advice to adopt uncertain/"swingy" strategies against stronger opponents or as 2nd player. The shuffles alone already provide sufficient randomness for weaker players to beat stronger players (or for P2 to beat P1). Picking a "risky" option just to do something different from your opponent is a sure-fire way to lower your chance of winning (very rarely will it increase it). The next best choice is often significantly worse than the best one. I would always suggest to pick what you think the strongest play on the board is, do your best and try to learn how to play it better. Playing mirrors against stronger players is a great way to reveal what mistakes you might be making.

I think deviations from strong general plans or strategies should happen when you get favorable/unfavorable shuffles for a certain strategy, not because you psych yourself out before the game even starts.

I strongly agree with this.





People make way too much about things like terminals colliding or not, or taking different strategies, or something like Treasure Map. (Treasure map is most often bad even if you get an early collision, by the way, for those of you who advocate this kind of thing). You definitely shouldn't play differently vs a 'strong' opponent than against a weak one (maybe you give them credit for polishing you off and resign sooner, but that's about it). It's actually important to realize that just having treasures line up differently or having a card miss a shuffle, or even something like getting your engine cards at the top of the shuffle rather than the bottom while you're building the engine up has at least as big of an effect and is just as high variance. Similarly, people get afraid of over-terminalling (cough cough Adam), because they see that time they collide and think of that as really bad. But they don't think of it as much when they don't collide, to realize they're very far ahead. Obviously there's a trade-off, but the way to assess it is (how often it happens)*(how good it is when it does), where 'how good it is' is based on what that does to your win%. Of course, that's hard to calculate, so you go on intuition.

Where it really comes in is in situations where one player has already established a pretty big lead in the game (position-wise, not necessarily VP). In this case, you as the player behind are going to lose unless you get lucky, so you play to maximize your chances to win if you do. Very often, this means building for quite a bit longer, possibly dipping into a different VP source, hoping that they choke at some point (hoping an opponent fails to draw their potion in an Alchemist stack is a very prototypical example of this). Conversely, you need to sometimes make a crazy stab for ending the game RIGHT NOW, because over the long haul, they will be able to both build AND have VP over you. It feels bad to do this, because "if they have any normal reasonable hand, I'm just going to lose next turn." But you need to realize, that's going to happen anyway. Pick your spot when they're most likely to dud out (and/or when it's best for you if they do), and just roll the dice. Conversely, if you're way ahead, balance having enough points/pile control/deck strength to minimize the chances of them being able to take any of these paths. I find it's actually harder to do the ahead thing a lot of the time, though you still usually win anyway, because by definition if you weren't super far ahead, you should basically just be playing 'normal'.

[silverspawn]

I think I agree with both WW and MicQ here. The thing about "if you're behind you need to take risks" is that, while it is not an inherently bad advice, it's incredibly overestimated/used. If you want to become a better player, there are a million things that you should care about first, and trying stuff like this probably makes more people lose than win overall.

On the top of my head, I remember exactly one game where I was far behind, made a high variety move, ended up winning, and after the game felt like it really was because I took that risk.

[GeoLib]

I think I agree with both WW and MicQ here. The thing about "if you're behind you need to take risks" is that, while it is not an inherently bad advice, it's incredibly overestimated/used. If you want to become a better player, there are a million things that you should care about first, and trying stuff like this probably makes more people lose than win overall.

On the top of my head, I remember exactly one game where I was far behind, made a high variety move, ended up winning, and after the game felt like it really was because I took that risk.

I think this comes from the fact that people are really pretty bad at estimating expected value, especially in something as complicated as dominion. That said, a specific example I can think of for this type of analysis is when to break PPR, which I think is more than the 1 in 1000 type deal you imply here.

[silverspawn]

I think I agree with both WW and MicQ here. The thing about "if you're behind you need to take risks" is that, while it is not an inherently bad advice, it's incredibly overestimated/used. If you want to become a better player, there are a million things that you should care about first, and trying stuff like this probably makes more people lose than win overall.

On the top of my head, I remember exactly one game where I was far behind, made a high variety move, ended up winning, and after the game felt like it really was because I took that risk.

I think this comes from the fact that people are really pretty bad at estimating expected value, especially in something as complicated as dominion. That said, a specific example I can think of for this type of analysis is when to break PPR, which I think is more than the 1 in 1000 type deal you imply here.

I didn't consider breaking PPR as a case of applying this concept, because the decision is too simple. You buy the province, the chance of your opponent not hitting 8$ and you hitting 8$ again is bigger than the chance of you winning via buying duchies, so you go for the province. To deserve the term "high variance strategy", it needs to be something more complex.

That said, if you do consider breaking PPR an example of this concept, and I can see why you would, then yea it becomes much more relevant. I break PPR all the time.

[Awaclus]

I didn't consider breaking PPR as a case of applying this concept, because the decision is too simple. You buy the province, the chance of your opponent not hitting 8$ and you hitting 8$ again is bigger than the chance of you winning via buying duchies, so you go for the province. To deserve the term "high variance strategy", it needs to be something more complex.

It isn't a high variance strategy, but it is a high variance tactic. Generally, I think that the advantages of doing something risky for a potential reward are mostly found in tactical plays.

[Seprix]

You have to realize that faster decks will get Greened quicker, and will slow while the slow deck will not get Greened as fast. You simply develop longer, then play a game of catch-up with the guy who bought a Province first. It depends on the board. There are some games where you want to get an early lead but still want to develop, then there are games with Counterfeit and Venture where you want to be the first to buy to keep up the pressure. In the end, it really depends on the board, but generally you want to develop while not falling more than 1 Province behind if you can help it.

[HiveMindEmulator]

I think this comes from the fact that people are really pretty bad at estimating expected value, especially in something as complicated as dominion. That said, a specific example I can think of for this type of analysis is when to break PPR, which I think is more than the 1 in 1000 type deal you imply here.

Yeah, I think in general, people are really bad at computing expected value. Like 99% of people can't do it well. In a room of 30 people, I'd expect only 5 could.

Seriously though. That's kinda how it works...