First of all: this response made the whole article worthwhile. I sense from your response that you enjoyed the article and found it thought-provoking.
I think the article will remain a touchstone for future discussions on many issues. I foresee myself referring to it many times in the future for supporting any given contention that might arise in the course of normal debate (complexity of vintage play being just one example, but another example the importance of the player - that is, given the sheer quantity of options, the ASL of the player means that you really can't know the win ratio of a particular matchup without holding skill constant - something that is a highly variable quantity).
I think this article causes me to ask the same question that I always asked in the past in response to any analysis of a Long variant - is this deck impossible to play perfectly, and are many of the decisions just a roll of the die? For instance, Steve asks the readers what their decisions would be at critical junctures. I have no idea if I can even identify *one* correct play in there (apart from anything obvious). The way I tend to play, for instance, is to always leave myself with outs whenever possible in case my current investment doesn't pay off. In the example of the first game, I would side with Chapin and Grim Tutor for a Mana Crypt over Lotus, and I don't think I would touch Brainstorm before the Windfall as i'm not accustomed to playing Brainstorms that aggressively. However, I do recognize that such rather conservative play might not be best.
I agree with Zvi: there is only one "optimal" play in the sense that one play is better than all others. As we split hairs seemingly equal plays will become inferior or superior depending upon the criteria we use to judge them. But as I said: many decisions will get you to the same result: winning the game.
I wonder then, if anyone thinks it is within the realm of possibilities in the immediate future for a Long player to be able to determine what the *right* play is in such situations? It seems like this should be coming down to a combination of actually calculating out what the statistically best play would be, and doing everything you can to extract information from the opponent as far as what he's holding or what archetype he's playing. This view is opposite from Steve's contention in the article:
In my view, the task of a good Vintage player isn't to be able to identify the optimal play among several excellent plays. The task of a good Vintage player is to be able to think about why a given line of play is good and why it might not be so good. In each case in which I thought there were genuinely close and debatable plays, I presented the reasons for and against those plays and made the decision that I felt was weighted, however slightly, toward the play I decided on.
I feel that this argument doesn't make sense. I feel that understanding the reasons behind the plays is naturally an important component, but I it doesn't stop there. The task of a good vintage player, a truly good player, is precisely try to identify the optimal play. As I understand it this is precisely what separated a player like Finkel from the rest. Is there a budding Finkel in T1's future?
I don't think I really articulated what I was trying to say very well.
What I was trying to say is this: What matters is understanding so that you make the better plays more often in the long run. Sometimes, in the short run, you can make a suboptimal play and win. However, to win most consistently you need to make the right play most of the time. An inadequate grasp of the pros and cons of a particular decision will lead you to make second or third or fourth best decisions some of the time, but you'll be making better decision - in the aggregate - and in the long run, if you have a more solid grasp of the situation.
Now, here's the trick: I don't believe that we, as human beings, are capable of making the optimal decisions within a reasonable time.
This is why I once stated that I don't think it is humanly possible to play meandeck Tendrils, apart from flaws in the deck. When I took all the time in the world to make decisions with that deck, I rarely lost games.
You know how I said that there were hundreds of permutations with the game one Grim Long hand? And there were. There were literally 7! options to Meandeck Tendrils. Almost every card could be played in any order.
It isn't humanly possible to make the optimal play. That's why I've joked that DEEP BLUE would play meandeck Tendrils and win tournaments (a point that Rich Shay vehemently denies).
In an earlier draft of this article I had extensive tree diagrams that actually listed some 30+ permutations of play. Pat Chapin criticized me for doing that.
Here was an exchange I had with Pat:
Pat: I would Grim Tutor for Crypt there.
Me: That's wrong, imo. I think the only debatable question is whether you Brainstorm and, if so, when.
Pat: Prove me wrong.
Me: Ok; I'll pretend I got Crypt there and goldfish the hand 30 times and see what happens
Pat: You'd have to randomly goldifsh the MDG hand that many times as well because it's hand will be different alot and you can't just keep using the hand that MDG has or you'll never know what the truly "right" play is - just what the right play is for MDG given hand.
Me: that's true.
The point Pat made is that it is inhumanely possible to really assess what the correct play is. You'd have to go through that scenario with both decks a statistically significant number of times and test each permutation. You could do that, but then that scenario - exactly as it is, would never arise again. So you'd pretty much be wasting your time. Esp. for a format like Vintage.
He's right.
For me, the only way we can get at the correct answer (since the math is just WAY too hard to run) is to understand the pros and cons for a given line of play and then make a reasonable decision.
I agree with you, in principle, that the task of a great player SHOULd be to find the optimal play every time. I just think, as a practical matter, its infeasible unless you have an IQ of 1,000,000.
I also told Pat that I thought that Black Lotus was almost certaintly the right play given my experience with Grim Long.
He said that if you polled the top 5 magic players in the world: Kenji, etc, that 3 of 5 would go for Mana Crypt there.
I think it goes to show that the best players don't actually rely on raw computer power and logical/mathmatical analysis of any given line of play so much as intuition.
In fact, of interest to you Peter, this conversation veered into Chess. I said that I thought Kasporov was capable of looking 30 moves ahead, etc. Pat corrected me and said that Kasporov could probably look entire GAMES ahead - in fact, he could probably mentally diagram hundreds of moves ahead and hundreds of permutations, but he wouldn't because he is playing under time constraints and pattern recognition will get you most of the way.
Steve continues:
In Vintage, there will be very clearly obvious reasons to pursue a given line of play. What separates the wheat from the chaff and the players who outperform the rest is an understanding of the more subtle risks and benefits of any particular line of play. This is why I say that understanding is more important than identifying the optimal line of play. Understanding will lead to better decisions in the long run.
Understanding pros and cons of a particular decision is fairly straightforward, at least at an above average Vintage level.
Ah: But I'm arguing that it is not. That there are the obvious reasons to make a given play but then there are the very very very minute reasons to make or not make a given line of play that begin to aggregate.
As players gain more experience with decks like Long or Gifts, they begin to appreciate the various subtleties of their deck and subtleties of various match-ups - the subtleties, an ambiguous term to begin with, is something that I would interpret as "game play options". What is far more difficult to do is to weigh the pros/cons accurately once you assess all of your options. I suspect that, like Steve, the majority do it almost entirely intuitively and base it on pattern based learning - a decision under particular circumstances might have been successful in the past, so it is repeated in the future.
Note that I strive *mightily* NOT to rely on pattern recognition. I think with Grim Long it is helpful, but dangerous. I am a super huge fan of Forward Thinking - that is, asking the logical question: if I do this, he could do That, etc. I do not like using pattern recognition (which is different from Intuition).
In Pat's view: intuition is a mixture of computer processing in the brain AND pattern recognition, although it is NOT full throttle forward thinking.
We therefore repeat that play because, despite our limited sample size, our hope is that the trend witnessed thus far equates to a statistically higher chance of success. Unfortunately, it doesn't work like that, because it is conceivable that you, for example, can make the right decision and end up losing 5 times in a row. Perhaps we are too quick to accept decisions as correct because of a positive outcome?
On the other hand, perhaps part of the fault is that we don't do enough to generate information necessary to weigh certain plays more accurately?
Ding ding. This is Definitely a part of the problem. But to get information sets large and relevant enough is extremely burdensome. In fact, I would say inhumanely so. You could play magic full time as a job and still not have large enough information sets for Vintage complexity and the variety of unique situations that arise that are so context specific.
For example, Steve mentions in this article that he wouldn't make a certain play unless he knew for sure that his opponent didn't have a FoW. It struck me at first that such a statement is not particularly useful unless the opponent by some stroke of luck decides voluntarily to reveal info to you. But then I thought that perhaps we as players (some of us anyways) don't do enough to extract such info such as engaging in banter specifically designed for that purpose, or study body language/eye movements/pacing both during our actual matches AND studying our opponents when they are playing in other matches. You hear the odd anecdote about such here and there on these forums, or see it in person (or first hand), but I wonder how often this is just a stroke of luck and how often it was carefully crafted beforehand to get an accurate read on your opponent?
That itself would be subject for an interesting article - drawing some parallels between the skills absolutely necessary in games like Poker and a game like (T1) Magic, where up to now you could probably largely rely on technical playskill and deckbuilding skill to outcompete other T1 players.
Without commenting on those points, I would like to reiterate what I felt the point I was trying to make:
The LONG RUN is what matters. You want to make more decisions correctly in the long run. Any given situations will not likely arise again. If you make a second best decision that wins you the game, no harm. If you make a second best decision that loses you the game, that sucks, but it will happen. Understanding is the information that will help us make better decisions most of the time in the long run.
You can have very little understanding and make the obvious plays, but that won't make you the Robert Vroman of Stax or the Stephen Menendian of MDG or Grim Long (to inflame myself a little bit).
I tried to expose the soft tissue of my brain a bit and get you into my thinking.
But note this: even with 30 pages on just three games of magic, I omitted very large thought processes. I have to balance detail against entertainment and utility. That's why I cut out my tree diagrams and also why I pared back some of my analysis. If anything, though, I erred on the side of more explication rather than less.
I felt like i could have written 100 pages just on the first game alone.
