SpielByWeb

[Alashar]

I love Wallenstein and for the most part I love the implementation of it here, with one glaring exception: the Chit Tower. From what I gather, the odds in this online version are:

1. An army going into the tower has a 30% chance of getting stuck.
2. An army previously stuck in the tower has a 30% chance of being dislodged during any battle.

I don't know where these percentages came from, but they don't seem to reflect the behavior of the chit tower in the boardgame version. As I understand it, the chit tower was designed to bring the element of randomness into the battle. Well and good and it seems to function quite well in the boardgame version. Not so here. I would argue that in its current implementation, it throws in so much randomness as to often negate all the planning and strategizing one puts in at the beginning of each season.

A couple of examples: In Big Brawl VI, I launched an attack with 12 armies to which 1 was added from the cup. Only 5 came out of the chit tower. In the same season, another player was defending a territory with 10 armies to which 2 were added from the cup: only 3 of his armies fell out of the tower. I've seen worse examples and they happen too often.

I enjoy the random element of the chit tower, but when it gets to the point of negating one's choices and making the consequences totally arbitrary, then I think it's time to make a change. I stayed away from this online version for awhile because of that, and I know I'd play it more now if it more faithfully reflected the boardgame chit tower.

I don't know what the percentages should be, but I'd wager the percentage of chits getting stuck in the tower should be lower, perhaps 25%. And of those being disloged during battle, I'd estimate it to be significantly higher, perhaps 50%.

If it takes several people doing statistics with their chit towers at home to find what percentages would be more accurate, I'd be very happy to participate and do my part.

Meanwhile, please vote and/or express your opinion.

[TheKeck]

WOW! Those voting to change the tower outnumber by 2 to 1!!!

[sbryantf]

I disagree that the online version is "more random". I've seen more strange events in real life than I have here.

Anyway, if what you're going for is a more faithful reflection of the real life tower, you really should use a more powerful statistical model. The underlying assumption of the current model (with or without your proposed tweaks) is that each event is independent. But it seems to me that in the real life tower, the events are actually highly dependent. For example, the more recently a cube has been added to the tower, the more likely it is that it will be in a position to be knocked out.

The problem is that a more powerful model will require more parameters. This means you will need a much larger sample size in order to produce meaningful statistics. It also means that it is going to be much more complex to implement the model.

To me it comes down to "if it ain't broke...." Short of some compelling statistics showing that the current implementation somehow gives consistently lopsided results, I'd much prefer for the developers to be spending their efforts on new games.

[Golux13]

The two examples you give are bad (in terms of outcome), but they are just that -- two examples. In the first case, 13 armies went into the tower and in the second case 12, so the expected average number of armies to emerge (assuming none already in) is 9.1 and 8.4 (70% of 13 and 12), respectively. In your two examples, obviously, the number that came out was significantly lower... but that's the thing about probabilities: sometimes, the less-probable outcome happens. There is nothing inconsistent with five, or even three, armies coming out when twelve go in; it's called "bad luck."

It's also the case that if SbW made the change you suggest -- 25% chance of getting stuck instead of 30% -- then your two bad scenarios could still come to pass. The expected average in both cases would go up to 9.75 and 9 (75% of 13 and 12), respectively, that's all. Same thing if you increase the chances of dislodging a stuck army: the outlying, less-probable outcomes can still occur.

I think it would be wonderful to have the SbW tower mirror the real-world tower as closely as possible. There are two considerations with that, though: (1) the real-world tower probabilities are dictated by physical factors that cannot be duplicated on any computer short of the NOAA's weather forecasting supercomputers; and (2) if you are going to mirror the real world, you need data from the real world, and not just a few hundred data points. Ideally, before setting the algorithm, the SbW admins would have watched hundreds of thousands, or millions, of runs of army cubes through the tower and compiled reasonably accurate probabilities. Obviously, that didn't happen, and they just picked numbers that seemed reasonable to them. The "reasonableness" can be debated -- I think it is, and a couple of bad examples are not going to change my mind -- but if you are going to propose a change, I think the only valid rationale for the change would be to conduct the kind of massive study I described. (Sorry, but voting on what everyone thinks is the "right" probability is unhelpful.) I don't see it ever happening, so we all have to live with the fact that whatever numbers we plug into the probability equation, we are never going to have a perfect simulation of the real-world tower.

So what I do when calculating whether to attack is create a simple equation: (0.7 * A1) + (0.3*A2) = At, where A1 is the number of armies going into the tower, A2 is the number of armies stuck in the tower, and At is the total number of armies that would be expected (on average) to emerge. I calculate that for both sides, and figure my odds of success on that basis, recognizing that even if I am heavily favored, I could still lose. The technical term for which, I believe, is "shit happens." And the best way to avoid having that be a game-turning moment for you is to plan your strategies and avoid attacking unless your calculated advantage is huge.

Sorry about the long-winded post, but it always nettles me when I see people complaining about probabilities being wrong or broken on the basis of a handful of anecdotal bad outcomes. (This may have something to do with playing online poker. You would not believe the whining.)

[Golux13]

Oh, and thanks, sbyrantf, for saying the same thing I said, but only taking 1/3 the space to do it.

[Alashar]

Yes, Golux, I only cited two examples, but I could in a very short period of time, produce 200 from games recently played, and given a little more a time, a thousand. Of course that wouldn’t necessarily prove anything because in the end, it’s all rather subjective. We can also bandy around statistical catch phrases all we like and while that may make us sound intelligent, it still doesn’t prove anything.

Yes, of course, it would be ideal to have sample size of 100,000 or better yet, a million, but I doubt that it really needs to be that large to be more accurate than picking “reasonable” numbers out of the air. It’s likely that a sample size of several thousand would suffice and be more accurate than that currently in use. Of course, if you’re happy with the way it currently functions, than it’s very easy to call for a sample size so large that it defies actual testing.

Yes, perhaps the SBW admins picked numbers that seemed reasonable to them when they designed the chit tower, but do they seem reasonable to those of us who play regularly? The “reasonableness” is the question here and is the subject of debate. Nor do I agree that the “only valid rationale for the change would be to conduct the kind of massive study I described”. If people are unhappy with the way the chit tower is functioning, then a poll is a way to find that out. If more players would be happier with a change, then perhaps that is a valid reason for making one.

Golux also said “the best way to avoid having that be a game-turning moment for you is to plan your strategies and avoid attacking unless your calculated advantage is huge.” And how do you go about planning precisely when and where someone will be attacking you? The point is your calculated advantage should not have to be huge. I don’t see that that accurately reflects the board game. Requiring it to be huge seriously undermines the strategy and the planning, which is the point of the game. By increasing the randomness to the point of arbitrariness, the game is robbed of all the careful planning and thought we put into it.

“If it ain’t broke…” In my opinion it is broke and it needs fixing. I am not suggesting the developers spend 6 weeks locked in a room testing a chit tower a million times. I too think that time can be better spent in developing new games. But if changing an algorithm in a game can significantly improve play, then why not do it?

I’d be happy to run a sample of a thousand myself, and if there are enough people willing to do smaller sample sizes, it might be worth doing. Another possible avenue is run several beta versions of Wallenstein with different algorithms and see how they play. It’s quite possible that the game in its current condition is not perfect and could be improved upon.

[sbryantf]

Alashar -- I encourage you to go ahead and compile a list of examples. My only suggestion would be that you define in advance what an unreasonable result would be. Is it any time the difference between the number of armies going in and coming out is at least 3? Maybe some sort of percentage? Once you have some objective criterion of what constitutes an unreasonable result, then it simply boils down to an exercise in counting. How many times were the results reasonable, and how many times were they unreasonable?

Of course, people may argue over your chosen criteria, but the great part is that you can test a variety of different criteria simultaneously. Once you get the raw data into some sort of spreadsheet, it's a relatively simple matter to see what happens under different definitions of "unreasonable".

I suppose one straightforward solution would be to add an option alowing person who creates the game to choose the parameters of the tower. List these numbers on the games waiting to start page, and each player will know what they're getting into. I suspect that this wouldn't be overly difficult to implement.

[Big Bad Lex]

A computer model will never totally replicate the action of the real tower and in the absence of a chimp willing to throw every permutation of cube into the tower a squillion times, how could we know.

So the accusation of the model not being realistic is unproven. Critisisms are purely subjective and the important thing is it cuts both ways. I have had some dire results in an online game; 6 armies attack an empty province and lose. However I have benifited from some totally unjustified victories too.

Like someone else said the board game does some strange stuff too. As long as the model is consistant for both sides this approach works fine.
_________________
It's not the winning, it's utter annihilation of your opponant that matters.

[dragondreams]

suck it up and deal with it.

[Golux13]

[LostSoul]

The only difference that there may be between the online and physical dicetowers is that of capacity. If the online game uses the formula of .7 and .3, there is a possiblity that all the cubes could be kept in the tower( this is obviously an extreme case). The physical tower must have some sort of capacity that once reached cubes fall out on a one to one ratio. A simple modification to the formula that adjusts for cubes in the tower could be all that is needed. I have never played the physical game so I appologize if I am talking out of my ass.

[HappyProle]

There are plans to turn the cube tower into a sub-game developed in Flash that will allow each player to "throw" his cubes into the tower at a certain angle and with a varying amount of "force." The tower will be mapped in a 3D space that will even approximate those pesky inner corners in the actual tower. It will use a sophisticated physics engine that we've spent the better part of the past six months tuning and refining.

[HappyProle]

Did I say there were plans to do this? I meant to say there are no plans to do this, now or really anytime in the foreseeable future.

[Golux13]

[Big Bad Lex]