Thursday, June 29, 2017

Recent playtests: Deities & Demigods & Joan of Arc

Deities & Demigods

There are 33 playtests recorded in my Deities & Demigods record sheet google doc. This covers several iterations of the game, going back to May 17, 2016. In the last few playtests at Origins 2017, I used the latest version of Hera cards to determine start player each round. The first time I just shuffled all the Hera cards together and laid out 5 for the game. It was kinda lame that for the first few rounds, those conditions were irrelevant because, for example, it's completely impossible to have done a quest before the first time the condition is checked. But I thought of an easy solution to that - I took all the cards that WERE potentially relevant in the first round, and labeled them "A," and I took all the cards that were not relevant in the first round and labeled them "B." Then I shuffled them and used 3 "A" cards for cycles 1-3, and 2 "B" cards for cycles 4 and 5. I've played a few games with that setup, and it's really worked well.

I've tried a few of the draft player powers, and I guess they're alright. I'm not sure if I think they're necessary, but I suspect people would dig them, so I ought to try them some more and make sure I have at least 4 that I like. One problem with including them is that if you have diverse player powers, it's boring if you only have 4 for a 4 player game. So I'd really want to have 6-8 for more variety.

I'd still like to try the Hades expansion module, but that need not be in the base game box, so it's not all that important. Still, it would be cool to try out.

Joan of Arc

Since my last post on the topic, I did get a chance to play Joan of Arc (twice) at Origins.

The new column format for the tiles is cool, but frankly, for the current implementation, cards would
be more appropriate than tiles. So I have two choices...
* Give up on it being a "bag building" game. This could be OK, even if it is intended to be related to Orleans, but as a product I'd like to keep the bag building. Hopefully that's not just me being stubborn.
* Change the format such that tiles become the appropriate component.

Daniel made a comment that sparked my interest - he said it would be neat if you laid out the tiles in a more organic way or something, and activated them like a neural network. I'm not entirely sure what he had in mind, but to me that inspired the idea to connect tiles in 2D space (as opposed to vertical columns), and instead of the tiles having 2 sections with income (one that gets covered up), maybe they have an icon in the center, an little 1/2-icons on some of the edges.

So instead of columns, maybe you have three 2x2 square areas to put tiles, and you want to combo up tiles so that they have matching 1/2-icons on the sides so you get extra income.

I actually can see several different ways to do this... 2x2 squares, plus shaped areas where the 4 tile spots are orthogonally adjacent to a center space with printed income, or even columns where the tiles have the 1/2-icons on the top and bottom only.

Today I showed the idea to David Short, and he had an idea for a single, bigger grid of spaces to put tiles into, and then you activate a row or column. I added that maybe you could activate a 2x2 square instead. So you add tiles to a 4x4(?) grid, then activate a row, column, or 2x2 square inside that grid, take the income, then discard the used tiles from the board. I'd have to figure out how exactly you add tiles to the board and activate them to make it non-trivial (add 3 in a row, activate that row). David suggested that you could have 3 activation tiles, 1 showing a ow, 1 showing a column, and 1 showing a 2x2 square, and when you use one to activate that configuration, it's gone till you've used them all. There's something interesting about that, but I'm not sure it solves any problems. However, maybe if you had to place in one of those configurations and activate a different one, that could start to work. For example, maybe you place in a column, then you activate a row or square. Then you can't possibly activate more than 2 of the tiles you just drew. Or maybe you can only place in a single row or column, and then activate the other, but you can do a 2x2 square if it's complete (4 tiles). Or only the center 2x2 square. I'll give this some thought, because it might be a better personal puzzle than the three columns thing.

Another, different idea that David also suggested, more similar to the current column idea with the tiles I already have, is to stack them... so you place 3 tiles into the column, but not overlapping, then you stack the next 2 tiles on top of those, offset so you cover the top half of 1 tile and the bottom half of the next. Finally, if it gets that far, you stack the last tile on top of the top 2, making a sort of pyramid. This would maintain the idea of getting 1/2 of the stuff on the tile sometimes, and it would make tiles the more appropriate component, but I'm not sure it would work as well as some of the other ideas.

Those are all juicy ideas that are very interesting, but they are almost completely separate to some of the other changes I want to make, such as adding static effects on spaces for when the battle tile is gone, or changing up the stat tracks, or adding cannons. I DID increase the costs of sieges, somewhat arbitrarily. I probably will need to revisit those costs a bit. And I'd like to maybe make some more end game goal cards, perhaps a series that wants you to advance a certain amount on the stat tracks, so you'd be extra interested in training.

That's all the updates I have for now.


Josh 'Dagar' Zscheile said...

Hey Seth,

no idea about what the tile laying aspect of Joan of Arc really means in the game or why it would be thematic, but I can at least tell you how neural networks work, so maybe that helps you understand what Daniel could have been meaning by that comment.

A single smallest component of a neural network is a neuron, just as it is with biological neural networks. You can think of a single tile here. The neuron has two kinds of connections to other neurons; inputs and outputs. You can see the inputs as prerequisites to meet in order for the output (the effect of the tile) to work. In a neuron, all the inputs have weights that represent the importance of that input to the output. If the combined inputs (usually added up) reach or surpass a certain threshold, the neuron fires (the effect occurs). This in turn is the input for other neurons in the net.

The most simple form is a convolutional neural network that is strictly hierarchical. More complex ones do not have that hierarchical form and can have feedback loops in them.
As an example for a game, picture square tiles laid out in a pyramid form. Consider each tile connected to the three below (the one right belo and the two diagonals). The tiles have three numbers from one to four at the lower edge that represent the weight given to the input of the according tile below (e.g. on a tile you have a 2 on the bottom left corner, a 1 on the bottom edge and a 3 on the bottom right corner). The tiles also have a threshold value that needs to be met or exceeded by the added inputs in order for the tile to trigger its effect (e.g. a threshold of 3). For our example tile, of the tiles on the pyramid step below either the left and middle ones need to be activated, or the one to the right, or some combination of that, as their combined values meet or exceed the threshold.
A system like that could lead to interesting decisions. Do I activate tile x for its effect, or do I rather use tile y the effect of which does not suit me, but that in turn activates (or prepares the activation) of a tile one or more levels up that have good effects? Combine that with boosting the effects the higher up the pyramid they are and with player interaction (all players activate and build on the same pyramid), and you have a pretty engaging mechanism all by itself.
Heck, now that I think about it, I want to try and design a game with this :D

Hope that helped!


Josh 'Dagar' Zscheile

Seth Jaffee said...

Thanks Josh!

I suppose in my version the "input" would be 1 half of the 1/2-icon, and the output would be the other half, and if they match then they "meet the threshold" and he effect (income from that icon) occurs.