Zendo-like (Prototype)
So inspired by the CT meetup last week I have taken 3 days to code up a prototype for an idea I had. The game is an inductive logic puzzle game, based largely on the physical game Zendo. The game works as follows:
There is an updated game description here: http://makegamessa.com/discussion/comment/5624#Comment_5624
The object of the game is to figure out a hidden rule that describes sequences of symbols. Symbols have two attributes - color (red, yellow, green, blue) and shape (circle, triangle, square, hexagon). So you can have a red hexagon or a blue square. Up to 5 symbols can be grouped to form a sequence. Every game has a hidden rule (determined randomly at the game start) that describes a valid sequence. A valid sequence is represented by a small white circle and an invalid sequence by a black one. Some example rules you can currently get:
- Valid sequences must have a red symbol.
- Valid sequences must have a square and a circle.
- Valid sequences must have an odd number of green symbols.
- The last symbol in a valid sequence must be a triangle.
- All circles in a valid sequence are followed by a blue symbol.
You can do a few actions in the game:
- Propose a custom sequence and have the game tell you if it is valid or not. Select symbols at the top, click on the plus to add them to the sequence and click on symbols in the sequence to remove them.
- Ask for a random valid or invalid sequence.
- Do a test to determine if you have figured out the hidden rule. The test is the user must determine the status (valid/invalid) of 8 sequences. Click on sequences to cycle their status.
The plan is to develop this with a focus on mobile. I am using Haxe/NME, which means I can target Android, iOS, Windows, Linux, Mac, Flash, HTML5 and others. For testing, it is probably easiest if I provide flash and html5 versions, but I can provide others if requested.
If you get really stuck, you can press 'r' in the test screen to show the solution, but this is obviously just for testing purposes. It sometimes take me more than 20 sequences to determine the hidden rule, and I know all the possible rules. This game is not meant to be an easy game.
Some things I still plan on doing:
- Improve the GUI in general (this is quite a rough but playable prototype).
- If you didn't pass the test, show one of your mistakes.
- Prevent you from adding duplicate sequences.
- Add more rules
At this stage I'm looking for just about any feedback. Do you enjoy trying to figure out the rule? Is it too easy/difficult? Was anything annoying/frustrating? What would you like to see added?
The current prototype is available here: http://leafcloud.com/zendo/flash/ (v0.0.12)
There is an updated game description here: http://makegamessa.com/discussion/comment/5624#Comment_5624
The object of the game is to figure out a hidden rule that describes sequences of symbols. Symbols have two attributes - color (red, yellow, green, blue) and shape (circle, triangle, square, hexagon). So you can have a red hexagon or a blue square. Up to 5 symbols can be grouped to form a sequence. Every game has a hidden rule (determined randomly at the game start) that describes a valid sequence. A valid sequence is represented by a small white circle and an invalid sequence by a black one. Some example rules you can currently get:
- Valid sequences must have a red symbol.
- Valid sequences must have a square and a circle.
- Valid sequences must have an odd number of green symbols.
- The last symbol in a valid sequence must be a triangle.
- All circles in a valid sequence are followed by a blue symbol.
You can do a few actions in the game:
- Propose a custom sequence and have the game tell you if it is valid or not. Select symbols at the top, click on the plus to add them to the sequence and click on symbols in the sequence to remove them.
- Ask for a random valid or invalid sequence.
- Do a test to determine if you have figured out the hidden rule. The test is the user must determine the status (valid/invalid) of 8 sequences. Click on sequences to cycle their status.
The plan is to develop this with a focus on mobile. I am using Haxe/NME, which means I can target Android, iOS, Windows, Linux, Mac, Flash, HTML5 and others. For testing, it is probably easiest if I provide flash and html5 versions, but I can provide others if requested.
If you get really stuck, you can press 'r' in the test screen to show the solution, but this is obviously just for testing purposes. It sometimes take me more than 20 sequences to determine the hidden rule, and I know all the possible rules. This game is not meant to be an easy game.
Some things I still plan on doing:
- Improve the GUI in general (this is quite a rough but playable prototype).
- If you didn't pass the test, show one of your mistakes.
- Prevent you from adding duplicate sequences.
- Add more rules
At this stage I'm looking for just about any feedback. Do you enjoy trying to figure out the rule? Is it too easy/difficult? Was anything annoying/frustrating? What would you like to see added?
The current prototype is available here: http://leafcloud.com/zendo/flash/ (v0.0.12)
Thanked by 1Ramperkash
Comments
Some rules are quite difficult, especially if you do not know the entire space of possibilities, and it does get quite frustrating when you simply don't have a clue. Perhaps you could introduce new rules gradually and deterministically, instead of randomly, so as to give the player a better chance of developing strategies, or you could give sequences to deliberately help the player refine their assumptions.
A would not prevent the player from adding duplicate sequences (it becomes difficult to remember the results of all your tries, so a duplicate may be necessary).
Look forward to see more!
At the moment difficulty is very random. For the future I plan on grading the rules by recording and using some play stats. For example, the rules could be ordered according to how many guesses it takes to solve and then easy problems are the 25% that are the easiest to solve, or something like that. I would also like to have some sort of tutorial with a few pre-determined rules, in order to introduce the game to the player.
My plan with duplicate sequences is to prevent them, but if the player tries to propose a duplicate, tell then that and remind them what the previous result was. I think this should address your concern.
Another issue that someone has shown me is that you can spam the valid/invalid sequences to quickly get an idea of what the rule is, without ever proposing a rule. I think that I may remove the option of asking for valid/invalid sequences all together. The game is then much more about probing the rule system and figuring out what sequences are valid or invalid. For now I would like it if people could try avoid using the buttons and see how it is.
I have implemented the feature that shows an incorrect sequence when you fail a test, and I will put it in the the next release. What I plan on doing next is:
1. Improving the propose sequence screen (at the moment it isn't very usable).
2. Add some in-game help, even if it is just text.
3. Improve the scrolling of sequence history.
4. Prevent duplicate sequences.
5. Add more rules.
I plan on releasing the next version later today. I will update this thread when I do.
EDIT: BTW, if you missed it, you can already scroll through your sequence history by dragging.
To me this feels like a gamified for of debugging (which as a coder, I enjoy).
That being said, it's super difficult, which alright, but I think there needs to be a 'give up' button that presents the rule. That way at least you get a better understanding of the solution space with every failure.
The valid/invalid functionality can essentially be derived through random proposals, The only difference is that it would ne more of a pain in the ass.
I would suggest that instead of removing the valid/invalid button entirely, you monitor the number of requests made. This metric can then be used to reflect performance.
Have you played SpaceChem? Instead of leaderboards, SpaceChem provides histograms that the elegantly show the distribution of every players performance for a given puzzle. This is really interesting because it allows the player to quickly see how their performance stacks up against the rest of the player population. For an infinitely better explanation, check out this interview with the SpaceChem dev, you can jump to "We Were Innovative With Our "Community Features".
If you provided histograms visualizing the number of valid/invalid requests, players could see how that stack up compared to everyone else. This could drive the metagame in that players would keep requesting puzzles to get a better average placement.
I really like this. I think the game will be much more compelling with some sort of campaign mode that marks progress and gets incrementally harder. You could also try a survival mode where there is a fixed number of valid/invalid example requests, and players must try to complete as many games as they can before exhausting their requests. Not sure it it would be fun, but its worth a try.
I plan to add an option to give up soon, but at the moment you can give up by cheating to see the rule (pressing R on the test screen). I do plan on adding a simple difficulty selector in the near future that has a few easy rule classes hard-coded in, mostly to ease people into the prototype.
Actually I got the idea of recording all the stats and making histograms from SpaceChem, I just haven't fully fleshed out the idea here. The idea is that I will record these stats for every class of rule. Not only will the player be able to see the stats to rate themselves, but I will use the stats to rate the problems. I would obviously also have to have some sort of global leaderboard that combines the stats of each player, but that is quite a way off.
I'm still not sure about whether having the valid/invalid buttons are a good thing, and there is of course other options besides just removing them completely, but I will leave them there now for testing. The idea is that you have to try propose some clever sequences to make educated guesses about what the rule is. So for example: if changing the colors, but nothing else, changes the status, then color is important (note the converse is not always true though).
Some sort of campaign or survival modes are good ideas, but for the moment I want to focus on the "normal" mode and make that as good as possible. Multiplayer is another idea that I foresee working later, maybe for version 2 or something.
I am currently focussing on making it easier to propose sequences, which is quite a chore at the moment. Expect an update later today (might be quite late).
Some changes:
- When you fail a test, show you one of your mistakes, and add it to the game history.
- Some graphics artifacts have been smoothed over.
- Scrolling is improved in look and functionality.
- Replace the new sequence screen with one that is hopefully much easier to use.
- Prevent duplicate sequences, but show you what the previous result was.
I find that I can have quite an enjoyable game without using the valid/invalid buttons, so I'm strongly considering removing them altogether, but I'm still not 100% sure. Some options I have considered:
- Removing completely.
- Adding some sort of cost, relative to proposing custom sequences.
- Cooldown
- By proposing valid sequences, you get tokens that can be used to generate invalid sequences, and vice versa.
Next things I plan on working on:
- Adding an option for easy rules, even if some are just hard-coded.
- More rules.
- Some built-in game help.
- An option to give up and see the rule.
I get what you are saying about the invalid / valid buttons, but they are very convenient...
Perhaps the following will make the proposing screen much easier to use: click (not drag) on the icon to put it in the proposed sequence, and click on it in the sequence to delete it. So proposing rules become very fast.
And why not put the propose screen on the same screen as the list of sequences? It will make it even less cumbersome to use.
Still not sure why you would want to not allow duplicate sequences... (even if you add a cost, you could just invisibly not add it)...
But these are all details, I think the most important thing is to somehow get the difficulty thing sorted. (I'm still stumped half the time). Ideally, the game should "teach" me (by the order of presentation) how to develop the logical tools to figure the rules out.
I can make it so that you can do either. That will probably be the best.
While this is true, I don't feel there is enough space for both on the same screen. You have to remember that I'm targeting mobile devices too. I find that the symbols are already quite small on my phone, and I'd prefer to not make that even worse. However, I do get what you're saying. I was thinking of putting a small history also on the propose screen, probably just two sequences, but then you could easily see what you were busy doing. Do you think this would be sufficient?
Do you want to do this so you can see them lower down on the history? I could do this quite easily, but wouldn't a better solution be to allow you to rearrange the history to your liking? This will obviously be a bit more work, but in the long term I feel it could give a better result. My reason for not wanting duplicates is basically I want the game to be as clean and simple as possible, and I feel that duplicate sequences are just a little messy. Maybe I'm just crazy though.
While I agree in principle, I have do idea how to actually do this. While some rules do have a natural progression between them, this is not always the case. For example, how do you introduce a rule where valid sequences must contain 1-4 yellows or a red circle but not both? I feel that introducing each of the parts before the whole might work, but not necessarily.
If on the other hand, you mean that I should grade the problems so you can do some level 1 problems before doing level 2 problems and so forth, then I fully agree with you. Then the problem is more how I grade rules. I was thinking of grading them by looking at how well users can solve them, but I'm starting to think this will require too much data for me to bootstrap the idea. I have a few other ideas with how I can generate a difficulty of a puzzle automatically, but I'm thinking they might not always work so well. I will have to see...
When I play, I first need to get some clue of what the rule may possibly be. So I spam a few valid, and a few invalid to form an opinion. Then I spam some more, and form various "hypotheses" [grrr, sounds so pretentious], and use the sequence to elimination wrong ones. I don't really care about the history, when my hypothesis change, I don't relook the history, I just spam away until it's disproven or I have enough confidence that it is true. In a sense, I don't really need the history, because it is faster to find the evidence I need right now by just hitting the button several times. If you take these spammers away, then similarly, the history is of little value [provided the proposal tool is fast enough], since any information I need is faster to re-obtain than to search for it.
Of course, I may not represent the average guy or girl, so best speak to more players about how they deduce rules.
----
[Start section of speaking out of my ass, so be warned!]
The way I would think of teaching the tools is as follows: First, form a sequence of (roughly) progressive rules (grouping together similar ones, for example _1_ red square is equivalent to _1_ yellow circle.)
Once you have this sequence, you start with the first one: How do you teach this group of rules? The first one should be fairly easy to teach, so I'll skip this. Now coming to the second group, at this stage the player will start of evaluating the sequences by whatever rules are in the first group. So you need to provide sequences that will directly negate this. So they see they are wrong, and not necessarily why. At the same time, you need to provide sequences that will give a clue to what the modification may be [And this will be similar to teaching the first rule group.
Moving on to group three, you consider how players will evaluate sequences based on the first two groups, and then contradict those assumptions.
And so on.
Maybe it turns out that this is not a good strategy, so you will have to test it, I guess...
But this assumes that you _do_ provide sequences... if you don't, then of course players literally have to teach themselves. In this case, I am sure there are things that can be done as well, although right now I can't think what they could be...
----
Another thing I would do is calculate roughly how long it will take to learn rules. (How much evidence does the player need?) This number will depend on the space of rules of course (for all possible rules the player will of course need to test all possible sequences). This will help you structure rule groups, and sequence them together in a good order. It will also show you whether the total set of rules is too large.
----
Just some of my 2c ;)
I like the idea of progressing by using rules that disprove previous assumptions. Like: "You thought order doesn't matter? Well now is does! Mwhahaha." I will try form a logical progression like this and see if it works as a little tutorial. I think this could work whether or not the [in]valid buttons are available - in the end, assumptions are built around what works and doesn't, so I think the rules might be almost directly correlated with what assumptions formed.
How would you determine this? By measuring how long it actually takes player in practice? This is what I was thinking of initially, but I'm thinking it will require too much data to be useful. Or do you mean by working it should be worked out automatically? How?
Another thing I just tried was to determine the valid:invalid sequence ratio and see if that was correlated with difficulty, but it seems it won't be as useful as I was hoping. I'm still busy analysing some numbers and I'm still hopeful I can use this for difficulty options, although there might only be easy and hard options.
Your 2c is much appreciated :)
- Click or drag on the propose screen.
- Give up to see the current rule.
- Recent history on the propose screen.
- Some of the rules are now more flexible.
- New tutorial with 7 hand-chosen rule classes.
I tried to design the tutorial levels to slowly break any assumptions the user might have and give them an idea of what possible rules there are. I will see if I can try it out on someone who hasn't played the game yet.
I still need to find a good way to determining rule difficulty and this will probably be my next big job. There are still quite a few smaller things I have on my todo list, so I will also be tackling some of them.
As usual, let me know if there is anything in the new version you like/dislike. I will be playing it myself in the mean time...
EDIT (2012-12-09 01:01): I did a bit of work today, so I am releasing a new version (v0.0.4). Changes in this version:
- Some GUI changes, mostly superficial.
- Moved the [in]valid and give up buttons into a small menu.
- Combined the [in]valid buttons into a hint button - 50% chance of a valid or invalid sequence.
- Added a quite few new rules.
Next I plan to do some more user testing, especially with the tutorial. I also want to start recording some statistics and uploading them to a server so I can make informed decisions, especially regarding rule difficulty.
As an extreme example, if you only had 2 rules, "requires red square in position 3" and "requires yellow square in position 3" the two rules will give exactly the same results for all sequences except for two. This means the player has to wade (worst case) through 16^4 - 1 sequences to find the right rule.
I think what you really need to do is actually see if there is a strategy that can always determine the rule in a decent number of steps...
----
I have played the new version, the clicking system is much nicer in the proposal window. (Hehe and your hiding away the valid/invalid options basically forced me to play with just the proposal options only!)
The tutorial is also nice and helpful... but I don't think helpful enough (I still had to give up a few...). I think it is still a bit too difficult. Perhaps you should give more than two sequences in the tutorial, and handcrafted ones. For example, to show that an odd number of circles are required, you could give the sequences 0 00 000 0000 00000.
Edit: I forgot to mention, there is a problem with tutorial 4/7;for some reason, nothing happens when you click on "Go". All others work fine.
Also, if you think you have found a rule, then guessing will either tell you you are correct, or show you a counter example. So the idea is not that you only guess once, but that you use the guess counter examples to guide your search. A problem I have discovered is when I pass a test, but I didn't actually have the correct rule, just a close one. I might be able to fix this be choosing better test sequences.
I like the idea of hand-crafted sequences in the tutorial, I will have a look into this. I don't want to make it too easy though. IMHO you shouldn't be able to see the rule just from the starting sequences.
I see in all my updating I have introduced a bug in tutorial 4. I will be sure to fix this in the next release, thank you for pointing it out.
I am curious, after about how many sequences do you tend to give up? I'm just looking for a ball-park figure.
EDIT: quickly updated to v0.0.5 to fix the tutorial bug.
I usually give up between 20-30 moves if I have not found the rule. The biggest difficulty is finding "new" rules. Once I know a certain rule-type, I can usually figure out what it is, but before not so much.
This is part of the reason why I keep adding new rule types. As I see the game, you should believe that the rule could be practically anything - there should be no reason to think it must be of a similar form to what you've seen before. The idea is to come up with a theory that works for everything you've seen so far, maybe test it a bit more, and then guess the rule. If you are wrong, you get given a bit of counter evidence you can use to formulate a new theory.
EDIT: I just had a game that went on for almost 30 moves. It got a bit frustrating trying to make new theories when none seemed possible, but also kinda fun. I found that it is useful to try guess weird theories that include a lot of pieces like: "contains no hexagons, except when there is a circle or a red square, and the number of symbols is odd." I'm pretty sure this isn't the rule (unless someone is adding working code to my game), but it means I have something I can guess, and then I get a counter-example that often sparks some ideas, or I get the test correct with a bit of luck and then I hit myself when I see how simple the rule is.
Any chance you could smooth out that "WTF is going on?" curve a little? Introduce rules slowly over time, give people a set and simple tutorial to go through that explains each step of action that they're expected to take? I had no idea what the different buttons were, couldn't tell what Guess was actually doing and had no clue how to suggest a set of rules...
I would suggest having rules laid out for the player via a card-like system: So you'd have something like "All valid sequences contain a ___ ___" and then give the player cards for triangle, square, circle as well as red, blue yellow and green. Then give them two of those mad-lib style rules for them to fill out. There's a strong chance you're already doing this, but I feel like I'm looking at an overly complicated superset of Mastermind with less information given to me to figure things out with.
Even a "1 rule passed, 2 rules broken, 0 rules don't apply" type of display would be more helpful, kinda like Mastermind's x correct in the correct place, y correct in the incorrect place.
Updated Game Description:
The object of the game is to figure out a hidden rule that describes sequences. Sequences are made up of one to five symbols. Symbols have a colour (red/yellow/green/blue) and a shape (circle/triangle/square/hexagon). The rule determines whether a given sequence is valid or invalid and this is signified with a white or black circle in the game. Rules are never dependant on time or the order of separate sequences. Some example rules include:
- A sequence is valid if it contains a red.
- A sequence is valid if it doesn't contain a circle or a green square.
- A sequence is valid if it contains an even number of symbols.
- A sequence is valid if there are more yellows than hexagons.
Each level has a different hidden rule. In order to determine what the hidden rule is, you can propose sequences of symbols and you will be told whether your proposed sequences are valid or invalid. Once you have formulated a theory of what the rule is, you can guess the rule. Guessing the rule means you must complete a short test where you see a list of a number of sequences and you must say whether they are valid or invalid. If you have worked out the correct rule, you will hopefully answer the test correctly and win the level. If not, you will be shown one of your mistakes (with the corrected validity) and you can use this information to start forming a new theory.
So a level might go something like this: You notice that there are no red in any of the invalid sequences. You guess the rule must be "... contains a red" and find out you are wrong because when you guess you get the counter example that {red circle, red square} is invalid (you marked it as valid in the test). Looking over the history you see that in fact all the valid sequences had an odd number of reds. So you guess the rule must be "... contains an odd number of reds" and are pleased to find that you are correct.
Some UI help: Starting menu and tutorial buttons are pretty simple. When starting a level, you will be greeted with the history screen and two sequences to get you started (one valid and one invalid). You can scroll through this history and use the 3 buttons at the bottom. Propose takes you to the propose screen. Here you can drag symbols from the palette below, into the current sequence, or click on the palette to add them to the end. Symbols in the current sequence can be dragged around, or clicked to remove. The last two sequences in the history are provided at the top for reference. Back on the history screen you can click the guess button to test your rule theory. This takes you to the test screen where you will see a number of sequences (8 at this point) and none of them are marked valid or invalid. You must then click once or twice on these sequences to mark them as valid or invalid according to your rule guess. When you are done, you can click the guess button at the bottom to see if you are correct. If you are, you've just won. If not, you are shown a mistake and it is added to your history. On the history screen there is also another button which shows a small menu with a few simple options to select.
---
I agree that the tutorial needs quite a bit of work to introduce new players and I need to work on that. A proper guided tutorial is definitely something I should do, but I've been trying to avoid it for the moment, especially when the game is in a fairly liquid state. I might "clobber" together some screenshots with text, and see if that will be enough for the moment.
For the moment, if you can't get past a tutorial level, you can use the "give up" option to see what the rule was. Then you can try the level again. You will get a similar but most likely different rule you can now try to figure out.
At the moment rules are generated by filling in the blanks similarly to how you describe, and I think giving the player some more help of this form in the tutorial might be helpful, especially for the first few tutorial levels. But if you mean the player should fill in such a card to determine if they know the rule (replacing the current test mechanism), then I disagree. Either the cards will be too complex to allow all possible rules, or there will be too many cards. The later option might seem ok, but I don't want the player's rule theories to be restricted by their ability to memorise the different types of cards.
Basically, I want the game to be inductive (like the scientific method) as opposed to deductive (like mastermind). The idea is that you are given very little information and you have to find some pattern or theory that holds by experimentation.
Sorry for the really long post, I hope it was helpful.
I strongly suggest putting some description text on the guess screen, at least the first time it appears (or maybe during all tutorials) that says "So you think you've figured out the rule? Score these sequences with what you think the rule is and we'll see if you're correct!" or something to that effect.
I also think that the white/black dots don't communicate very well. Ticks/crosses and or red/green read much better. Also, I think you should switch to the abstraction of two lights next to each other, one lit for valid and the other lit for invalid, that way players can set valid/invalid directly on the guess screen (also, left click a sequence for valid guess, space/right click for invalid, while you're at it).
Have you thought of a game mode that's simply a collection of guess screens, one after the other, with all the guesses correctly scored in your history?
Maybe the guess screen shouldn't be called that - it's more a Test Hypothesis" screen, after all.
I don't know if it's been mentioned before, just something I thought about.
Officially on the TODO list.
I've purposely stayed away from ticks or crosses because I want to try and minimise the influence I have on players' though process. Ticks and crosses come with many preconceived ideas. All of my rules are invertible, and at a later stage I want to remove the concept of valid or invalid, abstracting things to just white and black (or whatever they become). However, I do like the idea of lights that are next to each other, but not exactly that idea. I will think it over. In the mean time I will add right click on the guess screen for the flash/pc version (BTW: ultimately I want to focus on the mobile platform).
I don't quite understand this, could you please explain it a bit more?
It was first called the "test" screen, but I renamed it because I want to let the player know it is ok to test/guess early on, and not just when they are relatively sure already. Ultimately I want to replace the text on those buttons with icons.
Do you mean that you should be able to click on the sequence history shown on the top part of the propose screen? If so, then I do plan on doing this. If not, then could you please explain a bit more?
EDIT: @dislekcia: It seems it isn't possible to capture right click events in Haxe/NME, so I'm not going to add that right click option until I have a chance to find a workaround. I suspect it will be messy, so I am avoiding it for now.
If you're focusing on mobile, please make the guess screen have left- and right-swipes validate/invalidate sequences. That would be awesome... Of course, the validity indicators would have to be on opposite sides of the sequences to foster that kind of interaction, but it'd be worth it :) No propose mode. You'd start out with 8 sequences, scored for you as valid/invalid - then you'd get 8 guesses to score according to what you suspect the rule might be. Over time you'd build up a history of valid and invalid guesses, giving you enough information to figure out the rule. Personally, I think I'd find that a much more compact game mode than the current one.
What I mean by inverted is: if I change the rule so that everything that was valid is now invalid and vice versa, then I have inverted the rule. So changing the rule from "must contain no reds" to "must contain at least one red" is inverting the rule. So basically I'm saying I don't want to use things like ticks and crosses that may lead the player to think that valid is better than invalid or vice versa. Different sounds are something I definitely want to add, and maybe I should start to add them sooner rather than later. Different shapes can also work well, as long as the shapes are neutral. So does this mean you think black/white stones are not clear enough? This isn't something I had considered...
On another read of your comment it occurs to me that you might mean I should have more that just binary output - that I should somehow indicate how close to valid or invalid a sequence is. This sounds much easier than it is. While some of the rules are of the form "this condition and/or that condition," most of them aren't. For example, how would I indicate this for a rule like "the number of red is less than the number of circles"? Do you feel like more information in the rule output is necessary to make the game easier, or to give the player some direction and help speed up the process?
I can do this, but it might be a bit confusing. I'd have to see if it works. If it does it will be pretty cool.
I'm mostly focussing on mobile because I want a good game on my phone that can keep me mentally occupied when I'm bored. Also, I generally feel that this isn't something people will generally sit in front of a PC and play.
Sounds like an interesting game mode - could be a "quick play" mode. I'm not sure if 8 guesses would be enough, but it could work. At the moment, you could emulate this by asking for 6 hints at the start; then together with your starting 2 sequences you'd have 8. From there it would simply just be a matter of repeatedly guessing until you get the rule or guess 8 times. I'll give it a go and see what I think.
Thanx for all the feedback so far, it seems like you've given this a fair amount of thought, which I do appreciate.
I mean, a proposed sequence either passes the rule or doesn't, making that binary state ambiguous or in any way muddy doesn't make any sense to me. An inverted rule that differs to something the player has seen previously doesn't factor at all, as the rule is only ever revealed once the player has figured it out anyway. Even if you change the rule halfway through a session, the existing historical sequences are going to have to be re-evaluated and their pass/fail state is the most important thing in the game! Players need to be able to see that at a glance as quickly as possible.
Valid IS better than invalid, in terms of answering "did this sequence pass or not", I can understand that you might be trying to encourage players to not feel bad about negative results, but making that highly important binary outcome less obvious is not the answer to that. Besides, players will figure out how to handle negative results logically over time playing the game anyway, that's kinda the point of inductive logic, right? I think you're overloading the meaning of ticks and crosses far too much here, so much that you could easily sabotage your interface needlessly.
If anything, I'd add different states to valid/invalid modes when scoring guessed sequences: The player's valid or invalid mark should be visibly different from an actual game-scored valid/invalid mark. The best thing to do there would make the player's marks be obviously temporary, to emphasise that these sequences are going to be scored by the game and it's knowledge of the rule once the player's theoretical scoring is complete.
TL,DR: I could not figure out what was valid or invalid at a glance when I first saw your game. I had to be taught what the little dots meant and, given the similarity of this game to mastermind, I assumed completely incorrect meanings for black and white dots. Ticks and crosses (or other yes/no analogs) do not carry that mixed message and are immediately understandable in terms of your actual gameplay, anything that players don't have to be explicitly taught helps them get into the game much faster.
To be honest, even though I see the obvious similarity with mastermind (I often use it as a starting point when describing it to someone), I had never made the direct comparison with the black and white markings. It probably has to do with all the Go I play... ;)
The more I think about it the more I think I have been wrong with my fears of influencing players too much. Also, you are correct in saying that overcoming this is part of inductive logic. Now that I think back, I realise I saw someone I tested on overcome it just this morning. At the start he basically just decided white=yes, black=no and carried on, so it really was just another layer of understanding required.
So I am pleased so say that your logical argument has swayed me and the next version will reflect that. Thank you for persevering :)
---
Related to your earlier "quick play" idea: I tried it a bit and it seems quite interesting. On one hand I feel frustrated in not being able to propose sequences to minimise variables, but on the other hand I do feel it helps pattern recognition and probably also makes you better at the "normal" mode. Also, because games are inherently short, the reward:time ratio is pretty good. I think I'll add a proper in-game option in an upcoming version. I don't think it will replace the normal mode, but I think it could make a welcome alternative, thank you for the idea.
- You can use the mouse wheel to scroll in the history.
- Replace white/black with tick/cross.
- The sequences in a test don't regenerate all the time.
- UI tips display when for the three core screens.
- Hints for each tutorial level.
Next I am planning on adding:
- An in-game general introduction.
- A level where everything is done for you, you just click next (zero-choice tutorial level).
- Quick-play mode.
- Start collecting stats.
As always, any feedback is appreciated (not just feedback on recent changes). I still feel that the learning curve is very steep, but hopefully the recent changes make the UI a bit easier to use for the moment.
I have to admit I don't use the propose screen. I look at the sequences I've got and then I guess because if I'm wrong one of the guess sequences gets added to my history. I suspect you can drop the propose segment of the game without any complaints or lose of gameplay
This is a good point. At the moment you can reset all the tips in the settings menu, but I agree that it should be easier to get help when you need it. I will add small buttons, dedicated to showing the tip for the current screen.
This is interesting. I haven't found this to be the case with others. I personally find it much easier to use a combination of the two. For now I think I'll leave it in (it is easy to remove at a later stage). I did also plan on having the proposals cost a single point and the guesses cost a few (not sure on the number), because you can potentially get more information from a guess. I'll have to see if this idea works in practice, or if it is just more of a complication. When I have implemented some statistics, I should be able to see how many people propose vs guess and make an educated decision.
Anyway, I'll try and explain better what I wanted to say earlier. I don't feel the propose menu is necessary. I'm not sure how hard it would be to implement, but I would like to be able to click on shapes that I used in my history to create a new sequence to propose for testing. I'll try and explain on a picture(edit : dammit the pic is too small, I'll try to make it bigger)
edit2: Nope, I suck at doing what I want with pictures. But just click on the pic and you should be able to read the stuff I wrote.
So that is basically what I meant. It makes the propose process seem a little faster and more coherent to me. I can see all my previous options and build from them.
One other thing I don't quite understand is why you tell me if I have already tried a certain combination. Why doesn't the game just give me the result again?
This sometimes happens. I haven't found that it happens very often, so I haven't focussed much on trying to fix it. I have some ideas about how I could fix it, mostly revolving around generating sequences for the guess screen in a more clever manner, but we'll have to see if they actually work in practice.
The idea is twofold: I don't want to clutter up the sequence history unnecessarily and I want the number of sequences proposed to be a metric that you try to minimise. While it would be easy to just not count duplicate sequences, I feel this might not be very intuitive to the player. To be honest, this just seems like "The Right Way (tm)" of doing it, and I can't really think of a good reason why it shouldn't be like this. Do you find that, even though the game shows you what the previous result of a duplicate sequence is, it breaks up your flow of thought? Maybe I could just add a setting to allow duplicate sequences.
This is one solution, but I don't like that after 20 guesses there is essentially no more "limit" on guesses and the player can just guess willy nilly. I would also like to reward the player for making fewer proposals. I'm not sure how most players will like it seeing their score count down, and I'm not sure I want to induce more stress with the scoring system. Another way will be to just show the score at the end of the level, possibly converted to a star or letter rating. In this case, I can use a fairly complicated system to work out the score, because I won't have to show it to the player. This is all assuming that players don't like to minimise a score, which could just be nonsense...
We've had the same thing happen with DD: The more gameplay information we provide on the final score screen, the more people find different axes to compete on. We have some people competing on time, some for most tiles wasted, some for highest or lowest bonus experience, some for least kills, specific badges, etc.
Announcing a new version (v0.0.7). Changes:
- Add help tip button on core screens.
- The current game is saved and can be continued.
- A tip for the main menu that explains the game.
- Small UI changes.
- Game summary, showing some of the stats I will collect.
Next I will be looking at:
- Collecting stats.
- New game modes (quick-play and something else).
- Tutorial 0 - everything will be done for you, so you can see how a game could progress.
- Two new game modes (Quick and Master).
- Some stats are recorded with Playtomic's analytics (http://playtomic.com).
- Change the background of sequences to reflect their validity.
Quick Mode:
You start with 8 sequences and you only have 8 guesses (no proposals) to figure out the rule. Also, when you fail a guess, all the sequences are added to your history.
Master Mode:
Same as normal, but you only have one guess, so you have to be sure about things before you guess.
Next work:
- Move proposing onto the history screen (Rigormortis' idea).
- Collect and show some local stats.
- Tutorial 0.
- Some internal changes.
Hopefully in my next post I will have some stats I can share with you. In the mean time, I'm curious what others think of the new game modes and if the white/black sequence background works.
EDIT: Minor update (to v0.0.9). Includes local stats and a few bug fixes.
EDIT 2: Another update (to v0.0.10). Improve some of the UI and finish off the rule system. I won't be adding any more rules from now on, just minor tweaking and (hopefully soon) some sort of difficult ratings.
The interface improvements make a big difference; it's a lot easier (and more pleasant) to play, and the information is a lot clearer. (Your and @dislekcia's conversation confused me to the point where I wondered whether I actually had the meanings of black and white inverted!)
I agree with @Rigormortis that you must try to find a way to make it easier to propose; perhaps a solutions is a panel that can be dragged over the history (but still display the last few items), and stays on top (while you make proposals until you drag it away again).
I now frequently encounter situations where the rule I use to "win" is completely different from the one actually used (it just happened to be consistent with the test set). This is often the case with the more complicated rules (for example, I just played a game where I guessed the rule is "all triangles except blue ones must be followed by a hex" when it was in fact there must be fewer triangles than blue items). Not sure if this situation is desirable / acceptable.
Also, I still think you should just ignore a repeat proposal (or just silently reinsert it). Knowing that I guessed something before is not useful, and it breaks my flow.
Not sure if this information is useful: to me it feels like "quick" games take longer. Analyzing sequences takes longer than constructing ones to test potential rules.
Also, a few notes on my general experience (also not sure if it is helpful at all):
In general, I find it almost impossible to guess a new rule. Now that I know potential rules better, I can play much better. My strategy is always to go down my mental list of possible rules and try to eliminate using proposals. (So I still play deductively, inconsistent with your aim). (Of course I don't think this is a problem, but just to share what may happen as opposed to what you think will happen or want to happen [as I understand]).
Now that there are so many rules, I can also not guess a rule if I forget it that it is a possibility.
And finally, I often make up rules that are "half" or too complicated (they over-specify what is really going on). I think it may be helpful to explore this as intermediate steps to find the rule; perhaps there is a way that you can help the player discover a rule progressively. (Hmmm... written out it seems a but abstract... I'm roughly thinking of something equivalent to the kind of markings you can make in some Sudoku games or Mine Sweeper to denote partial information).
(You should also ask other players what strategies they use and what they are thinking, and please share it here. I would be interested to know what range of tactics are available. For example, it was surprising to me to learn that you guess "early", but it did explain some of your design choices).
The solution I'd like to try is as follows: There will be an area below the history where sequences can be proposed. You will be able to click on the symbols in the history to add them to the end of the area and you will be able to remove symbols by clicking on them. To rearrange symbols within the sequence, you can drag them around. To add symbols you haven't seen before, another button will popup a symbol palette you can click on, similar to the symbol palette currently in use. Hopefully this will work, and if not, I will change it again.
I have also encountered this quite a few times myself. I have some ideas on how I could possibly choose different test sequences so that this happens less, but I might find it doesn't work. In that case I can either leave it as-is, prune some of the rules, or add more sequences to the tests. I'm really not sure which is best, and I'll have to give it some thought at some point soon.
Ok, I think I will add a setting to change this behaviour and see how it plays out.
I have also found this myself. I think the mode is interesting, but another name would be better. Also, I find that after 3 or 4 guesses I either find the rule or give up because I can't think of a rule that works. This is a bit frustrating and maybe I need to tweak the mode a bit.
It is helpful, thank you. :)
I don't have a problem with first mentally going through some sort of checklist. This in itself isn't inconsistent with inductive logic. The idea is that after doing this, you should still find that a lot of the time you haven't solved the problem, and then you have to resort to "thinking outside the box."
This is a bit undesirable. I would hope that you could figure out rules again, but I do understand this is easier said than done. Part of the reason I have added so many rules is so that I can't easily keep a mental list of all of them and just run through the list. Maybe once I have managed to implement some form of difficulty rating, then this will mean that you can reduce the subset of rules that are chosen from by picking an easier difficulty.
The method of guessing overly-complicated rules is actually a known advanced strategy in the original Zendo game. By doing this you sometimes are correct, and other times you get a helpful counter-example you can use to spark your imagination. So personally I think it is great you have discovered this by yourself. Maybe I will include some text on advanced strategies at some point.
I wish I had some interesting information to share here, but unfortunately I find that most of the people I test on don't really like this sort of hardcore puzzle game. While this does mean I get a lot of useful information on how to improve the interface and flatten the learning curve, they don't usually play long enough to start forming strategies.
After I improve the proposing (and one or two other things), I plan on calming down with development, and mostly focus on getting people to play the prototype and get constructive feedback I can use. Hopefully I will have some interesting tactics to share then.
Something else I'm looking for help with is: finding a good name. I would prefer to use something other that Zendo because of the clash with the original. Two ideas on potential names: "Zenlike", "The Art of Logic." If anyone has any ideas on this, I would appreciate it.
You can click on previous sequence symbols to add them, or toggle a palette where you can do the same. Clicking on symbols in the custom sequence will remove them. I think this system is a lot easier to use. There are a few things I'd still like to improve, like allowing reordering of symbols and copying whole sequences, but I think the change is big enough to warrant a new version on its own. Any feedback on this new system, or the game in general, would be appreciated.
The next big thing I will be working on is some sort of difficulty rating for problems. I feel that although the game can be quite fun at the moment, the difference between easy and difficult rules can be quite large. Hopefully if I can separate some of the easier rules from some of the harder ones, I will be able to cater to a wider range of players.
I'm also still trying to find a new name. Any suggestions there?
I only read up on Zendo a few nights ago (that, and the references to the Cameron Browne games you gave sent me off on a journey of exploration that is still going on...wow thanks!) and now I think I understand your overall goal better. I see even with humans playing too-difficult-rules are sometimes a problem, so with the computer it is much more understandable. And with humans you have the additional advantage that the game master will scale up the difficulty as he judges the problems too easy. If you get some type of rating going, then trying out new rule types gradually should go a long way to simulate this.
Maybe this is an idea: One flash card app that I use dynamically alters the frequency of displaying a particular card based on your confidence with it (you provide it on a 1-5 scale), and it only introduces new cards so that the number of new cards is roughly fixed. Instead of getting the player to supply confidence, you could measure it based on how many moves the player takes to guess it (if at all). You can then similarly keep the number of new rules fixed. I think one advantage of this system is that it will allow you to add more rules, which will be nice for veteran players, without making the game more difficult for beginners. (And it may also help people train rules so they are less likely to forget them).
Another idea concerning difficulty: perhaps you could limit the number of symbols in sequences, the number of colours and shapes in play, and increase it gradually.
Then... do you think it is feasible to generate a test set that contradicts all non-rules that overlap with the actual rule (well, at least some percentage of the samples in it)? I think without this, other efforts to make the game easier may not work. To do this with a fixed number of sequences may be a tricky algorithm (and not even possible), but perhaps you could at least improve the chances of getting such a set.
[
Here is a clumsy and perhaps slow algorithm that may do the trick: Say R is the correct rule, and other rules are R1... RN. Then calculate (or lookup) the number of sequences that are valid for both. Order the rules by this number (so relabel then r1...rN, so that r1 is the one that overlaps most with R. For all sequences that pass R but not r1, choose the one that pass most other rules, and scratch those from the list. Scratch r1. Do the same with the next r that is not scratched, until you generated enough sequences. (Of course, you need to modify it some to have random sequences appear).
]
One final thought: I am not in love with Master mode... it feels a bit artificially-made-harder (perhaps you could balance the mode a bit by giving more sequences up front). Quick mode is cool, but normal is still my favorite.
I like the idea of limiting the rules that are used. If I can get some sort of initial difficulty rating going, then I can simply have a maximum rule rating. I can increase this maximum with every game, with some relation to the number of sequences used. When I select a rule for a game, it will be easy to make sure it is not too difficult. I can actually turn it into a "campaign" mode where you are shown your progress (current maximum rating vs. maximum rating) and you have to play enough games to unlock more difficult rules. I think it might be best if there isn't a separate campaign mode, but rather playing any of the modes will increase your progress.
This is something else that I could try, but I think the other idea might work just fine and be slightly "cleaner."
I did a count today. There are about 9388 possible rules and 421 if you group "equivalent" rules. Example of equivalent rules: contains a red; contains a hexagon; doesn't contain a blue. This is all from 11 overarching classes that have a number of parameters. I feel that this is plenty and I doubt I will add any more. The next thing I need to do is make an easy way of enumerating the 421 equivalent rule classes. Then I can try and order some/all of these by hand in terms of difficulty. I can then use this ordering as the initial difficulty rating. If this works well enough, I might be able to remove the tutorial.
This is close to the idea I had. The problem with the above algorithm is that the number of rules is quite large (see above), so I figure I will have to reduce this a bit. I plan on going through each overarching rule class (only 11) and hard-coding a few rules I think each rule could be confused with. Then I will do pretty much what you describe. So for "contains an odd number of X," I would have "contains X." I'll then have to try it out a bit and see if it works most of the time, without giving too much information when you see the test sequences.
More sequences upfront could make things a bit more manageable for master mode. I will try it out and see if it works. I personally find it quite enjoyable already, but I'm always looking to improve things.
- The addition of rule ratings (the biggest change).
- You now have a level, which dictates what the most difficult rule you might get is.
- Later I want to make it so that you need to advance enough to unlock the Quick and Master modes.
- Because there is now a much more gradual difficulty curve (due to levels), I have remove the tutorial.
- I have remove hints (I found nobody was using them).
- Master mode now starts with 6 sequences.
- A number of bug fixes.
I have also started work on the next generation UI. It is mostly surface changes, but I feel that together the changes could make for a much more usable interface. I have attached the current mockup:
The next stuff I plan on working on (in rough order):
- Improving test accuracy.
- New UI.
- Intro walkthrough (making use of the new UI).
- New stats system (I am running into Playtomic limitations).
So I am currently looking for the following feedback: Do you find that the level system makes things easier? Does it introduce new rules gradually enough? Do you enjoy trying to get to the next level? What do you think of the UI mockup? Besides the stuff I plan on working on soon, do you notice any shortcomings?
EDIT: this prototype has been re-posted for continued discussion: http://makegamessa.com/discussion/3344/prototype-luminare-a-game-of-inductive-logic