I think you need to decide whether you’re trying to use statistics and “science” to justify a position, or research a position.
As I’m sure some of you are probably thinking, there are easier ways to playtest with acceptable accuracy than to go to all this trouble building something around a “magic” number that was created through preset rules. But being the curious, stubborn person that I am, I want to solve it this way; it’s fascinating; it’s unique; and although it’s imperfect, it will work for me if it’s indeed possible.
Given that quote, I think you already know what I’m talking about. So why say something you already know? Because I’m saying, as a friend would, just a reminder about perspective. I’m not saying it’s wrong to do what you want to do how you want to do it. But is it really about detecting balance issues in the game, or is it about exploring a mathematical process you find interesting? (In reading through comments after starting this reply, I noticed CWO Marc already brought this up, pardon the repetition).
Is it about getting a product out the door, or about things you may find along the way? If the latter, then by all means, most certainly look into the round robin tournament algorithm. If the former, well.
As you may have noticed, I’ve been almost maintaining a bit of secrecy about what this game is, and that’s because it’s actually something unique we came up with. This is just me being hopeful and daydreaming, but I think the game may have market value, even if only in a digital format.
I wish you every success, but from what I’ve seen, my understanding is you can secure rights to things related to physical processes, but you cannot protect ideas. So you can copyright artwork and such, but you cannot protect, say, the idea that a certain number of dice are rolled and having a number or less indicates a “hit”. You can secure rights to the “Tap” symbol in Magic the Gathering but you can’t prevent other games from rotating cards to indicate conditions. And so forth. I am not a specialist in copyright or trademark, or a lawyer, nor do I have much familiarity with international law, but you may find what I’m saying to be the case.
Am I saying to not maintain secrecy? No. Maintain secrecy by all means. I think were you to open the idea to others, you would find few others that would be able to contribute meaningfully, and a lot of unasked for unproductive advice that even when accurate (which it often wouldn’t be) simply wouldn’t apply to your particular circumstance. When you have the product in release state, you’ll have the advantage of competitors needing to take time to reverse engineer the systems (so to speak) so even if they copy it, they may not be as successful. (Though be aware someone can take what you’ve done, slap on a few distinct features to try to make it more marketable and/or profitable, then overtake your market share. There are literally companies that do only this.)
All right, so the project is secret, but still, a few comments.
- I recommend using a modified Elo/Glicko formula, with, hm, I don’t know the word for the concept.
As I understand it, Elo/Glicko something like respective player ratings are plugged into a formula, and you get an output that you can interpret to get the expectation a particular player wins or loses. After a match is completed, a player’s rating rises or falls, which is how the system self-corrects.
You probably already know that the, is it “k” value changes depending on the situation? Particularly in I think chess federations FIDE and USCF (are those the names?), you have a high k value while playing your first I think it’s 25 matches, so your rating can correct more quickly. After you finish your provisional games you’re assigned a rating, and instead of using the provisional k value which is high, you then use a new k value for all future games, which reduces the ratings gain/loss you would otherwise get. Then after you reach - I don’t know, Master or something, your k-value decreases yet again. And I think Glicko maybe introduces something like a consideration for real time elapsed, like if you don’t play for a long time, you may get rusty, but your fundamentals are still strong so you can catch up, but maybe you don’t get the rust off because you’re not active . . . colloquially speaking.
So obviously it’s going to be a little more complicated because instead of using only two ratings, you’d ideally use multiple player ratings.
Well, that much is pretty predictable I guess, but I think the tricky part is assigning instead of a constant k-value (or even a k-value that varies under different conditions), you assign a non-straight-line curve to player actions.
Suppose I say you have a 10% chance to win a game, and that you typically play a “conservative” game. If you know your chances are 10%, then aren’t you more likely to play differently than your usual manner? This may increase your chances or decrease them, but functionally you’re using an entirely different playstyle.
So you see what I mean about using a non-straight-line curve, right?
Anyways, there’s a lot of hurdy-gurdy mysticism mumbo-jumbo in the mathematics because who really knows if the curve you’re using is correct? Just as I think initially for Elo systems they used a standard distribution for prediction but it was found the normal distribution was more accurate, or was it vice versa? I forget. And I won’t even get into the specific value of k; anyways you use data received to correct the formula . . . er . . . okay, hopefully you understand what I’m getting at.
Which is, instead of having a readily understandable system that perhaps breaks apart under examination (as I mentioned when talking about cabals of players, differing motivations, and changing playstyles) - instead of saying “well, if you take this PARTICULAR case then the whole testing regimen falls apart” - instead of having a more readily understandable process, instead you use some god-awful mutation of Glicko which is itself a mutation of Elo, which has all sorts of complicated variables and nonlinear this and that, and if investors dare to try to understand it, you just look them in the eye, nod, and say you understand their concerns and can they perhaps point out which specific part of the equation should be changed? You say major federations using ratings formula found it necessary to change formulas over time, you show how you’ve built in self-correcting measures so feedback from data self-corrects the specified variables in the formula, and I’ll bet you dollars to donuts they’ll suddenly find something very interesting just outside the window.
Mind, you do need some serious mathematics and methodology in case someone takes your report back to some serious analysts, but your system is even somewhat reasonable, then by using formulas instead of an extensive and rigid paradigm, you have more freedom.
Of course, you then say that the formulas need something like 5,000 inputs (or whatever) to arrive at a baseline, then say you expect the formulas to change depending on data, then you’re covered and everyone’s covered . . .
Right. It’s not that you are TRYING to be obscure (wink wink). It just works out that way (wink wink).
- For a digital format, I suggest multiple winners per game and multiple variable win conditions.
Say you have a 5 player game that ends up with 4 losers. If it’s a board game played live, you know the actual people, maybe you’re all watching a movie on TV or whatever, sure, fine, whatever, if someone loses again and again they don’t care, they’re there for the pizza and wings and football game. But digital? They’re not there for the pizza, wings, football game, if digital you’re taking them away from that. So as manipulative as it is, it’s even more important for players to feel their in-game actions matter, that they had fun, that they had a chance to not just win, but win on their terms, and the game needs to be designed that way.
- Corrective and catchup measures but cut it short. Let’s say game starts, you get screwed by dice (if there’s dice in the game or whatever). Then you get screwed some more. Then you’re like eff it, you don’t even want to play, but you’re stuck because the rules don’t just let you quit. (Or do they?) So how are you going to stay interested in the game? If there’s a comeback mechanic and/or multiple win conditions, that’s one way.
But you have to can’t get carried away with comebacks, you have to make sure the game ends before too long. Don’t get me wrong, there’s players that want the fourteen-hour-experience, and if you really want to cater to that, then by all means, totally ignore what I’m saying here (know your audience and cater to them!) But if you’re going for a casual market, then players don’t want to get into a game that offers every opponent a comeback, then the game just gets drawn out.
In closing - all right, so your process is secret, and you will do things your way. Sure. But I want you to realize cabal formation in multiplayer groups is natural and expected, that player motivation is going to be different to simply trying to “win” some abstract “game” by the rules in the box. You must not think of things in terms of one player “bullying” another. Nor should you assume every player is going to play it fair, above board, and try to win only by themselves and for themselves. It should be completely natural and expected in your predictive model that when there are two friends in a game, one of which is bad at the game and knows s/he is bad, that that friend will sacrifice themselves to assure the other’s success. It should be completely natural and expected that a player that is losing badly will try to eliminate a winning player, not because they are spiteful, it’s very natural for players to want to feel that they are personally significant and that their decisions matter; by changing the outcome of the game they feel they accomplish just that.
more philosophical in nature than I actually feel is necessary
If you already have your core clientele and marketing strategy all worked out, of course it is more philosophical than necessary.
But since you know I have no way of knowing you’ve done that based on what you’d written to that point, it’s a thinly veiled hint that I’m wasting my time - and perhaps yours, if I may be so bold.
It is indeed a valuable life lesson to remember that no matter how much time one may spend on something, if it is not exactly what someone else wanted, such effort may at best go unremarked, or even earn a reprimand.
I thank you for your reminder to value others’ time, and to spend one’s own time wisely. It really is good to be reminded of these things time to time.
being the curious, stubborn person that I am
Me too, you know? Me too. I hope you will view what I have written in this light, and pardon me for not having developed and presented an algorithm for you to get the special round robin tournament generator you wanted. (Edit - having read this after it was written, I feel that may have come off as having intended disrespect. That was not the intent; as I commented earlier I do think it’s just an application of combination and permutation mathematics; I could write a computer program to generate and test sequences myself if I had some more programming knowledge; I don’t imagine it would take more than half a day if I were familiar with the commands I’d need. (Though if I’m wrong about the core mathematics (which would become evident in the check step after data generation) it would take a lot longer.)) But you can see if it really isn’t a big project (which I think it might be), no disrespect was intended, if I could have done, I think I might have done it for a bit of diversion even. Well, I still have a ways to go with programming so eh.
I saw from a post in this thread you seem to have already found one solution; congratulations and I wish you every success with your project! I would imagine, that solution being found, that your objectives with this thread are met?