Rating discussion

[ufm]

Automa mode in multiplayer games
- Do not have any impact on ratings at all
- Ties are possible

Co-op games
- The final game score changes the rating of each player
- Ties are impossible (in almost all co-op games), you either win or lose as a team

I follow that. The most common "co-op" mode is solo mode, in both this and other games I've done. I've alway been consistent about a zero being a tie. While I hear you about ties being impossible in Daybreak, I'm still unlikely to make things inconsistent by carving out a special exception here.

Thanks for the scoring proposal. If your winning score would scale with the difficulty, would your losing score also scale? Or is it only upside for higher difficulty, but no real downside? It does feel pretty silly that your score only ever goes up, even if it has nothing to do with Elo.

The downside is lower win rate itself.
If players are not skillful enough, they will lose more time to earn ratings compared to the default difficulty.

I'm not suggesting to make Daybreak as a special exception:
Daybreak is not an automa mode of a competitive game, but a cooperative game. The category itself is different.
For now, it looks like an exception as you implemented two co-op games (Daybreak and Mini Rogue), but it won't be if you implement more.

I never reduce co-op rating on losses, as I consider co-op ratings XP meters, not skill indicators.
They can still roughly display some skill levels as better players will accumulate the rating faster, while removing significant issues raised constantly (for example, avoiding beginners or abandoning/conceding the game to prevent rating losses).
Also, attempts to measure player skills with artificial Elo ratings have created all kinds of loopholes and mishaps.
You can search 'Elo' or 'rating' topics here: https://boardgamearena.com/forum/viewforum.php?f=102

[Kayvon]

I'm not suggesting to make Daybreak as a special exception
...
For now, it looks like an exception

I hear you. I'm happy you have a system that's working well for you in the games you develop. A lot of people really enjoy that you do it that way.

I need you to hear me also, even if you don't decide to agree with me. Positive for win, zero for tie, negative for loss. A lot of people (not you, obviously, but others) prefer it this way, too. I appreciate your point of view, but the arguments have only pushed me towards not creating an exception.

Also, attempts to measure player skills with artificial Elo ratings have created all kinds of loopholes and mishaps.

As I said in my last post, this has nothing to do with Elo for me. It just seems silly to have absolutely nothing on the line for a loss. If I implement a system where you can gain lots of points, I may also implement the downside on it as well. Reading a bunch of Elo posts (again) has no bearing on this system for the exact reasons you indicated.

I sincerely hope you can find some acceptance in other ways of doing things, even if they're not what you would have personally chosen.

[ufm]

I'm not suggesting to make Daybreak as a special exception
...
For now, it looks like an exception

I hear you. I'm happy you have a system that's working well for you in the games you develop. A lot of people really enjoy that you do it that way.

I need you to hear me also, even if you don't decide to agree with me. Positive for win, zero for tie, negative for loss. A lot of people (not you, obviously, but others) prefer it this way, too. I appreciate your point of view, but the arguments have only pushed me towards not creating an exception.

Also, attempts to measure player skills with artificial Elo ratings have created all kinds of loopholes and mishaps.

As I said in my last post, this has nothing to do with Elo for me. It just seems silly to have absolutely nothing on the line for a loss. If I implement a system where you can gain lots of points, I may also implement the downside on it as well. Reading a bunch of Elo posts (again) has no bearing on this system for the exact reasons you indicated.

I sincerely hope you can find some acceptance in other ways of doing things, even if they're not what you would have personally chosen.

No offense taken. For cooperative games, the developer decides their own rating system, after all.

[Kayvon]

I've implemented a varying scale of VP based on several factors. These may be adjusted going forward, so feel free to recommend other values for the gains and losses.

If you feel that the new system is worse, add a comment. We can always go back to the simple +1 for victory and -1 for loss. I will not be implementing an entirely different system, so any suggestions that a different system is preferred to the new one will be counted as a vote to return to the old one.

A) Games are worth 5VP by default (6VP for 4player games)
B) Easy difficulty are worth 2VP fewer, hard games 2VP more. Extreme difficulty is +4VP.
C) Group challenge cards double or half the sum of A+B (for group challenge/handicap)
D) Individual challenge cards add or subtract 1VP to the value from step C (for challenge/handicap)

The result is your final score, positive (win) or negative (loss).

Edit: Corrected "of" to "or" in typo on step C

[Romain672]

Let me see:
B) From what I have seen, the difficulty of the game don't change much the result, it's slightly harder, but having more emissions at the start is really good.
C) Group challenge which double have some really harsh ones (those similar to +10 emissions turn 1) and can goes extremely fast out of control, so doubling look nice
D) Individual usually doesn't matter much like B, so +1 is nice too.

So yeah the balance look perfect, and +5 default is something I really like to be able to let enter people with 20 or 100 victories, thanks for that

[Jellby]

For an elo-like rating, I'd expect higher difficulty to mean larger gains for a victory and smaller losses for a defeat...

[Kayvon]

For an elo-like rating, I'd expect higher difficulty to mean larger gains for a victory and smaller losses for a defeat...

These may be adjusted going forward, so feel free to recommend other values for the gains and losses.

You'vesaid what you don't like, but without a constructive suggestion it's just complaining. What, specifically, would you like to see instead?

[Jellby]

I've only played this game once, so take my opinion for what it's worth

I like the approach in Hanabi, where the average elo of the players is compared with a "bot elo" (that depends on the difficulty of the game), and a victory is computed as a tie for elo purposes. That gives, among other things, a mathematical top to the maximum achievable elo.

Other than that, I could suggest a standard 5 points gain/loss for victory defeat. More difficulty would mean higher gain for victories (6, 10, whatever), but lower loss for defeats (4, 2.5, whatever).

[quietmint]

In co-op, I think the only point of ELO is a rough gauge of experience with the game. The concept of measuring an individual's skill at a co-op game basically makes no sense -- you are winning or losing as a team. There are too many variables to figure out who is "to blame" for losing or who is "to credit" for winning. That whole idea is completely contradictory to co-op gaming!

So when BGA says someone is a "Master" at Daybreak it should mean "they know how to play". Absolutely nobody assume "Master" means "they win 100% of the time" -- which is the system you built. It means that they can win -- not at all how often they win. I want the labels "Beginner" vs "Master" to clue me towards whether a person knows what they are doing vs. needs a helping hand or an explanation of how the cards work. The more you play, the better you get at a game. If you don't play a game for a year, you will get rusty so I'd like some kind of ELO decay over time, but that's irrelevant because it would need to be implemented site-wide by BGA, not by individual games. Since that does not exist yet, yes, in short, I expect ELO can only go up.

If someone plays a challenging mode and loses, why is that a deduction? A lot of the game is luck. The more challenging you make it, the more likely you are to lose. That doesn't mean you're a bad player, it just means you want an extra challenge. Why relegate these people (who are really the ones that deserve "Master" title) back down to "Apprentice" since they are trying something more difficult? In fact, you're rushing them back down. Once you know what you're doing, losing a game on easy is far "worse" than losing a game on hard. By definition, you should be able to win more often at lower difficulty... your scale is completely backwards for punishing an expected/likely loss (on hard mode) as far worse than an unexpected/unlikely loss (on easy mode). Really there should be no ELO loss at all, but definitely it shouldn't be like this where it's worse to lose harder games.

Going with a "bot" system like Hanabi is overkill. It's very difficult to gauge exactly exactly how difficult all the options are, especially in combination with each other where each player has a different starting power, each player can adjust their personal difficulty, and the group can adjust their overall difficulty, and again the entire game is co-operative (so everyone should be working together, so how can you attribute who did what). It's not worth your time to do something like Hanabi. Make a simple scale where ELO goes up for winning, going up more for a "more difficult" game. Don't punish people for losing. Definitely don't punish people for playing on harder difficulty settings.

[Kayvon]

I already responded to your reply on the suggestion ticket itself, but it seems like you didn't see that before responding here, so forgive me if I repeat myself a bit.

If someone plays a challenging mode and loses, why is that a deduction?

In all of our adaptations, we use negative numbers to indicate a loss, zero to indicate a tie, and a positive number to indicate a win. It's been discussed throughout the thread, so I won't rehash it here. We will not be deviating from that, but I will cordially welcome you to implement it differently on your adaptations if you feel strongly about it.

A lot of the game is luck.

If luck is a significant factor, might I suggest that you simply ignore the ELO altogether?

You may recall my earlier statement:
> If you feel that the new system is worse, add a comment.
> We can always go back to the simple +1 for victory and -1 for loss.

As noted earlier in this thread and in my reply on the suggestion ticket, I'm open to constructive suggestions (i.e. "here's what I would tweak"). I'll take this complaint as a vote to revert back to the previous, simpler system.