What about my other point? Do you dislike this? And if so, why?
The more the games the more accurate the elo, so 18-2 would be better than 90-10 depending of how reliable the initial calculation is. That's why I propose a lower K for the beggining, to avoid this impact. This shouldn't discourage players because when they take the normal K, they should be able to adjust their elo (if not close enough to their true level) relatively fast. In fact, there is no really that big of a deal with the artificial initial boost but the "runs". For a normal players who doesn't care about that, after the initial elo, they would reach their true elo eventually, being it lower or higher to the initial one after the 20-games. The only problem is that if it's a system than let you take a high elo "easy" it will encourage runs.
For the variable K, there is no need to change it a lot, or maybe to change it at all (after the initial games). In the chess example I used the elo is calculated per torunaments and not games so the impact it's bigger, maybe, going game by game it's good enough like it is, like in online chess. But, as I said before, it could still be good for the initial ranking.

What about my other point? Do you dislike this? And if so, why?

The more the games the more accurate the elo, so 18-2 would be better than 90-10 depending of how reliable the initial calculation is.

Through my eyes the K-factor only depends on the number of games one is able to make in a period of time. So I would compare the number of games per player per month in warlight with the number of games in chess and intialize the K-factor to that value.

There are also some psychological factors:
1st: How much luck is involved in a warlight game? It is frustrating to loose a game and loose a lot of points only because you had bad luck. Just imagine the following situation:
A, elo 2000; B, elo 1900 => A wins 2 out of 3 games (if I am right); But lets imagine in the game there will appear (with a very high probability) a situation where winning chances are 50%/50%; Than A will loose most probably loose elo against B.
2nd: Someone who plays activly, usally wants to see his elo increasing. But sometimes there are some barriers and he wont do it that fast. This is sometimes very demotivating.

I am not sure how, but I would definetly try to push activity activly and keep players motivated. Maybe a small inflation of the system is wanted as it will push activity.

One also has to think about the time it takes to balance the elo-system in the ladder. Wouldnt a higher K-factor lead to a faster balancing (I know it wouldnt be that accurate at the start, but at least one would see different elo-numbers)?

Edited 11/15/2016 23:52:59

so why do the amount of games matter in an Elo system? i googled what Elo was and i didn't see anything that had to do with it

so why do the amount of games matter in an Elo system? i googled what Elo was and i didn't see anything that had to do with it
I will put an easy example with aproximate numbers, not using the real fórmula:

Player A has 1700 elo now, but with a "real strenght" of 1800. If he plays against a 1700 he wins or loses 16 points (K=32) but he should win more than losing because he "is" a 1800. Let's see what happens if he go 1 game at a time:
A vs 1700, wins --> +16 --> 1716 elo
A vs 1700, wins --> (now he get a bit less, let's say 15 points) --> 1731 elo
A vs 1700, wins --> +13 --> 1744

Now let's see what would have happened if he played 3 games at the same time at the start:
A vs 1700, 3 wins, each one 16 points --> 1748

The difference it's only 4 points, but with more games and elo diferences it's a bit more. As I said with "game to game" calculations it's not so big a deal, that's why I focused more in the initial elo calculation ;)

Edited 11/17/2016 20:16:57

From what I read, a solution to much of the discussion above could be achieved by not sorting/ranking on the calculated (mean) rating, but to correct for activity (and accuracy) by subtracting the standard deviation (see TrueSkill, that subtracts 3 times the standard deviation, which is a more harsh version of this).

This is a conservative ranking that gives advantage to players whose rating is more accurate, often by playing more.

I don't see why you would let games expire. Elo was originally made for non-expiring games, I believe and all non-expiring updating rating systems simple do work better than those that use expiring games.

Very much agree about expiration with MW.

I don't see why you would let games expire. Elo was originally made for non-expiring games, I believe and all non-expiring updating rating systems simple do work better than those that use expiring games.

For shortening time calculations. He explained it above and that's one of our discussions :)

Edited 11/17/2016 23:26:28

I don't see why you would let games expire. Elo was originally made for non-expiring games

ELO was originally used for chess, which is a single game with no parameters. This MD ladder is a grab-bag of 35+ templates which may change over time.

We can expect that some players will have a higher likelihood of winning on certain templates. So in that regard, ELO might be poor choice. Expiring games doesn't improve on that, but helps ensure ratings stay relevant as the template grab-bag changes.

Tracking an ELO for each template and then somehow combining those ratings might be interesting.

Tracking an ELO for each template and then somehow combining those ratings might be interesting.

Runs into the problem, that you can blocking templates and the number of games on each template may be very different and therefore the elo may not be that comparable.

Furthermore, how do you combine these ratings. Also I think it isnt that intuitiv. How do you match players then? Right now you match them by elo and then choose the template.

Edited 11/18/2016 06:31:21

Player A has 1700 elo now, but with a "real strenght" of 1800. If he plays against a 1700 he wins or loses 16 points (K=32) but he should win more than losing because he "is" a 1800. Let's see what happens if he go 1 game at a time:
A vs 1700, wins --> +16 --> 1716 elo
A vs 1700, wins --> (now he get a bit less, let's say 15 points) --> 1731 elo
A vs 1700, wins --> +13 --> 1744

Now let's see what would have happened if he played 3 games at the same time at the start:
A vs 1700, 3 wins, each one 16 points --> 1748

The difference it's only 4 points, but with more games and elo diferences it's a bit more. As I said with "game to game" calculations it's not so big a deal, that's why I focused more in the initial elo calculation ;)

wouldn't that be the reason for using Elo? i was asking about why the amount of games matter

Edited 11/18/2016 14:47:58

Runs into the problem, that you can blocking templates and the number of games on each template may be very different

Yes, choosing/blocking templates further distorts the meaning of a single ELO rating.

However, blocking is a nice feature that makes the competition more fun. It allows players to specialize and avoid templates they don't like or are not good at. Actually, template blocking seems to increase the need for multiple ELOs and a composite metric.

Furthermore, how do you combine these ratings. Also I think it isnt that intuitiv.

Here's an attempt:

Consider a system with 1 elo per template, and you get points based on your ranking from each elo system. Say you are ranked 3 for Guiroma. One way is for points to be given by f(k) = 1/k or 1/3 points. If more people join the ladder, your points won't change.

If one person had rank 1 in all 35 templates, they would have 35 points. If somebody was ranked last in all templates, they would have just above zero points.

How do you match players then? Right now you match them by elo and then choose the template.

You would match based on points, and randomly choose the template. Then the outcome of that game effects only the elo for that template.

Edited 11/19/2016 01:35:37

The above point system is almost certainly not practical as is, but maybe it could be tweaked.

@J_Dog: the amount of games matters because of https://en.wikipedia.org/wiki/Statistics.

Also, "real strength" of a player for a certain game would be based on knowledge of the number of neural connections, FLOPS, experience, current brain/body state, etc. In the absence of that knowledge, the best we can do is try to estimate real strength based on game outcome data.

because he "is" a 1800

You can't really say that for sure. If he has another account with a rating of 1800, he should be using that account:)

Edited 11/18/2016 21:51:12

because he "is" a 1800

You can't really say that for sure. If he has another account with a rating of 1800, he should be using that account:)

Ratings derived from the games played are just estimates of a player's true skill. When he says the player "is" a 1800 he means their "true" rating which Elo is attempting to estimate.

Edited 11/18/2016 22:00:20

How did he compute 1800? Can he estimate the "true" rating better than ELO? After 1000 games if the ELO is still 1700, is his estimate wrong or is ELO wrong?

It was just an example -.- The example was though for someone that doesn't have a elo that reflected his/her true strengh like a new player...
But new players it's not the only issue, maybe someone who left and returns to the ladder later being an overall better player :/

And yes, one elo per template would be nice but the idea it's to use something way more simple than that. And at the end, if you match with that system, you could be matched against players with a skill level much different to yours in that template only because the average of both players it's similar :/ The idea it's try to avoid that.

server seems to be down

Yeah sorry about that. Pushing out a new update right now...


EDIT: Fixed.

Edited 11/18/2016 22:59:36

We have added a new feature to track the rank and rating of players over time! Check it out on the player profile pages. You can also zoom in to a specific portion of the chart by dragging over a region. Click again to zoom back out. All hail Muli!




Edited 1/4/2017 18:46:01

Shouldn't Red Dead Redemption be Straight Rounded?