Classicists League Table

[raphtown]

I think there should be more of a balance between points lost and gained in each game. Otherwise we will once again begin favoring those having played many games over those having played fewer games. What do you think?

[InterMPC]

Okay,

I've made a change to balance as you've suggested. Let me know if you think more or less.

[raphtown]

Looks good for a start. If no one has any objections, we could try it out for a bit and see how it goes.

[fellowes]

This sounds like a good idea, except that experience might not always match up to skill (some starters on playdip might be quite good). An idea similar to this one would be that everyone has a point value from 0 to 100 representative of their skill (this can be initialized based on the membership ranking) and you somehow factor the combined loser's score with the combined winner's score to determine points gained/lost. Thoughts?

I like the concept. Here are a few issues to be aware of.

1. If we have a system where you can lose points and there is a cap on how high your score can go, we may see a scenario where people think it’s in their best interest to not play classicist games. E.g. I’m on 99 points (because I’m so awesome). I can’t get higher than 100, but I can certainly drop considerably. Isn’t it best for me just to not play anymore so I can stay on top of the table?
2. We need to define at exactly what point in time we take peoples score for calculation. What I mean by this is; let’s say I start a game and I have 50 points. Just before the game finishes, I have dropped to 45 points because I lost a different game in the mean time. Do we use 45 or 50 to calculate scores for that game?
3. The spreadsheet to be able to handle the table is possible, but would require a lot of work. One solution being that we manually enter each game in to the spreadsheet, which players are participating and what their starting score is, that will be used to calculate the result. We then enter the result into the spreadsheet and the spreadsheet automatically allocates new scores and updates the table. The spreadsheet will be able to handle a limited amount of games using this method though.
4. How do you calculate peoples starting scores? Who decides? Is it fairer to just start everyone on 50 and let their new results determine how they climb? People’s appropriate ranking would quickly come to fruition.

Issue 1 is the biggest concern for me, as the others are all quite fixable.

I believe a solution has already been invented, and is being used from games from chess to Starcraft (both generally 1v1 matches) to League of Legends (5v5 with shifting teams). It is called the ELO rankings system. I don't know the details of the formula, but all users start at a fixed number of points, and their wins net them some function of their score and the defeated score, with stronger opponents resulting in a larger gain. The loss function is the inverse, with the loser losing as many points as the winner gains.

This solves issues 1 (the top score is a "soft cap" since as you get farther ahead of your rivals, even if you keep winning, you will get less and less points for those wins), and 4 (all players start at a fixed value). I don't know enough about excell to deal with 3, and 2 is really a matter of preference, but it seems the most sense to adjust scores after a finish. Any system with concurrent games is not going to be perfect.

[Iamnobody]

ELO is going to be the best system (IMHO).....I've looked at it and I can't make heads of tails of how to do it in Excel. Instead of everyone starting at the same point though we could rank everyone based on average number of points per game.

[raphtown]

If someone can implement this ELO system, we can concurrently try it along with the home-brew system we have and compare the two.

[InterMPC]

I can add another column to my testing spreadsheet that shows Elo ranking change.

[InterMPC]

Okay, I've added an Elo style rating system to the spreadsheet located here - https://spreadsheets.google.com/ccc?key ... y=COvq2tcK

Have a play and let me know what you think.

The 'K-Factor' and the 'Results Weighting' can be changed to get more desirable results. I think it's worth debating what these should be. In Chess the results weighting is 1 for a win, 0 for a loss and .5 for a draw. The K-Factor fluctuates between 10 and 25 depending on how many games you have played.

Also the Elo formula uses a 'predicted' result based on the two participating players. As we have 7 players per game I've had to use a different method for calculating the predicted results. (currently invisible in the spreadsheet.)

Again, the white cells have data you can manipulate while the coloured cells have formulas.

I'm eager to get everyone's input. - Enjoy

[raphtown]

Okay, I've added an Elo style rating system to the spreadsheet located here - https://spreadsheets.google.com/ccc?key ... y=COvq2tcK

Have a play and let me know what you think.

The 'K-Factor' and the 'Results Weighting' can be changed to get more desirable results. I think it's worth debating what these should be. In Chess the results weighting is 1 for a win, 0 for a loss and .5 for a draw. The K-Factor fluctuates between 10 and 25 depending on how many games you have played.

Also the Elo formula uses a 'predicted' result based on the two participating players. As we have 7 players per game I've had to use a different method for calculating the predicted results. (currently invisible in the spreadsheet.)

Again, the white cells have data you can manipulate while the coloured cells have formulas.

I'm eager to get everyone's input. - Enjoy

Must say, I am liking the Elo system. A few things though:

1) If I weight a 2-way at 0.10, it seems to subtract points from the winners as well. Is that supposed to happen?

2) What is the range of scores we are going to expect to be using? Will everyone start somewhere between 1000 and 0?

[InterMPC]

Hey Raphtown,

*EDIT* Anything less that 1/7th of 1 will mean you lose points with this system. 1/7 or 0.14285714285714.