Board Game Arena

Fiora_Laurent wrote: ↑24 February 2021, 23:00 And about those posts, why I edited them? When the adminstrator chated with me:

• He ask me to delete my "Announcement"(which admit I have multiple accounts) in my each profile page in order to reduce the influence.

• Then to show my sincerity, I ask him if I need to delete the post in forum. He said "You can".

• I tried to delete them but I find I have no authority, so I edited them.

• So I came back to tell the adminstrator the condition and ask him to delete it by his admin authority, if he thinks it is necessary.

• He said "That's ok for the forum."

That's all.

Thanks for clearing that up.

Wreckage wrote: ↑24 February 2021, 20:36 Chadmil, you have stated your case well. I have no problem with you playing how you want to play. Also, your win % is higher, I stand corrected. I’ll admit I didn’t look very deeply into the whole ELO issue before my first post. What I do know is, in multiplayer it is nearly impossible to exceed an ELO more than 200 above your average opponent. And even if the miracle of beating 99 opponents out of my last 20 games were possible (which it isn’t) against your last 100 opponents, my ELO would have only gone up 62.5% as much for my victories.

Thanks for replying Wreckage (I'm still waiting for the other guys ^^).
I didn't know about the history of the K factor, that's interesting.

Yes, your elo would go up 62,5% only but go down only 62,5% for each loss! Don't skip this!
It can't be possible to gain 100% as individual games.
We know and we said the ELO system isn't really transposable for 3+ players because we can't exactly determinate a probability of a win in a 3p game from h2h games. Except one particular case :equal elo, it means everybody gets the same chances to win. In a 5p game, 20% chances and 50% in an individual game.
4 individual wins = 40 pts and a probability of 0,0625 (0,5x0,5x0,5x0,5).
Should it be fair to win/lose also 40 pts for 1 victory /last place while the probability is much higher (20%)?!
That's why the % reducing exists, to correct this angle even if he's not accurate.

The question could be different for games involving players with high difference elo because the % chances of winning is difficult to evaluate, but maybe not equal to the multiplication of individual games. I finally agree with Bids on that point. After all, I think all depends on the subjective win odds you give in a 3+p game compared to the combinaison of individual probas.

It's true that in multiplayer we gain and LOSE only 62.5%, but since expert players win more often than we lose, our ELO ceiling would be higher.

Fiora_Laurent wrote: ↑24 February 2021, 21:41

Wreckage wrote: ↑24 February 2021, 20:36 Flora, that player has only played 5 games during the current ELO system. Before that they played with a 25 K factor.....

Sorry, my fault. But my main conclusion maybe still holds, because:
At least the win ratio is not fake. Even if his opponents have only 200-300 ELO, it is still an outstanding achievement to keep 98% win ratio in <multiplayer set>. I think neither you or me, has no capacity to keep 98% win ratio in any kind of game sets. So it is maybe impossible for him to reach 800 ELO once again, but still enough to let him reach his ELO up to 700. maybe also can surpass <Chadmild>.

Actually, my win ratio stays between 98-101.

Chadmild wrote: ↑25 February 2021, 02:30 Should it be fair to win/lose also 40 pts for 1 victory /last place while the probability is much higher (20%)?!
That's why the % reducing exists, to correct this angle even if he's not accurate.

Yes it would be fair. The probability of a win against each of these opponents is calculated the same as for 2 player. I can gain from one and lose to another at the same time. How was 62.5% chosen? if it were 95% we could at least compete, even though still at a disadvantage.

On a new account we start with a K factor of 38. This basically gives a new player 3X the gain and loss of the normal ELO for the first 20 games.
So let me ask a couple easy questions:
If that ELO bonus didn't end after 20 games, but we kept that bonus going continuously, would our ELO ceiling still be the same as it is now?
What if it were only 2X the current rate? Any effect on our ceiling?

(By the way, I think that new player bonus is extremely overpowered too.)

Wreckage wrote: ↑25 February 2021, 12:36 Actually, my win ratio stays between 98-101.

Sorry about that , my fault again.
I only check the win ratio of <Whatshisbucket> but not yours. His win ratio is only 76% and has almost the same ELO to you, That's why I made a wrong conclusion.

Wreckage wrote: ↑25 February 2021, 12:36 Yes it would be fair. The probability of a win against each of these opponents is calculated the same as for 2 player. I can gain from one and lose to another at the same time. How was 62.5% chosen? if it were 95% we could at least compete, even though still at a disadvantage.

How could you say that would be fair ? We agree BGA system allows (+/- by mistake) the same elo pts as for x individual wins. But you have to compare with the real probability to win/lose !
We know it : 33,3% vs 25% (for 3p). Another extreme example with 10 players : 10% vs 0,01% (1000x more !) => If you have the choice, will you prefer playing x individual games or 1 multiplayers game for the same gain ?!
So,that would be a clear advantage for multiplayers game winners in this case...and a disadvantage for losers !
That's the aim of the reducing %, to estimate more realistic % chances in a multiplayers game.
And why 75% for 3p ? Maybe because you win/lose 25% more (25/33,3) ?

On a new account we start with a K factor of 38. This basically gives a new player 3X the gain and loss of the normal ELO for the first 20 games.
So let me ask a couple easy questions:
If that ELO bonus didn't end after 20 games, but we kept that bonus going continuously, would our ELO ceiling still be the same as it is now?
What if it were only 2X the current rate? Any effect on our ceiling?

(By the way, I think that new player bonus is extremely overpowered too.)

You definitely don't want to see that it also applies for the losing points !

Yes, with a high K factor, you could win much more points and go faster to the top, of course.But you lose much more points too.
Example for a 90% probability of win (according to BGA) with extreme K factors :
K=1 : +0,1 pt win / -0,9 pt loss
K=100 : =10 pts win / -90 pts loss

Whatever the K factor, your mean value is still 0 (0,1*0,9 - 0,9*0,1 = 0,9*10 - 0,1*90). Perfectly fair !
And it's important to notice that we talk here about people who played hundreds/thousands games, so that's absolutely statistically significative. There would be a problem if a player reached #1 with only 20 games played.

Chadmild wrote: ↑25 February 2021, 14:36 There would be a problem if a player reached #1 with only 20 games played.

Ok, I Started playing a new account https://boardgamearena.com/table?table=151169591 after Chadmild's last post.
On this new account, I likely won't make it, but if I restarted 20 times, I'm confident that I could make it to 700 at least once, and into the 600s most of the time. The new player bonus is extremely overpowered.

So, I think we have established:
In 2 player games you can maintain 90% actual victories and get 100% of the allocated points on each game.
In 5 player games you can maintain 75% actual victories and get 62.5% of the allocated points on each game.

Your position is that this results in the same ceiling for all players...hmmm.