Player rating for tournaments across all installments of Heroes. — Rating systems.

AmberSoler

19 years ago

Auto-translated

You are both talking about the same thing... BotINKi is more accurate in terms of terminology, IronAxe uses simpler words... :) But the essence remains the same - we are simply making minor adjustments to the blitz results (for all participants, without infringing on their interests) in order to align the figures with the existing ranking...

Сначала было слово...
FizMiG v.2.0 *** Очередное обновление справочника! *** http://vk.com/fizmig

In reply to AmberSoler

Kyle

19 years ago

Auto-translated

AmberSoler;99508

You are both talking about the same thing...

No... I disagree...
As far as I understand, they are talking about different things...
BotINKi supports the idea of calculating the rating using the formula:
Game rating = (Blitz rating/5)*K
where K is the coefficient that brings the leader's rating closer to a rating of 100 points. i.e., K=100/96.89 = 1.032
In this case, it takes into account how much rating a participant gained during the Blitz Tournament.

IronAxe, as far as I understand, is talking about something slightly different...

IronAxe

What coefficient are you talking about? First place in the blitz - 100 rating points. The rest according to the formula. We calculate based on the final table for the blitz.

From this, I understand that:

Game rating = 100-[(100/N)*(R-1)]
where
R is the player's position in the final table,
N is the number of positions in the final table
i.e., this is the formula that was used to calculate the blitz rating in the preliminary stages...
But in this case, it doesn't matter how much rating the player gained... it only matters what place he took...

In short, we are getting bogged down in details again... that's why I suggested not reinventing the wheel... ;)

Below is a table comparing all the listed options (there are currently three of them).
From this table, it can be seen that:
my rating is the most conservative...
The rating with the coefficient is slightly more generous...
And the rating based on the position is the most generous...

Sincerely.

Attachments 1

1 file attached • Total size: 4.1 KB

zip

FT_Rating

4.1 KB 97

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров

AmberSoler

19 years ago

Auto-translated

Let's just evaluate all three options in terms of how they reflect the actual state of affairs... It is obvious that your option and the one I proposed (with a correction) are essentially the same thing... Only my version is adapted to the rating calculation rules, namely—the leader receives 100 points, and the others are calculated relative to them...

That means two options remain—mine and Iron's (a classic of the genre :D)...

Which one is closer to reality? Second place trails first by only one point... On what basis can we create a gap of four or five points between them? The current blitz rating more fully reflects the merits of each player, and therefore, recalculating it BASED ONLY on the position taken (without considering relative placement) is inappropriate...

I hope we don't create the preconditions for another argument over nothing.

Сначала было слово...
FizMiG v.2.0 *** Очередное обновление справочника! *** http://vk.com/fizmig

In reply to AmberSoler

Kyle

19 years ago

Auto-translated

AmberSoler;99584

Which one is closer to reality? The second place is only one point behind the first... On what basis can we make the gap between them four to five points?

Well, here you can ask, what reality are we talking about?

Further, to the point:
a gap of four points? Well, you shouldn't have taken a place lower than your opponent... :D
And on what basis do we introduce some kind of coefficient that gives the biggest plus to the leaders... and those who are in the middle of the table are neither hot nor cold from it..
i.e., if you look, the leader was given 3.11 rating points,
but, for example, a player in 20th place only 0.63 points?
Why such unfairness?

AmberSoler

I hope we don't create the preconditions for another pointless argument.

I'm afraid that's exactly what's happening... :D
and that's exactly why:

Kyle

That's exactly why I didn't want to reinvent the wheel...

Sincerely.

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров

AmberSoler

19 years ago

Auto-translated

1. Blitz has its own rating calculation method. 2. This rating is not identical to the General rating. 3. Combining the ratings involves adapting it to the General rating (the main idea is that the leader has 100 points, and all others have points depending on how far they are behind the leader). 4. Blitz does not contain the points of the General rating; they cannot be compared (!). These are just Arabic numerals that should be considered only as a criterion for the relative position of players on the rating scale. This is a dimensionless value. 5. We see the real rating numbers for each player for the first time only after adaptation, and the subtraction procedure (finding the delta) is inappropriate. It doesn't show anything. I will give an example again: we could consider the Blitz rating to be based on 1000 points, and then the leader would have 969 points at the moment, and the outsider would have 10. After converting to a 100-point system, your example loses its meaning because we take away 869 points from the leader, but only 9 points from the outsider... :)

Сначала было слово...
FizMiG v.2.0 *** Очередное обновление справочника! *** http://vk.com/fizmig

Kyle

19 years ago

Auto-translated

to AmberSoler: based on everything mentioned above, it follows that calculating the player rating based on the position occupied in a tournament is the most accurate and correct approach! :D
because it is better decoupled compared to the rating based on the blitz rating.
that is, it best fulfills point #4

AmberSoler

4. Blitz does not carry any points of the overall rating - they cannot be compared (!), these are just Arabic numerals that should be considered only as a criterion for the mutual arrangement of players on the rating scale. This is a dimensionless quantity.

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров

AmberSoler

19 years ago

Auto-translated

Well, the oscillator reflects this most accurately and precisely – it takes into account both the position (which you insist on) and the relative position of the players, as has already been discussed here... :)

I am not against this method of calculation.

Сначала было слово...
FizMiG v.2.0 *** Очередное обновление справочника! *** http://vk.com/fizmig

In reply to AmberSoler

Kyle

19 years ago

Auto-translated

AmberSoler;99592

Well, the oscillator reflects it most accurately and precisely – it takes into account both the location (which you insist on) and the relative position of the players, as has already been discussed here... :)

What about this... it's an idea... I think it could be added here...
I'll create a table right now... :D

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров

In reply to Kyle

Kyle

19 years ago

Auto-translated

In short, here it is....
a "Oscillator" column has been added.

however, the leader can't reach 100 points...

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров

In reply to Kyle

AmberSoler

19 years ago

Auto-translated

Kyle;99597

True, the leader doesn't end up with 100 points...

Nevertheless, the oscillator should give the winner 100 points — remember the beginning of all our discussions on this topic? :) It is not calculated quite correctly; you should probably use my rating version and Iron's version in the calculation... You have again taken abstract (parametric) figures and started adding them to the real rating... :)

And yet, this was already very vividly discussed by all of us once... Why complicate things and start this topic anew? After all, as of today, the rating is maintained according to three different systems, including the oscillator... I believed, and continue to believe, that the oscillator most accurately represents the result of a player's efforts in a tournament, but here again — everything is at the discretion of those who maintain and calculate the rating... Any version is essentially acceptable... The main thing is not to delay making a decision and to finalize the results.

Сначала было слово...
FizMiG v.2.0 *** Очередное обновление справочника! *** http://vk.com/fizmig

In reply to AmberSoler

Kyle

19 years ago

Auto-translated

AmberSoler;99599

Nevertheless, the oscillator should also give the winner 100 points – remember the beginning of all our discussions on this topic? :) It's not quite calculated correctly; you should probably use my rating formula and Iron's formula in the calculation...

but, according to the theory of the oscillator:
1) we take the result in the tournament
2) we take the place taken
3) we calculate the average
Is that right?

But, if we take your version with a coefficient, then in the first point, we will not get the "result in the tournament," but some kind of modified thing... that is, a pure oscillator will not work...

but this problem can be fixed...
the mistake crept in because neither my formula nor your formula can be used as the first point... but we need to use the results of the final... and imagine that this is not a rating, but, for example, experience (or the number of Armageddon swords earned :D)... and then everything will immediately fall into place...

In short, I recalculated it...

Attachments 1

1 file attached • Total size: 4.4 KB

zip

FT_Rating

4.4 KB 96

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров

Kyle

19 years ago

Auto-translated

I think that the oscillator is the most appropriate tool to use in this case... :smile01:

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров

AmberSoler

19 years ago

Auto-translated

Kyle;99601

but this problem can be fixed...
the error lies in the fact that you cannot use ... your formula as the first item.

I think you didn't understand the meaning of introducing my coefficient, which brings the result closer to a 100-point system... :) You just did the same thing (!), compare the numbers you just obtained with my suggestion:

It is not quite correctly calculated; you should probably use my rating option and Iron's option in the calculation.

There are exactly the same numbers... :)
What problem did you find in my suggestion and solve it? :)
By suggesting that the average blitz rating be considered as a kind of "experience" for the purpose of calculating it based on its final rating, you performed the same "adjustment" procedure... I just introduced this coefficient, which is taken into account in your latest formula in the form of "D4/$D$4" - so as not to go too deep into Excel... :)

Oh... mathematicians... ;)

Сначала было слово...
FizMiG v.2.0 *** Очередное обновление справочника! *** http://vk.com/fizmig

In reply to AmberSoler

Kyle

19 years ago

Auto-translated

AmberSoler;99603

The numbers are exactly the same... :)
What problem did you find in my suggestion and how did you solve it? :)
Ah... mathematicians...

Well... there are still some small differences for some players... :D

But, honestly, these are just rounding errors... :o

Yes, in principle, we arrived at the same numbers... but with slightly different formulas... :D

Кто состоянием своим доволен, тому жить весело. (с)Алекса́ндр Васи́льевич Суво́ров