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

Registered: 2/15/2006

Posts: 1,676

#181 1/8/2007 11:56 AM UTC

Auto-translated

Here's what it means to be absent for 4 days...
I've managed to grasp almost all the information, and I'll try to comment very briefly.

First, let me clarify that GRAILS and the GAME RATING are TWO DIFFERENT THINGS!
For calculating GRAILS, the scoring system must be strictly fixed!
I'm tired of explaining why...
They should be awarded according to the scheme that is currently in effect.

P.S. Soon, a special section dedicated to GRAILS will be created! Where you can discuss, criticize, and express your opinion, and much more.

Initially, I started responding to each post

AnberSoler, post #20:
1) I agree
2) For calculating the rating, yes, there are no Grails. Or do you mean something else?
3) Why do you think that a system based on the number of participants will be more effective than a fixed system? Personally, I don't think so, and I have repeatedly given examples:

Well, okay, I'll give another one:

Rating calculation according to your system

a Heroes 1 tournament was held, with 3 participants.

Player_1 took 1st place
Player_2 took 2nd place
Player_3 took 3rd place

We get:

Player_1 - 23
Player_2 - 21.5
Player_3 - 20 points

A Heroes 3 tournament was held, with 10 participants.

Player_1 - 65 points
Player_2 - 60 points
Player_10 - 20 points

Where is the fairness?

According to my system, or IronAxe's system, the result for 1st place is fixed, and the first places will receive the same number of rating points!
It is impossible to give preference to one tournament or another; all tournaments are good; choose to your taste!

P.S. Although, for the sake of experimentation, we can introduce your system and see which one works better.

4) I agree
5) I agree
6) Thank you for your concern, but it is more convenient for me to use a table with a number for each save when preparing the report. That's why I introduced this requirement.

Alan, post #27: Can you explain to me how your system differs from mine? Of course, apart from the numbers.
I understand that in your scheme, points are assigned fixed and do not depend on the number of participants.

7) I told IronAxe that you shouldn't age the rating, as this will lead to an imbalance...
To be honest, I don't like the aging either.

Sage, AmberSoler, you are absolutely right; the RATING SHOULD BE SEPARATE FOR ALL PARTS OF HEROES! I don't think there's anything to discuss here!

AmberSoler, post #32: Yes, you are probably right about the 50 participants and 2/3. But where were you before?
In any case, my system is best suited for awarding Grails, at least for now... Well, let's think about calculating the GAME RATING.

Regarding the two saves, it is up to the respected AlexeyPanky to decide; I think that one save is enough for this tournament.

,but then I realized that this is not realistic, so I will answer specific questions-answers.

Post 122, AmberSoler

1) Current rating - I agree
2) Cyclic rating - I agree
3) Tournament rating - either I didn't read it carefully, or I don't understand, but why introduce it if there is a Current rating and a Cyclic rating?
Please explain to me:o
4) Scale - I agree
5) Minimum guaranteed rating - not needed.
6) Tournament qualification. How will you determine who is strong and who is weak? Take the situation when a very strong, new player comes along and wins with a huge advantage! What then? I think this point is absolutely unnecessary. It will be very problematic to introduce such a coefficient, and there will be endless questions from participants related to a different coefficient... Why complicate things?

Regarding the two saves, there will still be 2 saves in the Best of the Best tournament. I will try to significantly change them in terms of difficulty next time.

That's all for now:o

Thank you for the lively discussion, dear AmberSoler, IronAxe, Sage!
It's very nice to read your posts! My brain is just boiling, so don't be offended that I didn't read this dialogue to the end; I'm just tired, I swear:o

P.S. Let's come to a common decision and calculate the rating according to this system!

With great respect, VDV_forever

Единый справочник по Героям Меча и Магии

ТУРНИРЫ ПО ГЕРОЯМ 1!:smile13:

In reply to VDV_forever

Registered: 5/22/2006

Posts: 1,396

#182 1/8/2007 2:41 PM UTC

Auto-translated

VDV_forever;60566

...but then I realized that it's not realistic, so I'll answer specific questions.

Thank you for the comments. I will try to be brief.

You did the right thing by stopping commenting on every idea. We modified and adjusted our ideas along the way, so my post #20 is no longer relevant. You can start with post #122.

VDV_forever;60566

AmberSoler, post #32, you're probably right about the 50 participants and 2/3! But where were you before?

I'll say this: we're not too late. We're all just at the beginning of the journey. And the fact that a month has passed and only one tournament has been held doesn't mean anything... In three years, this moment will not even be remembered, so everything can be changed five more times without any problems... It won't be difficult to redraw a couple of pages and recalculate the rating for the first BB - it hasn't affected other tournaments yet. And no one will throw stones at you...

VDV_forever;60566

1) Current rating - I agree
2) Cyclic rating - I agree
3) Tournament rating - either I didn't read it carefully, or I don't understand, but why introduce it if there is a Current rating and a Cyclic rating?
Please explain.:o
4) Scale - I agree
5) Minimum guaranteed rating - not needed.
6) Tournament qualification. How will you determine who is strong and who is weak? Take the situation when a very strong new player comes along and wins with a huge advantage! What then? I think this point is absolutely unnecessary. Introducing such a coefficient will be very problematic, and there will be endless questions from participants related to different coefficients... Why complicate things?

3) Tournament rating - this is the most important one and it contains the calculation of the rating for all participants for each tournament played. We put the most effort into calculating it. The cyclic rating consists of a simple arithmetic sum of tournament ratings for a specified period. The current rating consists of the sum of tournament ratings for the entire period, taking into account aging.

6) I will not determine who is strong and who is weak, but the current rating of the player will. And the fact that a newcomer will have a rating of zero will not significantly distort the picture against the background of other distortions. It is impossible to take everything into account. BUT: At the start of the rating, there will be inconsistencies. Over time, when there are enough players with a formed rating, such a player will not have a significant impact on the coefficient in the sum... over time, newcomers with zeros in the rating will be an overwhelming minority in the tournament, and the picture will even out...

VDV_forever;60566

P.S. Let's come to a common decision and calculate the rating according to this system!

That's what we're doing! :) Almost everything is ready. For example, you can recalculate the BB rating according to the specified scheme. I will do this and post the table in Excel here. You can take a look and maybe comment. And then - we'll decide...

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

In reply to AmberSoler

Registered: 5/22/2006

Posts: 1,396

#183 1/8/2007 3:18 PM UTC

Auto-translated

I am posting the calculation of the tournament rating for players who participated in the "Best of the Best" tournament (at the impossible level), based on all the rating calculation ideas that have been agreed upon so far. For those who do not have time to delve into the debate that has unfolded in the current thread, I will briefly outline the main points of the calculation here. The rating consists of two main components: 1. A non-linear rating R1, which depends only on the comparison of the participant's result with the result of the tournament leader. The leader always gets 100 points for the win. R1(n)=S(1)/S(n)*100, where R1(n) is the non-linear rating of the player S(1) is the result of the leader S(n) is the result of the player // Formula for calculating a speed tournament, where the best result is the lowest value. For calculating tournaments based on experience, resources, or troops, the formula will take the form R1(n)=S(n)/S(1)*100 // 2. A linear rating R2, which evenly distributes points among all participants in the tournament, depending on the place they took. The leader receives 100 points, and the player who took last place receives 0 points. R2(n)=100-100/(N-1)*(n-1), where R2(n) is the linear rating of the player N is the number of ranked places (usually coincides with the number of players, but not always) n is the place of the player being considered 3. And finally, the final rating R (oscillator), which is the arithmetic mean of the two parameters mentioned above. R(n)=(R1(n)+R2(n))/2 The essence of this calculation scheme lies in the combination of two independent parameters, reflecting all the main achievements of the players in the tournament, and the player's rating depends not only on his own efforts, but also on how his direct competitors performed in the tournament. The higher the result compared to the leader's result, the higher the first component of the rating will be, and the place the player took does not matter here; that is, only the player's own result is reflected, without taking into account the game of the other participants. Also, the higher the place the player took (regardless of his result), the higher the second component of the rating will be. And here, in addition to the importance of how the player performed, it is very important how all his opponents performed in the tournament. In future calculations of the tournament rating, the player's current rating will also be taken into account in comparison with the average current rating of all players who participated in the tournament. Moreover, the higher the average rating of all players (i.e., the higher the level of the competition), the more significant the player's result in the tournament will be. That is, a boosting coefficient will be introduced, which will be applied to the tournament rating of each participant. This coefficient will have a stronger effect (in relative terms) on a low rating and a lesser effect on a high rating. This is quite logical. This aspect still needs to be refined. *** The rating of the past "Best of the Best" tournament, taking into account all the points mentioned above.

Attachments 1

1 file attached • Total size: 3.8 KB

zip

R_BB_1

3.8 KB 143

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

In reply to AmberSoler

Registered: 2/15/2006

Posts: 1,676

#184 1/8/2007 4:14 PM UTC

Auto-translated

Dear AmberSoler, could you please provide a working link to this file:smile01:?

Who will be responsible for these calculations and monitoring the 3 rankings?:confused:
I already have a severe lack of time:(

Dear IronAxe: I don't even know which ranking to use to calculate the Gameplay rating for the Best of the Best tournament:confused:
Please clarify.

Sincerely, VDV_forever

Единый справочник по Героям Меча и Магии

ТУРНИРЫ ПО ГЕРОЯМ 1!:smile13:

In reply to VDV_forever

Registered: 5/22/2006

Posts: 1,396

#185 1/8/2007 4:31 PM UTC

Auto-translated

VDV_forever;60649

Dear AmberSoler, could you please provide a working link to this file:smile01:

I have attached the file to the previous post. In any case, I can see it there... :)

VDV_forever;60649

Who will handle these calculations and monitor the 3 rankings? :confused:

There will be one ranking, and everything will be calculated using a single formula. I have simply broken it down into sections for everyone who wants to do the calculation themselves.

The first two options will not be published anywhere – only the last one. I will create a template where you only need to substitute the values. IronAxe has agreed to handle these calculations.

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

In reply to AmberSoler

Registered: 2/15/2006

Posts: 1,676

#186 1/8/2007 4:36 PM UTC

Auto-translated

AmberSoler;60656

I attached the file to the previous post. In any case, I see it there... :)

I see it too, but it seems like it wasn't attached properly to that post. Please attach it to this one.

AmberSoler;60656

The rating will be one, and everything will be calculated using a single formula. I just broke it down into sections for everyone who wants to see the calculation. The first two options will not be published anywhere - only the last one. I will create a template where you only need to substitute the values. IronAxe has agreed to handle this calculation.

That's great, I'm all for it!
But first, it's necessary to evaluate your calculations based on the proposed system, and then decide whether to approve it or not, right?;)

AlexeyPank: Why isn't this system visible in the voting? Or am I blind:eek: . Although, in my opinion, this voting is not objective, unfortunately...

Best regards, VDV_forever

Единый справочник по Героям Меча и Магии

ТУРНИРЫ ПО ГЕРОЯМ 1!:smile13:

In reply to VDV_forever

Registered: 5/22/2006

Posts: 1,396

#187 1/8/2007 4:46 PM UTC

Auto-translated

VDV_forever;60657

But first, shouldn't we assess your proposed reward system and then decide whether to approve it or not? Isn't that the right approach?;)

Exactly. This system was only finalized yesterday...

I don't see the point in voting for it in the way the voting process is structured. The entire process of familiarizing oneself with all the methods is too long and complex... And there is no objectivity in the voting without detailed comments from each participant – you are right about that. Simply ticking a box and forgetting about it is not the kind of question that should be left to chance...

Everyone who was interested has already expressed their opinion on the matter in this thread, and the final version was created taking into account the opinions of everyone who wanted to contribute. We should simply gather the opinions of all participants here, rather than posting announcements everywhere... :)

So, take a look at the calculation table. Those who are interested can experiment with the numbers in the proposed table and see what happens when the number of tournament participants increases/decreases, when the composition of the leaders changes, when very strong players appear among the leaders, etc. Everything is clear and easy to understand. If you have any questions, feel free to ask.

Attachments 1

1 file attached • Total size: 3.8 KB

zip

R_BB_1

3.8 KB 150

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

In reply to AlexeyPank

Registered: 6/23/2006

Posts: 1,585

#188 1/8/2007 6:19 PM UTC

Auto-translated

AlexeyPank;60531

1 - The rating is calculated separately for each version (it is summarized into a general table only for those who have played at least 3/5 tournaments).
...
I think that's all. Any comments?

In general, everything is correct, but I don't like point 1. Who will decide who to include and exclude from the general table?
Calculate in separate tables. Simply summarize them into a general one for statistics, just to see how things are changing. Does this table obligate anyone to anything?

Added 53 minutes later

VDV_forever;60649

Dear IronAxe: I don't even know which rating to use for the Game Rating in the Best of the Best tournament:confused:
Please clarify.
Sincerely, VDV_forever

I think that AmberSolera's improvements make sense. However, I still don't like the second component in this formula, where there is a dependence on the place and the number of participants. I don't like it because in some cases it negates the first component. Although the place is, in general, nonsense, but the approximation in days to the leader's result is not nonsense. AmberSoler promised to somehow correct this situation, but I haven't seen it in his final post yet. Perhaps this will be taken into account with a tricky coefficient later, when we can already calculate the strength of the tournaments.
Publish the rating based on AmberSolera's calculations. We will develop the idea further if necessary, and who knows, maybe we will come to something :)

In reply to IronAxe

Registered: 5/22/2006

Posts: 1,396

#189 1/8/2007 6:30 PM UTC

Auto-translated

IronAxe;60664

I don't like it because in some moments, it consumes the first component.

In which "some moments"? Provide an example with numbers, not words. And don't confuse the concepts of "consuming" and "correcting." I haven't found a single instance yet...

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

In reply to AmberSoler

Registered: 6/23/2006

Posts: 1,585

#190 1/8/2007 6:44 PM UTC

Auto-translated

AmberSoler;60656

The first two options will not be published anywhere – only the last one. I will create a template where you only need to substitute the values. IronAxe agrees to handle these calculations.

I agree, but for now, it's just an experiment in the VDV tournaments for Heroes 3. I don't know what will happen at Kanighem's tournament, but I can calculate the Heroes 3 players in our rating, taking into account the best of the best. Then, AmberSoler will remain a consultant and assistant for as long as possible. Just in case I mess something up. If everything goes well, then that's fine.

Now, think about this: I didn't promise to maintain a rating for all versions of the game. Who will take on the rest? Or maybe we should take everything into our own hands, AmberSoler and Sage??? Maybe we can calculate the ratings together? I can give Heroes 3 to Amber, and I'll take on something else. What do you think? I think the administration will definitely reward the raters for this.

Added 5 minutes later

AmberSoler;60674

In what "certain moments"? Provide an example in numbers, not words. Just don't confuse the concepts of "eats up" and "corrects".

I did give you an example where a player in 1st place completed it in one day, and the player in 2nd place completed it in 1000 days. Do the math yourself. The first component of your rating is useless for the player in 2nd place; it won't correct anything.

I mean that under normal conditions, when there are many players, a player with 1000 days would hardly take 2nd place.
But here, he got lucky and got a lot for free. And it's strange to hear such a suggestion from AmberSoler when he was talking about the large aging factor and that he doesn't want randomness to greatly affect the change in the result. And my example with 1000 days and the second component shows when there was no strong player between the 1st and 2nd players by chance!

VDV wanted to say something similar when he jumped into this topic. If nothing changes with the second component and the ranking, then I will demand that only matches with at least 10 participants are recognized as rating matches. Although there may still be an awkward situation there, but the chances of it appearing are already lower.

Added 6 minutes later

Sage;60551

I want to correct you a little. With aging at 1/3, the rating ceiling will be only 300 points (300/3=100 + 100 (for a win)=300). I think you missed this point.

Yes, that's right. This example turned out to be unsuccessful. But essentially, what difference does it make where the limit will be – 300, 1000, or 10000? In my opinion, 300 is also good, but with 2 decimal places

In reply to IronAxe

Registered: 5/22/2006

Posts: 1,396

#191 1/8/2007 6:52 PM UTC

Auto-translated

IronAxe;60675

Yes, I gave you an example where the first-place player completed it in one day, and the second-place player took 1000 days. Do the math yourself. The second-place player doesn't need your first component; it won't correct anything.

The first component is yours, not mine. It's a non-linear rating that you suggested to everyone a month ago. I simply brought it to a more correct form, like an oscillator. Please be consistent in your statements. But if the second place isn't significant, it will be unfair to the player who took that place.

IronAxe;60675

But here, he got lucky and reaped a huge reward for free.

And if the second place is also the last place, then the rating doesn't need to be calculated. I agree with that too.

Just don't overreact... All your examples are unlikely. This leads to a very nice conclusion: you don't see any real problems in the scheme I proposed. And that's despite the fact that no one has delved into this task as deeply as you. I think that's a plus. And minor issues (the ones you're talking about) will appear in any scheme. The probability of such a scenario is zero, so it shouldn't be given much attention.
So, everything is fine.

But you said you didn't see any changes in my last post. This means I should conclude that you didn't look at the rating attached to the last post. Please don't make unsubstantiated claims. I ask you again to indicate what exactly bothers you in the latest version of the rating. Or don't say anything at all.

P.s. By the way, the player who took last place and was 1000 days behind the leader will receive a rating close to zero. And this is also visible if you put the values in the table.

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

In reply to AmberSoler

Registered: 6/23/2006

Posts: 1,585

#192 1/8/2007 7:30 PM UTC

Auto-translated

AmberSoler;60681

The first component is yours, not mine. This is a non-linear rating that you suggested to everyone a month ago. I just brought it to a more correct form in the form of an oscillator. Please be consistent in your statements.

I am consistent - I am saying that the first component, in my understanding, has more significance than the second. The first component will be 0.05 for a player with 1000 days, and what will the second be:

R2(n)=100-100/(N-1)*(n-1),
where
R2(n) is the linear rating of the player
N is the number of places assigned (usually coincides with the number of players, but not always)
n is the place of the player being considered

R2(1)=100-100/(3-1)*(1-1)=100
R2(2)=100-100/(3-1)*(2-1)=50
R3(3)=100-100/(3-1)*(3-1)=0

I don't understand what is linear and what is not. And I see that a guy who doesn't know how to play, but accidentally took second place, would get a rating of 0.05 in my system. The leader is 100.

And with your correction, the leader also has 100, and the newbie gets a rating of 25. That's a quarter of the maximum! And the guy doesn't know how to play at all - isn't that clear from the example? It's just that there were no strong players between them by chance.

AmberSoler;60681

Don't make unsubstantiated claims. I ask you again to indicate what exactly bothers you in the latest version of the rating. Or don't say anything at all.

I would rather play a game right now if I hadn't promised VDV to calculate the rating for his tournaments. But the option I proposed is simple. And now everything is complicated, although in my opinion it is more correct, but I said what I am not happy with. I don't know why you don't understand the example I gave. I am not considering a real calculation in this case. And by the way, you used to agree with this statement of mine, and now you don't remember?
What should I look for in this topic to quote your post?

Added after 8 minutes

AmberSoler;60681

And if the second place is also the last, then the rating can be ignored. I also agree with this.

P.s. by the way, a player who took last place and is 1000 days behind the leader will get a rating close to zero. And this is also visible if you put the values in the table.

And if the second place with 1000 days is not the last, there is also a third place with 1001 :)
Third place will get almost 0, second place will get about 25, the leader will get 100.
Is this correct? Yes or no?

It seems to me that something is wrong. These two components should not be added together and divided by two, but rather some kind of ratio between them should be used, so as not to completely nullify the first component.

And one more thing. Even comparing the approximation in days to the leader is not entirely correct. What if we assume that any map can be completed in ONE day! And compare each result to the leader's result relative to this number :)

In reply to IronAxe

Registered: 5/22/2006

Posts: 1,396

#193 1/8/2007 7:35 PM UTC

Auto-translated

IronAxe;60685

With your correction, the leader also has 100, and the other player gets 25 rating points. That's a quarter of the maximum! And the guy clearly doesn't understand – is that not obvious from the example? It's just that, by chance, there weren't any strong players among them.

Wow – by chance... :) If no one could play better than him (except the leader), then it's NOT by chance, and he deserves it. Even if it took 1000 days. ALL the others played even worse, which means this player SHOULD get a significant rating...

According to your system, the leader's super-result will determine everything, and he managed to do something amazing on a map where most people took 1000 days or more. And that happens. And that's why all the other players got zero rating because of the existence of one super-player. This is pure imbalance because the random probability of such players appearing on the map SHOULD NOT reset the tournament bar. I've said this ten times already.

And the randomness here is the appearance of a "dinosaur" in the tournament, not the fact that a player "randomly" took second place... The player competes not only with the leader but also with all the other participants in the tournament. And he beat ALL the others, earning a deserved rating. And the more participants he beat, the higher the rating. But in your calculation, it doesn't matter how many participants this player overtook.

Let's imagine that 100 people played in the tournament, and they all did worse than him, i.e., took more than 1000 days. According to your system, all 99 people (except the leader) will get a zero rating. Did they not compete with each other? And does the player who took SECOND place out of 100 players not deserve a significant rating? I'd like to see how you would present the results of such a tournament, where one monster appeared and took 100 points, and you would give the rest zeros, even though there was a real battle between them because the map was really difficult... Would you tell them that they don't know how to play? And your result, by the way, could be among them...

Should I show you a multi-month map that the experts created in 8-10 days? One such tournament would be enough to kill everyone's desire to play for a result that they don't deserve. And you will only understand this when a dozen tournaments have passed, and what we are discussing theoretically appears in reality.

I'm very sorry that all your theoretical models focus on the problem of a weak player getting an undeserved rating. And why don't you consider the opposite situation, when a worthy player, having put a lot of effort into the tournament and beating almost all the opponents, gets zero according to your method? We are talking about a general balance of calculation that takes into account the interests of all players in all tournaments, and your arguments are specific cases, unlikely events.

And I'm not against giving a beginner a little rating "for free" if, in doing so, I can preserve the tournament rating for a player who really put a lot of effort into completing the game and achieved a significant result. But you would give everyone zero. Because of dogma!

That's all.

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

In reply to AmberSoler

Registered: 6/23/2006

Posts: 1,585

#194 1/8/2007 8:03 PM UTC

Auto-translated

AmberSoler;60688

Let's imagine that 100 people participated in the tournament, and all of them performed worse than him, meaning they took more than 1000 days. According to your system, all 99 people (except the leader) would receive a zero rating. Did they not compete against each other? And does the player who took SECOND place out of 100 players not deserve a significant rating?

I'm talking about the case where only three people participated in the tournament. One super-monster, and two very weak players. The rest were on vacation. Let's discuss the number of participants in the tournament and when a tournament can be considered a rated one.

In reply to IronAxe

Registered: 5/22/2006

Posts: 1,396

#195 1/8/2007 8:07 PM UTC

Auto-translated

IronAxe;60692

I'm talking about a case where only three people participated in the tournament. One super-strong player, and two very weak ones. The rest were on vacation. Let's discuss the number of participants in the tournament when the tournament will be considered a ranked one.

That's reasonable. I don't object. This will resolve the issue and eliminate isolated, unlikely events. Just suggest your vision regarding the number of players, because it's difficult to determine everything here... My tournament experience in a three-player format shows that there are never fewer than ten (usually significantly more). But for a two-player format, 10 people is normal... I think we need to approach this flexibly.

But is it worth dwelling on

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