Connection ErrorBad RequestThis page went staleSession ExpiredSign in requiredForbiddenNot FoundToo Many RequestsServer ErrorBad GatewayService UnavailableGateway TimeoutHTTP Version Not Supported
Session Expired
Your session has expired. Please log in to continue.
Sign in required
You need an account to do that. Sign in or create one — it only takes a minute.
This page went stale
Your security token was rejected, usually because the page sat open too long or you signed in elsewhere. Nothing was saved. Reload the page and try again.
Request Failed
We encountered an issue while processing your request.
Please check your internet connection and try again.
Error Details:
Share Link
Posts from Player rating for tournaments across all installments of Heroes.Auto-translated
It's in the works, it's in the works. I suggest we return to a specific discussion about the ranking after the first round of the Heroes of Might and Magic tournament is over. Once we have data from five stages, each with its own version of Heroes, we can calculate everything and determine the pros and cons of all ranking versions. For now, I'll calculate based on both, and then we'll see.
My system is still under development :) I'm listening to the criticism :)
I read the discussion between AmberSoler, Kule, and Tokha. I'm completely on AmberSoler's side. However, I still disagree with AmberSoler regarding the second component of the ranking. It is more important than the first component, but I still think that the current ranking should be based not on the number of participants and the places they took, but specifically on the results of each day and how close they are to the leader based on these results.
I'm attaching a file with a slightly modified ranking system that doesn't include the second component, along with an explanation of why I think this way.
The point regarding the rating based on the leader's walkthrough time relative to some ideal result is quite interesting. However, we will never actually know what result would be ideal for each specific map. Therefore, such a system, while interesting, introduces a serious element of subjectivism. In this regard, I propose the following: calculate that same coefficient taking into account the results not only of the leader, but of the top three players. That is, K=(T1+T2+T3)/3. Rtur(i)=K*100/Days(i). This means the leader can score more than 100 rating points if the results of those following them are significantly worse. This K will be a kind of analogue to a tournament rating. Applying such an additional coefficient will avoid the flaw where the leader finished in 1 day, while second place took 1000. The leader will simply receive many more points because of this. And a linear result depending on placement will not eat away at the leader's rating. Then the oscillator will become more objective and can be successfully applied.
1. The most important thing is not to go too deep into the details so that we stop understanding each other. Because, as practice shows, even people who think logically and understand numbers can no longer keep track of our conversation... :) Kyle is an example – he understands the essence of all our calculations perfectly, but because he hasn’t been following the topic for some time, he has already lost the thread of our discussion. We three are still managing somehow... :)
2. As always, I strictly adhere to the following principles: solving the problem in a general form and avoiding any arbitrary figures. The first principle helps avoid wasting time analyzing specific and highly unlikely cases, and the second is simply the only correct and reasonable approach to balancing the system. Because even one figure taken out of thin air significantly disrupts the entire balance.
Therefore, I agree that the idea of averaging the top three results – which is itself new and interesting – will reduce the likelihood of a large gap appearing. However, the number "3" itself is taken arbitrarily and undermines the consistency of the system.
Let me explain. Why do we take the top five players or not just two, but specifically three? There is no hidden meaning in this number. And please note that this point clearly gives an advantage to the leader, and with some probability – to the player who came in second. If we take the top five results, the leader's advantage will be even greater, and there will also be chances for bonus points for players who finish from 2nd to 4th place. And so on – endlessly. That is, taking N top places already implies an unfair attitude towards the other players. Moreover, they have less influence on it :) Of course, you can say “nothing prevents them from taking the top places,” but that would be stretching things...
Also, there is quite likely to be a persistent gap between the top three players, which makes it impossible to solve the problem in a general form. The probability is lower, but consider the recent BB tournament. Two players significantly outpaced the rest. It could have been three...
2 IronAxe
A small comment about your "lyrical digression."
I was considering a situation where, out of one hundred people, one player completed the map very quickly (pulled off a trick - the original source is here) – let’s say, in 10 days, the second took 1000 days, and all the remaining 98 people did even worse. This example is exaggerated just to illustrate the point. In this case, all 99 people, compared to the leader (who will have 100 points), will receive a rating from 0 to 1, which is incommensurate with the amount of effort they put into competing for the prize. Moreover, the map was indeed difficult – it's just that the first player turned out to be an “overwhelmingly” strong competitor compared to the others.
With you:
The situation is that suppose the player who finished in 1000 days turns out to be the leader. The rest are even worse. And in this case, nothing prevents them from getting a full rating. The leader will get 100 ranking points, everyone else will get the corresponding rating, depending on how close their result is to the leader's result.
You probably just didn't grasp the idea of my example. I certainly don’t think you intentionally distorted it... In your comments, you left out the first "fast" player from the analysis and considered all the others; accordingly, in this case, the essence of my example is lost.
Let's read further:
However, such a situation cannot be resolved by AmberSolera's formula either, so he did not actively promote this argument against my formula during the discussion.
This is incorrect on your part because the oscillator smooths out this imbalance by distributing additional points for the places taken. In the case of the oscillator, the leader will receive 100 points, the player in second place will receive 50 points, and the rest will evenly receive from 0 to 50 depending on the position they took. You shouldn't distort reality after all... :) And if my lack of activity tells you that your statements are correct, then you can demonstrate the opposite... :) But do we need to? Let's remain reasonable and, above all, listen to each other, rather than pay attention to emotions. For now, it seems to me that you simply don’t hear what exactly I am saying and have distorted my thoughts more than once.
I completely agree. But this will give stronger players the opportunity to pull ahead. And it will make this particular metric more significant, rather than a linear one. The number 3 was chosen based on the three prize-winning places, and to avoid increasing the leader's advantage even further. In principle, you could calculate the average result of the tournament and compare it to that. And the leaders would pull even further ahead. But that way, there wouldn't be any arbitrary numbers. We could try recalculating the tournament results, taking into account the average result, and see what comes out.
I hope you noticed that in your "non-linear rating" calculation, a multiplier of =1.213 was simply applied to the existing results of all players... In other words, it provides nothing additional; the only thing is that it reduces the likelihood of one leader having a huge lead... Simply put, every player received a bonus of 21.3% of their earned points for their game in their pocket... :) (I am referring to the calculation of the average result among the three leaders and not touching upon the calculation based on the average of all players, as the distortion there is even greater)
The oscillator behaves exactly as it should in both cases—it slightly adjusts the final result relative to the place taken. Thus, this averaging gives an equal relative bonus to everyone; only in absolute terms is this bonus not equal. The higher a player is in the tournament table, the higher their bonus when converted into points.
Yes, it turned out to be a bit silly. We didn't bring anything new to the table. We just added an extra formula. But it does reduce the likelihood of mistakes like completing the first task in one day and the second in a thousand. Maybe we can develop this idea further sometime?
Guys, I'm amazed by you... one moment you're talking about the "current" rating, and the next about how to calculate the rating for the VK tournament... you need to make up your mind because these are not the same thing! Let's start... First: to AmberSoler, please excuse me, but I perceive this post as mere sophistry and therefore arguing here is pointless... Second: how to calculate the VK tournament? In my view, there are no problems here, since this tournament fulfills all the conditions I described here. Let's figure it out... if we calculate the VK tournament rating according to your system, what happens? It becomes more profitable to play only in the final stages because decay will destroy all the rating gained in the early stages... and that is wrong! All ten stages of this tournament are equal! And there should be no decay here! If (for example) we look at how ratings are calculated in sports, where there are similar stages... then the final rating (very often) is calculated based only on the participants' placed positions (i.e., according to the second component of the rating presented by the esteemed AmberSoler).
In my view, the rating for the VK tournament should be calculated as follows: First: for each stage individually! At the same time, let's say n-people participated in all stages. Consequently, for each stage, the rating must be calculated for n-participants, regardless of whether only n/10 (for example) people sent their saves! The rating of someone not participating in a given stage will be zero. And the rating must take into account the number of participants.
Second: after the final stage, you need to sum up the players' ratings across all stages, and whoever has the higher total wins!
P.S. to IronAxe, I have a request: if you write my nick, please write it in Latin or in Russian, but correctly. Thanks in advance.
We are talking about how to calculate the rating across ALL parts of Heroes. The VK Tournament has a convenient feature: it includes all versions, which is why it is often discussed. I am also against aging, because a player who has played in one tournament in the second round may be pushed down from their position, because their rating has not increased, while the ratings of those who played have increased. The rating should be calculated after each stage, and not once a year (because we will only know the total number of participants after the last tournament). Both the completion time and the place taken should be taken into account, as this will be much more objective. An oscillator is the best solution. As for comparing with the average result, this can help to avoid disputes about why we are calculating it this way.
We are talking about how to calculate the rating across ALL parts of Heroes.
I already said that this is not possible! only within the framework of some tournament. A good example: ВК where it is possible. Otherwise, it turns out like "a 100-meter race and a 400-meter hurdles!" And instead, I suggest introducing a purely informational rating that shows the strength of players with the greatest accuracy, but according to which you cannot compare and choose winners! It is described here.
I am also against aging, because a player who has played one tournament may be pushed from his place in the second, because his rating has not increased, while those who have played – their ratings have grown.
That's right! And I’m saying the same thing!
The rating should be calculated after each stage, and not once a year (because we will find out the total number of participants only after the last tournament).
We will be able to calculate the final rating only at the end. After each stage, you can give an approximate rating (for information)
Both the time taken and the position occupied should be taken into account, as this would be much more objective. An oscillator is the best solution.
1. The oscillator is applicable only to calculating the Tournaments ranking. (2 Kyle, I hope you understand the difference between the Tournaments, current, and cyclic rankings :). In my opinion (and how else could it be?) - this is the most balanced option for calculating the Tournaments ranking.
2. All stages of the ВК Tournament are definitely not considered for calculating the winner of this tournament. This is also self-evident... Aging is an attribute of the current ranking, which does not divide into BB, BK, and possibly other tournaments consisting of several stages... All ranking tournaments are taken into account in sequence. Or maybe no one will keep a current ranking at all... That will depend on how things develop.
3. The annual ranking is calculated throughout the year, adding up the results of each participant in all their tournaments... Who said that it is calculated "once a year"? The results are summarized once a year - that is clear. But the calculation and dynamics are visible constantly and updated after each stage...
Right now, we are working on the methodology for calculating various options. Which of them will be implemented will be shown by time... The most important thing is to distinguish and not confuse these methodologies with each other. The calculation of the Tournaments ranking (the ranking of each specific tournament) caused the most controversy. Here - it is up to the organizer. You can use a non-linear ranking, an oscillator, or something else...
Then, having the results of each completed tournament, you can create a current player ranking (result + activity), a cyclic ranking (annual, quarterly, or something else - in the form of the sum of the results of all tournaments in this version), and also a BK ranking - you will agree that the final BK table is also a ranking, just called "Tournament Results". In essence, this is a standard cyclic ranking, which takes into account the results of all stages in which the player participated...
If you understand these nuances, many disputes and misunderstandings will be resolved. Naturally, the annual and current rankings are calculated separately for each version. How Alex will calculate his BK ranking, he has not yet said.
Here, please provide more details... explain what tournament, current, cyclic, and annual ratings are. This is necessary so that we can speak the same "language."
Yes, it turned out to be a bit silly. We didn't bring anything new. We just added an extra formula. But it really does reduce the likelihood of errors like getting first place in one day and second place after a thousand days. Maybe we can develop this idea further?
Well, this oscillator fixes that error... :) What else is there to develop? The oscillator doesn't have this problem – I showed this error in IronAxe's method solely as an example of inefficiency…
So, if you are adopting the oscillator, you shouldn’t think about random breaks. This sometimes happens, but it won’t affect the final calculation…
Take the calculator that I suggested and conduct an experiment. Second place will always be rewarded with at least 50 points, regardless of any break from the leader, while the rest will receive scores evenly distributed from 0 to 50.
Added after 9 minutes Kyle;63395
Here, tell me more about this: tournament rating, current rating, cyclic rating, and annual rating. This is needed so that we can speak the "same" language.
Perhaps, before answering a question from anyone who wants to, I should first ask them, "Have you really read everything that was said in this thread?" :)
Sorry,:smile15: I think I've read the whole section, but somehow I missed this post...
Some minor corrections:
...BK is also a rating, it’s just called “Tournament Results”. In essence, it is a standard cyclic rating that takes into account the results of all stages in which the player participated...
In essence, it is a standard cyclic rating that takes into account the results of all stages in which the player participated or did not participate.
1. Current rating. It dynamically reflects the current state of affairs, taking into account the player's actual result in the tournament, as well as the player's activity.
Regarding this: I don't think combining the "player's result" and his "activity" into one number is the right solution! For this, it is better to divide it into two ratings, one shows the average result of the player, the other - his activity ... and combining everything into one number will not tell us anything. i.e., it may be able to show us the leader, but there will be complete confusion about the rest because it will be unclear whether the player is not playing, or he is playing but poorly, etc.
1.2. General current rating.
Regarding this: everything will depend on what we decide with the "current rating".
3. Tournament rating.
In my opinion, an oscillator should be used.
4. Scale.
A 100-point scale.
Minimum guaranteed rating.
In my opinion, it is not needed.
Coefficient that increases the qualification of the tournament and its impact on the calculation of the rating.
Some minor adjustments: Essentially, it's a standard cyclical rating that takes into account the results of all stages in which the player participated or did not participate.
If the player did not participate, what should be taken into account? Should we calculate the sum of zeros using a complex formula? The result will be very predictable, by the way... :) Kyle;63410
Regarding this: I don't think that combining the "player's result" and their "activity" into one number is the right solution! It's better to separate it into two ratings: one showing the player's average result, and the other showing their activity... Combining everything into one number won't tell us anything. That is, it might show us the leader, but there will be complete confusion regarding the rest because it will be unclear whether the player isn't playing, or is playing but poorly, or something else.
Instead, come up with a logically and mathematically sound mechanism for separating them... In my opinion, this is an unsolvable problem, so either leave it as it is or abandon it altogether. If you remove the aging factor (which is precisely the parameter that reflects activity), the rating will constantly increase and reflect not the player's strength, but the number of tournaments played. And if you again suggest dividing the sum by the number of tournaments, then the players who won one tournament and then stopped playing will remain at the top of the rating forever... And if you remove inactive players, where will you get the "activity criterion"? And in general, you will reduce the current rating to a cyclical one, which is not the same thing...
I dare to give some advice: take a piece of graph paper, a pen, and experiment with a dozen abstract characters in ten abstract tournaments... And then tell us what observations you made regarding the accounting of their results... It's a very interesting exercise, I assure you... :) Simply, so many options have already been calculated manually, in Excel, on graphs, and with other clever methods that I don't think you will be able to invent anything new... What has been proposed is the best option so far, taking into account the interests of all players... Kyle;63410
I think it shouldn't be introduced yet.
Justify it. And what does "not yet" mean? Does that mean it will become "necessary" later? And what needs to change for this to happen?
IronAxe
