The direction of New matchmaking system
Although I have pointed out some matchmaking, it seems that the MMR system is still having trouble being created, so I would like to suggest a few more systems here. Of course, some of the problems with matchmaking may have already been solved internally by developers, and various other public algorithms may be tested or introduced. It may also be a solution that has already been tried or other problems have been discovered, or that has been abandoned due to feasibility and difficulty in implementing practical development.
We have been using an emblem/rank system, an extremely crude matchmaking system, for more than two years. The downside to this matchmaking system is that it's already been argued on numerous post in forum. It would be a waste of time to discuss this, as developers all know. However, the new matchmaking system (MMR matchmaking) also has some problems. When it comes to matchmaking, we have to consider many variables; calculating asymmetric game-based survival/killer MMR scores, difficulty in determining winning criteria, how players influence each game's winning factors (and how to calculate it statistically), map wining odds (murder/survivor sided), and further, quality of connection (server host latency - accesing with VPN at remote and ruining the game connection quality), imbalance of the killer and survivor's pool of players (which has the same skills), SWFs, and the use of consumables (items, addons, offerings), ... etc. All variables, such this make it difficult to design a new MMR production algorithm, and make the new system difficult to operate.
At the heart of all the most deadly and major issues in the game is matchmaking, and many of the balance issues that are being talked about in the forum are all tied to matchmaking. Also, for competitive games, this matchmaking is almost a top issue in the quality of 'competitive' experience for all gameplay.
Ch. 1 Matchmaking System Design
First, developers seem to be designing the wrong concept for the new matchmaking. What does this mean? First, the reason why this new system is being created is to create a skill-based matching. The developers blocked the killer from changing and to determine each killer's ability by calculating each MMR.
But this is a very wrong idea. This seems to come from the idea that killer players will have a very large skill difference for each killer, but in reality, the actual difference between each skill is not that big. Unless it's a very special case, such as a nurse or a blight, the killer's ability to deal with survivors of similar skills represents about the same level. Therefore, calculating MMR separately creates a great deal of confusion. This allows the matching to blend randomly when a good killer player at one character plays another killer character. And confusion comes, such as many trials before MMR stabilizes. Of course, that is true. Each killer character has a different "tier" and even if it lacks skills, it can have a high kill rate because the killer is powerful. However, it is strange to set separated MMR for this reason. While briefly testing the new matchmaking briefly, The results of each MMR rating interacting with each other and affecting each MMR, also support this argument. Calculating the rating separately because of the strength and weakness of the killer brings us to the same conclusion as calculating the rate of each add-on separately. Some killer addons are powerful, some killer addons are weak. Will we calculate matchmaking for each add-on? This looks very strange.
Therefore, the point is that the killer's MMR should be calculated with only one. This will reduce the complexity of the problem as developers create algorithms for different killers. And it can reduce confusion in the MMR system. If different MMRs are used, each time a new killer is released on the live server, there will be tremendous confusion. And with each chapter released, it is inappropriate for this confusion to repeat. Thus, the conclusion is that the player's MMR is calculated with only one rate. However, as mentioned above, the power and kill rate of each killer vary, so when playing, the match search in such a way that the MMR is multiplied by the modification factor.
For example, a killer player's MMR was calculated as 1000. Since Spirit is a powerful killer, let's say it has a modification multiple of 1.5 (this would be a work to do, It is discussed below). If the players search the match with Spirit, he will be matched with survivors with a rate of 1500. Ghostface is a weak killer, so let's say it has a modificatio multiple of 0.7. When he search the match with Ghostface, he will be matched with survivors with 700 MMRs. The simplest way to calculate the modification factor is to calculate the factor based on the current average kill rate, depending on how much the killer deviates from this average.
But this also creates some confusion, such as Nurse. The current Nusre kill rate is much lower than average. Therefore, the modification factor will be much lower than 1. (Let's assume 0.5 for example) but this is a disaster in the upper MMR. Think of a skilled Nurse player. Let's say he has 2,000 full rates in killer, and 3,000 Nurse 'skill' rate and 1,500 walking killer skill rate. Even though he has 3000 nurse skills, 2000 rates will be multiplied by the modification factor to search lower rates, which will result in opponent survivors being swept away. Of course, if this continues, his killer rating will be very high. He continues to play Nurse, raising the killer rate to 3,000. On the contrary, if he plays a wraith at 3000 (*Will be a modification ratio of wraith will exceed 1, isn't it?), he will suffer from more than 3,000 survivors despite his skill being 1,500. Of course the assumed situation is a very special one, and we are given some choices here.
1. Adjust the modification factors by creating a more sophisticated factors. : This may be explained in more detail below, but it is simply to calculate the factor differently for each MMR series. This modification consists of the overall mean statistics of the MMR class.
2. Get rid of the correction factor at all, and create only a full-killer rating. : This is a great way to improve the overall matchmaking quality, ignoring extreme killer characters such as Nurse, Spirit, Blight and Ghostface. It can also solve the inconvenience of killer players who cannot change their killer.
3.Make each MMR as the developers did. However, this is also problematic, and the development team has made it possible for each of the different MMRs to interact. Number 1 and number 3 are actually very similar, number 3 is the change in the MMR for each player's killer character, but number 1 is the modificatoin for each grade is calculated by the statistics of the class as a whole, and the change in feedback of the game results is applied to only one rating score.
Each has its pros and cons. Now here, we are going to focus on number one and discuss this solution further.
First, the reason why we propose integration scores in this system is that they are more systematically reasonable. In the case of League of Legends, a system was created to calculate different MMRs for each line, and later developers realized it was a stupid idea and later deplecated it. Even games with a large pool of users such as LOL didn't work well, although DbD is a smaller pool of users than LOL, and even fewer killer players, these systems really don't fit. It may seem OK to have different calibrations for different MMR classes. For example, it would be in this form. If the killer MMR is 0–500, the nurse modifying factor will be 0.2, if the rating is 500-100 and 0.7 for 1000-2000..... and 1.5 for 3000-4000.... like this. This factor can be in the form of a table internally or in the form of a graph. Also, this average causes periodic changes. For example, if a patch occurs that nerf a killer character, the factor will be affected in the subsequent season. In addition, if a new type of killer trick or tech is developed (for example, let's say J-flick), this factor will be affected again as a whole.
The expected problem with MMR Skill based matchmaking is that the results of the average game flow and match give feedback back to the correction factor, which again affects the results and causes. Therefore, there is a possibility of unexpected problems.
Ch. 2. Developers should not focus on calculating killer rating, but rather on calculating survivor rating.
It's not that I don't know why developers are more interested in calculating the killer's rating. It is fuzzy Winning or losing this game and making it in system is very difficult and ambiguous, and it is very difficult to determine how much a survivor is contributing to escape and sacrifice. No matter one survivor was killed, if the survivor performed to continue the chase for five generators, and the other three escaped, the survivor contributed 100 percent to the remaining three. However, it seems to be more focused on killer skill-based MMRs because it is very difficult to calculate statistically and realistically difficult to apply it to the system, such as the escape/sacrifice of other players after a certain survivor's game.
Nevertheless, except for some killer-driving characters (and as mentioned above, Nurse, Spirit, and Blight), all kill/escape rates are generally proportional to the survivors' performance, rather than proportional to the killer's performance. Devs seem to have the wrong idea about this. As you saw in Chapter 1, killer skill is not the main factor leading the game. Also, the fair quality of the game is not whether the killer's skill rating is accurately calculated, but how accurately the survivor's rate is calculated and the survivors of the same skill meet. If the killer player's MMR calculates exactly 500 according to the skill, and the survivor player gets 100 and 900 caught in the same game in matching, we don't think it's a fair game. And even if killer MMR is calculated correctly, if 500 MMR is calculated even though skill is 100 survivors, this is an unfair game. This is why it is simply difficult to calculate escape rates. Of course, it is likely that the survival rate will eventually reach a similar level, but tremendous confusion will continue, and there will be many imbalances in skills among survivors. In other words, poor chases and genrush-based survivors may have similar survival rates to those with professional-level chases, which may lead to complaints from survivors with professional-level chases.
The quality of the game is guaranteed if the overall survivors' chasing(looping) skills are leveled. If the 500MMR killer, 100MMR survivor A, and 900MMR survivor B play earlier, the 500MMR killer will find it too easy to chase 100MMR, and it will not be fun to chase 900MMR survivors. Eventually, the 100MMR survivor will be tunneled and the game will end with a quick 4kill.
This is a very complex issue, and it is not simply accessible. Developers tried to solve this problem through the emblem system, but the emblem system, which put arbitrarily created indicators, eventually failed. The methods of calculating each indicator were very different, and it was difficult to calculate in a simple way as the level and meta of the game changed. And it created emblem inflation, and the problem became worse as it changed the rank reset criteria that slightly relieved it. Nevertheless, we need to look carefully at some points. First, there are categories such as how much the generator is repaired, how altruistic is, and how good at the chase. Of course, these indicators are very crucial to the escape of the survivor, but they may not be seen as indicators of skill. At a certain rank level (I mean, in the red rank) the most discriminating indicator is the skill to chase. Pro-level stealth play is not very difficult at a particular skill level, but good chase is highly skill- and experience-dependent. Many community members say that the level of player is generally determined at the chase level, and that even the chase performance evaluation emblem should be doubled. Of course, stealth play is not without merit, but according to past surveys, about 30 percent of players have fun in stealth play, and 70 percent of players have fun in being chased.
We need to pursue a more thrilling chase play experience in upper MMR games to make the game more fun. This was also the reason for the emblem penalty for the killer players camp. Unlike the weak criteria of Bloodpoint, this requires more sophisticated judging criteria. For example, the system's judgment of what the killer is doing (or what survivors are being chased). An evaluation model is needed to evaluate how survivors use pallets and windows, how good chase skills are, and how winning the chase affects escape rates.
Before we talk about the standards of a good chase, let's look at the simpler standards of unhook and generator. Except for HEX:Ruin and Pops, generators are always contributed according to the charge of repair generators. If these two perks are not present, these models are very consistent with the criteria for evaluating 'performance'. Also In the case of POPS, the 'repair contribution' is usually just a repair.
In this case, the survivor's time is divided into only tree: idle time (almost repair time), healing time, and chase time(or it could include moving time). Idle time can include stealth time and can be distinguished from stealth time and repair time.
The most difficult thing is to distinguish the time given to survivors. The simplest way is to calculate the time you don't run. It is very simple. healing time and chase time always occur in a running way, and idle time does not always run. The downside to taking this is time for stealth. However, in normal gameplay situations, it is a kind of penalty if you waste time doing stealth. The problem is that two survivors are killed first, and the rest of themselves struggle. Depending on the need, we can either exclude the time in this case or solve it by reducing the amount of apllied this, generator factor as they are sacrificed.
This is complicated. As survivors are sacrificed at some level of the game, setting the game's repair contribution differently as much as turning the generator is very complicated and difficult. So, how will this be reflected in the MMR? As I said earlier, this can also bring similar models. In the entire game's given idle time, how much of a survivor is repairing a generator is a matter of hardworking genrush. It is easy to calculate this on a relative basis. But as I said earlier, it is not good to simply reflect this in the MMR. This is because the skill ratio of the generator and the skill ratio of the chase are different from each other in the escape. Genrush does not require skill. Therefore, the proportion of the chase needs to be calculated much higher.
Then Let's go back to the HEX:Ruin exception. Of course, with one exception, it is simple to apply the model. And with HEX:Ruin, it is unlikely that there will be a significant difference between repair time and actual generator repair contribution. Nevertheless, destroying HEX:Ruin contributes very much to the game's main escape rate. Giving Advantage points for this destruction may seem reasonable. (But when i ponder about this, i realized it is very hard to give advantage in specific cleaning up HEX totem) Generators that are damaged by HEX:Ruin are, in fact, hard to say, real repairs. Therefore, it seems reasonable not to reflect this amount of destroyed generator in the rating as it is not a actual repair. It also seems reasonable to allow these cases to obtain additional compensation in order to determine that you did nothing at idle time and thus avoid penalties for falling rating.
In any case, the problem is very complicated by HEX:Ruin and other damaged generators, but in the end of the ideal model, the calculation of how much each survivor contributes to the repair until the actual generator completes the repair. Simply, it can be calculated as the amount of time (more precisely, as much as Charge) that the generator is repaired during idle time. As mentioned earlier, given the time to hide from the killer, this idle time can also be calculated only 'strict idle time', not just idle time.
Finally, it's a chase model. This is the surest way to determine how good survivors' skills are in the survival MMR, and to determine how good they are in the MMR system. Therefore, sophisticated systems and access are needed. So far, the system for the chase emblem is divided into several elements (Chaser emblem and Evader emblem). Hiding and being chased are different scores in the Evader category, and scores vary with distance. When camping, the killer gets a penalty at the Chaser emblem.
The survivor's hit (regardlless of injured or dying immediately) is considered to have ended with the killer's winning chase, and seems to be calculated separately as the survivor's victory with the end of the chase.(chasing BGM end) It's a pretty detailed system, but it's unsophisticated, and crude. In particular, it is highly unsuitable for insta-down killers, remote-hitting killer, and hideout killers who disrupt chasing system. Perhaps some reasons include this; that each killer has a different way of pursuing internally, and that each killer has a very different performance. These are likely to be the developers trying to create their own MMR system.
But chasing rating system is more important in terms of survivors than killers. The killer's chasing emblem is still problematic, and it is problematic to use it as a skill indicator. (Remember, in DFC rules, survivors are scored, but killers only evaluate their skills by hooks.) In other words, survivors are more important than the part about killers here.
Survivor>
Sol. 1) Simplified systems
A simplified system leaves everything to the statistical average, and can be implemented relatively easily, but less accuracy. This method begins with the assumption that the killer always tracks at least one survivor. In other words, the killer always assumes that he is in a chase with a certain survivor, and how he (or in some cases multiple survivors) behaves in a chase against the killer (how long does the agro last and win) will be an evaluation factor.
That is to say, for ex, if the killer is at the Pro level and the survivor is at the Noob level and the survivor goes down in just a few seconds, then the chasing Rating in this system subtracts the survivor's Rating.Conversely, if the killer is Noob and the killer drags a huge amount of time down to the Pro level and wins the chase, the survivor gets Rating. (*note: Although the comparison has been made, the rate variation becomes smaller, if the greater the rate difference between the two.)
Killer must search and chase survivors, and at least one of them is always receiving in situation. It may have to be excluded from the exit, but it is meaningful that it is a simple system. It has the advantage of having reasonable results for insta-down killer or semi-insta-down killer because it only cares about 'down' and not about injuries. And technically, it can have a simple form. Of course, it would be different from a Blood Point Chasing System or a Bloodlust Chasing System, and a new 'substantial' Chasing System might be needed. If you're running with the killer within a certain range, we need to make sure that the system is aware that it's in chase.
From the simple way, the survivor who is running when the killer looks is chased, and the survivor who is not running when the killer looks is easily regarded as stealth. The killer's view of the Spin Chill Perk mechanism may be recycled, but if the killer is backstepping, or intentionally approaching the killer while considering of the Spine Chill (In a professional-level game). Of course, the Blood Point Chasing System or the Blood Drust Chasing System should take precedence over 'insight'. Bloodpoint and Bloodlust mean they're being chased, of course. However, as you begin to consider other complex variables, the problem becomes complicated. I think this is the best way. feat. Occam's Razor.
Sol. 2) More sophisticated systems (maps & using pallets)
To refine the system, we have several variables to consider. It is the use of realm and maps. In fact, this is where the problem gets complicated. First of all, there are 36 different kinds of maps in Realm, and of course statistics on each survival rate, and statistics and pallets on average time, the number of windows, etc. It can be sort out a lot of things. But it makes me think again whether it's good to divide it into 36. And when it comes to using the map of survivors, you can compare the time to the number of pallets, etc. The longer of aggro drags and the less pallet usage, the better the survivor's skills are.
First, it is the devil's whisper to apply survival statistics for each map as a correction factor. It sounds really cool to hear, but when you actually apply it, you face a lot of problems. For example, for low-skill Newbie survivors, Gideon map is an indoor map, difficult to locate pallets and structures and can easily die due to difficulty in stealth. But for high-skill professional survivors, Gideon is heaven.
And it's different for each killer. Indoor maps such as Léry’s and Hawkins are difficult for Nurse, but favorable for Ghostface. Cannibal, Hillbilly, becomes very weak in many suffocation pits and Léry’s. Considering this for each realm creates a need to consider a lot of variables together, and we believe that if all 36 map statistics are applied, the problem can be very complicated. For example, in the MMR test match, Hillbilly had a problem, likewise in the realm included MMR system, it can be problematic in Badham 1 and There's no problem with Badham 2.
Anyway, it's almost impossible to consider all of this. Therefore, it seems better to give up this variable. When it comes to using the map of survivors, the number of pallets uses looks really cool. Especially if you were playing and a lot of potatoes throwing away the pallets and you were looking for a pallet and you died blaming pallets, level designer, or team member. Pallets are shared resource for the entire survivor, and the fewer pallets you use (and the more windows you use), its clear you are a professional survivor.
Of course there are some problems here as well. For example, are you better off choosing to die without using a god-pallets? Or is it good for a survivor who doesn't down even if he uses a lot of god-pallets in the first chase? In general, the second survivor is definitely better. But what about this in the match system? What statistically significant difference is there in the rate of use of pallets per second?
And, in the case of Carnival and Oni, they have no choice but to use pallets very quickly. Even if the pallets is used to hold out standing, even if the five generators take turns using all the pallets on the map, the killer will end up with a kill of camp unless multiple survivors down at once. Well, okay, maybe we can put aside some exception killer characters.
NOTE: I didn't consider the killer case because it was passive. Most survivors decide how the game goes. But you can't ignore the killer. This can also use the killer's correction at each chase rate.It is very difficult to deal with Spirit for survivors for a long time. On the contrary, it is very easy to deal with Ghostface for a long time (*Ch. 1)
And even when dealing with professional killers, a higher level of survival skills is needed. And anyone can deal with a Noob-level killer. Therefore, in the feedback of the results, the killer's ability factor is also important. If the killer is Nooob level, you don't get a high reward rating even if you win the chase perfectly, and if the killer is professional, the rating is less even if you lose the chase badly.
NOTE2: In fact, lying down may not be the norm. In the old emblem system, we may base our impact on something else, as it was on the standard. For example, you can base only on when the survivor hooks. If so, smart survivors will be considered for rating, whether they are down under a pallets, using an unbreakable, or even variables such as sabotage and flashlight.
It just mentioned that it's a lot of different judgments (e.g., the killer's camp right now, what time to unhook, where to fix the generator first, drag killer toward not fixing gens, and so on. Now that I mentioned, I'll wrap up the chase model here. These judgement are also an important skill factor, but they are likely to be ignored because the system is difficult to judge. However, survivors who are good at chasing, survivors who have a high chance of survival, are more likely to show better judgment and other performances, so even if this element is ignored, it is unlikely that there will be any major problems.
Ch 2.5 Setting the Standard(initial MMR)
Current emblem systems can bring statistical indicators contaminated by inflation. I mean, it's clear that 'deploying' the first pool of players will be a very difficult task. If, at a certain point, the MMR starts at zero completely, then the MMR will be calculated the same for survivors who have escaped five streak from the yellow rank and five streak escapes from the red rank.
This is also the main reason for confusion in testing MMR. In today's contaminated and forced matched rank-match, the MMR system would see they have forced matching of different MMRs, and these match pools would re-contaminate the MMR system or not have a significant impact on rate feedback. Thus, while the full record in the rank match may be of some help to the deployment of MMRs, it is extremely confusing to use as a main indicator.
Then, how can we calculate the skills at the foundation stage? We think this can be done in playtime.For example, let's say that survivors who play 1000 hours can typically have 1000 skills 'on average'. The rank matching so far can then yield a 'modified' rating, analyzing how survivors with B-level skills have resulted among different players.
Of course, I'm not talking about using Steam Playtime. Steam playtime is not a real playtime, but it is excluded because it includes matching and playing time.Therefore, in-game play times (each survivor and killer) are calculated and used as indicators. This is calculated as a 'first indicator' and not as a rating. For example, if the survivor's real playtime is 1000 and the killer's real playtime is 2000, each pre-MMR is calculated at 1000 and 2000. Then, the real MMR is calculated entirely in the past. For example, if a survivor with a pre-MMR of 1000 "wins" a killer with a pre-MMR of 2000, the survivor will have increased MMR and the killer will be deducted MMR.
Ch. 3 Is it really necessary not to change the killers?
If there is no MMR for each killer, and only a very strongly integrated(ideal) MMR exists, there is no need to stabilize the MMR system by preventing changes in each killer. We want to be able to deal with strong survivors with strong killers, and in the case of weak survivors, we want fun play by choosing other weak killers that we want.
By allowing killers to change, and if the MMR difference is large by the changed killer, we can solve it in a way that has no or less MMR change. Of course, in order to prevent Rating Abusing, all players will have to avoid displaying MMR. If there is no Rating Abusing concern, there is no reason not to display MMR.
Ch. 3.5 Indication of Rating
Many people will wonder if the new system works well, and if it shows the rate, even if it is a beta version, they will try to recognize their current skills and build better game competition and skills. In addition, if the system is incomplete, some feedback may be suggested by rating display. Of course, this problem is deeply related to Rating Abuisng, and may be burdensome to easily present Rating algorithms or display MMR. However, we believe it is reasonable to display the rating score in the new Rating (beta system). Displaying both the Rating algorithm and Rating seems to be more helpful for the entire game.
Ch. 4. SWF Issue
The SWF issue is complex. Of course, SWF is much more advantageous than solo que survivors, and it is an in-game feature without penalty. However, it is difficult to eliminate SWF or impose other types of penalties. Any professional level survivor can play with a friend, and we should not taboo this. That friend could be Noobs, and it's great to play and learn from each other.
However, in terms of fairness in the game, it is very important to select the survivor's que and the killer's Rating. This is the same for conventional rank matches, so it is natural to explore matches as criteria for the higher rank in SWF matches. We are not going to discuss this. Of course we have solutions such as showing the killer SWF survivors or giving bonus blood points, but here we'll only talk about matches.
In SWF, of course, SWF with many advantages will have a higher escape rate than solo cue survivors, and rating will be higher. But it's hard to match the right killer by calculating the right level. Matching too high a killer, the survivors of the same team will complain about the Newbie survivor, his rating will not rise or fall at all, and the falling solo Rating will return to "slaughter". Matching too low a killer will also lead to disaster.
The best method is to calculate the MMR 'average' between each survivor of SWF and then explore the match by multiplying the SWF correction factor. This correction factor increases as the number of SWF increases. That is, for example, if 1.5 for duo, 2 for trio, and 3 for squads are multiplied to explore the match. If the error is large, the mean of SWF can be calculated by various means (e.g., RMS, other forms of power mean, etc.) rather than by "general mean".
In SWF, survivors receive rating feedback. However, as the SWF correction factor is multiplied, it will need to receive weakened rating feedback. If the rating abusing method using SWF is not resolved, it can be resolved by limiting the high difference rating ques (not recommended, not wanted by many other forum users), or by adding various restrictions on SWFs rating feedback. Alternatively, simply applying match navigation to higher rates may also be an alternative.
Ch.5 Match Waiting Time and Accurate Skill rating Match, Proper Balance
In countries with large user pools such as North America (Canada, USA), and China, the waiting time for matches is very short. In these countries, the number of survivors is high and the number of killers is relatively high. Therefore, the match latency is low because it is easy to navigate the match latency is low. However, in some countries, particularly those that do not play well in DbD, it is difficult to find a match due to the lack of an absolute pool of players as they rise to a higher skill level. The absolute lack of killer player pools and matching imbalances have led developers to consider 'extended matching' and apply an extended matching system to current live servers.
However, this in most cases resulted in low-rank killers competing with high-rank survivors, resulting in a smaller user pool for the killer. Also, players should be able to get a variety of information from the match lobby. First, it is about the estimated waiting time. Algorithms that can produce expected latency by comparing (average) match pools. We also think that players need to know whether extended matching is being applied due to long latency. Despite the long match time, it is understood that expansion matching is now applied automatically if no match is made in balanced rank
Even if the Rating system is applied, it will not change much. Therefore, even if an extension match is explored, the player should be informed that the extension match is being explored. In the case of extended rate matches, the maximum fairness must be achieved by applying the penalty to the minimum in the case of defeat and giving additional rewards in the case of victory (such as a Bloodpoint Boost). I've suggested the other day about an ideal system; a system that makes toggle buttons small above the match lobby about whether players will focus on Rating themselves or find a fair match even if it takes a long time to wait.
Ch. 6. Connection quality, server region, latency issues
[While I was writing this, 5.1.0 added a hit verification system for Dedicated servers. I would appreciate it if you could read this chapter simply for trivia.]
This chapter is all about the same thing. Players should have the right to play games in optimal connectivity, and be able to check their or other players' connectivity and choose a seamless gaming experience. This relates to current Dedicated Server issues and to determining interactions such as pallets stun or killer strikes. We want to play games with players in the same legion(area), and we don't want to play games with players using VPN.
As a result, the ultimate goal will be to improve the efficiency of the server or game client's computation and communication speed, and to make a fair judgment. This can also block the killer to increase client-side communication delays, which can be such as Ping Abuse. In some other games, it also introduces a "ping kick" function that blocks access to ping that is too high (although it varies from game to game but more than 150ms) to game. Considering the current server latency and communication speed of DbD, there may be problems with smooth games if it is more than 100ms. But I'm not arguing that the ping kick function should be introduced into DbD.
The ultimate goal is to improve judgment, but for now, it seems that there is a very long way to go. This problem can be solved simply by bringing in the system of the past. When there was no Dedicated server, the killer was in charge of the host in a P2P, and both the survivor and killer players could see the opponent's "ping" and check the communication status and participate in the game. But when It came into Dedicated Server, this changed strangely. In a Dedicated Server system, both the killer and the survivor only check the connection ping with the server, and do not know what other players are communicating with each other. Such ping systems should be improved and all other players should always be display RTT in the lobby. Technically, this is done not only by pinging 'me' and 'server', but also by combining ping from 'other players' and 'server' and server latency. In other words, if you're a survivor, it's a system that displaying "the time it takes for a server to handle a hit and 'I' receive a "hit packet" in the lobby after the strike is entered in the killer side client. Conversely, if you are a killer, it also means displayig in the lobby "the time it takes for the survivor to receive the attack packet after the strike is entered in the killer client". The same applies to pallet stunts.
Since the introduction of Dedicated Servers, clients seem to automatically find and connect to the lowest (nearest) latency Dedicated server in the game-start phase.I deal more with the absence of Dedicated Servers. For example, if there is no Dedicated server in the area, it seems to automatically explore Dedicated servers in other countries. This leads to high latency access to servers in other countries even though there is no Dedicated server. As of now, there is no fundamental solution to this. Servers without Dedicated servers need to be added, but which is not possible in areas where AWS services are not allowed, such as China. Nevertheless, it is also impossible to apply the past P2P method to the system for some of them.
If they rent and use Dedicated servers in elevated ping, they and gameplayers in the original area will suffer from unstable connections. We can work this out without difficulty, either. Survivors with high pings connect on Dedicated servers, matching players with similar latency. This allows players with low latency to play smoother games, but the only problem is that players with high latency have to double the pain.
A great solution to this is that, first, the servers mediate between clients, and secondly, each clients do not have to be equal in this arbitration. If the server mechanically judges each client equally, high ping players and low ping players are always forced to take advantage. Therefore, when a server determines whether its judgment is justified, a player with a low ping must have priority. Of course, in the case of a very high ping player, this is bound to have very strange results, so it will be difficult to even establish a match in such a match.(I'm talking about a practical ping kick, but not strict ping kick.
Issues expected according to the new MMR system)
Ch. 1. Massive Match Latency Imbalance.
Naturally, the proportion of skilled killers is very small compared to actual skilled survivor players. In addition, the more skilled a player is, the less pool of player matches is explored. Therefore, I think we need the player will choose the 'match latency priority' match and 'rate accuracy priority' match as I mentioned before. In this new MMR system, this is only a temporary measure if extended matching is used continuously again. Then the killer with low rating will be kidnapped and forced to be sacrificed by match pools of high-rating killers, resulting in a smaller match pool of killer players.
And developers need the advantage of the killer role group to attract new killer players. It's not just about balance and it's about the killer's "mood". For example, offering more blood points or making it easier to get emblems like survivors. Lowering the survivor's match latency by forcing the killer to sacrifice "forcedly" is not a good option. No one wants to be sacrificed for the pleasure of others. This is a penalty for surviving players to choose a 'popular' role group, and an advantage for killer players to choose a 'unpopular' role group. (I mean, in latency of matchmaking)
Of course, reducing this width is absolutely necessary and entirely up to the developers.
Ch. 2. Always: Nurse, Spirit, Blight (and other OP killers)
This is also natural, but in the system we designed earlier, the performance of the 'killer character' is multiplied by the player skill points as a multiple. The higher the killer's performance, the stronger the actual in-game performance compared to the player's skill, which results in a match with professional-level survivors due to higher coefficients. Of course, this is no problem with the fairness of the match, but it is never good for high-level survivors to meet only one group of killers. This would be the statistical basis for the wrong balance patch. Some transitional problems will arise. But this will be the first time developers have been able to correct a game that has so far led to a strange balance foundation. Developers can then base their choices on balancing based on these statistical data.
Ch. 3. Rating Boosting Problem
There may be cases where the rate is forced to boost, or forcefully lowered to abuse the rate. Problems arise, such as deliberate suicide, or by fully enjoying the game with toxic playing and choosing suicide for end games, preventing the rate rise. Alternatively, a sniper queue can force one side to lower the rate and increase the rate on the other to falsify the rate. In community players' discussions on this, there was no clear answer. Balancing an appropriate level of pursuit/survival model seems important. In addition, during the anniversary event or Bloodhunt, five survivors often do score farming work with each other. In this case, the killer's rate will continue to fall, and the survivor's rate will continue to rise. Locking the rate variation to avoid fluctuating well during the event period may also be a helpful temporary solution.
Comments
-
Although I have pointed out some matchmaking, it seems that the MMR system is still having trouble being created, so I would like to suggest a few more systems here. Of course, some of the problems with matchmaking may have already been solved internally by developers, and various other public algorithms may be tested or introduced. It may also be a solution that has already been tried or other problems have been discovered, or that has been abandoned due to feasibility and difficulty in implementing practical development.
We have been using an emblem/rank system, an extremely crude matchmaking system, for more than two years. The downside to this matchmaking system is that it's already been argued on numerous post in forum. It would be a waste of time to discuss this, as developers all know. However, the new matchmaking system (MMR matchmaking) also has some problems. When it comes to matchmaking, we have to consider many variables; calculating asymmetric game-based survival/killer MMR scores, difficulty in determining winning criteria, how players influence each game's winning factors (and how to calculate it statistically), map wining odds (murder/survivor sided), and further, quality of connection (server host latency - accesing with VPN at remote and ruining the game connection quality), imbalance of the killer and survivor's pool of players (which has the same skills), SWFs, and the use of consumables (items, addons, offerings), ... etc. All variables, such this make it difficult to design a new MMR production algorithm, and make the new system difficult to operate.
At the heart of all the most deadly and major issues in the game is matchmaking, and many of the balance issues that are being talked about in the forum are all tied to matchmaking. Also, for competitive games, this matchmaking is almost a top issue in the quality of 'competitive' experience for all gameplay.
Ch. 1 Matchmaking System Design
First, developers seem to be designing the wrong concept for the new matchmaking. What does this mean? First, the reason why this new system is being created is to create a skill-based matching. The developers blocked the killer from changing and to determine each killer's ability by calculating each MMR.
But this is a very wrong idea. This seems to come from the idea that killer players will have a very large skill difference for each killer, but in reality, the actual difference between each skill is not that big. Unless it's a very special case, such as a nurse or a blight, the killer's ability to deal with survivors of similar skills represents about the same level. Therefore, calculating MMR separately creates a great deal of confusion. This allows the matching to blend randomly when a good killer player at one character plays another killer character. And confusion comes, such as many trials before MMR stabilizes. Of course, that is true. Each killer character has a different "tier" and even if it lacks skills, it can have a high kill rate because the killer is powerful. However, it is strange to set separated MMR for this reason. While briefly testing the new matchmaking briefly, The results of each MMR rating interacting with each other and affecting each MMR, also support this argument. Calculating the rating separately because of the strength and weakness of the killer brings us to the same conclusion as calculating the rate of each add-on separately. Some killer addons are powerful, some killer addons are weak. Will we calculate matchmaking for each add-on? This looks very strange.
Therefore, the point is that the killer's MMR should be calculated with only one. This will reduce the complexity of the problem as developers create algorithms for different killers. And it can reduce confusion in the MMR system. If different MMRs are used, each time a new killer is released on the live server, there will be tremendous confusion. And with each chapter released, it is inappropriate for this confusion to repeat. Thus, the conclusion is that the player's MMR is calculated with only one rate. However, as mentioned above, the power and kill rate of each killer vary, so when playing, the match search in such a way that the MMR is multiplied by the modification factor.
For example, a killer player's MMR was calculated as 1000. Since Spirit is a powerful killer, let's say it has a modification multiple of 1.5 (this would be a work to do, It is discussed below). If the players search the match with Spirit, he will be matched with survivors with a rate of 1500. Ghostface is a weak killer, so let's say it has a modificatio multiple of 0.7. When he search the match with Ghostface, he will be matched with survivors with 700 MMRs. The simplest way to calculate the modification factor is to calculate the factor based on the current average kill rate, depending on how much the killer deviates from this average.
But this also creates some confusion, such as Nurse. The current Nusre kill rate is much lower than average. Therefore, the modification factor will be much lower than 1. (Let's assume 0.5 for example) but this is a disaster in the upper MMR. Think of a skilled Nurse player. Let's say he has 2,000 full rates in killer, and 3,000 Nurse 'skill' rate and 1,500 walking killer skill rate. Even though he has 3000 nurse skills, 2000 rates will be multiplied by the modification factor to search lower rates, which will result in opponent survivors being swept away. Of course, if this continues, his killer rating will be very high. He continues to play Nurse, raising the killer rate to 3,000. On the contrary, if he plays a wraith at 3000 (*Will be a modification ratio of wraith will exceed 1, isn't it?), he will suffer from more than 3,000 survivors despite his skill being 1,500. Of course the assumed situation is a very special one, and we are given some choices here.
1. Adjust the modification factors by creating a more sophisticated factors. : This may be explained in more detail below, but it is simply to calculate the factor differently for each MMR series. This modification consists of the overall mean statistics of the MMR class.
2. Get rid of the correction factor at all, and create only a full-killer rating. : This is a great way to improve the overall matchmaking quality, ignoring extreme killer characters such as Nurse, Spirit, Blight and Ghostface. It can also solve the inconvenience of killer players who cannot change their killer.
3.Make each MMR as the developers did. However, this is also problematic, and the development team has made it possible for each of the different MMRs to interact. Number 1 and number 3 are actually very similar, number 3 is the change in the MMR for each player's killer character, but number 1 is the modificatoin for each grade is calculated by the statistics of the class as a whole, and the change in feedback of the game results is applied to only one rating score.
Each has its pros and cons. Now here, we are going to focus on number one and discuss this solution further.
First, the reason why we propose integration scores in this system is that they are more systematically reasonable. In the case of League of Legends, a system was created to calculate different MMRs for each line, and later developers realized it was a stupid idea and later deplecated it. Even games with a large pool of users such as LOL didn't work well, although DbD is a smaller pool of users than LOL, and even fewer killer players, these systems really don't fit. It may seem OK to have different calibrations for different MMR classes. For example, it would be in this form. If the killer MMR is 0–500, the nurse modifying factor will be 0.2, if the rating is 500-100 and 0.7 for 1000-2000..... and 1.5 for 3000-4000.... like this. This factor can be in the form of a table internally or in the form of a graph. Also, this average causes periodic changes. For example, if a patch occurs that nerf a killer character, the factor will be affected in the subsequent season. In addition, if a new type of killer trick or tech is developed (for example, let's say J-flick), this factor will be affected again as a whole.
The expected problem with MMR Skill based matchmaking is that the results of the average game flow and match give feedback back to the correction factor, which again affects the results and causes. Therefore, there is a possibility of unexpected problems.
Ch. 2. Developers should not focus on calculating killer rating, but rather on calculating survivor rating.
It's not that I don't know why developers are more interested in calculating the killer's rating. It is fuzzy Winning or losing this game and making it in system is very difficult and ambiguous, and it is very difficult to determine how much a survivor is contributing to escape and sacrifice. No matter one survivor was killed, if the survivor performed to continue the chase for five generators, and the other three escaped, the survivor contributed 100 percent to the remaining three. However, it seems to be more focused on killer skill-based MMRs because it is very difficult to calculate statistically and realistically difficult to apply it to the system, such as the escape/sacrifice of other players after a certain survivor's game.
Nevertheless, except for some killer-driving characters (and as mentioned above, Nurse, Spirit, and Blight), all kill/escape rates are generally proportional to the survivors' performance, rather than proportional to the killer's performance. Devs seem to have the wrong idea about this. As you saw in Chapter 1, killer skill is not the main factor leading the game. Also, the fair quality of the game is not whether the killer's skill rating is accurately calculated, but how accurately the survivor's rate is calculated and the survivors of the same skill meet. If the killer player's MMR calculates exactly 500 according to the skill, and the survivor player gets 100 and 900 caught in the same game in matching, we don't think it's a fair game. And even if killer MMR is calculated correctly, if 500 MMR is calculated even though skill is 100 survivors, this is an unfair game. This is why it is simply difficult to calculate escape rates. Of course, it is likely that the survival rate will eventually reach a similar level, but tremendous confusion will continue, and there will be many imbalances in skills among survivors. In other words, poor chases and genrush-based survivors may have similar survival rates to those with professional-level chases, which may lead to complaints from survivors with professional-level chases.
The quality of the game is guaranteed if the overall survivors' chasing(looping) skills are leveled. If the 500MMR killer, 100MMR survivor A, and 900MMR survivor B play earlier, the 500MMR killer will find it too easy to chase 100MMR, and it will not be fun to chase 900MMR survivors. Eventually, the 100MMR survivor will be tunneled and the game will end with a quick 4kill.
This is a very complex issue, and it is not simply accessible. Developers tried to solve this problem through the emblem system, but the emblem system, which put arbitrarily created indicators, eventually failed. The methods of calculating each indicator were very different, and it was difficult to calculate in a simple way as the level and meta of the game changed. And it created emblem inflation, and the problem became worse as it changed the rank reset criteria that slightly relieved it. Nevertheless, we need to look carefully at some points. First, there are categories such as how much the generator is repaired, how altruistic is, and how good at the chase. Of course, these indicators are very crucial to the escape of the survivor, but they may not be seen as indicators of skill. At a certain rank level (I mean, in the red rank) the most discriminating indicator is the skill to chase. Pro-level stealth play is not very difficult at a particular skill level, but good chase is highly skill- and experience-dependent. Many community members say that the level of player is generally determined at the chase level, and that even the chase performance evaluation emblem should be doubled. Of course, stealth play is not without merit, but according to past surveys, about 30 percent of players have fun in stealth play, and 70 percent of players have fun in being chased.
We need to pursue a more thrilling chase play experience in upper MMR games to make the game more fun. This was also the reason for the emblem penalty for the killer players camp. Unlike the weak criteria of Bloodpoint, this requires more sophisticated judging criteria. For example, the system's judgment of what the killer is doing (or what survivors are being chased). An evaluation model is needed to evaluate how survivors use pallets and windows, how good chase skills are, and how winning the chase affects escape rates.
Before we talk about the standards of a good chase, let's look at the simpler standards of unhook and generator. Except for HEX:Ruin and Pops, generators are always contributed according to the charge of repair generators. If these two perks are not present, these models are very consistent with the criteria for evaluating 'performance'. Also In the case of POPS, the 'repair contribution' is usually just a repair.
In this case, the survivor's time is divided into only tree: idle time (almost repair time), healing time, and chase time(or it could include moving time). Idle time can include stealth time and can be distinguished from stealth time and repair time.
The most difficult thing is to distinguish the time given to survivors. The simplest way is to calculate the time you don't run. It is very simple. healing time and chase time always occur in a running way, and idle time does not always run. The downside to taking this is time for stealth. However, in normal gameplay situations, it is a kind of penalty if you waste time doing stealth. The problem is that two survivors are killed first, and the rest of themselves struggle. Depending on the need, we can either exclude the time in this case or solve it by reducing the amount of apllied this, generator factor as they are sacrificed.
This is complicated. As survivors are sacrificed at some level of the game, setting the game's repair contribution differently as much as turning the generator is very complicated and difficult. So, how will this be reflected in the MMR? As I said earlier, this can also bring similar models. In the entire game's given idle time, how much of a survivor is repairing a generator is a matter of hardworking genrush. It is easy to calculate this on a relative basis. But as I said earlier, it is not good to simply reflect this in the MMR. This is because the skill ratio of the generator and the skill ratio of the chase are different from each other in the escape. Genrush does not require skill. Therefore, the proportion of the chase needs to be calculated much higher.
Then Let's go back to the HEX:Ruin exception. Of course, with one exception, it is simple to apply the model. And with HEX:Ruin, it is unlikely that there will be a significant difference between repair time and actual generator repair contribution. Nevertheless, destroying HEX:Ruin contributes very much to the game's main escape rate. Giving Advantage points for this destruction may seem reasonable. (But when i ponder about this, i realized it is very hard to give advantage in specific cleaning up HEX totem) Generators that are damaged by HEX:Ruin are, in fact, hard to say, real repairs. Therefore, it seems reasonable not to reflect this amount of destroyed generator in the rating as it is not a actual repair. It also seems reasonable to allow these cases to obtain additional compensation in order to determine that you did nothing at idle time and thus avoid penalties for falling rating.
In any case, the problem is very complicated by HEX:Ruin and other damaged generators, but in the end of the ideal model, the calculation of how much each survivor contributes to the repair until the actual generator completes the repair. Simply, it can be calculated as the amount of time (more precisely, as much as Charge) that the generator is repaired during idle time. As mentioned earlier, given the time to hide from the killer, this idle time can also be calculated only 'strict idle time', not just idle time.
Finally, it's a chase model. This is the surest way to determine how good survivors' skills are in the survival MMR, and to determine how good they are in the MMR system. Therefore, sophisticated systems and access are needed. So far, the system for the chase emblem is divided into several elements (Chaser emblem and Evader emblem). Hiding and being chased are different scores in the Evader category, and scores vary with distance. When camping, the killer gets a penalty at the Chaser emblem.
The survivor's hit (regardlless of injured or dying immediately) is considered to have ended with the killer's winning chase, and seems to be calculated separately as the survivor's victory with the end of the chase.(chasing BGM end) It's a pretty detailed system, but it's unsophisticated, and crude. In particular, it is highly unsuitable for insta-down killers, remote-hitting killer, and hideout killers who disrupt chasing system. Perhaps some reasons include this; that each killer has a different way of pursuing internally, and that each killer has a very different performance. These are likely to be the developers trying to create their own MMR system.
But chasing rating system is more important in terms of survivors than killers. The killer's chasing emblem is still problematic, and it is problematic to use it as a skill indicator. (Remember, in DFC rules, survivors are scored, but killers only evaluate their skills by hooks.) In other words, survivors are more important than the part about killers here.
Survivor>
Sol. 1) Simplified systems
A simplified system leaves everything to the statistical average, and can be implemented relatively easily, but less accuracy. This method begins with the assumption that the killer always tracks at least one survivor. In other words, the killer always assumes that he is in a chase with a certain survivor, and how he (or in some cases multiple survivors) behaves in a chase against the killer (how long does the agro last and win) will be an evaluation factor.
That is to say, for ex, if the killer is at the Pro level and the survivor is at the Noob level and the survivor goes down in just a few seconds, then the chasing Rating in this system subtracts the survivor's Rating.Conversely, if the killer is Noob and the killer drags a huge amount of time down to the Pro level and wins the chase, the survivor gets Rating. (*note: Although the comparison has been made, the rate variation becomes smaller, if the greater the rate difference between the two.)
Killer must search and chase survivors, and at least one of them is always receiving in situation. It may have to be excluded from the exit, but it is meaningful that it is a simple system. It has the advantage of having reasonable results for insta-down killer or semi-insta-down killer because it only cares about 'down' and not about injuries. And technically, it can have a simple form. Of course, it would be different from a Blood Point Chasing System or a Bloodlust Chasing System, and a new 'substantial' Chasing System might be needed. If you're running with the killer within a certain range, we need to make sure that the system is aware that it's in chase.
From the simple way, the survivor who is running when the killer looks is chased, and the survivor who is not running when the killer looks is easily regarded as stealth. The killer's view of the Spin Chill Perk mechanism may be recycled, but if the killer is backstepping, or intentionally approaching the killer while considering of the Spine Chill (In a professional-level game). Of course, the Blood Point Chasing System or the Blood Drust Chasing System should take precedence over 'insight'. Bloodpoint and Bloodlust mean they're being chased, of course. However, as you begin to consider other complex variables, the problem becomes complicated. I think this is the best way. feat. Occam's Razor.
Sol. 2) More sophisticated systems (maps & using pallets)
To refine the system, we have several variables to consider. It is the use of realm and maps. In fact, this is where the problem gets complicated. First of all, there are 36 different kinds of maps in Realm, and of course statistics on each survival rate, and statistics and pallets on average time, the number of windows, etc. It can be sort out a lot of things. But it makes me think again whether it's good to divide it into 36. And when it comes to using the map of survivors, you can compare the time to the number of pallets, etc. The longer of aggro drags and the less pallet usage, the better the survivor's skills are.
First, it is the devil's whisper to apply survival statistics for each map as a correction factor. It sounds really cool to hear, but when you actually apply it, you face a lot of problems. For example, for low-skill Newbie survivors, Gideon map is an indoor map, difficult to locate pallets and structures and can easily die due to difficulty in stealth. But for high-skill professional survivors, Gideon is heaven.
And it's different for each killer. Indoor maps such as Léry’s and Hawkins are difficult for Nurse, but favorable for Ghostface. Cannibal, Hillbilly, becomes very weak in many suffocation pits and Léry’s. Considering this for each realm creates a need to consider a lot of variables together, and we believe that if all 36 map statistics are applied, the problem can be very complicated. For example, in the MMR test match, Hillbilly had a problem, likewise in the realm included MMR system, it can be problematic in Badham 1 and There's no problem with Badham 2.
Anyway, it's almost impossible to consider all of this. Therefore, it seems better to give up this variable. When it comes to using the map of survivors, the number of pallets uses looks really cool. Especially if you were playing and a lot of potatoes throwing away the pallets and you were looking for a pallet and you died blaming pallets, level designer, or team member. Pallets are shared resource for the entire survivor, and the fewer pallets you use (and the more windows you use), its clear you are a professional survivor.
Of course there are some problems here as well. For example, are you better off choosing to die without using a god-pallets? Or is it good for a survivor who doesn't down even if he uses a lot of god-pallets in the first chase? In general, the second survivor is definitely better. But what about this in the match system? What statistically significant difference is there in the rate of use of pallets per second?
And, in the case of Carnival and Oni, they have no choice but to use pallets very quickly. Even if the pallets is used to hold out standing, even if the five generators take turns using all the pallets on the map, the killer will end up with a kill of camp unless multiple survivors down at once. Well, okay, maybe we can put aside some exception killer characters.
NOTE: I didn't consider the killer case because it was passive. Most survivors decide how the game goes. But you can't ignore the killer. This can also use the killer's correction at each chase rate.It is very difficult to deal with Spirit for survivors for a long time. On the contrary, it is very easy to deal with Ghostface for a long time (*Ch. 1)
And even when dealing with professional killers, a higher level of survival skills is needed. And anyone can deal with a Noob-level killer. Therefore, in the feedback of the results, the killer's ability factor is also important. If the killer is Nooob level, you don't get a high reward rating even if you win the chase perfectly, and if the killer is professional, the rating is less even if you lose the chase badly.
NOTE2: In fact, lying down may not be the norm. In the old emblem system, we may base our impact on something else, as it was on the standard. For example, you can base only on when the survivor hooks. If so, smart survivors will be considered for rating, whether they are down under a pallets, using an unbreakable, or even variables such as sabotage and flashlight.
It just mentioned that it's a lot of different judgments (e.g., the killer's camp right now, what time to unhook, where to fix the generator first, drag killer toward not fixing gens, and so on. Now that I mentioned, I'll wrap up the chase model here. These judgement are also an important skill factor, but they are likely to be ignored because the system is difficult to judge. However, survivors who are good at chasing, survivors who have a high chance of survival, are more likely to show better judgment and other performances, so even if this element is ignored, it is unlikely that there will be any major problems.
Ch 2.5 Setting the Standard(initial MMR)
Current emblem systems can bring statistical indicators contaminated by inflation. I mean, it's clear that 'deploying' the first pool of players will be a very difficult task. If, at a certain point, the MMR starts at zero completely, then the MMR will be calculated the same for survivors who have escaped five streak from the yellow rank and five streak escapes from the red rank.
This is also the main reason for confusion in testing MMR. In today's contaminated and forced matched rank-match, the MMR system would see they have forced matching of different MMRs, and these match pools would re-contaminate the MMR system or not have a significant impact on rate feedback. Thus, while the full record in the rank match may be of some help to the deployment of MMRs, it is extremely confusing to use as a main indicator.
Then, how can we calculate the skills at the foundation stage? We think this can be done in playtime.For example, let's say that survivors who play 1000 hours can typically have 1000 skills 'on average'. The rank matching so far can then yield a 'modified' rating, analyzing how survivors with B-level skills have resulted among different players.
Of course, I'm not talking about using Steam Playtime. Steam playtime is not a real playtime, but it is excluded because it includes matching and playing time.Therefore, in-game play times (each survivor and killer) are calculated and used as indicators. This is calculated as a 'first indicator' and not as a rating. For example, if the survivor's real playtime is 1000 and the killer's real playtime is 2000, each pre-MMR is calculated at 1000 and 2000. Then, the real MMR is calculated entirely in the past. For example, if a survivor with a pre-MMR of 1000 "wins" a killer with a pre-MMR of 2000, the survivor will have increased MMR and the killer will be deducted MMR.
Ch. 3 Is it really necessary not to change the killers?
If there is no MMR for each killer, and only a very strongly integrated(ideal) MMR exists, there is no need to stabilize the MMR system by preventing changes in each killer. We want to be able to deal with strong survivors with strong killers, and in the case of weak survivors, we want fun play by choosing other weak killers that we want.
By allowing killers to change, and if the MMR difference is large by the changed killer, we can solve it in a way that has no or less MMR change. Of course, in order to prevent Rating Abusing, all players will have to avoid displaying MMR. If there is no Rating Abusing concern, there is no reason not to display MMR.
Ch. 3.5 Indication of Rating
Many people will wonder if the new system works well, and if it shows the rate, even if it is a beta version, they will try to recognize their current skills and build better game competition and skills. In addition, if the system is incomplete, some feedback may be suggested by rating display. Of course, this problem is deeply related to Rating Abuisng, and may be burdensome to easily present Rating algorithms or display MMR. However, we believe it is reasonable to display the rating score in the new Rating (beta system). Displaying both the Rating algorithm and Rating seems to be more helpful for the entire game.
Ch. 4. SWF Issue
The SWF issue is complex. Of course, SWF is much more advantageous than solo que survivors, and it is an in-game feature without penalty. However, it is difficult to eliminate SWF or impose other types of penalties. Any professional level survivor can play with a friend, and we should not taboo this. That friend could be Noobs, and it's great to play and learn from each other.
However, in terms of fairness in the game, it is very important to select the survivor's que and the killer's Rating. This is the same for conventional rank matches, so it is natural to explore matches as criteria for the higher rank in SWF matches. We are not going to discuss this. Of course we have solutions such as showing the killer SWF survivors or giving bonus blood points, but here we'll only talk about matches.
In SWF, of course, SWF with many advantages will have a higher escape rate than solo cue survivors, and rating will be higher. But it's hard to match the right killer by calculating the right level. Matching too high a killer, the survivors of the same team will complain about the Newbie survivor, his rating will not rise or fall at all, and the falling solo Rating will return to "slaughter". Matching too low a killer will also lead to disaster.
The best method is to calculate the MMR 'average' between each survivor of SWF and then explore the match by multiplying the SWF correction factor. This correction factor increases as the number of SWF increases. That is, for example, if 1.5 for duo, 2 for trio, and 3 for squads are multiplied to explore the match. If the error is large, the mean of SWF can be calculated by various means (e.g., RMS, other forms of power mean, etc.) rather than by "general mean".
In SWF, survivors receive rating feedback. However, as the SWF correction factor is multiplied, it will need to receive weakened rating feedback. If the rating abusing method using SWF is not resolved, it can be resolved by limiting the high difference rating ques (not recommended, not wanted by many other forum users), or by adding various restrictions on SWFs rating feedback. Alternatively, simply applying match navigation to higher rates may also be an alternative.
Ch.5 Match Waiting Time and Accurate Skill rating Match, Proper Balance
In countries with large user pools such as North America (Canada, USA), and China, the waiting time for matches is very short. In these countries, the number of survivors is high and the number of killers is relatively high. Therefore, the match latency is low because it is easy to navigate the match latency is low. However, in some countries, particularly those that do not play well in DbD, it is difficult to find a match due to the lack of an absolute pool of players as they rise to a higher skill level. The absolute lack of killer player pools and matching imbalances have led developers to consider 'extended matching' and apply an extended matching system to current live servers.
However, this in most cases resulted in low-rank killers competing with high-rank survivors, resulting in a smaller user pool for the killer. Also, players should be able to get a variety of information from the match lobby. First, it is about the estimated waiting time. Algorithms that can produce expected latency by comparing (average) match pools. We also think that players need to know whether extended matching is being applied due to long latency. Despite the long match time, it is understood that expansion matching is now applied automatically if no match is made in balanced rank
Even if the Rating system is applied, it will not change much. Therefore, even if an extension match is explored, the player should be informed that the extension match is being explored. In the case of extended rate matches, the maximum fairness must be achieved by applying the penalty to the minimum in the case of defeat and giving additional rewards in the case of victory (such as a Bloodpoint Boost). I've suggested the other day about an ideal system; a system that makes toggle buttons small above the match lobby about whether players will focus on Rating themselves or find a fair match even if it takes a long time to wait.
Ch. 6. Connection quality, server region, latency issues
[While I was writing this, 5.1.0 added a hit verification system for Dedicated servers. I would appreciate it if you could read this chapter simply for trivia.]
This chapter is all about the same thing. Players should have the right to play games in optimal connectivity, and be able to check their or other players' connectivity and choose a seamless gaming experience. This relates to current Dedicated Server issues and to determining interactions such as pallets stun or killer strikes. We want to play games with players in the same legion(area), and we don't want to play games with players using VPN.
As a result, the ultimate goal will be to improve the efficiency of the server or game client's computation and communication speed, and to make a fair judgment. This can also block the killer to increase client-side communication delays, which can be such as Ping Abuse. In some other games, it also introduces a "ping kick" function that blocks access to ping that is too high (although it varies from game to game but more than 150ms) to game. Considering the current server latency and communication speed of DbD, there may be problems with smooth games if it is more than 100ms. But I'm not arguing that the ping kick function should be introduced into DbD.
The ultimate goal is to improve judgment, but for now, it seems that there is a very long way to go. This problem can be solved simply by bringing in the system of the past. When there was no Dedicated server, the killer was in charge of the host in a P2P, and both the survivor and killer players could see the opponent's "ping" and check the communication status and participate in the game. But when It came into Dedicated Server, this changed strangely. In a Dedicated Server system, both the killer and the survivor only check the connection ping with the server, and do not know what other players are communicating with each other. Such ping systems should be improved and all other players should always be display RTT in the lobby. Technically, this is done not only by pinging 'me' and 'server', but also by combining ping from 'other players' and 'server' and server latency. In other words, if you're a survivor, it's a system that displaying "the time it takes for a server to handle a hit and 'I' receive a "hit packet" in the lobby after the strike is entered in the killer side client. Conversely, if you are a killer, it also means displayig in the lobby "the time it takes for the survivor to receive the attack packet after the strike is entered in the killer client". The same applies to pallet stunts.
Since the introduction of Dedicated Servers, clients seem to automatically find and connect to the lowest (nearest) latency Dedicated server in the game-start phase.I deal more with the absence of Dedicated Servers. For example, if there is no Dedicated server in the area, it seems to automatically explore Dedicated servers in other countries. This leads to high latency access to servers in other countries even though there is no Dedicated server. As of now, there is no fundamental solution to this. Servers without Dedicated servers need to be added, but which is not possible in areas where AWS services are not allowed, such as China. Nevertheless, it is also impossible to apply the past P2P method to the system for some of them.
If they rent and use Dedicated servers in elevated ping, they and gameplayers in the original area will suffer from unstable connections. We can work this out without difficulty, either. Survivors with high pings connect on Dedicated servers, matching players with similar latency. This allows players with low latency to play smoother games, but the only problem is that players with high latency have to double the pain.
A great solution to this is that, first, the servers mediate between clients, and secondly, each clients do not have to be equal in this arbitration. If the server mechanically judges each client equally, high ping players and low ping players are always forced to take advantage. Therefore, when a server determines whether its judgment is justified, a player with a low ping must have priority. Of course, in the case of a very high ping player, this is bound to have very strange results, so it will be difficult to even establish a match in such a match.(I'm talking about a practical ping kick, but not strict ping kick.
Issues expected according to the new MMR system)
Ch. 1. Massive Match Latency Imbalance.
Naturally, the proportion of skilled killers is very small compared to actual skilled survivor players. In addition, the more skilled a player is, the less pool of player matches is explored. Therefore, I think we need the player will choose the 'match latency priority' match and 'rate accuracy priority' match as I mentioned before. In this new MMR system, this is only a temporary measure if extended matching is used continuously again. Then the killer with low rating will be kidnapped and forced to be sacrificed by match pools of high-rating killers, resulting in a smaller match pool of killer players.
And developers need the advantage of the killer role group to attract new killer players. It's not just about balance and it's about the killer's "mood". For example, offering more blood points or making it easier to get emblems like survivors. Lowering the survivor's match latency by forcing the killer to sacrifice "forcedly" is not a good option. No one wants to be sacrificed for the pleasure of others. This is a penalty for surviving players to choose a 'popular' role group, and an advantage for killer players to choose a 'unpopular' role group. (I mean, in latency of matchmaking)
Of course, reducing this width is absolutely necessary and entirely up to the developers.
Ch. 2. Always: Nurse, Spirit, Blight (and other OP killers)
This is also natural, but in the system we designed earlier, the performance of the 'killer character' is multiplied by the player skill points as a multiple. The higher the killer's performance, the stronger the actual in-game performance compared to the player's skill, which results in a match with professional-level survivors due to higher coefficients. Of course, this is no problem with the fairness of the match, but it is never good for high-level survivors to meet only one group of killers. This would be the statistical basis for the wrong balance patch. Some transitional problems will arise. But this will be the first time developers have been able to correct a game that has so far led to a strange balance foundation. Developers can then base their choices on balancing based on these statistical data.
Ch. 3. Rating Boosting Problem
There may be cases where the rate is forced to boost, or forcefully lowered to abuse the rate. Problems arise, such as deliberate suicide, or by fully enjoying the game with toxic playing and choosing suicide for end games, preventing the rate rise. Alternatively, a sniper queue can force one side to lower the rate and increase the rate on the other to falsify the rate. In community players' discussions on this, there was no clear answer. Balancing an appropriate level of pursuit/survival model seems important. In addition, during the anniversary event or Bloodhunt, five survivors often do score farming work with each other. In this case, the killer's rate will continue to fall, and the survivor's rate will continue to rise. Locking the rate variation to avoid fluctuating well during the event period may also be a helpful temporary solution.
0