Stix wrote:Id go after one that was distracted (preferebly a pro with no better than a steel sword) and then after slaying them id retreat to heal up then go after someone with a mese/diamond sword (if i didn't already have one that is).
Only works some of the time, when the other player is distracted. I already covered this ("unless you get lucky", which doesn't count).
Stix wrote:its so low that atm complete noobs can gain access to it.
Not true. Explain how you went from (implicitly) saying "all the pros have a high K/D" (which is impossible unless a) there are a lot of noobs and b) the pros feed off of the noobs) to saying "everyone has a high K/D" (which is outright impossible).
Stix wrote:And if you cant then your an even bigger noob and have no excuse.
I've already said multiple times, give me a steel sword and I can kill someone else who has a steel sword with roughly a 50/50 chance. That's good enough for me and it should be good enough for anyone else. Expecting more is unreasonable. If you don't believe me I'm open to a 1v1 next time we meet on the server (you provide the swords, of course).
Xudo wrote:I agree that you can't earn good weapon. But it is the case if you are alone. Thats why you need to team up with someone and try to outnumber your enemy.
2 stone swords > 1 steel one.
That does work, sometimes. That's one reason why my K/D is 0.9, not 0.1.
Xudo wrote:There also tunneling strategy and "sky bridge" strategy to steal flag.
I have used and seen other players use those strategies before. Works until two players from the other team catch on and kill you while you're alone with nothing but a wooden sword.
Xudo wrote:micheal65536 wrote:What is actually needed, but which is never going to be implemented and which I don't really expect to be implemented, is a player matching system. Players with similar scores (i.e. a measure of skill) should be matched together in a match (multiple matches run in parallel). Then you don't have one good player getting frustrated in a team full of noobs, or noobs screwing up a team of good players, and so on.
I agree with you. Do you have any specific model in mind? Feel free to describe your specific idea in issues on github.
Not sure what you mean by "model" here. Basically, teams would be assembled between players with similar scores or rankings, and matches would be held between similar teams. Pairings could be performed based on the distribution of rankings between candidate players for a team, or by dividing the rankings into a number of "bins", or any other pairing algorithm.
It would however seriously complicate things. There would need to be multiple matches at the same time so that everyone gets a chance to play. Once a player has finished a match, they would need to wait in a lobby area until there are enough players to pair together to form a new match (rankings change after each match so the players would need to be re-paired after each match, and there would need to be a balance between waiting for closer pairings with players that are still playing another match and going with a wider pairing with a player that is already available). The maximum number of players on the server would need to be increased so that enough players are available to divide into multiple matches without the teams being too small. I imagine that altogether this would require a lot of re-structuring of code and have significant implications on server load.
Xudo wrote:micheal65536 wrote:1.5 simply does not make sense if the intention is to separate good players from bad players, or to separate players who waste weapons from players who use them productively.
Why, exactly? Good players do have K/D higher than 1.5. See rankings table.
Players are not equal. Some are good. Some are bad. K/D distribution is not
normal. It is
long-tail one. Some players are having unnattainable high K/D while vast majority are having really low K/D.
I am well aware that there are many players with a K/D higher than 1.5, and that K/D on the server does not have a normal distribution. This is part of the problem. Let me try to explain.
First suppose that the distribution of raw player skill levels (not scores, actual skill which cannot really be measured in practice but will be expressed on an arbitrary scale for the present purposes) is normal. In reality it probably isn't but as you will see later this actually doesn't matter too much and it's easier to explain if it is.
Now suppose that a player with a higher skill level will always kill a player with a lower skill level. In reality this is not the case but it simplifies the calculation. Now we can calculate the expected K/D of a player with a particular skill level as follows:
Code: Select all
kills(skill) = P(z < skill)
deaths(skill) = P(z > skill)
kdratio(skill) = P(z < skill) / P(z > skill)
(P is the cumulative distribution function for the distribution of player skill levels)
Now suppose that we expect a "good" player to be somewhere around the mean skill level of all (regular) players on the server. If we define the average (mean) players as "good" players then most of the players are likely to be satisfied but the below-average players (i.e. noobs) will still be classified "bad" players. We can calculate the expected K/D of a player with the mean skill level as follows:
Code: Select all
kdratio(mean) = P(z < mean) / P(z > mean)
Using the definition of mean:
Therefore
So it may be concluded that an average player on the server will have a K/D of 1. To appeal to an average player, the K/D requirement should be set around 1.
Now, I am well aware that in reality the K/D distribution on the server looks a lot more extreme, and that apparently all of the "good" players have a high K/D and there are a lot more players with a low K/D. This is indicative of a few issues.
The first issue is that it is assumed that players with a K/D lower than 1.5 are "bad" players, when in reality some of them may be good players who simply don't have a high K/D. Unless you want to exclude everyone but a smaller portion of players at the top of the skill range from being "good" players, instead of including the entire top half of all players, then this is a flawed assumption. As I have demonstrated, the top half of all players will have a K/D greater than 1. Assuming a fair environment, less than half of the players will have a K/D greater than 1.5 (the actual portion depends on the distribution of player skill levels).
The second issue, a development of the first issue, is that the few (relatively speaking) players that have a high K/D are now preventing more potentially-good players from achieving a high K/D, because they have superior weapons. This is very likely one reason why there are so many more players with a low K/D than a high K/D.
Xudo wrote:micheal65536 wrote:how anyone can reasonably expect every good player on the server to simultaneously hold a K/D greater than 1.5.
Thats because not everyone is good. Bad players has K/D much less than 1. Probably around 0.5. Thats why it is possible to require 1.5 from good players.
Depends how you define "good". If you define "good" as "average", which will include the majority of active players while excluding the below-average noobs who join once and then give up, then the expected K/D could be as low as 1. See previous quote for detailed explanation.
Xudo wrote:micheal65536 wrote:The only other solution would be to avoid score/ranking entirely, and to have a system based on "give back what you take out".
I agree with this idea. Thats why I submitted issue about
Chest points.
That does seem like a good system although I disagree with the idea of resetting every match, because this can easily create a situation where one or two players snap up all the loot chests at the start of the round and then the other players have nothing to trade for. To me the best method would be to have a total "score" of items taken and a total "score" of items given for each player, and if the items taken score is greater than the items given score by more than a decided threshold then they cannot take any more items. The score is a weighted count of items, where something like a pistol or a steel sword counts as 1, lower-valued items count as a lower value (say, a stone sword could be 0.5), and higher-valued items count as a higher value (say, a diamond sword or a rifle could be 2). If players "saving up" the score is a potential problem, then the score of items given could be capped so that it cannot increment beyond the items given score plus a decided threshold.