Destiny 2 Tracker Network

This thread is closed

MMagMaR
MMagMaR

1 Credits 5 Posts
1 Credit(s) Received
Does anyone know what the DTR score is based off of?


Mod edit:
For those looking for more information on DTR Score, give our Site Score help thread a look.
Morarn
Morarn

180 Credits 295 Posts
10 Credit(s) Received
McVeigh
McVeigh

10 Credits 25 Posts
MMagMaR said:Does anyone know what the DTR score is based off of?


Everything I think. I'm guessing its just a formula that takes into account kills, k/d etc and gives each of them a weight which is used to calculate the DTR. But thats a complete guess but @xorth would know the answer.
Permisslon
Permisslon

0 Credits 2 Posts
No clue iv'e been here a while never do anything though.
RhippiN
RhippiN

377 Credits 157 Posts
4 Credit(s) Received
DTR Score is based of all the data we collect from PvP such as kills / medals / assists / deaths ect

The front page score is currently just the medal score from Bungie which favours game types such as control as the score limit is higher.

DTR is a much more robust and 'skill' based system
Yin
Yin

311 Credits 336 Posts
5 Credit(s) Received
It's based on your stats! The exact formula is for me to know, and you to (hopefully) never find out!
XxdumpalumpxX
XxdumpalumpxX

0 Credits 2 Posts
okay cool thanks for explaining that rippin heads my DTR score Rumble playlist is at 45 on psn so i guess thats a good thing then =p
HTK Styx
HTK Styx

0 Credits 1 Posts
MMagMaR said:Does anyone know what the DTR score is based off of?


Just play a lot, and play good. That's what I'm gathering from this.
Subject22
Subject22

3 Credits 2 Posts
RhippiN said:DTR Score is based of all the data we collect from PvP such as kills / medals / assists / deaths ect

The front page score is currently just the medal score from Bungie which favours game types such as control as the score limit is higher.

DTR is a much more robust and 'skill' based system


Old thread I know, but I was wondering if you could clear something up. There are four things which bug me about skill ranking scores.

1) Scores which are based on number of wins or win/loss ratios are much less interesting/useful to me, as they favour players who often play in non-matchmade teams (i.e.: in pre-made teams).

2) I'm also less interested in scores which can be increased simply by playing for longer (eg: take two hypothetical players of exactly equal skill and have one play twice as much as the other - without changing skill - a less useful score would assign a higher value to the player who played longer).

3) Skill ranking systems which favour higher scoring playlists (eg: control) are also much less useful.

4) Some scores decay with time (i.e.: if you stop playing for a week or two your score starts to drop) which, again, makes the system less accurate (for my purposes, anyway).

You've already said that DTR is robust to 3), but how about the other three?

Thanks!
RabidSTD
RabidSTD

0 Credits 4 Posts
So, what's considered a GOOD DTR score?
Yin
Yin

311 Credits 336 Posts
1 Credit(s) Received
Subject22 said:1) Scores which are based on number of wins or win/loss ratios are much less interesting/useful to me, as they favour players who often play in non-matchmade teams (i.e.: in pre-made teams).

2) I'm also less interested in scores which can be increased simply by playing for longer (eg: take two hypothetical players of exactly equal skill and have one play twice as much as the other - without changing skill - a less useful score would assign a higher value to the player who played longer).

3) Skill ranking systems which favour higher scoring playlists (eg: control) are also much less useful.

4) Some scores decay with time (i.e.: if you stop playing for a week or two your score starts to drop) which, again, makes the system less accurate (for my purposes, anyway).


Super late on the reply, but maybe other curious minds would like to know the answers to these questions.

1) When DTR score looks at wins and losses for it's calculation, the formula will give or take away points for the result. It is not based on accruing a massive number of wins, nor is it based on averages. However, with any competitive statistic, playing in a team will always yield a better score. That said, the formula also awards points for individual performance, so you don't need a team to gain score.

2) When you create a formula with site score, you essentially need to focus on the end goal: do I want a score that shows your career accomplishment that continues to advance while I play, or do I want a formula that gives a value based on averages. We chose the best of both worlds in that your score will be higher the more you play, but it also bottoms out and becomes a numbers game towards the end (earning score becomes more difficult as you advance). I must say that I disagree with the score being less useful when two players of equal skill do not have the same score if one plays for longer. If I have a 2.0 K/D after 50 games and you have a 2.0 after 200, you deserve a slightly higher score because you've kept your performance at my level for a longer period of time. Very basic example, but I think it illustrates the point.

4) Score shouldn't decay as a result of inactivity, there is nothing in our calculations that considers your play time investment as we don't feel this is relevant to calculating someone's skill.
Yin
Yin

311 Credits 336 Posts
RabidSTD said:So, what's considered a GOOD DTR score?


It really depends. In the early stages of a Destiny career (much like yours), you will continue to accumulate score simply by playing more Destiny. Those who might be more skilled than you will earn that same amount of score at a faster rate, so the best way to use the statistic is for comparisons against other players. Does X person have more score because Y? That sort of thing. If you have a higher score with less games, particularly as early as you are, then your performance is good.

You could also use your leaderboard position for DTR score and compare number to other stats that will naturally climb like games played.
Subject22
Subject22

3 Credits 2 Posts
3 Credit(s) Received
Darkened Night said:Super late on the reply, but maybe other curious minds would like to know the answers to these questions.


Hey, no worries. I'm still subscribed :P Thanks for taking the time to respond. It was an interesting read, and I appreciate that there are bound to be differing opinions on this topic.

I do still view the bias towards playing in a team as a weakness, but a formula which didn't have this bias would likely have other problems, I suppose.

I see your point about maintaining a higher KD over a larger number of games, but I'm not sure I agree with your conclusion. In scientific terms a larger sample size tends to increase your confidence in a predictive model (to a point), although you're better off just estimating bias (error) and using that information to improve your model. There are probably more scientifically accurate methods for measuring player skill, but I don't know how easy their results would be to communicate on a site like this, which is valued for its conciseness (among other things).

Good to hear that inactivity doesn't reduce score.
Yin
Yin

311 Credits 336 Posts
1 Credit(s) Received
You're absolutely right regarding sample sizes impacting your confidence in rating someone at a certain point, that sort of formula is reminiscent in Halo 3's TrueSkill formula style, where more games increase the confidence in the existing rating, slowing the fluctuation of the skill value until they zero in on where the player should reside. It is difficult given the constraints I am forced to work within (given then I don't have access to as many numbers as a developer would) to create a model that can accurately estimate a skill level and still allow the formula to allow players to improve as they go (and not stick people in a certain place based on averages).

The concern we focus on is allowing players to "redeem" themselves as they improve. Of course, low-skilled players will end up with a precise cluster of data points on the lower end of the spectrum and the confidence would likely restrict them to the lower bounds for the remainder of their time on the game (or at least reduce them to a snail's pace) without considering outliers (we don't maintain enough data from each collection to keep track of such a range).

I'm not a statistician by any stretch, but I am interested in learning more. So if you have any opinions or other methods you have in mind, I'm open to look into. Any considerations would be greatly helpful when I design or refine the features around our sites.
thdestroyer11
thdestroyer11

0 Credits 1 Posts
Im the lighthouse master
Hello Im Swat
Hello Im Swat

438 Credits 1521 Posts
thdestroyer11 said:


No, I'm the lighthouse keeper!
narcosys
narcosys

0 Credits 1 Posts
It is obvious that dtr does not reflect true skill but rather who can grind it the most. I see players with a dtr of 22 that have a kd of . 7 and a win/loss of. 3.
Whereas my stats, though not entirely impressive, are much better than this and I'm sitting at 14 dtr.
I don't play often, but when I do I play well. This shoes me that you have to grind and play alot to get your dtr up. Severely flawed
Yin
Yin

311 Credits 336 Posts
narcosys said:It is obvious that dtr does not reflect true skill but rather who can grind it the most. I see players with a dtr of 22 that have a kd of . 7 and a win/loss of. 3.
Whereas my stats, though not entirely impressive, are much better than this and I'm sitting at 14 dtr.
I don't play often, but when I do I play well. This shoes me that you have to grind and play alot to get your dtr up. Severely flawed


I think you mean TrueSkill. There's more information available regarding it in our Help Thread.
redshawn717
redshawn717

0 Credits 1 Posts
MMagMaR said:Does anyone know what the DTR score is based off of?

Mainly W/L ratio then K/D ratio and then other stats


Mod edit:
For those looking for more information on DTR Score, give our Site Score help thread a look.

This thread is closed