Tuesday, December 11, 2007

My Perfomance Rating Formula Suggestion

We all have been playing with the idea of some sort of efficiency index of some sort. Here is my suggestion:

PR=100*(((K*0.1)+(A*0.05)+(Sp*0.01)+(AL*0.001))-((D*0.1)+((B+S)*0.08)))
Where: K=Kills, A=Assists, D=Deaths, B=Betrays, S=Suicides, Sp=Best Spree and AL=Average Life (in seconds)

Essentially this equals out to a number between -500 and 500. 500 being a 50 kills streak with no deaths, and -500 being 50 deaths and no kills these extremes are obviously highly unlikely. Most PR ratings are between -50 and 150. A PR of 0 or below is pretty bad, where as a PR of 100 or over was a very good game, with the majority of our games falling in between the 0 and 100 marks.

I used this formula on the 7 games, Pablito, SB, Adeclipse and I played last night. The spreadsheet can be found here.

The highest two PR's of the night were by SB and Adeclipse. Adeclipse had a 115.3 on Guardian (13K/7A/6D/4Sp/63AL) and SB had a 114.9 on Snowbound (10/5/3/4/159).

The lowest of the the night was also by Adeclipse who post a lack luster -31.8 on Isolation (9/7/16/1/22), but to be fair that is his least favorite map.

I however never broke the 100 mark, unlike all of my teammates. Adeclipse had two games over 100 and SB and Pablito both had one. Everyone had at least one game at or below 0.

Here were the average PR's from the 7 games last night.
Adeclipse: 50.3
The Daymonster: 43.8
Squatting Bear: 30.5
Pablito Neal: 26.6

I am definitely open to suggestions to the formula, but once we are set on it, we can never explain it to anyone other than "It's complicated, and it takes into account a bunch of stuff", that way it will seem way cooler. The one thing I do like about this is how it works pretty easily with bench marks (i.e. Negative=Bad, Over 100 really good, etc.)

30 comments:

adeclipse said...

Freaking Isolation.

adeclipse said...

And freaking hangovers.

Pablito Neal said...

that formula looks really good

locopuyo said...

hey that is pretty good, I don't think the average life should be in the equation though, considering you already factor that skill in when you do kills and deaths. Or if you keep that in there you would need a counter balance with time between kills.

Pablito Neal said...

i like how the results look and ya puyo's right. having avg life and deaths is a bit redundant. what about all of dammer's stuff with double kills, triple kills... ?

Daymonster said...

Yeah, I thought about that, the reason I figured that into the equation is if you had a game where you had 2 kills, 0 assists and 2 deaths or so, I wanted you to not get a score of 0. As you did have some sort of positive impact on the game, either by even distracting the other team by being on the map. I wanted the avg. life to put you a little bit over 0.

Granted, the lowest averge life I have ever seen has been like 19 seconds or so, that would give you 1.9 on your PR. Not too much of an impact, so I don't know if there needs to be a counter balance.

I wouldn't be against it, but what stat could you used from bungie.net to go against it?

Daymonster said...

I want to do double and triple kills but we cant find that info from bungie.

And I did mean to say 1 or 2 not 0 for my previous comment.

Daymonster said...

Actually on second thought, I don't have any issue with getting rid of the avg. life factor.

It doesn't change it much. The spread sheet I linked to now has both, PR* is without avg. life factored in to it.

Pablito has brought to my attention that you can get the double and triple kill stats on bungie.net. So I am working on how to get that factored in. Any suggestions.

Jeff said...

Yeah I would throw out the avg. life also. I'm still curious about using the percent of total when related to kills, deaths, and assists though. I guess if you calculate on a game by game basis comparing against your teammates that would already be indicated by your efficiency score. Hmm. If we were to to do an overall rating taking into account a multitude of games I think it might be beneficial, but not if you're just looking at a game by game average. Can you pull medals off of Bungie to get the Killionaire and stuff? I haven't looked.

Pablito Neal said...

yes you can see all medals for every player in each game on bungie if you go to the match and mouse over the player's badge it comes up. formula now looks like:

PR=100*(((K*0.1)+(A*0.05)+(Sp*0.01)+||(2xK*0.02)+(3xK*0.03)||)-((D*0.1)+B+S)

the part between the vertical bars is for the double kill, triple kill (overkill etc).

how about other medals? maybe just add the number of different medals * 0.01 as well? someone with 8 sniper kills would get another 1*0.01 but if you add a sniper kill, plasma stick, hijacker you'd get a 3*0.01.

should we just pick out certain medals that would qualify here? getting incineration certainly isn't helping the team because you're using a shitty gun.

Jeff said...

I don't think other medals are direct indicators of skill persay. If you use a sniper you have a shot at the sniper spree, same with shotgun, sword, etc. A lot of us use the BR and more and there is no spree medal for that. I think it gets too specific to your type of play as opposed to quality of play.

That being said, it would promote more variety of skill, and also promote the proficient use of power weapons.

locopuyo said...

Perhaps we could include multikill medals and killing spree medals.

I don't know about the spree medals though. You kind of get that factored in with number of deaths and longest spree.

Jeff said...

I think for the spree medals it's more of a how many times thing. You only get to see the longest spree, so in a game you could feasibly have 3 Killing Sprees and a Running Riot but only get credit for the 15 spree and not the 3 other 5's, although one would be included in the RR. Perhaps it should be set up to take the longest spree and if there are any extraneous sprees not counted into the longest factor them in? In the above example you would have the spree of 15 and then 2 extra Killing Sprees factored in (one being removed since it was part of the RR).

IMHO I would prefer to not use the longest spree and go with medals instead, further complicating the formula. That opens up the discussion about how much certain sprees are worth though, and more importantly the difference between sprees (if it should grow linear or exponential).

Jeff said...

Or.....take the longest spree, subtract the largest medal value, and factor in what's left. Then factor in the medals. This way a 19 and a 15 spree won't get counted the same.

Pablito Neal said...

the post game carnage report shows us longest streak so i think we should take that (minus the medal for it) and add (KillingSprees * 0.025) + (KillingFrenzies * 0.05) + (Riots * 0.075)...

The problem with this is someone who gets three streaks of 9 will get the same points as someone with a streak of 9 and two of 5: (9*0.01) + (2Sprees*0.025) = 0.14.

This doesn't bother me at all and there is no way to look at stats for second, third longest streak so i don't think we really have a choice unless you're counting them as you're playing.

It makes sense that we took out the avg life factor but that also made it so the ratings all turned into whole numbers which makes them look less precise and our formula less complicated. Perhaps my previous suggestion with adding up other medals could have a multiplier of like 0.007 or 0.009 (instead of 0.01) so that we get tenths in the rating and the other medals are a little less valuable.

Jeff said...

I agree with paragraph 3 pablito. I think that's as accurate spree-wise as we can get. If we had access to Bungie's servers and could parse through their game logs it would be different, but alas we cannot to my knowledge so I think this is the best solution we have so far. (At the least the most accurate)

Whole numbers are lame. All attempts at eliminating them should be taken.

I would like to throw (1.0 - BAC) back out there as a final multiplier just for fun. We could all have 0 alcohol levels unless otherwise specified for a certain game. This way it won't do anything unless you specifically tell it to.

Daymonster said...

Where are we at in the equation, im trying to read these comments but im getting confused here. send me link to the python code.

Jeff said...

I sent you the original stuff. I think right now we're sitting at:

- No average life.
- Spree - largest spree medal calculated.
- Spree medals not included in largest spree factored in
- Other medals not sure on

Anonymous said...

I'm too lazy to read through all of the comments so I don't know if anyone has mentioned this already but I think we should account for the Killjoy medal just for the fact that you allowed a player from the other team to get a killing spree. Of course that may be unfair if you have horrible teammates. Any thoughts?

Daymonster said...

Well, I don't know, I don't think it should be a negative attribute for an individual. As it could definetly have nothing to do with you in certain games. And them getting a killing spree will no doubt show up in your K's and D's. I don't think it should be included at all, but if it was it should be a small positive.

Pablito Neal said...

i think it's a small positive, like any other stupid medal (hijack, laser/sniper kill). the person might be camping in a good spot or have a shotty, sword or sniper that you can now take and help your team.

Jeff said...

Tilla and I discussed this during lunch/through work emails after he posted. We've come to agree that it wouldn't be pertinent for a personal rating, but more for a team rating. Here's the convo:

Me - I think it's more of a team rating thing than a personal rating. If you think of it as times that you ended a spree as opposed to times you were spreed upon it might be worth something, but there's a lot of luck involved with that. I need to ponder it more.

Tilla - I agree…what if you allow the other team to get killing sprees but your teammates get the killjoy medals??? Then it doesn't directly affect you.

Me - Good point….that makes it really hard to trace and apply to a specific individual rating. Especially since more often than not it means the person on the spree ran out of Snipes, BR, rocks, etc. and you happened upon them as they were reloading after a fight or something.

Tilla - If we had a "team" or "clan" efficiency rating, that would be a different story.

Daymonster said...

Alright so here is what I am thinking...

First I want to get rid of the term effeciency in this conversation. I think it implies how he efficient you are, and I think that is missleading. I think this has to do more with perfomance than effency... thoughts?

Also, I really dont think the things like killing sprees and runningriots should be given too much weight. Does someone that gets 10 kills right away and then dies 5 times in a row, have a better game than someone that has 10 kills, 5 deaths more spread out. I don't like how we are at the mercy of when they decide to make the cutt off... like toby said, 10 you get credit for 10, 9you only get credit for 5?

I really want it to be complicated but it should aslo make sense... I have avision of us posting this on bungie and people really liking to compare their games ratings weather it be BTB or TS or doubles.

I have a dream.

Jeff said...

If we do use sprees, like you said they should not be what drastically increases a rating. They are an added bonus of performance. If the other team knows you're owning they will either A: try to take you out or B: run away. In either case you're either creating predictability or fear in the other team. If there was a death spree then I would have that factored in also (hence why I wish we could get our hands on the official game logs).

I look at it this way. If you're in the postgame screen comparing lines, and two people have identical games, what are you going to look at to break the ties. As you go down the line piece by piece, that's how much weight should be placed on the factors.

My thinking is that the bulk of this rating should be determined in this order.

- Kills
- Deaths
- Assists
- Doublekills/Triplekills/etc.
- Longest spree
- Team victory (higher if this was a team rating instead of individual)
- BAC

Daymonster said...

haha you really want BAC in it don't you... well we all need to get calibrated breathalyzers then so we can figure that out...

as for the "how to break a tie argument" if they have identical k/a/d and one has a killing spree, they would also then have a longer death spree (probably)...

But I agree that they should be taken into consideration but i just dont want it to dominate

Pablito Neal said...

for BAC we'd need to steal a calculator from some other place and take body weight and # of drinks as inputs. i don't know how many beers makes 0.06 BAC or whatever

Daymonster said...

http://www.ou.edu/oupd/bac.htm

but the drinks only go up to 8 and the weight only goes up to 240 :(

Jeff said...

I'll get in contact with Luther's Res Life office. Each time I was busted for underage drinking I had to go to a "virtual bar" on the CD they gave me and drink myself to death using various factors such as weight, sex, type of drink, how fast you're drinking it, etc. It was actually quite fun and didn't hinder my drinking at all. If I remember right I took it both times with a drink in-hand.

Daymonster said...

Haha was that the Alcohol 101 cd? I had to do that a few times myself.

I would pound those grain alcohol drinks at the virtual bar until i died.

Jeff said...

Yeah I think that was it. Did you have to go around a party and watch people play quarters and fake being drunk?

I think I made a 400 pounder or something slam beers until he died. I think he made it to 40? maybe only 30.