[OFFICIAL W:ET] Discussion new rating method

[giriel]

Don't start with their xp at zero? (Que veux-tu dire exactement? )

Il y a des serveurs ou les joueurs commencent avec un xp plus elever que d'autre. (Comme franceclan) Et pour calculer les differences on doit stocker trop donc c'est pas possible.

I would like to point out that the system is still going to use a PPM basis. As such the rate calculated will be divided by the amount of time played.
Therefor you should not use it in your formula just yet.
Next to that you can use averages as such:

We can calculate the total amount of damage done by every player(with a minimum threshold cause otherwise it won't be very fair either), damage received, and then divide it by the number of players we used in this calculation.

As such I was thinking more in the direction of :

 { (DamageGiven / AvDamageGiven) / (DamageReceived / AvDamageReceived) } * (Hits - Headshots)     +
 { (DamageGiven / AvDamageGiven)² / (DamageReceived / AvDamageReceived) } * (Headshots)

Scarhand thanks for your suggestions they were useful, some pointers :

- Your left and right side of the formula aren't very proportionate. (Damage_Received / Damage_Given) will be really small and as such it won't have much influence in your calculation.

- Same thing in the right hand side: (HS / Hits) * (Hits / Shots) will become (HS / Shots) which is in most cases also quite small.
Even more so when we have to sum all the shots of all weapons that do headshots and all the produced headshots.

Still it was a great attempt with logical explanation! That's how we can solve this .

For those who want to join in the thought process but have no idea about numbers here are some examples:

MP-40 : hits : 68, shots : 333, accuracy: 20.420420, kills : 13, deaths : 2, headshots : 9
Thompson : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 4, headshots : 0
Panzer : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 1, headshots : 0
Grenade : hits : 1, shots : 5, accuracy: 20.000000, kills : 0, deaths : 0, headshots : 0
Airstrike : hits : 0, shots : 1, accuracy: 0.000000, kills : 0, deaths : 1, headshots : 0
Artillery : hits : 12, shots : 6, accuracy: 200.000000, kills : 4, deaths : 0, headshots : 0
Syringe : hits : 0, shots : 1, accuracy: 0.000000, kills : 0, deaths : 0, headshots : 0
Damage given : 3005, Damage received : 1657, Team damage : 0

Battle Sense : 52808 points
Engineering : 230130 points
First Aid : 50633 points
Signals : 43701 points
Light Weapons : 55480 points
Heavy Weapons : 54792 points
Covert Ops : 50427 points

Colt : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 1, headshots : 0
MP-40 : hits : 169, shots : 852, accuracy: 19.835681, kills : 16, deaths : 33, headshots : 25
Thompson : hits : 131, shots : 574, accuracy: 22.822300, kills : 12, deaths : 11, headshots : 9
Sten : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 1, headshots : 0
Panzer : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 2, headshots : 0
F.Thrower : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 2, headshots : 0
Grenade : hits : 1, shots : 1, accuracy: 100.000000, kills : 0, deaths : 1, headshots : 0
Mortar : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 5, headshots : 0
Airstrike : hits : 1, shots : 5, accuracy: 20.000000, kills : 0, deaths : 1, headshots : 0
Artillery : hits : 10, shots : 7, accuracy: 142.857143, kills : 7, deaths : 1, headshots : 0
MG-42 Gun : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 1, headshots : 0
Damage given : 7564, Damage received : 10004, Team damage : 0

Battle Sense : 52058 points
Engineering : 8571 points
First Aid : 50715 points
Signals : 20084 points
Light Weapons : 53895 points
Heavy Weapons : 50928 points
Covert Ops : 50182 points

Luger : hits : 2, shots : 10, accuracy: 20.000000, kills : 0, deaths : 0, headshots : 0
MP-40 : hits : 36, shots : 156, accuracy: 23.076923, kills : 7, deaths : 33, headshots : 5
Thompson : hits : 132, shots : 501, accuracy: 26.347305, kills : 12, deaths : 11, headshots : 11
Panzer : hits : 86, shots : 79, accuracy: 108.860759, kills : 54, deaths : 7, headshots : 0
F.Thrower : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 1, headshots : 0
Grenade : hits : 0, shots : 4, accuracy: 0.000000, kills : 0, deaths : 2, headshots : 0
Mortar : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 9, headshots : 0
Artillery : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 2, headshots : 0
Damage given : 20334, Damage received : 12394, Team damage : 0

Battle Sense : 52162 points
Engineering : 326 points
First Aid : 53695 points
Signals : 50 points
Light Weapons : 55494 points
Heavy Weapons : 53208 points
Covert Ops : 50011 points

Colt : hits : 11, shots : 54, accuracy: 20.370370, kills : 2, deaths : 0, headshots : 5
MP-40 : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 2, headshots : 0
Thompson : hits : 0, shots : 0, accuracy: 0.000000, kills : 0, deaths : 1, headshots : 0
Mortar : hits : 129, shots : 221, accuracy: 58.371041, kills : 64, deaths : 1, headshots : 0
Damage given : 16742, Damage received : 702, Team damage : 0

Battle Sense : 3241 points
Engineering : 1100 points
Signals : 38 points
Light Weapons : 358 points
Heavy Weapons : 3824 points
Covert Ops : 182 points

More can be found by yourself by playing ET and doing a statsdump

[Scarhand]

We can calculate the total amount of damage done by every player(with a minimum threshold cause otherwise it won't be very fair either), damage received, and then divide it by the number of players we used in this calculation.

As such I was thinking more in the direction of :

 { (DamageGiven / AvDamageGiven) / (DamageReceived / AvDamageReceived) } * (Hits - Headshots)     +
 { (DamageGiven / AvDamageGiven)² / (DamageReceived / AvDamageReceived) } * (Headshots)

Scarhand thanks for your suggestions they were useful, some pointers :

- Your left and right side of the formula aren't very proportionate. (Damage_Received / Damage_Given) will be really small and as such it won't have much influence in your calculation.

- Same thing in the right hand side: (HS / Hits) * (Hits / Shots) will become (HS / Shots) which is in most cases also quite small.
Even more so when we have to sum all the shots of all weapons that do headshots and all the produced headshots.

Still it was a great attempt with logical explanation! That's how we can solve this .

In your calculation you are figuring it in relation to other players already? I was rating you with a number, and that number would be compared to other player's rate and rate for that session. You are comparing them right from the start, I'm not sure how you will be able to factor in the rating of other players for experience calculation if you already use session comparison to determine rate. Also, you give no credit to accuracy.

Damage received was intended to not make a large impact. However, this was just a thought experiment and I didn't plug in numbers, it will be too small of a number. I'll work on this when I get more time.
(HS / Hits) * (Hits / Shots) is not the same as (HS/Shots), it gives as much credit to accuracy as headshot ratio. I believe both should be of equal value. This is also multiplied by your dmg per minute which may make it more significant. Again, I haven't had the time to run tests for my formula.

I'll work on this.

[giriel]

Hmm true, already using the averages is a bad idea.

Still my other remarks stay the same.
HS/Hits * Hits /Shots = HS/shots that's just pure math.

Also we need to be careful with those headshots since not all weapons perform them.

[Scarhand]

Hmm true, already using the averages is a bad idea.

Still my other remarks stay the same.
HS/Hits * Hits /Shots = HS/shots that's just pure math.

Also we need to be careful with those headshots since not all weapons perform them.

My bad on that math part. Only had like a minute to make a post and it was early morning, my brain read HS/Shots * Hits/Shots which is why it didn't add up.
Yes, not all weapons perform them, this is true. However, for weapons that don't use headshots, there is less skill involved in them. Do you think you should get as much rating for a hit with a panzer than a headshot with a luger?

As for my formula, it seems I have forgotten Occam's Razor, or at least the inaccurate summary of it. By multiplying damage (Later became Damage/Time) by your aim percentages, it guaranteed that trying to get good hs/acc will always benefit the player. If it were static, Players with high damage would be relatively unaffected while players with low damage would do better by trying to never miss a shot, rather than trying to kill people. Seemed good, but there is a problem. That will give more of a bonus to players who are already doing better. The Damage_Recieved/Damage_Given does the same thing, and though I had an idea to fix it, I may get rid of that one as well, even though it helped aggressive strategy, because it also gives more rating to players who already have more rating. While this itself wont break a rating system, I would rather it be a more simple formula.

As of now, I am thinking more of this simplistic idea:
(Damage_Given - (.5 * Damage_Received)) * (HS / Shots)
This gives a problem still of messed up ratings from never shooting, or from never getting a headshot (although even the worst players seem to always get many accidental headshots, it cannot be assumed they always will). If anyone can think of a more elegant solution than HS + 1 and Shots + 1, please contribute.
Also, I feel this still encourages camping too much, I need a better way of handling damage received, so that there isn't a disadvantage to playing aggressively.

Still no time to start testing numbers, for now its still a thought experiment.

[giriel]

I like your experiments and thought about some changes:

We want players to make as much damage as possible without losing their accuracy, as such trying to minimize the camping.

(Damage Given) / (Shots - Hits) 

Punishing them for being hit, but minimizing this if they themselves hit a lot of other players :

- (Damage received / Hits) 

Headshots should also be rewarded, that's the tricky part:

+ (Damage Given * Headshots) / (Shots - Headshots) 

-> I took Shots instead of Hits because otherwise I think this will have to big of an impact upon the formula.

Total :

{(Damage Given) / (Shots - Hits)} + {(Damage Given * Headshots) / (Shots - Headshots) } - {Damage Received / Hits} 

About the remarks you made:

-> We could make it so that if someone hasn't done a single shot they will not be rated using this new method.
-> About the headshots and damage received, that is fixed in my suggestion but perhaps doesn't satisfy you just yet .

[Le_Furet]

You are just explaning my formula or make it different form lol...


Pour les serveurs ou les joueurs commencent avec un xp plus elever que d'autre c'est une autre affaire, qui peut surement être régler parallèlement.

[Scarhand]

We want players to make as much damage as possible without losing their accuracy, as such trying to minimize the camping.

(Damage Given) / (Shots - Hits) 

Headshots should also be rewarded, that's the tricky part:

+ (Damage Given * Headshots) / (Shots - Headshots) 

-> I took Shots instead of Hits because otherwise I think this will have to big of an impact upon the formula.

I see where you are going with this, but rather than divide by how often you miss, I multiplied by how many times you hit. Either works. However, I wanted to combine your headshot percentage and accuracy into a single component, your "aim component" as it is solely dependent on your aim. I see you are not wanting to treat headshot percentage the same as accuracy, and I have thought about this; heavy weapons users will not get accuracy bonuses at all if I multiply those values together, and they are entitled to the accuracy bonus. I am really considering your approach to this.

I don't entirely agree with your approach to headshots, however. By multiplying damage by the number of headshots rather than the headshot percentage, you are favoring longer play times. People who play longer may have only 1 headshot per kill, and have 70 headshots over the map, while a person who joined recently may have 2 headshots per kill, but only get 40 headshots total, getting less headshot bonus (Less Damage * Less Headshots) even though he is better at getting them. Shots - Headshots is headshot ratio and accuracy combined, and because accuracy is already accounted for, this should be changed. I guess this was just your attempt at soving problems of 0% accuracy glitching the rate.

Punishing them for being hit, but minimizing this if they themselves hit a lot of other players :

- (Damage received / Hits) 

I REALLY like this idea. I was trying to think of how to do this and this is a good way to not penalize aggressive players too much.

Total :

{(Damage Given) / (Shots - Hits)} + {(Damage Given * Headshots) / (Shots - Headshots) } - {Damage Received / Hits} 

About the remarks you made:

-> We could make it so that if someone hasn't done a single shot they will not be rated using this new method.
-> About the headshots and damage received, that is fixed in my suggestion but perhaps doesn't satisfy you just yet .

I may be changing mine again, but I am trying to think of the proper solution to combining the "aim factor" to the damage given. When multiplied, it favors longer play time. If static aim factor, then a value must be determined for it and it would not hold its rate value compared to damage given. Your suggestions may be very helpful in this part Giriel, you have some good ideas. I'll edit this post and add my latest formula after I think about it more. Edit: Actually Giriel, you are close to having it right, I'm going to change a few things, but it will be based off of what you have.

[giriel]

Great that would be nice.
Still don't really understand the problem with the duration of the players session.
Since it will always be divided by the number of minutes played that shouldn't be an issue according to me.
Or am I missing something here

[Scarhand]

((Damage_Given / Time_Played) * (Hits / Shots)) + ((Damage_Given / Time_Played) * (Headshots / Hits)) - (Damage_Received / Hits)

Headshot ratio and accuracy must not be divided by time played, or if two players have equal aim, the one who played longer would get less rating. Because of this, frag rating will not be given in points and ppm. Damage_Given needs to be divided by time played, and a single rating is issued for that session, which will then be compared with your rating and other players ratings and session ratings.

I don't see any easy way to fix divide by 0s without throwing off accuracy of the rate, without conditional statements. Conditional statements should be possible though. (If hits = 0, then...)

Giriel's first stat example with this formula, low hs, damage, and acc for play time used for calculation:
Damage Given : 3005
Damage Received : 1657
Head Shots : 9
Hits : 81
Shots : 349
Time Played : will use 30 min on these because unknown, though probably much lower with this sample
((3005 / 30) * (81 / 349)) + ((3005 / 30) * (9 / 81)) - (1657 / 81)
23.247851 + 11.1296296 - 20.4567901
13.9206905
13.92

Stats example 2, smg with low damage and more received than given:
Damage Given : 7564
Damage Received : 10004
Head Shots : 34
Hits : 312
Shots : 1439
((7564 / 30) * (312 / 1439)) + ((7564 / 30) * (34 / 312)) - (10004 / 312)
54.666852 + 27.4760684 - 32.0641026
50.0788178
50.08

Example 3, panzer yielding relatively very high damage:
Damage given : 20334
Damage received : 12394
Head Shots : 16
Hits : 256
Shots : 740
((20334 / 30) * (256 / 740)) + ((20334 / 30) * (16 / 256)) - (12394 / 256)
234.482162 + 42.3625 - 48.4140625
228.430599
228.43

Example 4, mortar, high damage but did worse than panzer, but almost no damage received:
Damage given : 16742
Damage received : 702
Head Shots : 5
Hits : 140
Shots : 275
((16742 / 30) * (140 / 275)) + ((16742 / 30) * (5 / 140)) - (702 / 140)
284.106667 + 19.9309524 - 5.01428571
299.02

Obviously the heavy weapons did much better, but smg is more skill dependent, and I would like to try this out with my own stats some time. I don't think it is overly powered to the high-damage heavy weapons, I think it is just the sampling, and uknown times.

Anyone else have thoughts on the formula?

Edit: Played The last 9.73 minutes of Oilraid to get some stats really quick, even though there was very few players playing on the server and the session was short, but I wanted to get a realistic calculation from an accurate play-length:

>>> Map: oilraid

Overall stats for: -=PDP=-Scarhand (0 Rounds)

Weapon     Acrcy Hits/Atts Kills Deaths Headshots
-------------------------------------------------
Knife    :   0.0    0/6        0      0         0
MP-40    :  35.0   21/60       4      1         5
Thompson :  30.7  107/348     20      4        34
Dynamite : 100.0    1/1        1      0
K43 Rifle:                     0      1         0
Foot Kick: 100.0    1/1        1      0

Damage Given: 3679    Team Damage: 0
Damage Recvd: 1627

Damage Given : 3679
Damage Received : 1627
Head Shots : 39
Hits : 130
Shots : 414
((3679 / 9.73) * (130 / 414)) + ((3679 / 9.73) * (39 / 130)) - (1627 / 130)
118.729861 + 113.432682 - 12.5153846
219.647159
219.65

[giriel]

Hmm i dont really like that idea. Then the rating will be influenced by the amount of playtime. It would also require plenty of changes. Dont think that is completely what we want . Perhaps paul could shed some light upon the matter.