Wins Above Replacement - The Replacement for GS and Efficiency

[sensei_tanker 3,098]

So, lately there is buzz going around that TO devs need help thinking of a system to critique how good a player is, based on skill/performance/equipment. They just switched from Efficiency to GS, which makes more sense, but is still not a good system of keeping track how well a player does in battle and how good the player is, not to mention how you don't know how MU'ed up the opponents are exactly. I think with my method, there will finally be an accurate way of finding how good a player is statistically.

 

A statistic used in professional baseball is WAR, which stands for Wins Above Replacement. The calculations in this are extremely complex; however, in baseball, it effectively measures the impact/valuableness a player has on his team compared to the average player. The higher your WAR, the better player you are. If you WAR is negative, you are hurting your team by playing for it. The unit of the WAR is a win. It kind of directly shows how many wins in all of the games you've ever played your team got because of you.

 

For my TO version of WAR, I will simplify the calculations to make it more understandable and accurate.

 

My equation is not final, and there will be edits, so here I go. Right now it's simple. After I perfect it, I will turn it into an I&S:

 

Please note that the calculations only are based off of how much you are helping your team. Camping and not doing anything will not help your WAR rating.

 

11 basic components of WAR:

 

1. Gear Score

2. Shot Efficiency (Hits vs. Shots Taken) Note that hitting 2 tanks with 1 shot counts as 2 hits vs 1 shot taken so it's possible to go above 100%. Shot efficiency is split into multiple pieces for Railgun, Shaft, Thunder, Smoky, Magnum, Striker, and Ricochet. 

3. Flags Captured (CTF)

4. Points Captured  (CP)

5. Nearly-Capped Flags Returned  (CTF)

6. Goals Scored Per Game (RGB)

7. Important Kills per Game (An important kill is basically any kill that kills an enemy connected to a point, or holding your flag, or any enemy with a current 4+ tank killstreak, or a player with at least 2 supplies activated)

8. Average Points Per 10-Minutes (MMS only)

9. Average Placing Per Full Match on your winning team (MMS only)

10. Average Placing per Full Match on your losing team (MMS only)

11. Isida Teammate HP restored (Only counted when you are using Isida)

 

Our Variables:

 

GS = (Your Gear Score - Average Gear Score for Your Rank) / 100

SE = 100 * (Your Shot Efficiency % - Average Shot Efficiency %)

CF = (# of Flags You Captured / 10) - (Average # of Flags Captured per 10 minutes)

CP = (# of Points You Capped / 10) - (Average # of Points Capped per 10 minutes)

RF = (# of Flags You Returned that was within the radius of sqrt(Map Area)/8 meters)/10 - (The average of same stat per 10 minutes)

GOAL = (# of Goals You Scored / 10) - (Average # of Goals Scored per 10 minutes)

IK = (# of Important Kills You Made / 10) - (Average # of Important Kills per 10 minutes)

PP10 = sqrt(2 *( Your Average Points per 10 Minutes in MMS - Average Player's Points per 10 Minutes in MMS))

WIN = (Your Average Placing on a Winning Team - 2.5) ^ 2

LOSE = (Your Average Placing on a Losing Team - 4) ^ 1.2

HEAL = (Your Average # of Teammates Health Restored per 10 minutes  - 1st Quartile of average players of that stat per 10 minutes) if HEAL > 0. Otherwise HEAL = 0.

 

WAR = GS + SE + CF + CP + RF + GOAL + IK + PP10 + WIN + LOSE + HEAL

 

In theory, the average WAR should be 0. I might have to make all the WARs positive in order to encourage less-skilled players because they might get sad if they see that they are hurting their own team by playing. 

 

Also, with this system, we can remove the GS score from appearing when we press v and go back to the old system where we could see the modification levels of turrets/hulls. 

 

Please critique all you want, your suggestions are highly needed, I feel like if this goes right, this system could basically fix all of our matchmaking woes. The possibilities are endless. When I'm done, this will be turned into an I&S. 

 

Edited September 10, 2018 by sensei_tanker

[Maf 32,300]

I honestly don't understand why something like this doesn't exist yet. Yes, we had the effectiveness/efficiency rating, but I doubt that it was particularly good at showing a player's actual ability. Something like this would not only show who you're up against, but it could be used to create well balanced MM battles with teams of similar power.

 

Your variables definitely have flaws. For example:

• Some turrets simply aren't made for accuracy (Twins/Magnum), no matter how good of a shot you are
• What counts as an "important kill"? How would the game know?

[Pythor 1,354]

WAR in Tanki is actually a very good idea. Being a hard core MLB fan, I see firsthand how WAR is a really effective stat

[DEVlL 80]

The shot efficiency doesn't really work as it puts people with turrets like thunder and magnum miles above turrets like smoky or ricochet, even though said thunder and magnum players may be much less skilled than the others.

[Guest]

The shot efficiency doesn't really work as it puts people with turrets like thunder and magnum miles above turrets like smoky or ricochet, even though said thunder and magnum players may be much less skilled than the others.

I thought about it but if you think about it, turrets like ricochet and twins with no splash or penetration have a large amount of shots that will be fired and are relatively easy to aim with, so their percentage may be very high. Turrets like Railgun or Magnum, who can hit numerous enemies, only fire single shots and are relatively harder to aim with, so their percentage may be dragged down, where their compensation is the ability to hit multiple enemies. The only turret which may suffer in this system is Smoky

[railguniz4noobz 2,394]

I thought about it but if you think about it, turrets like ricochet and twins with no splash or penetration have a large amount of shots that will be fired and are relatively easy to aim with, so their percentage may be very high. Turrets like Railgun or Magnum, who can hit numerous enemies, only fire single shots and are relatively harder to aim with, so their percentage may be dragged down, where their compensation is the ability to hit multiple enemies. The only turret which may suffer in this system is Smoky

Fire rate means absolutely nothing, if you fire 100 shots and hit 90% of them you have the same efficiency as if you fire 10 shots and hit 9. Twins is meant to suppress enemies with constant fire, the game would heavily punish people for being smart about their infinite ammo. Firebird/freeze/isida/vulcan wouldn't even work in this system since it's almost impossible to be accurate with melee turrets and vulcan is also meant for suppression.

[Guest]

 

I honestly don't understand why something like this doesn't exist yet. Yes, we had the effectiveness/efficiency rating, but I doubt that it was particularly good at showing a player's actual ability. Something like this would not only show who you're up against, but it could be used to create well balanced MM battles with teams of similar power.

 

Your variables definitely have flaws. For example:

• Some turrets simply aren't made for accuracy (Twins/Magnum), no matter how good of a shot you are
• What counts as an "important kill"? How would the game know?

 

The shot efficiency doesn't really work as it puts people with turrets like thunder and magnum miles above turrets like smoky or ricochet, even though said thunder and magnum players may be much less skilled than the others.

Fire rate means absolutely nothing, if you fire 100 shots and hit 90% of them you have the same efficiency as if you fire 10 shots and hit 9. Twins is meant to suppress enemies with constant fire, the game would heavily punish people for being smart about their infinite ammo. Firebird/freeze/isida/vulcan wouldn't even work in this system since it's almost impossible to be accurate with melee turrets and vulcan is also meant for suppression.

Hmm... I think I will split the shot efficiency into multiple variables for the turrets railgun, shaft, thunder, magnum, striker, smoky, and rico. 

 

For example, now your Railgun Shot Efficiency is compared to all player's Railgun Shot Efficiency, your Shaft compared to all Shafts, Smoky compared to all Smoky, etc. 

 

As for Maf's second question, an important kill is simply when you kill an enemy who is carrying your flag, connected to a point, holding a RGB ball (ASL not included). It should be pretty easy for the game to know whether or not the enemy's death counts as an important kill. 

[Tani_S 736]

Well you could also take the average inflicted damage (includes splash) within one minute instead of the shot accuracy.

[a7x_Pedro48 13,999]

Yes, we had the effectiveness/efficiency rating, but I doubt that it was particularly good at showing a player's actual ability.

You can still see it on Profiles.

 

But efficiency is garbage. You can be awesome, play a battle like a pro, kill 55 tanks and earn 1k, but if you lose, you'll get poor crystals reward and your efficiency will drop. The Exp/Cry ratio is too important in the calculation of efficiency.

[ThirdOnion 4,898]

I do not think any individual rating system will work in Tanki. Make a competitive mode that forces players to play matches in full, assign players special ranks based on how many matches they win, give people winrates.

Edited September 10, 2018 by ThirdOnion

[wolverine848 13,097]

How does one measure ability to turn turret?

Is it pass/fail = a static measurement?  Determined once?

 

Shot accuracy way to variable to be used as an indicator.  Too many turrets are so different.

 

So a firebird entering your base with OD active may not count as "important kill"? Even though your base will likely be wiped out?

"Important kill" is impossible to measure.

[wild004 35]

GS currently is better, way easy to tell tank's level.

 

8000 GS means good decent MUed m3 combo and protection module. (m4 max = 7000, drug asset max 1800, total max wo paint, tx = 8800)

 

6000 GS means acceptable, 1500+1500 m3 combo + 750*3 m3 module =  5250.

 

if anything like 4500, means your teammate is too weak

[Tani_S 736]

How does one measure ability to turn turret?

Is it pass/fail = a static measurement?  Determined once?

 

Shot accuracy way to variable to be used as an indicator.  Too many turrets are so different.

 

So a firebird entering your base with OD active may not count as "important kill"? Even though your base will likely be wiped out?

"Important kill" is impossible to measure.

Yes, the ability to turn the turret is not really measurable.

I think one could philosophize about what is an important kill. You could measure it depending of the past or on the future moments. For the past I'd prefer to take the points he got after his last respawn as an indicator. For the future you could take being drugged up as important.

[DEVlL 80]

GS currently is better, way easy to tell tank's level.

 

8000 GS means good decent MUed m3 combo and protection module. (m4 max = 7000, drug asset max 1800, total max wo paint, tx = 8800)

 

6000 GS means acceptable, 1500+1500 m3 combo + 750*3 m3 module =  5250.

 

if anything like 4500, means your teammate is too weak

GS isn't very accurate though, for example my GS is 5900 when I run 10/10 twins, 10/10 viking, 3x26% modules, synesthesia and a drone.  My GS is low because I only have 60 repair kits (even though I have 1000 of all the other drugs) so a relatively low GS (5000-6000 at my rank) doesn't mean that the player will weigh your team down.

[r_Issimo2 1,495]

And why do we need a WAR number, instead of a GS number, for MM?

 

If it is for matching players for battle, then the exp and rank system could become useless and that would be problematic for game economy.

 

What we need is to know the MUs of other tanks, and to make it less obstrusive than before, if that was the reason behind the update, it could be like this:

 

35 - 30 - 20 (turret - hull - drone)

 

But a win/loss ratio could be used after ranks to place players in MM.

[Guest]

How does one measure ability to turn turret?

Is it pass/fail = a static measurement?  Determined once?

 

Shot accuracy way to variable to be used as an indicator.  Too many turrets are so different.

 

So a firebird entering your base with OD active may not count as "important kill"? Even though your base will likely be wiped out?

"Important kill" is impossible to measure.

I removed the ability to rotate a turret as one of the variables. I realized that whether or not you turn your turret should reflect on your overall gameplay, so I let the other stats speak for itself.

 

I split Shot Accuracy over multiple turrets so it is now many different variables, 1 for each turret. I hope that is better. Or else, I will have to get rid of it and go with average damage output per 10 minutes.

 

I simply set Important Kill as for when you kill an enemy who is trying to be a playmaker (so basically an enemy who is trying to cap). But, I see your argument, and I will change it so that if you kill an enemy who is on a 4+ tank kill streak, that also counts as an important kill.  

[wolverine848 13,097]

I removed the ability to rotate a turret as one of the variables. I realized that whether or not you turn your turret should reflect on your overall gameplay, so I let the other stats speak for itself.

 

I split Shot Accuracy over multiple turrets so it is now many different variables, 1 for each turret. I hope that is better. Or else, I will have to get rid of it and go with average damage output per 10 minutes.

 

I simply set Important Kill as for when you kill an enemy who is trying to be a playmaker (so basically an enemy who is trying to cap). But, I see your argument, and I will change it so that if you kill an enemy who is on a 4+ tank kill streak, that also counts as an important kill.  

It would be even more "important" to kill that tank before it has a kill-streak and wipes out your base

 

IMO I don't see how this stat can be quantified properly. Way too many variables.

[Guest]

It would be even more "important" to kill that tank before it has a kill-streak and wipes out your base

 

IMO I don't see how this stat can be quantified properly. Way too many variables.

Fine, i will also include tanks with at least 2 supplies activated. 

 

There aren't way too many variables. Look at how the original WAR is calculated in baseball. 

[wolverine848 13,097]

Fine, i will also include tanks with at least 2 supplies activated. 

 

There aren't way too many variables. Look at how the original WAR is calculated in baseball. 

Sorry I should have been more specific - too many variables to determine what an "important kill" is.

I don't see how it can be implemented into WAR properly.

It's a subjective stat, and different people have different definitions of what "important" is.

[Guest]

Sorry I should have been more specific - too many variables to determine what an "important kill" is.

I don't see how it can be implemented into WAR properly.

It's a subjective stat, and different people have different definitions of what "important" is.

To be honest, I don't think it really matters what variables determine what an important kill is. Regardless you are always comparing your # of important kills to the average player's # of important kills, so if you think about it, the difference between those 2 should almost always be the same, regardless what we define what is an important kill is. 

 

However, I will still think of a better way to turn playmaking into an actual stat. 

[Potdindy 866]

I love this. It could be something that could reset when it is created and then it can adapt up and down with time. You could even tweak values to reward players for team play. Meaning k/d could influence this so players who have lower k/d from capping flags would be put in lower/easier battles while those who camp would be put in harder battles

[thesquarerootof2 165]

The name is fitting for this game so therefore this system must be implemented