392 replies to this topic

[face4star]

Sorry for having so many threads dedicated to wotbstars.com. This one will be the last one I create in a near time future.

 

Many people ask to implement WIN8 system based on the data collected on wotbstars.com. We currently have more than 5000 samples what is enough to do a basic analysis.

 

However implementing a full WIN8 algorithm would be too hard as it is required a huge effort that currently I am not able to handle.

 

So there is a simplified version of the algorithm that I am going to implement in the nearest future. Let call it STAR1 rating

 

The core concepts are bellow:

 

1. All tanks that player owns and have more than 50 battles are evaluated individually but not equally. Tier I tank ratings will carry the weight of 1 when tier II will carry the weight 1.1 and tier X will have a weight of 2.

 

2. The final rating will be the average of all the tank ratings for the given player adjusted to the tier coefficient. Higher tiers would have the higher weight but all tiers ratings will affect the final number.

 

3. For each tank a baseline would be determined for a)win rate b) damage c) kills d) spots WDKS. Baseline would be set as the average values for the top 50% of the players who own this tank.

 

4. Players would be evaluated by deviations of their stats from the given baseline. For example if baseline for Panther is 1 kill per game, then players who in average make 1 kill per game will get 1 point. Those who make 2 kills will get 2 points. Those who make 0.5 kills will get 0.5 points.

 

5. Each factor of WDKS will have their weight depending on the tank category with W(winrate) have a constant weight of 0.3 for all tanks. Damage will be 0.4 for TDs and 0.3 for mediums. Exact values to be determined.

 

6. So your overall tank rating will be determined by how close you were to the base line in each of four categories W(win), D(damage), K(ill), S(pot) and weight of this category in the overall rating for this tank type.

 

7. Now the main point. Platoon players would have their coefficients downgraded by 0.7 for all platoon wins they have. If a player has 50% platoon win rate then it is expected he plays 50% of games in platoon. So we will take 50% of his wins and reduce them by multiplying by 0.7. It goes from assumption that playing in platoon player scores approximate 30% better winrate, 30% better damage etc.

 

That it.

 

 

 

Edited by face4stas, 05 February 2015 - 12:31 AM.

[concep]

face4stas, on 04 February 2015 - 07:24 PM, said:

7. Now the main point. Platoon players would have their coefficients downgraded by 0.7 for all platoon wins they have. If a player has 50% platoon win rate then it is expected he plays 50% of games in platoon. So we will take 50% of his wins and reduce them by multiplying by 0.7. It goes from assumption that playing in platoon player scores approximate 30% better winrate, 30% better damage etc.

 

That's not fair whatsoever. Why be penalized for platooning? Why be penalized for playing with friends or clanmates? Platooning is fun, albeit I don't platoon often. But that means you HAVE to play solo which means worse ratings anyways.

Edited by concep, 05 February 2015 - 12:31 AM.

[face4star]

concep, on 05 February 2015 - 12:30 AM, said:

 

That's not fair whatsoever. Why be penalized for platooning? Why be penalized for playing with friends or clanmates? Platooning is fun, albeit I don't platoon often. But that means you HAVE to play solo which means worse ratings anyways.

 

Platoon wins is a function of the performance of two players. When STAR1 is focusing on the individual performance. People play platoon because they expect to get a higher winrate and they do get it.

 

We have to recognize that platooning improves statistics and make provisions for it when check individual stats.

[DisordeR_]

face4stas, on 05 February 2015 - 12:24 AM, said:

Sorry for having so many threads dedicated to wotbstars.com. This one will be the last one I create in a near time future.

 

Many people ask to implement WIN8 system based on the data collected on wotbstars.com. We currently have more than 5000 samples what is enough to do a basic analysis.

 

However implementing a full WIN8 algorithm would be too hard as it is required a huge effort that currently I am not able to handle.

 

So there is a simplified version of the algorithm that I am going to implement in the nearest future. Let call it STAR1 rating

 

The core concepts are bellow:

 

1. All tanks that player owns and have more than 50 battles are evaluated individually but not equally. Tier I tank ratings will carry the weight of 1 when tier II will carry the weight 1.1 and tier X will have a weight of 2.

2. The final rating will be the average of all the tank ratings for the given player adjusted to the tier coefficient. Higher tiers would have the higher weight but all tiers ratings will affect the final number.

3. For each tank a baseline would be determined for a)win rate b) damage c) kills d) spots WDKS. Baseline would be set as the average values for the top 50% of the players who own this tank.

4. Players would be evaluated by deviations of their stats from the given baseline. For example if baseline for Panther is 1 kill per game, then players who in average make 1 kill per game will get 1 point. Those who make 2 kills will get 2 points. Those who make 0.5 kills will get 0.5 points.

5. Each factor of WDKS will have their weight depending on the tank category with W(winrate) have a constant weight of 0.3 for all tanks. Damage will be 0.4 for TDs and 0.3 for mediums. Exact values to be determined.

6. So your overall tank rating will be determined by how close you were in the base line in each of for categories W(win), D(damage), K(ill), S(pot) and weight of this caregory in the overall rating for this tank type.

7. Now the main point. Platoon players would have their coefficients downgraded by 0.7 for all platoon wins they have. If a player has 50% platoon win rate then it is expected he plays 50% of games in platoon. So we will take 50% of his wins and reduce them by multiplying by 0.7. It goes from assumption that playing in platoon player scores approximate 30% better winrate, 30% better damage etc.

 

That it.

 

 

 

 

First of all, thanks very much facestas. If I understand the last point, I personally believe 30% is far too great a percentage. I have about a 70% win rate while platooned which means...the formula would assume I am normally a 40% win rate? About 2/3 of my wins are non-platooned and my win rate there is approx. 58%. As to damage, kills, spotting you often are splitting that with the platoon mate and do not get as much as solo. I know those who never platoon will give me grief over this, but 

Edited by jmstichler, 05 February 2015 - 12:36 AM.

[PhiliPTran]

concep, on 04 February 2015 - 04:30 PM, said:

 

That's not fair whatsoever. Why be penalized for platooning? Why be penalized for playing with friends or clanmates? Platooning is fun, albeit I don't platoon often. But that means you HAVE to play solo which means worse ratings anyways.

 

Exactly. Might as well take away the tank companies in PC battles also or teambattles from xvm. There should be a seperate sheet of stats for platooners

[concep]

face4stas, on 04 February 2015 - 07:34 PM, said:

 

Platoon wins is a function of the performance of two players. When STAR1 is focusing on the individual performance. People play platoon because they expect to get a higher winrate and they do get it.

 

We have to recognize that platooning improves statistics and make provisions for it when check individual stats.

 

WN8 focuses individual performance. That's why there's a rating for each player and not a pair of players. You're basically penalizing people who win more often than others. That's ridiculous. Also, in my experience, platooning does NOT bring more damage to platoon. You usually split the damage.

[twosp2]

Yeah I am not sure about 30 percent but it should be a percent. It is for sure easier to get a higher win rate with especially the same platoon mates.

[Serapth]

Why not make the platoon modifier dynamic?  Take the percentage of wins, percentage of wins platooned, then use the difference between to figure out how to weight platooned wins.

[tpcshadow]

Somewhat unrelated, but can you crunch some numbers on the tanks themselves? It would be interesting to see the averages for tanks themselves, not just to compare with your stats, but to compare against each other. For example, a number of months ago on the boards I did an informal survey and the Tiger I generally had the lowest avg dmg between it an the T29 and IS (those were the only tier 7 heavies, that's how long ago).

[thunderthies]

Excellent, face. Thank you again for your continued hard work on this. It's sincerely apprrciated and look forward to seeing how this evolves.

[face4star]

jmstichler, on 05 February 2015 - 12:35 AM, said:

 

First of all, thanks very much facestas. If I understand the last point, I personally believe 30% is far too great a percentage. I have about a 70% win rate while platooned which means...the formula would assume I am normally a 40% win rate? About 2/3 of my wins are non-platooned and my win rate there is approx. 58%. As to damage, kills, spotting you often are splitting that with the platoon mate and do not get as much as solo. I know those who never platoon will give me grief over this, but 

 

Not 40%. See bellow your stats and how adjusted win rate will be calculated.

battles	wins	number of platoon wins	Win Rate	Platoon wins adjustment coefficient	Platoon wins penalty (2889 * 0.3)	Adjusted wins (8502-867)	Adjusted win rate
13560	8502	2889	62.69911504	0.3	867	7635	56.30530973

[originalmadkilla]

Penalizing platooning is totally fair IMO!

 

The stats are supposed to be an accurate assesment of one player's individual performance. 

 

If somebody platoons 100% of the time and has stats that are way above anybody else, then what's the point of stats in the first place?

 

Platooners need to be brought down to earth and down to reality. The worst player in the world could constantly platoon with a great player, and the worst player in the world would have really nice win stats.

[face4star]

Serapth, on 05 February 2015 - 12:41 AM, said:

Why not make the platoon modifier dynamic?  Take the percentage of wins, percentage of wins platooned, then use the difference between to figure out how to weight platooned wins.

 

Only wins made in platoon are penalized. Not all wins. If you have 10000 wins and only 1000 wins made in platoon then your adjusted number of wins will be 9700. You only lose 300 of wins made when played in platoons. It is perceived that you have 0.7 wins playing solo for 1 win you booked playing platoon. So your platoon wins will be taken with a bit of salt. I do not buy the story that player who win 70 of 100 times with platoon would win the same 70 times playing the same battles solo. 

[DisordeR_]

Can you clarify if the downgraded coefficient is used in the calculation of the baseline average values?  

face4stas, on 05 February 2015 - 12:46 AM, said:

 

Not 40%. See bellow your stats and how adjusted win rate will be calculated.

battles	wins	number of platoon wins	Win Rate	Platoon wins adjustment coefficient	Platoon wins penalty (2889 * 0.3)	Adjusted wins (8502-867)	Adjusted win rate
13560	8502	2889	62.69911504	0.3	867	7635	56.30530973

 

Can you do someone with a high percentage of platoons and someone without to see how they match up?

 

How about fullestnuke and originalmadkilla?

Edited by jmstichler, 05 February 2015 - 12:50 AM.

[originalmadkilla]

BTW - nice job Face4stats! I look forward to seeing how this works out, when it goes live.

[thunderthies]

originalmadkilla, on 05 February 2015 - 12:49 AM, said:

Penalizing platooning is totally fair IMO!

 

The stats are supposed to be an accurate assesment of one player's individual performance. 

 

If somebody platoons 100% of the time and has stats that are way above anybody else, then what's the point of stats in the first place?

 

Platooners need to be brought down to earth and down to reality. The worst player in the world could constantly platoon with a great player, and the worst player in the world would have really nice win stats.

 

I for one see no problem in having stats adjusted based on platooning. I'll be the first to admit that my WR would be lower if I never platooned. In fact it was solely BECAUSE I platooned with players better than myself that I became a more proficienct player. 

 

Anyone saying platooning doesn't increase your chances of winning isn't being honest with themselves. 

[Catainium]

Anything but the platooning penalty... That's just not fair

[thunderthies]

Catainium, on 05 February 2015 - 12:53 AM, said:

Anything but the platooning penalty... That's just not fair

 

Specifically, why?

[Serapth]

Catainium, on 05 February 2015 - 12:53 AM, said:

Anything but the platooning penalty... That's just not fair

 

... and without it the entire thing is meaningless.  So basically damned if you do, damned if you don't.

[face4star]

 

jmstichler, on 05 February 2015 - 12:50 AM, said:

Can you clarify if the downgraded coefficient is used in the calculation of the baseline average values?  

 

Can you do someone with a high percentage of platoons and someone without to see how they match up?

 

How about fullestnuke and originalmadkilla?

 

	battles	wins	platoon wins	Platon wins adjustment	Platoon wins penalty	Adjusted wins	Win Rate	Adjusted win rate	Diff
jmstichler	13560	8502	2889	0.3	867	7635	62.7	56.31	6.39
Serapth	3384	1848	23	0.3	7	1841	54.61	54.4	0.21
fullestnuke	9402	7659	7362	0.3	2209	5450	81.46	57.97	23.49

Edited by face4stas, 05 February 2015 - 12:58 AM.