392 replies to this topic

[Serapth]

Hey Face4stas... do you have draw rates?

 

Just a point of curiosity, but people commonly say that 50% player wins more than they lose, which due to the fact a draw is effectively a loss when it comes to Winrate, this isn't actually true.  Well... it's true, but 50% isn't the actual cutoff.  

 

Factor in draws and I wouldn't be surprised that the median player is around 46%.  Basically saying anyone with a win rate above that is technically above average.  That would assume that about 8% of games end in a draw, which is probably about accurate to my observations anyways.

Edited by Serapth, 12 February 2015 - 07:55 PM.

[Unicorn143]

Serapth, on 12 February 2015 - 07:54 PM, said:

Hey Face4stas... do you have draw rates?

 

Just a point of curiosity, but people commonly say that 50% player wins more than they lose, which due to the fact a draw is effectively a loss when it comes to Winrate, this isn't actually true.  Well... it's true, but 50% isn't the actual cutoff.  

 

Factor in draws and I wouldn't be surprised that the median player is around 46%.  Basically saying anyone with a win rate above that is technically above average.  That would assume that about 8% of games end in a draw, which is probably about accurate to my observations anyways.

 

I think it is possible calculate draw rates but why ? what the point ? it is a loss.

[Serapth]

face4stas, on 12 February 2015 - 08:00 PM, said:

 

I think it is possible calculate draw rates but why ? what the point ? it is a loss.

 

I want to know what the median point is.  Like I said people talk about 50% like its that point, but when doing averages among a community, it's actually a net sum game.  Basically I want to know what the point a player officially becomes "above average", and draw rates are going to skew that down.  Well that, or you may have already calculated the median win rate... in which case, I'll settle for that.

 

Just a matter of curiosity really.  But as I said earlier, I think a 46win rate is really the median point, or at least, thats my guess.

[Unicorn143]

Aventre3, on 12 February 2015 - 07:13 PM, said:

 

You are making a dangerous assumption. You are assuming that only good players platoon with good players. Certainly I platoon with good players, but I also platoon with players who are not (like my brother). He negatively affects my win rate but I enjoy playing with him. Also some clans good player platoon with those still learning to try and teach them some new skills. That would negatively affect the WR of the better player. I think platooning should be left out. Like in wn8. If you only platoon with dddqqq then you are still bringing something to the battle: dddqqq.

 

I do not see problem with that. If you platoon with a noob and bring a less than average player in the battle you have to be penalized for this. It is the same penalty as for the failed platoons but instead of bringing a low tier tank you are bringing a junk player - result is still the same.

 

There is a real problem however pointed indirectly by concep who said that bringing a platoon mate might improve a winrate but might not be so much helpful for inflating the other stats. Let me explain why. For example in order to win a battle each player must make 1000 damage. 6 players in your team together fell short and made 5000. So there is 2000 that you have to make. Assuming that you are a skilled player and able to do this the profit all goes to your stats.  However if you had a misfortune of bringing in a skilled team mate you would now have to deal with a shortage of 800 that you have to split between each other. Instead of making 1000 of extra damage you and your friend will need to make only 400 each.

 

So for the battles you could win anyway but chose to play in platoon, you and your platoon mate will cannibalize each other damage, spots, hits etc. So ideally you only make a full advantage of platooning when it comes to the winrate. When it comes to other stats the improvements are not that straightforward.

 

To make it even more clear. Assume concep is a 60% solo player and 70% platoon player.

 

So for every 10 battles he players in platoon, the 6 battles he can win alone and grab all damage, kills, hits, spots etc. Only 1 from 10 battles is the one when the platoon mate helps him to survive and get the victory.  Yes in terms of wnirate it is a substantial boost but price for this is the fact that he has to share stats with his platoon mate.

 

Again here we are dealing with assumption that other team members would still be able to make their share of damage irrespectively of you play solo or you play platoon. In reality when platoon mate comes to the battle, then you and him are getting bigger share of the total damage by cutting work left to the rest of the team.  So it is still very well possible to make the bigger share of damage even in the battles you platoon but were able to win solo.

Edited by face4stas, 12 February 2015 - 08:21 PM.

[Serapth]

face4stas, on 12 February 2015 - 08:10 PM, said:

 

I do not see problem with that. If you platoon with a noob and bring a less than average player in the battle you have to be penalized for this. It is the same penalty as for the failed platoons but instead of bringing a low tier tank you are bringing a junk player - result is still the same.

 

There is a real problem however pointed indirectly by concep who said that bringing a platoon mate might improve a winrate but might not be so much helpful for inflating the other stats. Let me explain why. For example in order to win a battle each player must make 1000 damage. 6 players in your team together fell short and made 5000. So there is 2000 that you have to make. Assuming that you are a skilled player and able to do this the profit all goes to your stats.  However if you had a misfortune of bringing in a skilled team mate you would now have to deal with a shortage of 800 that you have to split between each other. Instead of making 1000 of extra damage you and your friend will need to make only 400 each. So for battles you can win anyway, you and your platoon mate will cannibalize each other damage, spots, hits etc. So ideally you only make a full advantage of platooning when it comes to the winrate. When it comes to other stats the improvements are not that straightforward.

 

This would certainly be a huge factor if it was 7 person groups.  But when you have 2 top tier players and 5 whatever... there is still a ton of damage to go around.  Yes, it could certainly have an impact, especially at the top tiers, but nowhere near the level it would if you had 7 good players on a team.

[Anachro12]

Are we trying to make a statistical analysis, or create an opinion? In the end opinions are like a body part we all posess, everybody has one. In the end to truly get to the advantage that platoons give and then adjusting win rate because of that is not a statistical analysis. Whether you want it to be or not. There are too many factors that make the platoon difference in relation to win rate different for different people. It will never be as simple as saying, if someone platoons 60% of the time their win record will be +X%. You may find a correlation, but you cannot say that it is due to platooning and not several other things.

I do believe that platooning makes for more wins in my case. But I don't think it is quantifiable. Many other things come along with platooning, I learn from other players. I see what they do and model it. I protect them sometimes at the cost of my own safety.  This also happens sometimes with someone I am not in a platoon with. They call for help, we go, and we smash the reds, working in perfect harmony. I think stats analysis should stay stats analysis and not subjective changing of stats based on what we see from a correlative relationship.  As seen from my previous dddqqq example and my noob example. How can we come up with a number to subtract from dddqqq's win rate that will also work for the noob's win rate? It is quite impossible and very subjective.  Let the stats say what they say and not inject some subjective changes. If we do that the amount of other changes that are just as valid as a platoons effect on win rate must be brought to bear as I said before. Premium ammo, stock grind vs free exp grind, and there are many more.

 

Otherwise do tell me how an adjustment for dddqqq is going to work the same as it is going to work for someone with a 56% win rate and 80% platoons? Does this 56% person receive the same deduction? different? how will you adjust? Why will it mean anything? Please let me know.

Edited by Anachro12, 12 February 2015 - 08:42 PM.

[CarterK]

Excellent work so far stas! Hey whats the deal with the Ru server having such a high % of super unicums - only 700 players checked, but 50 with >70% w/r! 

[Richie_McCaw]

How about not applying any penalties but putting that players platoon win% in brackets next to the rating. You will quickly be able to make your own assumptions when reviewing two players with the same rating but drastically different platoon victories.

 

Also do you think 5000 battles is a little high for the top players rating? There are a few very good players who fit the tierX condition but not the battles. 

[GoBack2ScrubLand]

CSTCARE, on 12 February 2015 - 10:21 PM, said:

Also do you think 5000 battles is a little high for the top players rating? There are a few very good players who fit the tierX condition but not the battles. 

I concur. (possibly because I'm under 5000 but would be on the list if I had more battles :b)

[mehliveat]

CSTCARE, on 13 February 2015 - 08:21 AM, said:

How about not applying any penalties but putting that players platoon win% in brackets next to the rating. You will quickly be able to make your own assumptions when reviewing two players with the same rating but drastically different platoon victories.

 

Also do you think 5000 battles is a little high for the top players rating? There are a few very good players who fit the tierX condition but not the battles. 

 

But if people know platoon adjustment is included in the rankings, then they can make their own assumptions as well.

 

In any case, the raw data is right there, and the raw data will tell the story much better than any ranking system.  Therefore its better to use the raw data when comparing two players.

Edited by mehliveat, 12 February 2015 - 11:12 PM.

[Richie_McCaw]

mehliveat, on 12 February 2015 - 11:02 PM, said:

 

But if people know platoon adjustment is included in the rankings, then they can make their own assumptions.

But you would not know how much the rating has been adjusted unless you looked up their platoon WR separately.

 

I just think it would be easier to have no adjustments. The people whose WR is highly inflated would be the minority of players. By placing a penalty you probably incorrectly adjust the majority where platooning had a much smaller influence on their WR. 

 

I think the top players WR just continue to get better due to platooning with other good players. Two potatoes will struggle to reach 60% regardless of how many games they play together. It's these people the penalty will punish the most.

[mehliveat]

A couple more graphs.  Take them as you will.

 

I have used the Top 100's data, plotted WR vs damage/battle, and WR vs kills/battle.  

 

Then I calculated an adjusted WR.  The adjustment is just based on my previous graph where a 0.1701 coefficient was calculated.  (1.7% increase in WR for every 10% increase in platoon win percentage)  Using this I plotted the same vs damage/battle and kills/battle graphs.

 

Spoiler

 

 

Spoiler

 

 

Spoiler

 

 

Spoiler

 

 

Spoiler

 

 

[Richie_McCaw]

Thanks for that, it just emphasises that Damage ratio and kill ratio are the best measures for performance imo. Either leave WR out altogether or give it something like a 10% weighting. That way an inflated WR due to platooning will only have a small influence on the final rating.

[Serapth]

The ideal stat would actually be average placement.  This would be consistent and would ignore win and loss rate.  Therefore those people that are consistently the top scorer during a loss get recognized.  of course, the top scorer in a win also get recognized.

[Unicorn143]

CarterK, on 12 February 2015 - 10:20 PM, said:

Excellent work so far stas! Hey whats the deal with the Ru server having such a high % of super unicums - only 700 players checked, but 50 with >70% w/r!

 

[Unicorn143]

Not sure what the deal with Russian forum. It is not possible to start any discussion there about the project. Moderators delete posts without giving a warning. May be they have antispam rules that forbid links to other websites but I saw a thread been deleted that was just about a list of top 100 players. Long story short - I would not really care but it needs more user input from Russia to establish a baseline for the median performance there. Those top players we got are probably devs or people who played WOT on pc for years

Edited by face4stas, 13 February 2015 - 12:54 AM.

[mehliveat]

face4stas, on 13 February 2015 - 10:54 AM, said:

Not sure what the deal with Russian forum. It is not possible to start any discussion there about the project. Moderators delete posts without giving a warning. May be they have antispam rules that forbid links to other websites but I saw a thread been deleted that was just about a list of top 100 players. Long story short - I would not really care but it needs more user input from Russia to establish a baseline for the median performance there. Those top players we got are probably devs or people who played WOT on pc for years

 

They do have this up though

 

http://wotmobilize.ru/stats/

 

No idea on how they collected the data, their forum is fairly empty.

 

[Rocket_na]

Hey Face,

The top player list. Is it real time?

[OGFiery]

mehliveat, on 12 February 2015 - 08:03 PM, said:

 

They do have this up though

 

http://wotmobilize.ru/stats/

 

No idea on how they collected the data, their forum is fairly empty.

 

 

They probably collect the data when it is searched, it takes a bit (up to 20-30 seconds) for a name to go through sometimes.

Edited by concep, 13 February 2015 - 03:20 AM.

[Spekulatius]

The 0.3 penalty coefficient for platooning is not just too high,there is no justification for it. I feel, if an adjustment is going to be made, it should be made based on data analysis from the game and it certainly shouldn't be  fixed coeffficient. In reality, the impact of platoon victories will very much depends on the player skill as well as the skill of the platoon partners. it has to be some dynamic data that can only be determined if sufficient data is available from platooned vs solo games.