Need to stop blow outs - Matchmaker

[Gemlin]

I've tried to get my win rate up the past few weeks, I've come to the conclusion its nearly impossible with the current matchmaker. The teams are so lopsided its not even funny.

Lets look at Match Maker Matchup today.

48.86 VS 55.79 = (DIFF 6.93) RED WIN - BLOW OUT
47.37 VS 51.30 = (DIFF 3.93) RED WIN - BLOW OUT
53.40 VS 47.67 = (DIFF 5.73) GREEN WIN - BLOW OUT
43.10 VS 48.14 = (DIFF 5.04) RED WIN - BLOW OUT
53.92 VS 45.90 = (DIFF 8.02) GREEN WIN - BLOW OUT
51.19 VS 43.89 = (DIFF 7.30) GREEN WIN - BLOW OUT
43.35 VS 47.99 = (DIFF 4.64) RED WIN - BLOW OUT
43.96 VS 51.51 = (DIFF 7.55) RED WIN - BLOW OUT
50.32 VS 52.08 = (DIFF 1.76) RED WIN - CLOSE GAME
48.31 VS 50.36 = (DIFF 2.05) RED WIN - CLOSE GAME

Now mind you, I don't mind losing, but when you stack the team, its just a stomping. a 4.5% or greater difference from what I've seen using Match Maker Matchup = Slaughter.

Wargames - Why don't you balance out the teams via win rate:

Keep Matchmaking selection the way it is right now. However balance the players out between the two teams somewhat even between the two teams. (You already selected the 24 players, just place them on teams that are balanced and NOT stack one side.)

[Dr_Venture]

Do you have ANY idea how much of a coding cluster**** that would be? That would eat up so much dev time just to even get the basic mechanics behind it done alone. 

Are we talking overall WR? Ship WR? There are so many variables and ways to break that system. 

[RipNuN2]

Blowouts still happen even with teams of closely matched skill due to the snowball effect since this game has no respawns or comeback mechanics. 

[Gemlin]

5 minutes ago, Dr_Venture said:

Do you have ANY idea how much of a coding cluster**** that would be? That would eat up so much dev time just to even get the basic mechanics behind it done alone. 

Are we talking overall WR? Ship WR? There are so many variables and ways to break that system. 

It would NOT take that much more time. The game has already selected the 24 ships, just even them out between the teams.

[Helstrem]

Just now, Gemlin said:

I've tried to get my win rate up the past few weeks, I've come to the conclusion its nearly impossible with the current matchmaker. The teams are so lopsided its not even funny.

Lets look at Match Maker Matchup today.

48.86 VS 55.79 = (DIFF 6.93) RED WIN - BLOW OUT
47.37 VS 51.30 = (DIFF 3.93) RED WIN - BLOW OUT
53.40 VS 47.67 = (DIFF 5.73) GREEN WIN - BLOW OUT
43.10 VS 48.14 = (DIFF 5.04) RED WIN - BLOW OUT
53.92 VS 45.90 = (DIFF 8.02) GREEN WIN - BLOW OUT
51.19 VS 43.89 = (DIFF 7.30) GREEN WIN - BLOW OUT
43.35 VS 47.99 = (DIFF 4.64) RED WIN - BLOW OUT
43.96 VS 51.51 = (DIFF 7.55) RED WIN - BLOW OUT
50.32 VS 52.08 = (DIFF 1.76) RED WIN - CLOSE GAME
48.31 VS 50.36 = (DIFF 2.05) RED WIN - CLOSE GAME

Now mind you, I don't mind losing, but when you stack the team, its just a stomping. a 4.5% or greater difference from what I've seen using Match Maker Matchup = Slaughter.

Wargames - Why don't you balance out the teams via win rate:

Keep Matchmaking selection the way it is right now. However balance the players out between the two teams somewhat even between the two teams. (You already selected the 24 players, just place them on teams that are balanced and NOT stack one side.)

Your post is useless.  You think you provide data, but due to not providing definitions we have no idea what your criteria are.  Define "BLOW OUT", define "CLOSE GAME".

As to your demand, it is absurd.  Blow outs aren't caused by the MM.  They are caused by the snowball effect and lack of effective come back mechanisms for the losing side.

[TobTorp]

5 minutes ago, Gemlin said:

It would NOT take that much more time. The game has already selected the 24 ships, just even them out between the teams.

...ill give you an example...you have 24 players...one of them is a unicum, five are total scrub lords, the rest is average... the mm (as you want it) would now stick the unicum with the 5 scrubs in one team, and the rest would be average...who would win?

[ArIskandir]

17 minutes ago, Gemlin said:

Wargames - Why don't you balance out the teams via win rate:

Keep Matchmaking selection the way it is right now. However balance the players out between the two teams somewhat even between the two teams. (You already selected the 24 players, just place them on teams that are balanced and NOT stack one side.)

Balancing on WR won't solve a thing, blowouts will still happen.

Best course to prevent blowouts is players learning to play, so... just get used to blowouts

[Compassghost]

Win Rate isn't consistent between ships and ship classes. Deconstructing players into a singular number doesn't accurately gauge a player's performance. It also undervalues some classes against others.

 

If the 1000 game 85% Midway gets put up against a 1000 game 20% Midway, any one other person being balanced against this will have almost no impact on the game. Could probably give the 85% Midway an AFK person and they'd clobber through it.

[khorender_1]

47 minutes ago, Gemlin said:

I've tried to get my win rate up the past few weeks, I've come to the conclusion its nearly impossible with the current matchmaker. The teams are so lopsided its not even funny.

Lets look at Match Maker Matchup today.

48.86 VS 55.79 = (DIFF 6.93) RED WIN - BLOW OUT
47.37 VS 51.30 = (DIFF 3.93) RED WIN - BLOW OUT
53.40 VS 47.67 = (DIFF 5.73) GREEN WIN - BLOW OUT
43.10 VS 48.14 = (DIFF 5.04) RED WIN - BLOW OUT
53.92 VS 45.90 = (DIFF 8.02) GREEN WIN - BLOW OUT
51.19 VS 43.89 = (DIFF 7.30) GREEN WIN - BLOW OUT
43.35 VS 47.99 = (DIFF 4.64) RED WIN - BLOW OUT
43.96 VS 51.51 = (DIFF 7.55) RED WIN - BLOW OUT
50.32 VS 52.08 = (DIFF 1.76) RED WIN - CLOSE GAME
48.31 VS 50.36 = (DIFF 2.05) RED WIN - CLOSE GAME

Now mind you, I don't mind losing, but when you stack the team, its just a stomping. a 4.5% or greater difference from what I've seen using Match Maker Matchup = Slaughter.

Wargames - Why don't you balance out the teams via win rate:

Keep Matchmaking selection the way it is right now. However balance the players out between the two teams somewhat even between the two teams. (You already selected the 24 players, just place them on teams that are balanced and NOT stack one side.)

How about keeping Random Random.

[weiwaynezhang]

if the numbers don't lie, why even play, just compare win rate and leave...

I always learn something from blown out unless I am last guy standing in my team

[Final8ty]

12 minutes ago, Compassghost said:

Win Rate isn't consistent between ships and ship classes. Deconstructing players into a singular number doesn't accurately gauge a player's performance. It also undervalues some classes against others.

 

If the 1000 game 85% Midway gets put up against a 1000 game 20% Midway, any one other person being balanced against this will have almost no impact on the game. Could probably give the 85% Midway an AFK person and they'd clobber through it.

Then the most obvious thing would to make an exception in regards to CVs.

[Midnitewolf]

49 minutes ago, Gemlin said:

I've tried to get my win rate up the past few weeks, I've come to the conclusion its nearly impossible with the current matchmaker. The teams are so lopsided its not even funny.

Lets look at Match Maker Matchup today.

48.86 VS 55.79 = (DIFF 6.93) RED WIN - BLOW OUT
47.37 VS 51.30 = (DIFF 3.93) RED WIN - BLOW OUT
53.40 VS 47.67 = (DIFF 5.73) GREEN WIN - BLOW OUT
43.10 VS 48.14 = (DIFF 5.04) RED WIN - BLOW OUT
53.92 VS 45.90 = (DIFF 8.02) GREEN WIN - BLOW OUT
51.19 VS 43.89 = (DIFF 7.30) GREEN WIN - BLOW OUT
43.35 VS 47.99 = (DIFF 4.64) RED WIN - BLOW OUT
43.96 VS 51.51 = (DIFF 7.55) RED WIN - BLOW OUT
50.32 VS 52.08 = (DIFF 1.76) RED WIN - CLOSE GAME
48.31 VS 50.36 = (DIFF 2.05) RED WIN - CLOSE GAME

Now mind you, I don't mind losing, but when you stack the team, its just a stomping. a 4.5% or greater difference from what I've seen using Match Maker Matchup = Slaughter.

Wargames - Why don't you balance out the teams via win rate:

Keep Matchmaking selection the way it is right now. However balance the players out between the two teams somewhat even between the two teams. (You already selected the 24 players, just place them on teams that are balanced and NOT stack one side.)

Oh God....Predicting if your going to win or lose a match by player win rate has finally reared its ugly head.  Heaven help us the absolute worst thing ever has made it to WoWS.  Save us, Save us.

Seriously, stop with this stuff.   Random means Random.  You have just as much a chance of being on the winning side as that losing side.  There is no grand conspiracy against you.  Sometimes one side or the other will randomly have more better players than the other, It's OK.  Over 1000 matches it will work itself out.

I mean don't get me wrong, It is annoying when your team just throws the match over and over but it goes the other way too.

[Crucis]

1 hour ago, Dr_Venture said:

Do you have ANY idea how much of a coding cluster**** that would be? That would eat up so much dev time just to even get the basic mechanics behind it done alone. 

Are we talking overall WR? Ship WR? There are so many variables and ways to break that system. 

While I disagree with the OP's desire for SBMM by WR, I disagree with your comment above.  I don't think that it would be difficult at all to code a WR driven SBMM.  I say this as a former professional programmer.  I just happen to think that trying to build a SBMM that's driven by WR is foolish for many reasons.

[Compassghost]

14 minutes ago, Final8ty said:

Then the most obvious thing would to make an exception in regards to CVs.

But how? Sometimes there's only 2 people in queue. What if another person of equal skill never arrives? Do we have to evaluate individual CV win rates? What if they have an 80% WR with T4s but a 20% WR with T10s because they never learned to club? The problem pares deeper and deeper depending on these requirements.

[Crucis]

14 minutes ago, weiwaynezhang said:

if the numbers don't lie, why even play, just compare win rate and leave...

I always learn something from blown out unless I am last guy standing in my team

You can't use WR to program a skill based MM, because the SBMM will by its very nature draw all players closer to a 50% WR.  But that doesn't mean that the better players are suddenly getting worse or the worse players are getting better.  All that's happening is that SBMM is making WR pointless as a skill metric, not that it's that great at it now.

No, if one wants to do an SBMM properly for WoWS, it needs to be divorced from Win Rate.  And IMO, the best skill metric for this purpose would be Average Base XP (overall, not per ship). ABXP wouldn't go up or down appreciably when teams got evened out in terms of skill.  It would be affected a little bit because winning and losing does affect the Base XP one receives for a battle. But in the end, the better players would still earn higher Base XP than the lesser players, win or lose.

 

[Navalpride33]

50 minutes ago, Gemlin said:

48.86 VS 55.79 = (DIFF 6.93) RED WIN - BLOW OUT
47.37 VS 51.30 = (DIFF 3.93) RED WIN - BLOW OUT
53.40 VS 47.67 = (DIFF 5.73) GREEN WIN - BLOW OUT
43.10 VS 48.14 = (DIFF 5.04) RED WIN - BLOW OUT
53.92 VS 45.90 = (DIFF 8.02) GREEN WIN - BLOW OUT
51.19 VS 43.89 = (DIFF 7.30) GREEN WIN - BLOW OUT
43.35 VS 47.99 = (DIFF 4.64) RED WIN - BLOW OUT
43.96 VS 51.51 = (DIFF 7.55) RED WIN - BLOW OUT
50.32 VS 52.08 = (DIFF 1.76) RED WIN - CLOSE GAME
48.31 VS 50.36 = (DIFF 2.05) RED WIN - CLOSE GAME

From a statistics POV. I thank you for posting this up... SO what do we have here boys and girls? We have an example of the bell curve example of the current population of WOWS. This is a perfect example to the following explanation.

• The %45 to %55WR block represents, the %90 percentile of the server population of all regions,
  • What is demonstrated above is perfect reflect if this law/math principle in action... 7 out of 10 battles were with players with in this range
• The %44WR and below represent %5 percentile of the WOWS population..
  • 3 out of 10 matches had players from this percentile. (which is in the standard of deviation).
• The %56WR and above are in the other %5 percentile. Because they dont play huge amounts of volume of games, it doesn't surprise me not to see any %56 or above WR players

Right off the bat, we can tell the MM is picking ships in the population of the queue at the time these matches were created...  MM has no say as to who is available in queue when the fleets are made up. While the OP dont mention this, I am willing to suspect... MM waited 3min and just dumped ships into a match.

We also dont know if this is high tier, mid tier or low tier matches. The tier you play will greatly influence or intensify, the Snowball affect cause by OP or over capable ships.. 

As you can see above boys and girls... This is not a MM issue.. Its the MM scapegoat because we the community love OP WOWS products.

[Asym_KS]

3 minutes ago, Crucis said:

While I disagree with the OP's desire for SBMM by WR, I disagree with your comment above.  I don't think that it would be difficult at all to code a WR driven SBMM.  I say this as a former professional programmer.  I just happen to think that trying to build a SBMM that's driven by WR is foolish for many reasons.

I was as well.......  The MM is working as built.  It relies of a "random distribution" of "skill" to ever work !   That 68% in the middle of the bell curve need to be "average players" or the whole process stops working.   This isn't an ELO platform nor do we even require any form of "skill" to do anything.....................sad, but true.

The coding is suspect is one MM that drives all modes of play with exception statements.  I'd bet it would take a lot of time to work through all of the exception partitions to insure each mode works......  Again, not insurmountable, nor that complicated BUT, not remotely a "value added" project.    Customers complaining make no difference to the profit margin nor income streams..........

[Crucis]

1 hour ago, Gemlin said:

I've tried to get my win rate up the past few weeks, I've come to the conclusion its nearly impossible with the current matchmaker. The teams are so lopsided its not even funny.

Lets look at Match Maker Matchup today.

48.86 VS 55.79 = (DIFF 6.93) RED WIN - BLOW OUT
47.37 VS 51.30 = (DIFF 3.93) RED WIN - BLOW OUT
53.40 VS 47.67 = (DIFF 5.73) GREEN WIN - BLOW OUT
43.10 VS 48.14 = (DIFF 5.04) RED WIN - BLOW OUT
53.92 VS 45.90 = (DIFF 8.02) GREEN WIN - BLOW OUT
51.19 VS 43.89 = (DIFF 7.30) GREEN WIN - BLOW OUT
43.35 VS 47.99 = (DIFF 4.64) RED WIN - BLOW OUT
43.96 VS 51.51 = (DIFF 7.55) RED WIN - BLOW OUT
50.32 VS 52.08 = (DIFF 1.76) RED WIN - CLOSE GAME
48.31 VS 50.36 = (DIFF 2.05) RED WIN - CLOSE GAME

Now mind you, I don't mind losing, but when you stack the team, its just a stomping. a 4.5% or greater difference from what I've seen using Match Maker Matchup = Slaughter.

Wargames - Why don't you balance out the teams via win rate:

Keep Matchmaking selection the way it is right now. However balance the players out between the two teams somewhat even between the two teams. (You already selected the 24 players, just place them on teams that are balanced and NOT stack one side.)

 

Balancing out the skill between the two teams would be helpful and would probably somewhat reduce the likelihood of blowouts.  But at the end of the day, blowouts can and will still happen because they're not entirely driven by large or small skill differentials.  Instead, they're driven by the "snowball effect".

[Final8ty]

Just now, Compassghost said:

But how? Sometimes there's only 2 people in queue. What if another person of equal skill never arrives? Do we have to evaluate individual CV win rates? What if they have an 80% WR with T4s but a 20% WR with T10s because they never learned to club? The problem pares deeper and deeper depending on these requirements.

As the OP said you don't sort in queue you sort after the 24 players have been picked.

And the exception is you don't balance out the teams against the CV.

[Sovereigndawg]

1 hour ago, RipNuN2 said:

Blowouts still happen even with teams of closely matched skill due to the snowball effect since this game has no respawns or comeback mechanics. 

Like RipN said there will always be blowouts, even if there was a skill based MM. What I would like to see is a skill based aiming system, rather than the RNG based one we have.

[Final8ty]

4 minutes ago, Crucis said:

You can't use WR to program a skill based MM, because the SBMM will by its very nature draw all players closer to a 50% WR.  But that doesn't mean that the better players are suddenly getting worse or the worse players are getting better.  All that's happening is that SBMM is making WR pointless as a skill metric, not that it's that great at it now.

No, if one wants to do an SBMM properly for WoWS, it needs to be divorced from Win Rate.  And IMO, the best skill metric for this purpose would be Average Base XP (overall, not per ship). ABXP wouldn't go up or down appreciably when teams got evened out in terms of skill.  It would be affected a little bit because winning and losing does affect the Base XP one receives for a battle. But in the end, the better players would still earn higher Base XP than the lesser players, win or lose.

 

 

2 minutes ago, Crucis said:

 

Balancing out the skill between the two teams would be helpful and would probably somewhat reduce the likelihood of blowouts.  But at the end of the day, blowouts can and will still happen because they're not entirely driven by large or small skill differentials.  Instead, they're driven by the "snowball effect".

Just like everyone will die but that does not mean we should not try to prevent people from dieing early.

Reducing blow out is the point and not flat out preventing them.

[Compassghost]

Just now, Final8ty said:

As the OP said you don't sort in queue you sort after the 24 players have been picked.

And the exception is you don't balance out the teams against the CV. 

 

We pick 2 CVs. WRs 75% and 35%

8 BBs. WRs 50% across the board. (Simplifies the example)

6 DDs. WRs 50% across the board. (Simplifies the example)

8 Cruisers. WRs 45%, 50%, 50%, 55% two of each
Are you telling me the team comp will become:

 

 

CV 75% vs CV 35%

4 BBs evenly matched

3 DDs, evenly matched

4 Cruisers, each side has 45%, 2 50%, and 1 55% evenly matched

We have ignored balancing around the CV.

 

So both sides have virtually identical WR except for CV. One CV is one of the best NA players and the other CV thinks it's cool to play in a plane but zero capabilities to do so in a competitive environment.

 

Is this what you are stating is an acceptable outcome? Because all we've done is made "the targets" more balanced for the CVs. In fact, this directly benefits the CV with the higher WR because they're guaranteed a flat even team distribution as opposed to possibly having to deal with a lower overall WR.

[Crucis]

2 minutes ago, Asym_KS said:

I was as well.......  The MM is working as built.  It relies of a "random distribution" of "skill" to ever work !   That 68% in the middle of the bell curve need to be "average players" or the whole process stops working.   This isn't an ELO platform nor do we even require any form of "skill" to do anything.....................sad, but true.

The coding is suspect is one MM that drives all modes of play with exception statements.  I'd bet it would take a lot of time to work through all of the exception partitions to insure each mode works......  Again, not insurmountable, nor that complicated BUT, not remotely a "value added" project.    Customers complaining make no difference to the profit margin nor income streams..........

Asym, the highlighted portion above is more of a software design function than a software programming function.  This isn't to ding you.  I was responding to the person above who was trying to say that a SBMM would be difficult to program, which is a statement that I disagree with.  Designing a good SBMM is a different story.

Regarding customers, I'm not sure that I can fully agree with you there.  If you have enough unhappy customers that has the potential to hurt revenues, particularly if enough of the "whales" decide to leave or close their wallets.  

[Final8ty]

1 minute ago, Compassghost said:

 

We pick 2 CVs. WRs 75% and 35%

8 BBs. WRs 50% across the board. (Simplifies the example)

6 DDs. WRs 50% across the board. (Simplifies the example)

8 Cruisers. WRs 45%, 50%, 50%, 55% two of each
Are you telling me the team comp will become:

 

 

CV 75% vs CV 35%

4 BBs evenly matched

3 DDs, evenly matched

4 Cruisers, each side has 45%, 2 50%, and 1 55% evenly matched

We have ignored balancing around the CV.

 

So both sides have virtually identical WR except for CV. One CV is one of the best NA players and the other CV thinks it's cool to play in a plane but zero capabilities to do so in a competitive environment.

 

Is this what you are stating is an acceptable outcome? Because all we've done is made "the targets" more balanced for the CVs. In fact, this directly benefits the CV with the higher WR because they're guaranteed a flat even team distribution as opposed to possibly having to deal with a lower overall WR.

Yes the CV is ignored unless its a 4 CV battle.

[ArIskandir]

14 minutes ago, Crucis said:

No, if one wants to do an SBMM properly for WoWS, it needs to be divorced from Win Rate.  And IMO, the best skill metric for this purpose would be Average Base XP (overall, not per ship). ABXP wouldn't go up or down appreciably when teams got evened out in terms of skill.  It would be affected a little bit because winning and losing does affect the Base XP one receives for a battle. But in the end, the better players would still earn higher Base XP than the lesser players, win or lose.

For AB XP to be useful it must be filtered for the Premium/Standard bonus discrepancy. I'm nowhere near sure but I'm under the impression the API register the BXP including the Premium time modifier. Then, you need to correlate it with average tier. Even then, you'll need to calculate some exponential value for the impact of a very good or very bad player on the match outcome as any of those can skew the match result far beyond what their avg XP indicates. 

All in all I don't think all of that would bring any real improvement to the enjoyment of the game for the majority of players.