Lazy MatchMaking

[Daniel_Allan_Clark]

I was excited to play my new Renown tonight after getting through the HMS TIger grind.

Loaded in and was top tier.

Oh, but the enemy team had a triple division of gold qualified players and a double division. We had no divisions.

That went as poorly as you would expect. Complete and utter stomp. I had pretty much two thirds of the enemy team chasing my Renown to the map border mere minutes after initial contact after my allied Pensacola died showing broadside and my friendly DD died because he smoked up immediately upon entering the cap.

Yup. Done with the game for the night.

There was no reason that the divisions could not be placed on opposite teams. It was already a fail matchmaking because we were already in a 9v9 game.

Sigh. Time to play something else so I don't get tilted at the absolute stupidity of the game.

[Wolfswetpaws]

 
 

[Unlooky]

 I don't understand why the proposed solution in Clyde's video on MM is not feasible for implementation. Changes nothing about the searching parameters for MM as it is, but adds a sort of "after balancing" by attempting to distribute skilled players evenly. Perhaps too much technical debt with the matchmaker? 

Either way these matches are rather unfortunate especially for players who can only play 1 match occasionally. 

[Daniel_Allan_Clark]

3 minutes ago, Unlooky said:

 I don't understand why the proposed solution in Clyde's video on MM is not feasible for implementation. Changes nothing about the searching parameters for MM as it is, but adds a sort of "after balancing" by attempting to distribute skilled players evenly. Perhaps too much technical debt with the matchmaker? 

Either way these matches are rather unfortunate especially for players who can only play 1 match occasionally. 

Indeed.

It's always funny to me how touchy WG is when people insinuate that they could improve...and then they tell us with a straight face that things like the MM can't be improved.

"Here's your sign."

I ran a Motorsport Manager race instead...got 12th and 14th in the F3 race. Not bad for the 9th best car on the 10 team grid.

[torino2dc]

4 hours ago, Unlooky said:

I don't understand why the proposed solution in Clyde's video on MM is not feasible for implementation.

I watched Clyde's video and I think it is a reasonable proposal. My suspicion, given that the matchmaker hasn't been touched in forever, is that WG isn't confident that they can improve it without breaking it. 

This is code that was probably written in ~2014-15, quite possibly based on a framework that was developed for WoT. We are now almost 10 years of personnel turnover and one company translocation removed from the people who understand how it works. 

If you compare the quality of life improvements on Mir Korabley vs. the lack of same in WoWS, I think we can reasonably assume that the guys who truly understand the spaghetti code stayed in St. Petersburg.

[thornzero]

I've been trying to figure out a strategy for dealing with up tiered matches in randoms. So far, not so good.

When it is even tiered I play decent and win my fair share.

When my team does poorly, I usually top the board but we lose.

When top tiered, my performance is poor, and we mostly lose. I feel that the lower tiered ships are focused on which leads me to think I should play more cautiously; but this does not work either.

It doesn't matter if I play DD or BB. The criteria for a win or a loss is about the same given what I am capable of during a match.

How does your strategy change when up tiered? [Short of calling it a night before you tank your stats.]

[desmo_2]

8 minutes ago, thornzero said:

How does your strategy change when up tiered?

Honestly, it usually doesn't change too much.  Destroyers are pretty unaffected by tier.  Battleships cope pretty well, but you may have to protect your broadside more and use map terrain to limit your exposure a bit.  If I have a bit of support I tend to push up in battleships in an attempt to intimidate and bully red ships into backing off.  I likely won't be as aggressive with that in a bottom tier battleship.  Cruisers require a bit more care.  Ultimately, how the minimap is developing plays more of a role in my strategy than my ship tier does.

Edited July 3 by desmo_2

[Daniel_Allan_Clark]

26 minutes ago, thornzero said:

I've been trying to figure out a strategy for dealing with up tiered matches in randoms. So far, not so good.

When it is even tiered I play decent and win my fair share.

When my team does poorly, I usually top the board but we lose.

When top tiered, my performance is poor, and we mostly lose. I feel that the lower tiered ships are focused on which leads me to think I should play more cautiously; but this does not work either.

It doesn't matter if I play DD or BB. The criteria for a win or a loss is about the same given what I am capable of during a match.

How does your strategy change when up tiered? [Short of calling it a night before you tank your stats.]

I lower my expectations for personal performance...basically.

I try to work more with the team and work to the limited strengths of my ship.

As a CV, if bottom tier...I focus a lot on scouting and spotting until targets that I know I can take on make themselves known (ships alone or with lower AA capabilities...or I can mask my approach with islands).

As a Battleship, I use the right ammunition choice for the ships I'm firing against. Often, your guns are going to not have enough AP pen against the other battleships or even heavy cruisers...so I end up using HE more and relying on fires. Focus from a positioning perspective is staying alive and not going it alone in unless necessary.

As a cruiser, positioning based on how I can stay alive longer and fight non-fair fights. This could mean island camping...or sailing with / near a friendly battleship. It also means not firing and hiding when required. Again, I'm not trying to carry a team in a bottom tier ship...I'm just trying to contribute.

As a DD, it means not getting into gun fights my HP can't take. It means respecting the enemy teams spotting consumables and capabilities and playing within my ships capabilities...

[HogHammer]

Please do your best to keep it civil, folks. The issue in this discussion is "MatchMaking"—not individual players or modes played. We have had several messages reported already for personal attacks.  

Now, back on topic...

@Unlooky brought up @clydeplays and his discussion on the current state and possible fixes to MM.  Here is the link (full video and bookmarks on topics)

Book Marks:

00:00 Intro 00:29 Video Outline 01:04 How the Matchmaker Works Today 03:46 The Complaints of the Player Base 04:57 Clyde's Beliefs: Never Give Up, Never Surrender! 05:59 "SBMM would increase queue times." - WG 06:40 The Questions to be Answered 07:08 The Experiment 07:53 List of Issues with the Dataset (for transparency) 11:11 Dataset Overview 12:13 Battle Performance Indicators 13:08 Which Indicator is best? 14:21 Battle Performance Indicator Results (Basic) 15:38 A More Elegant Way to Measure the Indicator Accuracy 16:25 Example Graph, Good Indicator 17:16 Example Graph, Bad Indicator 18:16 Battle Count in Ship Indicator (Detailed) 19:22 Account Battle Count Indicator (Detailed) 20:13 Win Rate in Ship Indicator (Detailed) 21:27 Account Win Rate Indicator (Detailed) 23:54 Average Base XP in Ship Indicator (Detailed) 24:36 Average Account XP (Detailed) 25:44 Cumulative Account Stat Charts 27:00 Three Problems with Matchmaking 28:36 Steamroll Charts 30:55 Why Players Don't Really Wan't SBMM 32:49 So - what do we do now? (My Proposal) 35:58 Why WG Won't Fix Anything 38:59 Final Thoughts

@clydeplays also discussed MM in a "Fireside Chat"  with WOWS Community Contributor FoamingJetty at about the same time.  Here is the link (full video and bookmarks on topics)

Book Marks:

00:00:00 Intro & Welcome 00:02:25 Who is FoamingJetty? 00:04:10 Why do players ask for SBMM? 00:08:05 How does Random Battle Matchmaking Work? 00:11:56 Queue time is Important 00:14:50 "Final Balancing" 00:17:55 Wargaming's Matchmaking Improvement History 00:19:27 Balancing the the Fleets, not the Players 00:21:43 Other Methods of Matchmaking (Elo, MMR) 00:25:38 Ranked 00:44:27 Can you use an Elo Rating for a Team-Based Game? 00:50:00 Effects of SBMM on Players and Stats 01:09:30 Clyde's Matchmaking Study and Results in Brief 01:22:10 Steamrolls 01:34:15 Why WOULDN'T we want SBMM? 01:40:45 WOWS is a Small Game 01:50:58 What kind of game is WOWS trying to be? 01:55:00 Some challenges WG faces - IF they wanted to fix this 02:00:28 Jetty Sums Everything up & Clyde is Difficult 02:03:30 Clyde's Proposal for Improving Matchmaking 02:09:10 This method vs problems with SBMM 02:16:35 Answering the question, "How do You Know?" 02:23:00 Wargaming House: Austin Edition! 02:23:39 Question: What about adjusting team size? 02:25:48 Question: Would removing max randomness cost something? 02:35:37 Question: SBMM in all modes? Or just Ranked/CBs? 02:38:20 Question: Does a fully random MM better serve all player skills? 02:48:01 The Fear of the World's Greatest Swordsman 02:48:45 Myriad ways to Implement the Solution (The Purple Box) 02:56:18 Does using Base XP Cause Issues over Time 03:05:28 Question: Should WG create separate queues for Premium and F2P Players? 03:13:31 Thanks for watching, go follow Jetty!

I will add that in my dealings with many official WoWs CCs, one recurring issue(s) that has really picked up steam in the last year is MatchMaking and Player Skill/Experience. (The two do go hand-in-hand)  Many generally may just comment on it during a game, but I have also witnessed (close to shaming individual players) during and after matches - leaving downvotes (unsporting behavior) and even digging into the player's WoWs Numbers' stats, and in-game profile while live on a stream.

@torino2dc comments above regarding the overall age of the current MM code is probably a very real issue for WG.  Despite the complexity it may present, I would really like to see WG make an attempt to improve it and at least communicate to players that they are taking a hard look at the current MM method employed in the game.