TW matchmaking - what's the logic behind this?

Now I already saw enough threads about people complaining that TW matchmaking is broken. I thought it's supposed to be improved since like a year ago or so?

The last two match-ups of my guild looked like this, according to C3PO:
(Left is ours, right is opponent guild)
second-last match:
Total : 254,263,617 vs 360,816,319
GL's : 62 vs 167
Jedi Master Kenobi (9 vs 39)
Jedi Master Luke Skywalker (11 vs 33)
Lord Vader (1 vs 12)
Rey (8 vs 18)
Sith Eternal Emperor (19 vs 29)
Supreme Leader Kylo Ren (14 vs 36)
Executor (12 vs 38)
(omitting any other important chars, like CAT, GAS, etc)

last match:
Total : 278,740,925 vs 358,045,324
GL's : 64 vs 148
Jedi Master Kenobi (10 vs 29)
Jedi Master Luke Skywalker (11 vs 34)
Lord Vader (1 vs 9)
Rey (9 vs 18)
Sith Eternal Emperor (20 vs 25)
Supreme Leader Kylo Ren (13 vs 33)
Executor (12 vs 34)

I know, the bot reporting the opponent's values don't take into account how many actually signed up for the TW, but when facing a wall of JMKs it's clear that the majority of guild members signed up.
In both cases (and frequently in the past) we received a FC almost instantly, while we're struggling cutting through the first line.

How on earth was this MM algorithm developed? The motivation in our guild is drastically going down. Next time we're going to refuse setting any defence at all.

Quoting CG_Crumb in a "state of galaxy" from Sep 29, 2021:
With a greater incentive placed on victory, it is important that we address matchmaking and ensure Guilds are paired against fair opponents. Today’s Update reworks the matchmaking system in Territory Wars to more accurately calculate the amount of power a Guild is bringing to the holotable. Our goal with these changes is to create a more fair and balanced experience long term. This system better accounts for guilds that bring a smaller number of powerful players. Additionally, the system takes loss streaks into account during matchmaking. You might also notice that cross-divisional matchups can now occur to help smooth the transition between Divisions. (Note that the number of Defenses needed is determined by the lower Division’s Guild but each Guild’s prize pool is determined by its own specific Division.)
While these changes are intended to have immediate effects, we also expect this to be the starting point for a greater ability to fine tune the system in the future. We will be monitoring the impact of these changes very closely and will likely make adjustments in the future.

Would be great to see some "fine tuning" very soon, instead of wasting valuable developer time on rubbish like Datacrons

Replies

  • Legend91
    2266 posts Member
    Their "solution" is to hand out free wins after you lose 2 TWs in a row. So theoretically you end up with like a 33% win rate. They can't be bothered to come up with a propper TW matchmaking update because of laziness and/or autopilot mode not allowing them to invest resources into anything else than new chars, ships and packs to sell (or creating new things to milk on like datacrons).
    Legend#6873 | YouTube | swgoh.gg
  • Artumas
    321 posts Member
    What I've noticed, personally, is this.
    You'll have a long streak of decent matchups where you win a ton, then you'll reach a tipping point where your winning streak matches you against someone significantly out of your league, you'll lose, THEN you'll get a SECOND match-up that's out of your league, and lose again, and then you'll get back to normal matchups.

    I don't know why it's like this. But this has consistently been how my guild's matchups have worked since they changed TW matchmaking. For whatever reason, it takes 2 matches in a row for matchmaking to "fix itself" after a winning streak pushes you into opponents you don't stand a chance against.

    Now, I can't really say anything if your streak is longer than 2, but you specifically mentioned 2 battles here.

    We've had a like 70-80% win rate, with those 2 specific battles being pretty much the only ones we usually lose. It'll be like 8 wins in a row, then 2 losses, then 7 wins in a row, then 2 losses, etc.
  • But even this doesn’t happen consistently. Our most recent TW, our 3rd win in a row, was a complete mismatch and the opposing guild had absolutely no chance. They’d lost their last 2, and their reward was to get absolutely stomped.
  • Dashrunner
    25 posts Member
    edited August 4
    Artumas wrote: »
    What I've noticed, personally, is this.
    You'll have a long streak of decent matchups where you win a ton, then you'll reach a tipping point where your winning streak matches you against someone significantly out of your league, you'll lose, THEN you'll get a SECOND match-up that's out of your league, and lose again, and then you'll get back to normal matchups.

    Fair enough, I've seen that in the past in our guild as well. However in the last 5 match-ups we won exactly 1. That one was the only one that was fair, all others all had way more GLs, Exes etc.
    But let's see. So theoretically our next match should be an auto-win for us.
    Although, to be honest, we never had an auto-win iirc, the ones we won were almost always even setups in terms of GP, GLs and meta fleets.
  • TW is supposed to be end game level PvP at whatever your team is in their phase of building the rosters but the rewards are still a few zetas and some gear. I’m not in a top guild but we are mostly players with 8.5M + accounts and two or three zetas aren’t really worth the effort when we have 200+ zetas banked waiting for the next release.
    I have given up on CGs ability to match up two guilds that will offer a fair fight. CG just is not able to figure it out despite all the available data ranging from who joined, what toons they have, history in other PvP events, times they tend to play… CG has the data, thry lack the data scientists.
    Interesting is waning. The rewards are not worth the effort if you have been playing for years.
    Rant complete.
  • Lumiya
    443 posts Member
    Range1974 wrote: »
    TW is supposed to be end game level PvP at whatever your team is in their phase of building the rosters but the rewards are still a few zetas and some gear. I’m not in a top guild but we are mostly players with 8.5M + accounts and two or three zetas aren’t really worth the effort when we have 200+ zetas banked waiting for the next release.
    I have given up on CGs ability to match up two guilds that will offer a fair fight. CG just is not able to figure it out despite all the available data ranging from who joined, what toons they have, history in other PvP events, times they tend to play… CG has the data, thry lack the data scientists.
    Interesting is waning. The rewards are not worth the effort if you have been playing for years.
    Rant complete.

    It's the relic mats they want guilds to hunt.
    We are all made of star-stuff
  • But in all fairness, why is it so difficult to just match two guilds that are roughly even in terms of GP, GLs and Win/Loss streaks?
    I mean, it should be possible to rate based on GP (of those who participate), #GLs, #Exe and #wins in the last 5 wars. Weigh each of these factors, possibly weigh GLs differently and there you have it.
    But maybe I'm seeing this too easy?
  • Dashrunner wrote: »
    But in all fairness, why is it so difficult to just match two guilds that are roughly even in terms of GP, GLs and Win/Loss streaks?
    I mean, it should be possible to rate based on GP (of those who participate), #GLs, #Exe and #wins in the last 5 wars. Weigh each of these factors, possibly weigh GLs differently and there you have it.
    But maybe I'm seeing this too easy?

    If you count number of GLs and number of execs, then match up even numbers, what is the incentive to unlock GLs or exec?

    Your new unlock will just mean you are matched against a guild who is even with you anyway.

    This would take away an incentive to unlock things, so CG is unlikely to do that.

    GP matching doesn't do the same thing, because the quality of gp can be vastly different between guilds. I think I saw a post from someone showing a GAC opponent with 6M gp and 6 GLs. A whole guild of players like them could be 300M gp, but 300 GLs, where most 300M gp guilds would have what, maybe around 100 GLs?

    So I think closer active GP matching would be the way to go. Guilds could work to build better overall rosters if they really want to win. And if a guild is outmatched in GLs that's on them.
  • MaruMaru
    2716 posts Member
    Lumiya wrote: »
    Range1974 wrote: »
    TW is supposed to be end game level PvP at whatever your team is in their phase of building the rosters but the rewards are still a few zetas and some gear. I’m not in a top guild but we are mostly players with 8.5M + accounts and two or three zetas aren’t really worth the effort when we have 200+ zetas banked waiting for the next release.
    I have given up on CGs ability to match up two guilds that will offer a fair fight. CG just is not able to figure it out despite all the available data ranging from who joined, what toons they have, history in other PvP events, times they tend to play… CG has the data, thry lack the data scientists.
    Interesting is waning. The rewards are not worth the effort if you have been playing for years.
    Rant complete.

    It's the relic mats they want guilds to hunt.

    They want to make us into zombies that chase brainz.
  • Artumas wrote: »
    What I've noticed, personally, is this.
    You'll have a long streak of decent matchups where you win a ton, then you'll reach a tipping point where your winning streak matches you against someone significantly out of your league, you'll lose, THEN you'll get a SECOND match-up that's out of your league, and lose again, and then you'll get back to normal matchups.

    I don't know why it's like this. But this has consistently been how my guild's matchups have worked since they changed TW matchmaking. For whatever reason, it takes 2 matches in a row for matchmaking to "fix itself" after a winning streak pushes you into opponents you don't stand a chance against.

    Now, I can't really say anything if your streak is longer than 2, but you specifically mentioned 2 battles here.

    We've had a like 70-80% win rate, with those 2 specific battles being pretty much the only ones we usually lose. It'll be like 8 wins in a row, then 2 losses, then 7 wins in a row, then 2 losses, etc.

    I've heard multiple guilds describing the same for their TW experience. Sadly this has been far from my own. It's been a tad better recently, but usually we get outmatched after just 1 victory. If we're lucky we can get a streak of 2, but that's about it. After a max of 2 wins we immediately get an opponent that's way too strong for us and we have absolutely no chance of winning. We're a decent guild for our GP, and we win a good amount of TWs against guilds of equal size, but sadly those matchups rarely happen. It usually goes like: 2 TWs where we're outmatched (one by a bit and one by a lot) and we lose, then 1 TW where it's an equal match, or we're very slightly higher and we win, and then 2 TWs where we're outmatched again. When the opponent is 70M above us there's not much we can do. I seriously have no idea how other guilds are able to win even more than 3 in a row. For my guild it's pretty much impossible because after 2 wins we already get an opponent that's way out of our league that we can't beat. But I guess for some guilds to have a 66%+ win rate there needs to be guilds that have to be sacrificed for the ratio and only get 33%. Ever since the TW changes, we only fought like a handful of guilds below us. Most of the time they are either a bit above us, or way above us. Fun times...
  • Legend91 wrote: »
    Their "solution" is to hand out free wins after you lose 2 TWs in a row. So theoretically you end up with like a 33% win rate. They can't be bothered to come up with a propper TW matchmaking update because of laziness and/or autopilot mode not allowing them to invest resources into anything else than new chars, ships and packs to sell (or creating new things to milk on like datacrons).

    Not for my guild it isn't. Win rate of more like 1 in 5. Even winning 1 gets us elevated to then getting kerb stomped again and again.
  • NicWester
    8827 posts Member
    edited August 6
    The logic is simple—you have won about as many matches and lost as many as your opponent and have similar skill rankings.

    The logic is simple: I was thinking of the wrong game mode. Ignore me.
    Post edited by NicWester on
    Ceterum censeo Patientia esse meliat.
  • TVF
    31990 posts Member
    NicWester wrote: »
    The logic is simple—you have won about as many matches and lost as many as your opponent and have similar skill rankings.

    What's your source on that?

    This isn't a GAC thread.
    The CGDF is no more. Now we hate CG because of the change to the shipment tab. Say hi in our Discord! https://discord.gg/AmStGTH
  • Artumas wrote: »
    What I've noticed, personally, is this.
    You'll have a long streak of decent matchups where you win a ton, then you'll reach a tipping point where your winning streak matches you against someone significantly out of your league, you'll lose, THEN you'll get a SECOND match-up that's out of your league, and lose again, and then you'll get back to normal matchups.

    I don't know why it's like this. But this has consistently been how my guild's matchups have worked since they changed TW matchmaking. For whatever reason, it takes 2 matches in a row for matchmaking to "fix itself" after a winning streak pushes you into opponents you don't stand a chance against.

    Now, I can't really say anything if your streak is longer than 2, but you specifically mentioned 2 battles here.

    We've had a like 70-80% win rate, with those 2 specific battles being pretty much the only ones we usually lose. It'll be like 8 wins in a row, then 2 losses, then 7 wins in a row, then 2 losses, etc.

    I've heard multiple guilds describing the same for their TW experience. Sadly this has been far from my own. It's been a tad better recently, but usually we get outmatched after just 1 victory. If we're lucky we can get a streak of 2, but that's about it. After a max of 2 wins we immediately get an opponent that's way too strong for us and we have absolutely no chance of winning. We're a decent guild for our GP, and we win a good amount of TWs against guilds of equal size, but sadly those matchups rarely happen. It usually goes like: 2 TWs where we're outmatched (one by a bit and one by a lot) and we lose, then 1 TW where it's an equal match, or we're very slightly higher and we win, and then 2 TWs where we're outmatched again. When the opponent is 70M above us there's not much we can do. I seriously have no idea how other guilds are able to win even more than 3 in a row. For my guild it's pretty much impossible because after 2 wins we already get an opponent that's way out of our league that we can't beat. But I guess for some guilds to have a 66%+ win rate there needs to be guilds that have to be sacrificed for the ratio and only get 33%. Ever since the TW changes, we only fought like a handful of guilds below us. Most of the time they are either a bit above us, or way above us. Fun times...

    That's exactly what happens to my guild.
    "You could warn me when I do something bad. Blink once for dark side, twice for light"
  • NicWester wrote: »
    The logic is simple—you have won about as many matches and lost as many as your opponent and have similar skill rankings.

    I'm calling crap on that.
  • NicWester
    8827 posts Member
    edited August 6
    TVF wrote: »
    NicWester wrote: »
    The logic is simple—you have won about as many matches and lost as many as your opponent and have similar skill rankings.

    What's your source on that?

    This isn't a GAC thread.
    AlexanderG wrote: »
    NicWester wrote: »
    The logic is simple—you have won about as many matches and lost as many as your opponent and have similar skill rankings.

    I'm calling crap on that.

    As well you should! I misread the question and thought we were talking GAC. My bad. Sorry about that. Edited my post so new folks to the thread will know it’s wrong, but didn’t delete it so your replies still make sense in context.
    Ceterum censeo Patientia esse meliat.
  • Quick update on our next opponent:
    While the last two were actually way out of reach, by far, (GLs 62 vs 167, JMK 9 vs 39; then GLs 64 vs 148,
    JMK 10 vs 29), now we our new opponent is like GLs 60 vs 91, JMK 8 vs 17, Exe 10 vs 24.
    Actually some of our folks are already cheering because this one sounds actually possible, that's how far this ridiculous MM algorithm actually brought us. Crazy.

    The only thing that's fairly consistent is roughly the same win-loss rate. Everything else seems to be random stuff.

    So my assumption how this algorithm is set up is:

    1) check Win-Loss rate of last couple matches
    2) find guild with roughly same W-L rate
    3) if it's within max 2x GP and 2x GL count (irregardless which GL), it's a match.

    I guess the devs were told to spend as small amount of time on this as possible and move on to other stuff that brings more cash in the long term. Let the poor community managers tell something fancy about highly sophisticated rules, monitoring and improving, then the community will be happy.
  • In my opinion there should never be a 50M GP plus difference in matching. The fact that they allow 70-100M GP gaps is s broken system. It’s not only GLs, it’s mods, omicrons, datacrons, relics, etc. What’s the point in having match ups like that?
  • Sometimes I see people say that their guild was “sandbagged”. I don’t know exactly what they’re implying. I think there was a time when the GP matchmaking based the match ups on active GP so a guild could control the active GP for the MM process. But I thought that ended. Is there still a way for a guild to “sandbag”?
  • NicWester
    8827 posts Member
    Dashrunner wrote: »
    The only thing that's fairly consistent is roughly the same win-loss rate. Everything else seems to be random stuff.

    Wait. Hang on. Does that mean that even though I thought we were talking about an entirely different mode I might have still been right?
    Ceterum censeo Patientia esse meliat.
  • NicWester
    8827 posts Member
    Justokay wrote: »
    Sometimes I see people say that their guild was “sandbagged”. I don’t know exactly what they’re implying. I think there was a time when the GP matchmaking based the match ups on active GP so a guild could control the active GP for the MM process. But I thought that ended. Is there still a way for a guild to “sandbag”?

    If it's how the OP is indicating--that it's recent win/loss record first, then search for a guild with a similar GP--then I think you could still sandbag. The matchmaking algorithm would look at the GP you have sign up for the TW, not the total GP of the guild so sandbagging is still an option.

    But, as stated, that's only if the OP is right, and we can't prove that. They have a solid hypothesis, but there's no way to test it.
    Ceterum censeo Patientia esse meliat.
  • NicWester wrote: »
    Dashrunner wrote: »
    The only thing that's fairly consistent is roughly the same win-loss rate. Everything else seems to be random stuff.

    Wait. Hang on. Does that mean that even though I thought we were talking about an entirely different mode I might have still been right?

    Absolutely
  • NicWester wrote: »
    If it's how the OP is indicating--that it's recent win/loss record first, then search for a guild with a similar GP--then I think you could still sandbag. The matchmaking algorithm would look at the GP you have sign up for the TW, not the total GP of the guild so sandbagging is still an option.

    But, as stated, that's only if the OP is right, and we can't prove that. They have a solid hypothesis, but there's no way to test it.

    Whatever the folks at CG developed, it should be working. Currently it's not. Would be nice to get an explanation how the algorithm works, if not, I would want CG to please improve it – as promised almost a year ago.
  • nottenst
    516 posts Member
    edited August 7
    Dashrunner wrote: »
    Whatever the folks at CG developed, it should be working. Currently it's not. Would be nice to get an explanation how the algorithm works, if not, I would want CG to please improve it – as promised almost a year ago.

    I would like an explanation. It really seems hard to believe that it is working as designed.

    We just went up a guild that was "only" 26M GP above us, but after the match I checked and saw that the active GP was about 50M GP above. They were two tiers higher and were given a gift because they lost two in a row. In terms of their last 8 they had won 5 out of 8, while we had won 4 out of 8. I guess they are only looking at the last two matches? Of course, they get rewards two tiers higher even though they are hitting down.

    Another mystery that I mentioned in another thread is that we have twice been up against guilds that have won at least 8 in a row. I know we were not a higher GP guild in both of those matches, so how were we up against them? Did their algorithm give up or ignore their records completely?
  • Dashrunner
    25 posts Member
    edited August 8
    I have no clue, given up on making any sense out of it.
    We now got 3 times in a row opponents that were way out of our league (see details above). Most people in our guild are really frustrated, including myself.
  • StarSon
    6194 posts Member
    NicWester wrote: »
    Dashrunner wrote: »
    The only thing that's fairly consistent is roughly the same win-loss rate. Everything else seems to be random stuff.

    Wait. Hang on. Does that mean that even though I thought we were talking about an entirely different mode I might have still been right?

    Only kind of. It's not looking at overall win/loss, but it does check streaks. If you lose 2 in a row, your next match will be much easier. At the higher end (350M+) you are almost guaranteed a win on the 3rd match.
  • NicWester
    8827 posts Member
    StarSon wrote: »
    NicWester wrote: »
    Dashrunner wrote: »
    The only thing that's fairly consistent is roughly the same win-loss rate. Everything else seems to be random stuff.

    Wait. Hang on. Does that mean that even though I thought we were talking about an entirely different mode I might have still been right?

    Only kind of. It's not looking at overall win/loss, but it does check streaks. If you lose 2 in a row, your next match will be much easier. At the higher end (350M+) you are almost guaranteed a win on the 3rd match.

    “Only kind of?” 🤔

    I’ll take it!
    Ceterum censeo Patientia esse meliat.
  • Dashrunner wrote: »
    But in all fairness, why is it so difficult to just match two guilds that are roughly even in terms of GP, GLs and Win/Loss streaks?
    I mean, it should be possible to rate based on GP (of those who participate), #GLs, #Exe and #wins in the last 5 wars. Weigh each of these factors, possibly weigh GLs differently and there you have it.
    But maybe I'm seeing this too easy?

    If you count number of GLs and number of execs, then match up even numbers, what is the incentive to unlock GLs or exec?

    Your new unlock will just mean you are matched against a guild who is even with you anyway.

    This would take away an incentive to unlock things, so CG is unlikely to do that.

    Out of curiosity, do you believe that is true?

    If TW matchmaking took GLs into account, would you really stop unlocking them and progressing your roster in a meaningful way?

    I’m not saying I agree with that method, but I am quite sure that I personally would keep chasing new GLs even if it worked that way.
  • At least the TW MM algorithm should take the number of GLs and fleet metas into account. I mean they don't have to do an exact match, but what I saw in the last 3 matches, where the opponent had 2-3 times as many GLs is just ridiculous.
    And for that matter, GLs are not the same. There's a difference between setting JML on def vs. a JMK+CAT lineup. The former can easily be mopped up by a few other GLs, the latter definitely not. The current algorithm certainly doesn't take this into account.
Sign In or Register to comment.