|
Post by amidf on Nov 18, 2017 13:18:16 GMT -5
This is just a quick mock-up of the random walk a win rate would take over 2000 games if wins happen 50% of the time. I used a 50 game sliding window. It is not clear to me how win rate is actually handled, since it doesn't move 2% after every match but is capable of moving quite quickly. I suspect it has some kind of buffer to changing. Edit: this is solved down thread. Note the peaks and troughs. Imagine if you were flipping a coin and got your heads rate up to 70%. Your lucky rabbit's foot must be paying off! Now it drops to 50% and keeps going down near 30%. It's time to sacrifice a chicken! Wins and losses in War Robots are of course not random. But there are many random elements. Who will your teammates be (among the set of "eligible" people near your league)? Who will be your opponents? When you get around a corner, will your shield bot face a plasma bot? Or will it be a death button? As you improve your equipment and hone your skills, you'll climb the leagues. But you'll also face tougher competition. Overlaid on all of this are the random elements. Now, Pix could have an algorithm that punishes you with bad teammates and good opponents when your win rate is high. And they could give you a boost when it is low. But if the same things could happen by random chance, which explanation makes more sense? We need to chart behavior that is very unlikely by chance to start to suspect other unseen factors at play. Can we really say we have done that? The chart below is typical, but if you were to generate it 1000 times, you could get some more extreme winning and losing streaks. That one person with an extreme random chart might come to the forums and declare the game rigged. Doesn't mean that it is. Caveat: I used a spreadsheet program to generate this chart. Pseudorandom numbers should be good emulators, but if someone wants to generate more random data (with clever seeding and sophisticated algorithms), please post the results here? I could do something in R, but it doesn't seem worth it to make the point.
|
|
|
Post by Thunderkiss on Nov 18, 2017 13:40:15 GMT -5
.....It is not clear to me how win rate is actually handled, since it doesn't move 2% after every match but is capable of moving quite quickly. I suspect it has some kind of buffer to changing. ..... If value 50 is a loss, and value 1 is a loss, your win % won't change, and vice versa.
|
|
|
Post by amidf on Nov 18, 2017 13:58:01 GMT -5
.....It is not clear to me how win rate is actually handled, since it doesn't move 2% after every match but is capable of moving quite quickly. I suspect it has some kind of buffer to changing. ..... If value 50 is a loss, and value 1 is a loss, your win % won't change, and vice versa. I don't follow. Can you explain further?
|
|
|
Post by Thunderkiss on Nov 18, 2017 14:10:22 GMT -5
So, we're only keeping track of the last 50 here. If game 50 (oldest game) was a loss, and it gets bumped off the last 50 list by a loss, your win rate won't change, the loss bumps a loss, you're last 50 didn't change.
If your newest game is a win, and your oldest was a loss, you'll gain win %. If your newest game is a loss, and your oldest game was a win, you'll lose win%.
So on and so forth.
Basically the newest game replaces the oldest game, and whatever those 2 values were will inform the change to your win%.
This is why you can go on a win streak, but your win % doesn't change. Your wins were replacing wins at the other end.
|
|
|
Post by zer00eyz on Nov 18, 2017 15:14:36 GMT -5
This is easier to understand with a small chart:
Lets say we track your last 10 games:
WLWLWLWLWL would give you a %50 win rate.
If you win your next game it would look like:
WWLWLWLWLW and you would have a %60 win rate.
Your win rate will only go UP or DOWN if you replace a different result in the stack of 50 games
|
|
|
Post by amidf on Nov 18, 2017 21:22:18 GMT -5
Thanks. I see what you're saying. Half the time, on average, your win rate will change. But you won't know if you're repeating the pattern from 50 games ago if it doesn't.
All the win rates I can remember from recent inspections are even. I guess I'll pay more attention to it. I bet someone on here has recorded the outcome of their last 50+ and could tell us going forward if their win rate matches up. I could track it but the payoff isn't until 50 games from now and I don't have that level of commitment.
|
|
ma'ElKoth
GI. Patton
Posts: 138
Karma: 87
Pilot name: ma'ElKoth
Platform: Android
Clan: [F₂P] Ø Ca$h = Ø Daϟh
League: Expert
Server Region: Europe
|
Post by ma'ElKoth on Nov 19, 2017 3:28:43 GMT -5
Tracked 200 matches and the calculation is exactly as stated and explained above. Always the last 50 matches are considered. If rate changes, it changes by 2%. As for overall change in win rate.... I started tracking at 42%. After 25 matches I had reached 50%. Next 25 matches brought me up to 60%. It kept fluctuating between about 54% - 64% for the next 120 matches then. The remaining ~30 matches got it down to 44%-46% area. There is another thread What Is Your Victories Percentage? concerning the topic as well.
|
|
|
Post by amidf on Nov 19, 2017 9:35:47 GMT -5
Tracked 200 matches and the calculation is exactly as stated and explained above. Always the last 50 matches are considered. If rate changes, it changes by 2%. As for overall change in win rate.... I started tracking at 42%. After 25 matches I had reached 50%. Next 25 matches brought me up to 60%. It kept fluctuating between about 54% - 64% for the next 120 matches then. The remaining ~30 matches got it down to 44%-46% area. There is another thread What Is Your Victories Percentage? concerning the topic as well. Thanks for the confirmation. My win rate has at times remained very stable. At other times, it has soared or dipped. This could be due to Pixonic having a complicated "correction" mechanism to force win rates to 50%. It could also be a random walk around 50% due to the MM using only League Points as described. So which should we believe?
|
|
|
Post by Thunderkiss on Nov 19, 2017 10:23:00 GMT -5
Tracked 200 matches and the calculation is exactly as stated and explained above. Always the last 50 matches are considered. If rate changes, it changes by 2%. As for overall change in win rate.... I started tracking at 42%. After 25 matches I had reached 50%. Next 25 matches brought me up to 60%. It kept fluctuating between about 54% - 64% for the next 120 matches then. The remaining ~30 matches got it down to 44%-46% area. There is another thread What Is Your Victories Percentage? concerning the topic as well. Thanks for the confirmation. My win rate has at times remained very stable. At other times, it has soared or dipped. This could be due to Pixonic having a complicated "correction" mechanism to force win rates to 50%. It could also be a random walk around 50% due to the MM using only League Points as described. So which should we believe? Personally I don't believe in the forced 50%. I just thin k there's an abundance of players who think more of themselves than is due. As such, I attribute losing streaks and such to bad runs of less than skilled players. My win rate has only gone under 70% since the event, and prior I was never under 80%, reaching as high as 92%. Luck of the draw on randos which is why I don't do randos anymore, more often than not they just suck. The difference? Games with randos.
|
|
|
Post by amidf on Nov 19, 2017 10:34:53 GMT -5
Personally I don't believe in the forced 50%. I just thin k there's an abundance of players who think more of themselves than is due. As such, I attribute losing streaks and such to bad runs of less than skilled players. My win rate has only gone under 70% since the event, and prior I was never under 80%, reaching as high as 92%. Luck of the draw on randos which is why I don't do randos anymore, more often than not they just suck. The difference? Games with randos. Sounds like that is the biggest factor.
|
|
|
Post by mechtout on Nov 19, 2017 10:50:17 GMT -5
For solo, it seems you get the weaker randoms of the match pool on your squad if you have a higher win percentage.
|
|
|
Post by Thunderkiss on Nov 19, 2017 11:09:00 GMT -5
For solo, it seems you get the weaker randoms of the match pool on your squad if you have a higher win percentage. My problem with that is the "seems like" part. I don't think there's code to do that kind of "thinking" in a timely fashion. No, sadly i think that most players really just are 50% or less. Look at how many see canyon or yam and get all excited and drop the worst first bot imaginable (long range) and then do the worst thing possible; stay in it all game. I'm sure they think they're doing a helluva stroke of business and blame the rest of the team for the loss, when in reality, the fault lies squarely with them. And there's THOUSANDS of people with this mindset. You see em all the time. Canyon? Yay treb time! Ummmm no. But this is the players base and mindset we're stuck with when we go with randos. They never fail to disappoint. Many times my squad will be down one and I'll joke "let's roll 5, how bad can the randos be?" And then they show me.
|
|
|
Post by ᎶƦ℮℮ƊᎽ ƤΛƝƊΛ on Nov 19, 2017 11:17:48 GMT -5
I have never believed in the forced 50%. The fact that most people tend to gravitate towards 50 (and even lower) is the way the league system is set up. It keeps moving you up until you can compete. So you can go through winning streak, but you will eventually start losing. It has nothing to do with win rate though, only your league placement.
|
|
|
Post by amidf on Nov 19, 2017 11:58:36 GMT -5
For solo, it seems you get the weaker randoms of the match pool on your squad if you have a higher win percentage. I agree with Thunderkiss and ᎶƦ℮℮ƊᎽ ƤΛƝƊΛ on this one. The whole point of this thread is that the behavior is indistinguishable from random chance.
|
|
|
Post by ΒΣRΖΣRKΛ²³ on Nov 19, 2017 13:03:39 GMT -5
So i thought about it, and this is My conclusion
It’s not a coincidence. I don’t believe in coincides. Pixonic actively enforces a 50% win percentage. In the background, they have tons of options to work your win percentage.
if I have 12 players to select 2 teams, I can put together 1 team that is likely to win. Now I don’t know how exactly this matchmaking works, But I am certain that your league points are not the only stat that matters.
Now if you squad all the time, you shouldn’t really talk about this subject, because this obviously only works with random solos.
The difference between a 30% and a 70% win percentage is way too noticeable...to be a coincidence. i dropped from expert to diamond prior to the event. If you reach 30% it is really tough to loose. In team deathmatch though it was easy - i ejected all my bots so my team started 0-5 every time. Tough hole to dig out of. 20% of the time my team still won...
When you win all the time, all of a sudden you might lose although you scored 1.3 mill and killed 10 robots. It’s like your whole team is made of paper ..
I admit though that I don’t have actual data to back my claim up.. Im sure though.
|
|
|
Post by Thunderkiss on Nov 19, 2017 13:14:15 GMT -5
So i thought about it, and this is My conclusion It’s not a coincidence. I don’t believe in coincides. Pixonic actively enforces a 50% win percentage. In the background, they have tons of options to work your win percentage. if I have 12 players to select 2 teams, I can put together 1 team that is likely to win. Now I don’t know how exactly this matchmaking works, But I am certain that your league points are not the only stat that matters. Now if you squad all the time, you shouldn’t really talk about this subject, because this obviously only works with random solos. The difference between a 30% and a 70% win percentage is way too noticeable...to be a coincidence. i dropped from expert to diamond prior to the event. If you reach 30% it is really tough to loose. In team deathmatch though it was easy - i ejected all my bots so my team started 0-5 every time. Tough hole to dig out of. 20% of the time my team still won... When you win all the time, all of a sudden you might lose although you scored 1.3 mill and killed 10 robots. It’s like your whole team is made of paper .. I admit though that I don’t have actual data to back my claim up.. Im sure though. Matchmaker can't adjust for player skill and experience.
|
|
|
Post by amidf on Nov 19, 2017 13:19:33 GMT -5
I hear you, ΒΣRΖΣRKΛ²³. But if you think there is no such thing as a coincidence, we have very different world views and are going to have a tough time seeing eye to eye. If I roll two dice and they both come out to the same number (e.g. two 5s), is that a coincidence? What if it's on my niece's birthday, and she just turned 5? Does it matter how many times I roll the dice that day? On January 13, 2016, the largest ever US lottery jackpot of $1.6B was split by 3 winners from California, Florida, and Tennessee. They didn't know each other, but they all picked the same numbers: 4, 8, 19, 27, 34, and 10 (powerball). Was it a coincidence that they all chose these numbers (or had a machine choose them)?
|
|
|
Post by Ron Gaul on Nov 19, 2017 13:57:59 GMT -5
What I used to see was this:
Against equally leveled or slightly superior gear, I win 70% of the time. Or more. Back before the new MM was introduced, I had a win rate upwards of 75% and climbing. My gear was L8-9 with a few 10's mixed in. Routinely faced enemies with gear L7-11. Could hold my own and then some. Had the occasional match against L12 squads, which would result in generally brutal defeats, but it was the exception and not the rule.
New MM and Leagues system introduced, and my win rate suddenly plummets. Why? Not because I got a lot worse, or the opposition got a lot better. Because the opposition became much more highly leveled. I went from facing L7-11 enemies, stuff in my league, to facing maxed hangars. When I lost bad, I'd face easier enemies at or below my own gear levels. So I'd start winning again. But, two or three matches later, the MM would suddenly start pitting me against those same maxed hangars I'd been losing to.
So, was my skill level fluctuating? Was I playing worse against some enemies and better against others?
Against gear equal to or slightly better than my own, I routinely finished in the top two or three. Even on losses.
Against maxed gear, it didn't matter what I did - their gear was at a minimum 24% stronger than mine, and usually much more than that - I'd regularly finish in the bottom two or three on either team. I had only two L10 items at this point; the rest were 7's, 8's, and 9's.
Conclusion: I was "only" good enough to kick the snot out of stuff that was equal to my own, but not good enough to stand a fighting chance against equipment vastly superior to mine.
Bear in mind that my own teams fluctuated at the same time. I'd get better teammates and win after a bunch of losses, and I'd get worse teammates and lose after a number of wins.
Did I ever track any of this? A couple times, but I got bored and gave it up. Because the pattern was too familiar. Win some games, suddenly run into a stone wall. Get buried under that stone wall, and suddenly it's lifted and I can play as normal. "Normal" being matched with enemies who do NOT possess gear that I cannot overcome with any amount of skill.
Coincidence? Can't prove it isn't without seeing the actual MM algorithm...but it certainly didn't look like it.
If it walks like a duck and quacks like a duck, chances are it's a duck.
|
|
|
Post by amidf on Nov 19, 2017 14:24:07 GMT -5
If you look at the graph I posted, you can see multiple instances of beat downs and usually at some point these are followed by win streaks. Coincidence?
Do you suppose that relates to Excel's rand() algorithm? If we were to toss coins instead (boring), I bet we could come up with some explanations regarding juju or mojo.
I'm not about to start tracking thousands of coins tosses because I can see that some people aren't easily convinced.
|
|
|
Post by zer00eyz on Nov 19, 2017 14:42:51 GMT -5
If you look at the graph I posted, you can see multiple instances of beat downs and usually at some point these are followed by win streaks. Coincidence? Do you suppose that relates to Excel's rand() algorithm? If we were to toss coins instead (boring), I bet we could come up with some explanations regarding juju or mojo. I'm not about to start tracking thousands of coins tosses because I can see that some people aren't easily convinced. I'll be frank, the MM is really Match Manipulator. It is using your league score (and I suspect it's velocity) as well as how you finish (1-6) to determine how you get mixed in with opponents. Stop chasing damage for a few games and go for beacons or run a capping hanger. It will frequently make you the strongest player league score wise in your matches. MM doesn't factor in your win rate, it is a byproduct of the system not a stat it uses.
|
|
|
Post by ΒΣRΖΣRKΛ²³ on Nov 19, 2017 14:57:58 GMT -5
You don’t need to:
Every time you flip a coin, you have a 50/50 chance
But I don’t think you can compare this or lottery to war robots.
In war robots I ASSUME „skill“ plays a role. This is why I believe not everybody can have a 50% win percentage.
If i assume wrong and skill doesn’t play a role in the outcome of games: well ok I guess in this scenario the 50% could be a result of statistics. It would shift almost all parameters that determine the outcome into pixonics hands though..even easier to manipulate.
if 50% is not enforced, and skill plays a role, everybody is equally skilled. But if everybody is not equally skilled, skill plays a role , and everybody is at 50%: it is enforced.
|
|
|
Post by moses on Nov 19, 2017 15:18:08 GMT -5
For solo, it seems you get the weaker randoms of the match pool on your squad if you have a higher win percentage. I agree with Thunderkiss and ᎶƦ℮℮ƊᎽ ƤΛƝƊΛ on this one. The whole point of this thread is that the behavior is indistinguishable from random chance. Could not agree more.
|
|
|
Post by amidf on Nov 19, 2017 15:35:32 GMT -5
I didn't say that the outcome of every match is the result of chance alone. No, obviously how you play, which is a product of your skill, is a factor. However, that factor and your hangar strength are compensated over time by your league placement. Who you are teamed with (when not in a full squad) and against is a major component of the outcome and is influenced by chance constrained somewhat by league position.
My point is that if it were ONLY chance, that doesn't preclude winning and losing streaks. It doesn't preclude 30% win rates and 70% win rates. So if the chance component by itself can cause these fluctuations, do we need to invoke complicated hidden algorithms at play? Sure, Pixonic could do this, and they aren't particularly trustworthy. But if they say they are just matching on league points and matching on league points can reasonably explain the behavior, that's good enough for me. The rest is just confirmation bias and regression to the mean.
I've already explained how league point matching can produce consistent win rates > 50% to support and beacon capping players and < 50% for strict damage dealers. At the top of the leagues, you get exceptions, but the middle chunk will find their place and be near 50% depending on their average damage placement.
|
|
|
Post by moses on Nov 19, 2017 15:54:07 GMT -5
I didn't say that the outcome of every match is the result of chance alone. No, obviously how you play, which is a product of your skill, is a factor. However, that factor and your hangar strength are compensated over time by your league placement. Who you are teamed with (when not in a full squad) and against is a major component of the outcome and is influenced by chance constrained somewhat by league position. My point is that if it were ONLY chance, that doesn't preclude winning and losing streaks. It doesn't preclude 30% win rates and 70% win rates. So if the chance component by itself can cause these fluctuations, do we need to invoke complicated hidden algorithms at play? Sure, Pixonic could do this, and they aren't particularly trustworthy. But if they say they are just matching on league points and matching on league points can reasonably explain the behavior, that's good enough for me. The rest is just confirmation bias and regression to the mean. I've already explained how league point matching can produce consistent win rates > 50% to support and beacon capping players and < 50% for strict damage dealers. At the top of the leagues, you get exceptions, but the middle chunk will find their place and be near 50% depending on their average damage placement. Yes, this is exactly it. No guarantees that Pixo is not up to something, but the logic supports what they state and there is no factual evidence to the contrary that would stand up to even the most basic statistical analysis so not worth wasting any more time over.
|
|
|
Post by Pulse Hadron on Nov 19, 2017 16:09:20 GMT -5
I ran a beacon capping hangar of mostly lights, and after my league score stabilized my win rate stayed hovering around 70% for the 2 weeks I played that hangar.
However, I’m not convinced but I think it’s possible a win rate >90% has special code to basically bump you up a league.
|
|
|
Post by zer00eyz on Nov 19, 2017 16:43:13 GMT -5
I ran a beacon capping hangar of mostly lights, and after my league score stabilized my win rate stayed hovering around 70% for the 2 weeks I played that hangar. However, I’m not convinced but I think it’s possible a win rate >90% has special code to basically bump you up a league. %70 is "league score stable" if your going to finish last. Your experience is my experience -- capping hangers will let you fall down till you have a very high win rate. Our sorting system leans into damage so heavily that this is a fairly predictable outcome.
|
|
|
Post by Replicant on Nov 19, 2017 17:09:22 GMT -5
If you look at the graph I posted, you can see multiple instances of beat downs and usually at some point these are followed by win streaks. Coincidence? Do you suppose that relates to Excel's rand() algorithm? If we were to toss coins instead (boring), I bet we could come up with some explanations regarding juju or mojo. I'm not about to start tracking thousands of coins tosses because I can see that some people aren't easily convinced. I'll be frank, the MM is really Match Manipulator. It is using your league score (and I suspect it's velocity) as well as how you finish (1-6) to determine how you get mixed in with opponents. Stop chasing damage for a few games and go for beacons or run a capping hanger. It will frequently make you the strongest player league score wise in your matches. MM doesn't factor in your win rate, it is a byproduct of the system not a stat it uses. The big-picture experience with MM is largely explainable by randomness plus the inflation factor in the way league points are handed out, but short term runs (e.g. switching to a beacon capper hanger) result in some weird MM behavior that is better explained by a velocity of damage-placement metric than anything else I can think of. The only thing I liked about the old hanger-based MM was that we knew how it worked.
|
|
|
Post by cmdrperalta on Nov 19, 2017 17:16:54 GMT -5
I find that as my winning percentage increases to 70% or so, playing best hangar I can with no focus on light robot beacon grabbing other than what I normally do with 1 beacon runner, that strength of opponents also increases and that often initial move to next league results in brief period of much stronger competition than you would expect in that league. League and/or league position may be factors but so are level of hangar, current performance of player, etc. I think they have variety factors that influence matches they place you in that make it hard to fully quantify w/out code b/c game is also dependent on who happens to be on when you are playing.
|
|