36 replies to this topic

[Flyby215]

Feral Clown, on 29 July 2019 - 04:43 PM, said:

Sure...but what's the point in going for improvement on one metric? What does getting a higher matchscore especially at the expense of the other stuff?

AMS isn't even particularly helpful. I say that and you don't recognise this name, but I play a lot when you do and most often running around in a dual AMS crabbie this month.

The only reason I do this is it helps me nsr Veagles, as there isn't enough AMS in the world to defeat Supernova's or Novacats.

Point taken.

Perusing through my mech-stats page, I noticed that my highest K/D and W/L mechs don't field AMS (with one obvious exception). It's actually rather interesting to me; I can't imagine why 1 ton of AMS/ammo would hurt a mech in any substantial way, while at the same time offer nothing in return considering how many ATMs are fielded now-a-days.

[SlippnGriff]

Damn, that must mean i can farm with lrms and get those numbers im statistically good

[Feral Clown]

Flyby215, on 30 July 2019 - 10:04 AM, said:

Point taken.

Perusing through my mech-stats page, I noticed that my highest K/D and W/L mechs don't field AMS (with one obvious exception). It's actually rather interesting to me; I can't imagine why 1 ton of AMS/ammo would hurt a mech in any substantial way, while at the same time offer nothing in return considering how many ATMs are fielded now-a-days.

Dude I find stuff like this interesting too. I am also an alt account nerd and used to be heavily invested in CW. So much so that there have been months and months where I have not done any quick play matches at all. This has given me isolated stats and helps me assess things better.

Why I mention CW? Because I know from experience having faced a team where every guy is running at least one AMS, this is ineffective against LRM's but will shut down two or three guys running ATM's.

So currently I am playing the CRB 27 with 5mpl and 2 AMS with one ton of ammo. Reason I am doing this is because it is fits a bunch of things for me. One being able to run 5mpl with 16 dbl hs. It's overall durability and being nice and fast, plus the AMS being pretty effective against the current plague of Veagles. Next season I am going to use the Vulcan, which is far less durable but has JJ's and is faster but still runs 5mpl and 16 dbl hs. I am betting that I will get a similar if not better avg. ms in the Vulcan despite not having AMS bonus points just because it is a better mech overall in my, and probably most people's opinion. Crabbies stats so far are 1.57 WLR, 2.4 KDR, and avg. ms 373.

Other thing I have done in the past was mess around with my version of the triple AMS Nova. What was weird with that one though, and not sure if it was just a string of bad luck, but the WLR on it is bad compared to my other mechs. The worst in fact being sub 1. Average damage, KDR and of coarse avg ms is all great, but it doesn't seem to help me get wins like other mechs I play do. So far it's a strange outlier I plan to revisit soon.

SlippnGriff, on 30 July 2019 - 11:37 AM, said:

Damn, that must mean i can farm with lrms and get those numbers im statistically good

Shut up and join ISMO already.

Jay won't let you do the lurms though.

Edited by Feral Clown, 30 July 2019 - 03:00 PM.

[ShaneoftheDead]

Nightbird, on 28 July 2019 - 11:22 AM, said:

lol more newbies that think match score actually corresponds to performance

your W/L is your performance, nothing else. Average is 0.9-1.1, good >1.5, excellent >2

The problem with W/L is that any given match outcome is largely determined by the 23 other people in the match. It is a weak correlation at best. W/L is more of a statistic about who you drop with than how good you are. Modern sports analysis has not taken "Wins" into account when judging players for decades now.

[Nightbird]

ShaneoftheDead, on 31 July 2019 - 08:32 AM, said:

The problem with W/L is that any given match outcome is largely determined by the 23 other people in the match. It is a weak correlation at best. W/L is more of a statistic about who you drop with than how good you are. Modern sports analysis has not taken "Wins" into account when judging players for decades now.

wrong, proof
https://mwomercs.com...thread-we-need/

in modern sports, teams are fixed for a season, apples and oranges

[thievingmagpi]

ShaneoftheDead, on 31 July 2019 - 08:32 AM, said:

The problem with W/L is that any given match outcome is largely determined by the 23 other people in the match. It is a weak correlation at best. W/L is more of a statistic about who you drop with than how good you are. Modern sports analysis has not taken "Wins" into account when judging players for decades now.

good players win more than bad players.

sure, sometimes you do 1.8 damage and get 8 kills and your team still loses. but that level of performance will guarantee you get more wins than losses, by a considerable margin.

if modern sports were a random assortment of players stuck together without any kind of sorting, then yes, wins would be counted.

[Kubernetes]

ShaneoftheDead, on 31 July 2019 - 08:32 AM, said:

The problem with W/L is that any given match outcome is largely determined by the 23 other people in the match. It is a weak correlation at best. W/L is more of a statistic about who you drop with than how good you are. Modern sports analysis has not taken "Wins" into account when judging players for decades now.

Wtf are you talking about? WAR? Also, when the sample size gets big enough, W/L becomes all about you.

[Prototelis]

Theres a reason why some of the best players in the game have high WLR.

Even if they separated GQ and SQ stats the strongest players are still going to have the highest WLR.

[ShaneoftheDead]

Nightbird, on 31 July 2019 - 08:50 AM, said:

wrong, proof
https://mwomercs.com...thread-we-need/

in modern sports, teams are fixed for a season, apples and oranges

Does not take into account bad players with good teammates that carry them. See it every day in Quick Play. WLR is boosted based upon the relative quality of your teammates vs your opponents. Talk your way into a group of good players and your WLR is going up and it has nothing to do with your 'Mech or piloting skills. Contrariwise, a good player or group of players can lose games when they get teammates that TK or Disconnect or do large amounts of team damage despite the fact that they played well. Or even they play well, but lost because it was conquest on Polar and did not Cap at all. WLR is a quick and dirty approximation, but it is not something you want to use as the basis of your Match Making engine. I would not use it to compare one 'Mech to another. It's not good enough, in my opinion. If it is good enough for you, then I am OK with that.

thievingmagpi, on 31 July 2019 - 08:53 AM, said:

good players win more than bad players.

sure, sometimes you do 1.8 damage and get 8 kills and your team still loses. but that level of performance will guarantee you get more wins than losses, by a considerable margin.

if modern sports were a random assortment of players stuck together without any kind of sorting, then yes, wins would be counted.

See above. Remember, the outcome of any given match is largely determined, 95.8% (23/24), by the other players of the match. 95% of your Win was based upon what other players did or didn't do. Yes, good players can influence more than 5%. But even if you inflate to 33%, reflecting a 4 kill game, that still leaves everyone else 67%. Again, the match is more in their hands than yours. You are better off using Match Score to rate individual performance.

I do agree that good groups win more than bad groups. Once you have a lance or more of good players working together on a voice chat then their chances for a Win go up. And that is how we usually see them, we see good players in a group with other good players or perhaps even sync dropping. While I don't have stats to back that up, I think we can agree that helps increase Win probability.

Kubernetes, on 31 July 2019 - 10:27 AM, said:

Wtf are you talking about? WAR? Also, when the sample size gets big enough, W/L becomes all about you.

Well, Wins Above Replacement is a counting stat. I would not base a MatchMaker off of WAR or any other counting stat. I have written other posts on this topic before. I would prefer a Battle Value calculation based upon the the Player, Mech, Loadout, Map, and Game Mode. Probably closer to how xFIP is calculated. Try to strip out other players influence, luck, and randomness of performance, etc. to get to a number that can get a close approximation of how you expect that player to do in this mode on this map with that 'Mech with that loadout. That would be my goal.

As for sample size, increasing the sample sizes will not change the underlying ratios of influence to match outcome. It "may" normalize with very large sample sizes, but relying on randomness for accuracy...I would not use that for a system. Also, if a player groups with good players all the time then a larger sample size will only skew his WLR even further. The fundamental flaws in using WLR still remain irregardless of sample size.

[Nightbird]

ShaneoftheDead, on 05 August 2019 - 11:42 AM, said:

Does not take into account bad players with good teammates that carry them. See it every day in Quick Play. WLR is boosted based upon the relative quality of your teammates vs your opponents. Talk your way into a group of good players and your WLR is going up and it has nothing to do with your 'Mech or piloting skills. Contrariwise, a good player or group of players can lose games when they get teammates that TK or Disconnect or do large amounts of team damage despite the fact that they played well. Or even they play well, but lost because it was conquest on Polar and did not Cap at all. WLR is a quick and dirty approximation, but it is not something you want to use as the basis of your Match Making engine. I would not use it to compare one 'Mech to another. It's not good enough, in my opinion. If it is good enough for you, then I am OK with that.



See above. Remember, the outcome of any given match is largely determined, 95.8% (23/24), by the other players of the match. 95% of your Win was based upon what other players did or didn't do. Yes, good players can influence more than 5%. But even if you inflate to 33%, reflecting a 4 kill game, that still leaves everyone else 67%. Again, the match is more in their hands than yours. You are better off using Match Score to rate individual performance.

I do agree that good groups win more than bad groups. Once you have a lance or more of good players working together on a voice chat then their chances for a Win go up. And that is how we usually see them, we see good players in a group with other good players or perhaps even sync dropping. While I don't have stats to back that up, I think we can agree that helps increase Win probability.



Well, Wins Above Replacement is a counting stat. I would not base a MatchMaker off of WAR or any other counting stat. I have written other posts on this topic before. I would prefer a Battle Value calculation based upon the the Player, Mech, Loadout, Map, and Game Mode. Probably closer to how xFIP is calculated. Try to strip out other players influence, luck, and randomness of performance, etc. to get to a number that can get a close approximation of how you expect that player to do in this mode on this map with that 'Mech with that loadout. That would be my goal.

As for sample size, increasing the sample sizes will not change the underlying ratios of influence to match outcome. It "may" normalize with very large sample sizes, but relying on randomness for accuracy...I would not use that for a system. Also, if a player groups with good players all the time then a larger sample size will only skew his WLR even further. The fundamental flaws in using WLR still remain irregardless of sample size.

Wrong, we're talking about solo queue. All your team's players and the other team's players will be average on average. If you're above average, your team will win more than 50% of the time, vice versa if you're below average.

This is because you're the only constant in all your matches

[Kubernetes]

ShaneoftheDead, on 05 August 2019 - 11:42 AM, said:

As for sample size, increasing the sample sizes will not change the underlying ratios of influence to match outcome. It "may" normalize with very large sample sizes, but relying on randomness for accuracy...I would not use that for a system. Also, if a player groups with good players all the time then a larger sample size will only skew his WLR even further. The fundamental flaws in using WLR still remain irregardless of sample size.

This is some fascinating illogic. You think a player's WLR is not meaningful because it's dependent on the other 23 players. What determines the WLR of the other 23 players? If Player A and Player B both play 1000 matches apiece, what are the chances that Player A consistently gets better teammates than Player B? If team composition is mostly random, what are the chances that the MM will, consistently over 1000 matches, give Player A a better team than Player B? Say Player A has a 2.5 WLR and Player B has a.0.5 WLR over those 1000 matches-- what is the likelihood that this discrepancy can be explained by the MM? Think about it and then slap yourself in the face for being dumb.

Edited by Kubernetes, 05 August 2019 - 12:48 PM.

[RickySpanish]

Kubernetes, on 05 August 2019 - 12:46 PM, said:

This is some fascinating illogic. You think a player's WLR is not meaningful because it's dependent on the other 23 players. What determines the WLR of the other 23 players? If Player A and Player B both play 1000 matches apiece, what are the chances that Player A consistently gets better teammates than Player B? If team composition is mostly random, what are the chances that the MM will, consistently over 1000 matches, give Player A a better team than Player B? Say Player A has a 2.5 WLR and Player B has a.0.5 WLR over those 1000 matches-- what is the likelihood that this discrepancy can be explained by the MM? Think about it and then slap yourself in the face for being dumb.

I believe the argument is something along the lines of the MM being so bad at balancing teams, that a match's outcome is largely determined by luck of the draw, to the point where most players are not skillfull enough to really affect that outcome, since most players who are in T1 should not be.

IMO you will see the effect you have on match outcomes eventually.

Edited by RickySpanish, 05 August 2019 - 03:49 PM.

[Swamp Butt]

I'm not the best around, and some games just complerely ****. However, the games that I achieve 600 match score, on any weight class, this is what I typicaly do. 1: used 2 arties, 2: used both coolshots as needed. 3: hit targets that are currently being attcked as I am not usually piloting anything thay can take a hit too well. Mediums and lights mostly. Hit the damn r button right after the currebt target is dead, and go after it. I am also normally limited to typical cspl's, cerul's, and/or srms. It's my preffered play style.

If I were to run AMS, it's for the sole purpose is to nerf the lurm boats, nothing more

Active probes to counter ems, ems set to counter, and hitting r, can actyally grt you match score.

Using a tag, or narc especially when you team is heavy on lurms can get a decent match score even if damage in neligible.

But, ulimatly damage output regardless of how, Kills, KMDD's, solo kills, are the main sorce of match score.

[Pixel Hunter]

Swamp *** MkII, on 07 October 2024 - 07:42 AM, said:

Active probes to counter ems, ems set to counter, and hitting r, can actyally grt you match score.

ECM prevalence is one of the biggest sins of this game. It still needs a nerf. I'd love to see an event with it only blocks the paper doll and see what happens. given how common it is now it needs looked at

[Tarl Cabot]

Swamp *** MkII, on 07 October 2024 - 07:42 AM, said:

I'm not the best around, and some games just complerely ****. However, the games that I achieve 600 match score, on any weight class, this is what I typicaly do. 1: used 2 arties, 2: used both coolshots as needed. 3: hit targets that are currently being attcked as I am not usually piloting anything thay can take a hit too well. Mediums and lights mostly. Hit the damn r button right after the currebt target is dead, and go after it. I am also normally limited to typical cspl's, cerul's, and/or srms. It's my preffered play style.

If I were to run AMS, it's for the sole purpose is to nerf the lurm boats, nothing more

Active probes to counter ems, ems set to counter, and hitting r, can actyally grt you match score.

Using a tag, or narc especially when you team is heavy on lurms can get a decent match score even if damage in neligible.

But, ultimately damage output regardless of how, Kills, KMDD's, solo kills, are the main sorce of match score.

Hmm, this thread is about 5 years old, prior to the PSR Update/Tier Reset and the Soup Queue, though your post would fit in many of the most recent threads. The original PSR/Tier was more of an experience bar, and lots of players who were in Tier 1/2 are now currently in Tier 3/4.

[Void Angel]

You found an ancient topic,
that's truly obsolete!
But you just had to comment,
rather than be discreet!

Necro, necro, necro!
What in the world is wrong?
Necro, necro, necro!
Why must I sing the song?

[Swamp Butt]

Tarl Cabot, on 07 October 2024 - 04:30 PM, said:

Hmm, this thread is about 5 years old, prior to the PSR Update/Tier Reset and the Soup Queue, though your post would fit in many of the most recent threads. The original PSR/Tier was more of an experience bar, and lots of players who were in Tier 1/2 are now currently in Tier 3/4.

Oops, I guess that's the price of browsing goofy topics on your phone while waiting to be loaded. Yeah, google and topics aren't really "date friendly"!