97 replies to this topic

[Rampancy]

Mercules, on 08 June 2015 - 02:28 PM, said:

People who sit in cover and hide lose. I don't suicide rush but if I notice 3 mechs are separated from the enemy team and can bait them while guiding my team into them I'll do it, even at the cost of my mech.

Knowing when and where to start tanking damage is a learned skill that will absolutely help your team win.

Tanking damage alone, however, is not.

[Mercules]

RampancyTW, on 08 June 2015 - 02:42 PM, said:

Knowing when and where to start tanking damage is a learned skill that will absolutely help your team win.

Tanking damage alone, however, is not.

Yep, and in the PuG queue when you should tank seems to be... "Never because no one is aware enough to take advantage of it, just use everyone else on the team as fodder." Which brings us back around to, "Stats don't actually measure anything because a teammate can basically use his team as fodder while he pads his stats other than W/L." Again, W/L means very little as well because you can end up with, "Beware my 14 Flamer Nova! I call it the SuperNova because it lights up and melts."

[Rampancy]

Mercules, on 08 June 2015 - 02:48 PM, said:

Again, W/L means very little as well because you can end up with, "Beware my 14 Flamer Nova! I call it the SuperNova because it lights up and melts."

Oh? And how do you end up with that?

Over a large sample size, the odds of a terrible player being carried to a positive W/L are infinitesimal. The inverse is also true.

[Mercules]

RampancyTW, on 08 June 2015 - 03:17 PM, said:

Oh? And how do you end up with that?

Over a large sample size, the odds of a terrible player being carried to a positive W/L are infinitesimal. The inverse is also true.

Yep, but given the matchmaker the odds of such a player being in my game is not infinitesimal. If said player is on an off night where he is not playing up to his Elo but screwing around it becomes even more likely.

Let's not forget that "I've never played this game before I hope this <insert current worst trial mech> is good!" players can end up in the same match with high Elo players.

[Rampancy]

Mercules, on 08 June 2015 - 03:22 PM, said:

Yep, but given the matchmaker the odds of such a player being in my game is not infinitesimal. If said player is on an off night where he is not playing up to his Elo but screwing around it becomes even more likely.

Let's not forget that "I've never played this game before I hope this <insert current worst trial mech> is good!" players can end up in the same match with high Elo players.

...I really don't see how that's relevant. Over a good sample size (say 100 games? The more the better) your relative contribution to your team will be reflected in your W/L. If you're a low Elo player and your perform better than the other team's low Elo players, your team will have a better chance of winning.

[Mercules]

RampancyTW, on 08 June 2015 - 03:27 PM, said:

Over a good sample size (say 100 games? The more the better) your relative contribution to your team will be reflected in your W/L.

You are missing another crucial factor. IF the Matchmaker was designed to give you 50% wins and 50% losses then yeah, over time it would probably average out. However the matchmaker gives F-all about what your W/L is, it just tries to get something close to similar Elo balance, calculates who the Winner and Loser of the match should be, then adjusts Elo based on what actually happened compared to the prediction.

The point you are missing is that the Matchmaker could actually set you up with the odds stacked so that you are determined to lose every single one of those 100 matches. So then W/L isn't actually measuring crap because your wins would only come from exceptional play by someone on your team. The opposite could be true also, you could be slated to win every single one of those 100 matches and only lose if your team messed up.

[Rampancy]

Mercules, on 08 June 2015 - 03:32 PM, said:

You are missing another crucial factor. IF the Matchmaker was designed to give you 50% wins and 50% losses then yeah, over time it would probably average out. However the matchmaker gives F-all about what your W/L is, it just tries to get something close to similar Elo balance, calculates who the Winner and Loser of the match should be, then adjusts Elo based on what actually happened compared to the prediction.

The point you are missing is that the Matchmaker could actually set you up with the odds stacked so that you are determined to lose every single one of those 100 matches. So then W/L isn't actually measuring crap because your wins would only come from exceptional play by someone on your team. The opposite could be true also, you could be slated to win every single one of those 100 matches and only lose if your team messed up.

No, you're missing it. Because the MM isn't designed to give you 50% wins and 50% losses, your relative ability is reflected in your W/L.

Again, the odds of that happening are infinitesimally low. Which is why over time, despite poor matchmaker performance in specific matches, over time your W/L will reflect your ability to contribute to team victories.

[Mercules]

RampancyTW, on 08 June 2015 - 03:35 PM, said:

No, you're missing it. Because the MM isn't designed to give you 50% wins and 50% losses, your relative ability is reflected in your W/L.

Again, the odds of that happening are infinitesimally low. Which is why over time, despite poor matchmaker performance in specific matches, over time your W/L will reflect your ability to contribute to team victories.

Match maker assigns you 100 matches you are supposed to lose. You play well in every match and win 2 of them. How has your true skill been reflected in a 2/98 W/L record? It hasn't.

You are correct, over thousands of players playing hundreds of thousands of matches the mean average of all those players and matches would even out... but for any given player they would literally have to play thousands of matches to even come close to it averaging out the flukes. Believing otherwise is giving into Gambler's Fallacy.

[Rampancy]

Mercules, on 08 June 2015 - 03:41 PM, said:

Match maker assigns you 100 matches you are supposed to lose. You play well in every match and win 2 of them. How has your true skill been reflected in a 2/98 W/L record? It hasn't.

You are correct, over thousands of players playing hundreds of thousands of matches the mean average of all those players and matches would even out... but for any given player they would literally have to play thousands of matches to even come close to it averaging out the flukes. Believing otherwise is giving into Gambler's Fallacy.

Uh

No

The vast, vast, vast majority of people's expected wins vs. expected losses will be roughly even

Declaring W/L broken due to the possiblity of statistical anomaly is asinine.

[ScarecrowES]

RampancyTW, on 08 June 2015 - 03:27 PM, said:

...I really don't see how that's relevant. Over a good sample size (say 100 games? The more the better) your relative contribution to your team will be reflected in your W/L. If you're a low Elo player and your perform better than the other team's low Elo players, your team will have a better chance of winning.

Your first statement is not actually true at all. It's exactly why Elo cannot be used to accurately judge the performance of individual players on a team relative to other players. What you've stated is a fundemental misunderstanding of how Elo works.

Your second statement is truer. But you've fundementally just expressed that if your personal performance is above what is typical for your Elo rating, your team will benefit from your improved play. Clearly that statement will always be true regardless of what your Elo says. However, the same could be said of any player on either team, and oppositely if an individual player is playing below his level. And certainly the generally random nature of matches cannot be taken into account here.

Expressing the problem with W/L or Elo further... you can be the world's best quarterback, but if you're saddled with a team that can't catch or run the ball, you will lose. You will lose consistently, despite your spectacular individual play. Will your play make your team more likely to win than if you were sitting on the sideline? Sure. However, how much your individual play can effect the outcome of the match will depend on the circumstances of the match. It isn't entirely random, but for all intents and purposes it's little more than trying to overcome chaos.

You win/loss ratio might as well be entirely random in pug play, and your Elo is only ever partially under your own influence.

[Rampancy]

ScarecrowES, on 08 June 2015 - 03:45 PM, said:

Your first statement is not actually true at all. It's exactly why Elo cannot be used to accurately judge the performance of individual players on a team relative to other players. What you've stated is a fundemental misunderstanding of how Elo works.

No it isn't. This is only the case if you assume completely static skill levels across the entire playerbase. But that isn't the case. If everybody's ability sat stagnant, then sure, everybody would eventually trend to 50%. But this isn't the case, so your ability to affect a match, relative to the guys on the other team's abilities to affect the outcome of a match, will be reflected in your W/L.

Quote

Your second statement is truer. But you've fundementally just expressed that if your personal performance is above what is typical for your Elo rating, your team will benefit from your improved play. Clearly that statement will always be true regardless of what your Elo says. However, the same could be said of any player on either team, and oppositely if an individual player is playing below his level. And certainly the generally random nature of matches cannot be taken into account here.

Which is why you require a large sample size to iron out the kinks. The same is true of any statistic.

Quote

Expressing the problem with W/L or Elo further... you can be the world's best quarterback, but if you're saddled with a team that can't catch or run the ball, you will lose. You will lose consistently, despite your spectacular individual play. Will your play make your team more likely to win than if you were sitting on the sideline? Sure. However, how much your individual play can effect the outcome of the match will depend on the circumstances of the match. It isn't entirely random, but for all intents and purposes it's little more than trying to overcome chaos.

This isn't an Elo problem, since the matchmaker structures roughly even teams and it's no more likely that your team will be idiots relative to the other team than that the teams will be relatively evenly distributed. Pandering to a statistical anomaly doesn't invalidate the relevance of the W/L statistic.

Quote

You win/loss ratio might as well be entirely random in pug play, and your Elo is only ever partially under your own influence.

Well, it isn't. Since all of the randomness over a large sample size will trend towards your true level of performance.

This is basic stuff, people.

Edited by RampancyTW, 08 June 2015 - 03:54 PM.

[Jman5]

Really at the end of the day, the only tried and true statistic would be your Solo Elo if group and solo were counted separately. However, since none of us can see our Elo and it's not split, I don't pay much attention to the mech stat page.

W/L is useless, KDR is useless. Comparing your damage to people in your game is better, but still imperfect. My rough line in the sand is over/under 300 damage.

[Karl Marlow]

I occasionally get curious but I don't really care. I don't think I've even looked at my stats in about a year. Of course you bringing it up has made me curious

[Soulscour]

Jman5, on 08 June 2015 - 04:07 PM, said:

Really at the end of the day, the only tried and true statistic would be your Solo Elo if group and solo were counted separately. However, since none of us can see our Elo and it's not split, I don't pay much attention to the mech stat page.

W/L is useless, KDR is useless. Comparing your damage to people in your game is better, but still imperfect. My rough line in the sand is over/under 300 damage.

I pretty much agree with this. My kdr and win ratio was much higher before they introduced elo. One stat that is relevant however is when you compare win loss ratio on different maps and game modes. I think its interesting when your winning much more games on one map vs another and may be an indicator of play style or map familiarity.

[Dingo Battler]

I will be brutally honest and say KDR does matter to me, more so than WLR. I put WLR down to matching and ELO, which is something I cannot influence to a great degree (how many of your matches have been close, compared to 2-12 stomps?), but KDR can be influenced.

Also, KDR greatly influences my playstyle. I play close range brawlers, so I can get my kills up. I don't play assault, as I cannot pursue. I do commit to a fight, contribute damage and push. However, if I spot a dying mech nearby, I will go out of my way to get my UAC20 into it.

That said, to maintain a decent KDR, certain good behaviours are learnt in the process. Deaths are far more penalising than kills. I rather have 0 kills, 0 deaths than 2 kills 1 death (Ratio for current mech is 2.5) Therefore, I fight conservatively, and try to survive longer, being alive to put out more damage. I learnt to torso twist to last longer and survive. I push with assaults, because that's when you get the majority of your kills. I try to flank, because that's when you can catch a mech out and kill it.

I do agree though, that KDR is a flawed ratio. I think its not favourable to assaults, mediums and lights. It penalises long range, and LRM boats, which are very important to beat down mechs enough to kill.

I think other performance indicators, like damage tanked, shots fired, damage dealt, etc. should be used to accumulate points. I think the best indicator is damage dealt, maybe divided by tonnage?

Until they start giving points per match for other indices, KDR will remain the dominant thing influencing the behaviour of pilots.

[Karmen Baric]

I use a Funometer ... if im having fun its important & counts for a lot.

[Davegt27]

I been looking at how much I die

I die 70% of the time in my lights so I suck pretty bad

Time in game and damage done would be stats I would look at
Edit
Avg time in game/match tells me if you are some nob Rambo charging over the hill or
A productive member of the team

At the end of a match I look at damage done more than anything else, if I get top damage
On both sides I take a screen shot and save it
I have folders for each Mech and how many top damage I get per month for each Mech
If any
This also tells me if my builds are productive
Even in CW the guy/gal that puts out high damage consistently is well respected

Anyone can get lucky and get kills but wrecking face constantly gets high marks

Edited by Davegt27, 09 June 2015 - 02:56 AM.

[PhoenixFire55]

Every stat matters IMO. I often check mech specific stats to see the long-term perspective on how my mechs work with certain builds etc. Would be nice if I didn't have to calculate it all manually and can sort out last x matches etc. KDR and W/L mean little really, I never cared about having or not having that final killing shot and I don't believe that the stupid Elo MM works and I'm getting matches with balanced teams. There are far more important things, like average damage per match and average damage per shot for certain weapons. W/L for different maps and game modes is interesting but hardly gives you any releveant information.