120 replies to this topic

[Kin3ticX]

Tesunie, on 26 February 2018 - 09:18 PM, said:

Thing is, you may know but I do not. I don't know if X account is Y's alternate, which makes figuring out their combined stats (which then would include their "learner" matches, for example) far more difficult.

Overall, I will comment that I like stats and crunching numbers, but I also don't stop at just stats. Stats will only tell you so much and does sometimes depend upon what you are looking at, how you are looking at it, and how you interpret the numbers.

For instance I find K/D to be flawed as a statistical analysis tool. It's fun to track, but doesn't normally tell you much actual information besides "this person knows how to time their shots to get the kill" more than "this person knows how to play the game and drop targets effectively". I know I've seen many matches with players with 500+ damage, and not a single kill. I've also seen people with matches where they do less than 20 damage, but somehow managed 3 kills...

Even W/L has it's issues, if looked at only and specifically. It can be gamed to some extent, as well as can be luck dependent. Even high damage (if looked at on it's on in particular) can be misleading. I've seen people do 500+ damage, we win by kills and they only have 3 assists... Makes you wonder a bit on how they managed to spread 500+ damage to only three mechs in the match... (etc)


Okay... I think I'm going way off topic now... sorry guys. Just being a wordsworth at the moment.

Yeah sure its not the be all end all but you also cant explain away a .9WL or a .9KDR with any amount of mental gymnastics. Which BTW is 90% of the discussion about why stats dont matter. aka "my stats are horrible so stats dont matter".


Its very true though that stats arnt the be all end all, but with that said, most people play normal and the stats fall where they fall. There are a few people that try to game their KDR by hiding on assault and conquest, but in the long run, you cant really fool anyone.

Edited by Kin3ticX, 26 February 2018 - 09:35 PM.

[Tesunie]

Kin3ticX, on 26 February 2018 - 09:32 PM, said:

Yeah sure its not the be all end all but you also cant explain away a .9WL or a .9KDR with any amount of mental gymnastics.

I find, personally, K/D to be near useless. W/L is far more explainable and more impactful.

A 0.9 K/D can be explained away easily. Wont tell you how many times I've nearly killed an opponent 1v1 (and paid the price in damage too), only to have someone else tap the target and "secure" the kill (because I believe in the team effort more than "kill stealing") at the last moment.

W/L is far less able to be manipulated, but lets face it it can be sabotaged. A single person disconnecting can do it, or yourself disconnecting from some error or connection issue. If one wanted to sabotage their own W/L, it's even easier, as one could make a show of participating, but actually take no real worthy effort to win. (Not saying people do/will, just saying it's a possibility.) More reliable, less likely to be explained away.

Stats I like to look at personally is damage per ton of mech invested into a match, average match score, and W/L averages. Stats I'd love to have access to would be damage per assist/kill, personal stat archiving (so I could archive a mech stat when I change builds), individual mech stats (compared to chassis/variant only stats). If I could have access to some of those other stats, I could then make a far more comprehensive analysis of mechs, builds and how well they each work for me. Right now, only the Huntsmen Prime(S) has reliable build to chassis stats, as that mech has had the same build on it unaltered since the (S) designated variants started to develop their own stat data. (Otherwise, my two LRM15 and four ERML build would have stats mixed with dual UAC5 or other prime builds from the base variant...)



Also, overall stats don't count in for new mechs, new builds, different mech builds you aren't use to and voiding experimentation out of your "normal" stats. As such, for example, this season I can see my stats have gone down, probably because I've done some experimentation on Light builds with dual AMS and assorted assault mechs I don't normally play. Actually, I don't normally play assaults practically at all. I'm a medium mech jock... Basically, many of us don't always play nor bring in our best at all times. Probably would be boring if we did.

[Kin3ticX]

Tesunie, on 26 February 2018 - 10:25 PM, said:

I find, personally, K/D to be near useless. W/L is far more explainable and more impactful.

A 0.9 K/D can be explained away easily. Wont tell you how many times I've nearly killed an opponent 1v1 (and paid the price in damage too), only to have someone else tap the target and "secure" the kill (because I believe in the team effort more than "kill stealing") at the last moment.

W/L is far less able to be manipulated, but lets face it it can be sabotaged. A single person disconnecting can do it, or yourself disconnecting from some error or connection issue. If one wanted to sabotage their own W/L, it's even easier, as one could make a show of participating, but actually take no real worthy effort to win. (Not saying people do/will, just saying it's a possibility.) More reliable, less likely to be explained away.

Stats I like to look at personally is damage per ton of mech invested into a match, average match score, and W/L averages. Stats I'd love to have access to would be damage per assist/kill, personal stat archiving (so I could archive a mech stat when I change builds), individual mech stats (compared to chassis/variant only stats). If I could have access to some of those other stats, I could then make a far more comprehensive analysis of mechs, builds and how well they each work for me. Right now, only the Huntsmen Prime(S) has reliable build to chassis stats, as that mech has had the same build on it unaltered since the (S) designated variants started to develop their own stat data. (Otherwise, my two LRM15 and four ERML build would have stats mixed with dual UAC5 or other prime builds from the base variant...)



Also, overall stats don't count in for new mechs, new builds, different mech builds you aren't use to and voiding experimentation out of your "normal" stats. As such, for example, this season I can see my stats have gone down, probably because I've done some experimentation on Light builds with dual AMS and assorted assault mechs I don't normally play. Actually, I don't normally play assaults practically at all. I'm a medium mech jock... Basically, many of us don't always play nor bring in our best at all times. Probably would be boring if we did.

eh, lets agree to disagree.

I am pretty unhappy with a mech if I cant strive to reach at least a 3.0 KDR playing solo queue. In the shortrun, KDR doesnt mean anything, but in the long run, KDR says something about a mech being able to finish the job (and a mech with a healthy KDR will probably have a pretty health WL rate too).

+ w/r, + kdr, +avg dam tend to be very co-linear in my own stats. Meaning, the more damage I do the more I win, the more kills I get the more I win, the more damage I do, the more I kill. Co-linear

Edited by Kin3ticX, 26 February 2018 - 10:39 PM.

[Tesunie]

Add-on about stats:

Taking from my own personal stats as recorded by the Jarl's list...
I can see what seasons I experimented more and what ones I probably used "good old reliable" builds and mechs. Can't and wont say I'm great, but I think it's reasonable. As for W/L and K/D, I see one season I ended up with a 2.0 W/L, but only a 0.8 K/D. Highest seasonal K/D is actually only a 1, but that season's W/L was one of my lower times.

I'm gonna guess that seasons 1-5 I probably was still on direct fire practice, as I was doing that when PSR first came out and for some time afterwards. Man, my tier crept back then from T4 starting to mid T3. Was season 7 when the Huntsmen came out? I'm guessing around that time? I see my stats jumped for a few seasons then... Probably when I was running (almost exclusively) my LRM Huntsmen. (Speculating here, as I don't exactly know what time frames the seasons are in.)


I might want to comment that I play probably as much QP as I do GP now. I'm sure that also has an impact on stats...

Kin3ticX, on 26 February 2018 - 10:37 PM, said:

eh, lets agree to disagree.

I'm fine with that. I just find that my data doesn't seem to match yours. I find I can have a high K/D (for me at least) and those times I may very well have a low W/L. Other times, I have a high W/L, but miserable K/D...

I also like to look at specific mech stats, rather than overall stats personally. Then we can discuss build efficiency, over possible experimentation or poor performance in newer mechs...

[Kin3ticX]

Tesunie, on 26 February 2018 - 11:11 PM, said:

Add-on about stats:

Taking from my own personal stats as recorded by the Jarl's list...
I can see what seasons I experimented more and what ones I probably used "good old reliable" builds and mechs. Can't and wont say I'm great, but I think it's reasonable. As for W/L and K/D, I see one season I ended up with a 2.0 W/L, but only a 0.8 K/D. Highest seasonal K/D is actually only a 1, but that season's W/L was one of my lower times.

I'm gonna guess that seasons 1-5 I probably was still on direct fire practice, as I was doing that when PSR first came out and for some time afterwards. Man, my tier crept back then from T4 starting to mid T3. Was season 7 when the Huntsmen came out? I'm guessing around that time? I see my stats jumped for a few seasons then... Probably when I was running (almost exclusively) my LRM Huntsmen. (Speculating here, as I don't exactly know what time frames the seasons are in.)


I might want to comment that I play probably as much QP as I do GP now. I'm sure that also has an impact on stats...



I'm fine with that. I just find that my data doesn't seem to match yours. I find I can have a high K/D (for me at least) and those times I may very well have a low W/L. Other times, I have a high W/L, but miserable K/D...

I also like to look at specific mech stats, rather than overall stats personally. Then we can discuss build efficiency, over possible experimentation or poor performance in newer mechs...

Dont forget KDR is zero sum. That means if you have a .6KDR you are feeding. I mean, I dont have a great KDR either but I wouldnt be satisfied with a .6KDR in MWO.

(but that also means in order for me to have a ~3-4 KDR someone has to have a poopy one)

[Tesunie]

Kin3ticX, on 26 February 2018 - 11:25 PM, said:

(but that also means in order for me to have a ~3-4 KDR someone has to have a poopy one)

This statement I feel is one that should be remembered, which is why I'd much rather look at other stats beyond W/L and K/D. For one to win, one much lose. What might happen to people's stats if those with a positive W/L always fought against each other? Why, no matter how good those players may be, one of their W/L would have to drop if the other is to rise... Even if evenly matched, then their W/L would normalize to around 1 eventually of only they fought each other (statistically speaking).

Still kinda supports statistics as a poor judge on if you should listen to someone or not. (Just to loop back into a part of the original conversation.) That was one point we all seemed to basically agree on.


On note of mix "bracketed" builds, they also can sometimes be good at pacing one self. For example, I have an Adder with 4 SRM4s and a single ERLL. Not only does the ERLL help boost damage over a smaller laser, but it also comes in handy in the opening moves of a match. Sure, it wont be as effective against a sniper build, but if I see (for example) an AC20 Annihilator, then I can keep my distance even though it isn't a strength. Personally, I like that flexibility. Otherwise... I find myself charging in too recklessly "because I want to help and do stuff"... (Personal flaw, I know.)

[Kin3ticX]

Tesunie, on 27 February 2018 - 12:16 AM, said:

This statement I feel is one that should be remembered, which is why I'd much rather look at other stats beyond W/L and K/D. For one to win, one much lose. What might happen to people's stats if those with a positive W/L always fought against each other? Why, no matter how good those players may be, one of their W/L would have to drop if the other is to rise... Even if evenly matched, then their W/L would normalize to around 1 eventually of only they fought each other (statistically speaking).

I don't think thats how it works and I think you are missing the point.

The point is, don't be the person with the .6KDR, let someone else hold it. Besides, that is one hell of an ez low hanging fruit. Tbh, with a .6 KDR, you got problems with a)losing in the mechlab, b)positioning out of whack c) aim. People usually start with builds first since those are copy paste and then focus on aim and positioning. Mental gymnastics of like, well if all the good players played the good players then everyone would have a 1.0 W/L....thats just a cop out.

Edited by Kin3ticX, 27 February 2018 - 12:44 AM.

[jss78]

IMO, W/L is really the bottom line. Do something helpful -- whatever it is -- and your W/L goes up. That's across a fairly large sample of games, to see your impact in that chaotic madness of 12-vs-12 random team queue.

However I would expect a strong correlation between K/D and W/L. It's just that various "support" roles aren't quite important enough in this game. Conquest isn't quite enough about the conquest. Scouting isn't really needed since the maps are small and we tend to go to the same spots. Ultimately we always find the enemy -- and then we shoot until the other team is dead.

And the guy who deals focused damage (lots of K) is going to be even more helpful (lots of W) because he's more likely to give an early numbers advantage to your team.

There ARE useful non-damage-dealing support functions. Watch your slow heavy-hitters' back a bit, letting them rack up that 1000+ damage, bring that 3xAMS for your main group, well-placed UAV's. But I think you'd still have to combine that with respectable damage and kills, otherwise it's a hard way to carry a team.

[denAirwalkerrr]

Koniving, on 26 February 2018 - 08:22 AM, said:

Consider this: More than 30 of the consistently top 100 players over the past 10 seasons now have purple names. It would not surprise me exactly how many purple names could be found within the top 10 thousand. Shame you can't do a direct search for it.

I’m rank 100 and this post insults me and violates CoC by accusations of hax without proof plz ban.

Edited by denAirwalkerrr, 27 February 2018 - 02:30 AM.

[Spare Parts Bin]

I have experimented with so many builds on so many Mechs including a bunch,I don't even have anymore. I don't look at my stats although I try to look at each game results individually. Each match I strive to do as good as my,better matches.

[Horseman]

jss78, on 27 February 2018 - 01:20 AM, said:

However I would expect a strong correlation between K/D and W/L. It's just that various "support" roles aren't quite important enough in this game. Conquest isn't quite enough about the conquest. Scouting isn't really needed since the maps are small and we tend to go to the same spots. Ultimately we always find the enemy -- and then we shoot until the other team is dead.
There ARE useful non-damage-dealing support functions. Watch your slow heavy-hitters' back a bit, letting them rack up that 1000+ damage, bring that 3xAMS for your main group, well-placed UAV's. But I think you'd still have to combine that with respectable damage and kills, otherwise it's a hard way to carry a team.

Caveat... a dead support mech doesn't contribute. So while you won't score many kills with such builds, you should also avoid getting killed often - and thus from losing KDR.

Spare Parts Bin, on 27 February 2018 - 03:54 AM, said:

I have experimented with so many builds on so many Mechs including a bunch,I don't even have anymore. I don't look at my stats although I try to look at each game results individually. Each match I strive to do as good as my,better matches.

I keep track of how mine change on a month-by-month basis, which is more helpful than the lifetime values.

Edited by Horseman, 27 February 2018 - 04:05 AM.

[Eisenhorne]

Horseman, on 27 February 2018 - 04:03 AM, said:

Caveat... a dead support mech doesn't contribute. So while you won't score many kills with such builds, you should also avoid getting killed often - and thus from losing KDR.

Yup. I've been playing a fire support Bushwacker lately with *GASP* LRM's on it. I am fast enough to stay with the group, so I normally just chill in the middle as we nascar around. It doesn't do great, but its a consistent 400-500 damage and 1-2 kills per game, as long as I live until the end. And because I'm just fire support, I'm usually one of the last to die. Its really the only type of mech LRMs belong on in my opinion, fast mediums. Over the past 39 games with it, I'm at a 1.44 W/L ratio, with a 1.0 K/D.

I tried to play the BSW-P1 with the classic SRM configuration, but good lord that was just a disaster for me in the solo queue. My stats for this chassis would be a lot higher if I could just count the LRM games.

[Tesunie]

Kin3ticX, on 27 February 2018 - 12:43 AM, said:

I don't think thats how it works and I think you are missing the point.

The point is, don't be the person with the .6KDR, let someone else hold it. Besides, that is one hell of an ez low hanging fruit. Tbh, with a .6 KDR, you got problems with a)losing in the mechlab, b)positioning out of whack c) aim. People usually start with builds first since those are copy paste and then focus on aim and positioning. Mental gymnastics of like, well if all the good players played the good players then everyone would have a 1.0 W/L....thats just a cop out.

Not what I'm exactly going at with my statement. I'm just saying that, for one person to win another much lose. That is the nature of a PvP game.

For example, to remove a lot of the random elements, lets just say you and ASH played each other in 1v1 matches. Now, stat wise as far as I can tell each of you are probably about equally skilled. However, in those matches (theoretically speaking), you got only 1 win out of every 10 matches. That's also 1 kill for every 9 deaths.

Now, in that situation, you would appear to be "a bad player" if you only looked at those stats. That may not be the case, but you may be slightly lower in skill than your opponent or may not even have as good hardware/software. (I know the last month I've been having horrible lag, which happens only when my brother's computer is on. We think something is seriously wrong with it. Just as a real life example.) That should not have any impact on your forum suggestions.


Now, yes. I know that this game isn't 1v1, so theory crafting above doesn't exactly apply. But the principal is still there. There is a reason I like to look at other stats in conjuncture with W/L and K/D. Which is why I wish I could have access to how often I do KMDDs (for example), because I may not get kills but I seem to average 1-2 KMDDs a match. Now, it's no K/D, but that indicates to me that I tend to almost drop a target when someone else does instead. (Wish I had more solid information with this, instead of "patterns I see".)

I'd also comment, as far as my overall K/D rate, I probably have too many matches in with a low K/D rate to effectively be able to bring it up. Though, my main stat page tells me I have a K/D of 0.74, but that does have the archived stats excluded and counts from before the Jarl's list and seasons/leaderboards.

Another reason though I actually prefer to look at individual mech/build stats is it will say more overall. For example, my Huntsmen Prime(s) (my LRM build mentioned previously, which is a mixed build) has 136 matches played (since time when they separated (S) records from normal variants), a W/L of 1.31, a K/D 1.05, damage per match of 457*, damage per ton per match of 9.14 (DPTPM of 5 is typically considered good) and average match experience earned of 1,349. (* If you wish to account for "effective damage", you could divide the damage per match in half to account for half laser and half LRM damage. Then divide the LRM half in half again and add to laser damage numbers. Meaning, effective damage would be about 350ish.)

As we all know, mech build and playstyle can greatly effect stats. As I don't always play this one mech build and have a love for experimentation (some of which are builds that should never be mentioned again), it can have a negative effect on my overall stats. I'd also make the example of something Ash mentioned... He said that we should practice with Gauss because it is a good weapon. He mentioned it may take 40-60 matches to really get the hang of it. If someone is really bad with Gauss compared to other weapons and they take that advice, that learning time will have a negative effect on a player's stats. Same with learning any new build/weapon/mech. A mech with no skills statistically performs worse than a fully skilled mech.


I'm just really saying that advice shouldn't be related to stats. Maybe weigh your own acceptance of advice based on the player's stats, but I wouldn't ignore it though either. I've had top tier players tell me to drop LRMs as they are "garbage", but yet they continuously (even in T1) are my better performing builds. I've even had some top tier players tell me some really "odd" advice before, which I tested and it failed utterly (at least for me) and then I rejected. I've gotten golden nuggets of advice from T5 players before with worse stats than myself, and it proved correct.

(I think I'm blabbing here on things we've essentially already agreed on. So I'm just going to drop rambling on now...)

[Tesunie]

jss78, on 27 February 2018 - 01:20 AM, said:

There ARE useful non-damage-dealing support functions. Watch your slow heavy-hitters' back a bit, letting them rack up that 1000+ damage, bring that 3xAMS for your main group, well-placed UAV's. But I think you'd still have to combine that with respectable damage and kills, otherwise it's a hard way to carry a team.

I will mention that I've been experimenting with these roles as of late. I've been trying to figure out if it is worth while to fill in such a roll for C-bills, experience and match score. My experiments have lead to dual AMS Panthers and Wolfhounds, as well as a Grasshopper. So far, in a light mech, it's just a match score sky rocket if you can intercept missiles while being nearby a heavy or assault mech. Don't need to deal much damage and I'm sometimes getting 300+ match scores. Protected light bonuses combined with AMS bonuses really can stack up a match score. Plus you protect your slower and heavier assets for your team, letting them deal more damage with less fear of a random light in their back and from incoming missiles...

It's not as rewarding as being the 1000+ damage dealing with 4+ kills, but it can get a reasonable match score rather consistently. At least the way I play it...

[Metus regem]

jss78, on 27 February 2018 - 01:20 AM, said:

IMO, W/L is really the bottom line. Do something helpful -- whatever it is -- and your W/L goes up. That's across a fairly large sample of games, to see your impact in that chaotic madness of 12-vs-12 random team queue.

However I would expect a strong correlation between K/D and W/L. It's just that various "support" roles aren't quite important enough in this game. Conquest isn't quite enough about the conquest. Scouting isn't really needed since the maps are small and we tend to go to the same spots. Ultimately we always find the enemy -- and then we shoot until the other team is dead.

And the guy who deals focused damage (lots of K) is going to be even more helpful (lots of W) because he's more likely to give an early numbers advantage to your team.

There ARE useful non-damage-dealing support functions. Watch your slow heavy-hitters' back a bit, letting them rack up that 1000+ damage, bring that 3xAMS for your main group, well-placed UAV's. But I think you'd still have to combine that with respectable damage and kills, otherwise it's a hard way to carry a team.

/shrug

I know I don't maintain a fantastic KDR on my BSW-X2 I pilot most often when I log in (1.36KDR), but she does have a good W/L raito of 4.5. Granted what I do most often with her is escort my heavies/assault mechs to where they need to be and bully other mediums/lights that get too close, while providing what fire support I can to the mechs doing the heavy lifting. So I guess I am doing something right with her, even if what I do, isn't really supported or shown in my match scores....

[jss78]

Metus regem, on 27 February 2018 - 11:04 AM, said:

/shrug

I know I don't maintain a fantastic KDR on my BSW-X2 I pilot most often when I log in (1.36KDR), but she does have a good W/L raito of 4.5. Granted what I do most often with her is escort my heavies/assault mechs to where they need to be and bully other mediums/lights that get too close, while providing what fire support I can to the mechs doing the heavy lifting. So I guess I am doing something right with her, even if what I do, isn't really supported or shown in my match scores....

Well, I didn't really mean to imply that doing support jobs is incompatible with doing well. It's a false dichotomy that doing support means that you must otherwise sit on your thumbs (not claimed by you here, but elsewhere by lots of people). Like you say yourself -- while you're doing the job of keeping your team's heavy hitters safe, you're also yourself killing more than getting killed. I'm not surprised the W/L is high.

On my clanner alt I have my 3xAMS 1xECM Kit Fox -- it wins a lot of games, but there too the key is that the little bug maintains a nicely >1 K/D, tends to survive in late game, and racks up decent damage over time. I try to play careful (makes no sense dying first if you're trying to be the support, as noted by Kineticx above), but don't hide behind a rock.

[Spare Parts Bin]

I took my Catapult K2 added a 275 LFE, a AC-10+4 tons of ammo, a HPPC,and 3 MLs. I also tried exchanging an ERPPC and a LPPC for the HPPC. It seemed to work better with the HPPC.

7 games( 4 losses,3 wins) 5 Kills,2 Kill Most Damage Done, 31 Kill Assists, 8 Component Destructions, and 1409 points of damage. My warhorn went off I swear I should have another kill. I think the fusion of Ash's & Baradul's designs works well.

[Tesunie]

Spare Parts Bin, on 27 February 2018 - 06:12 PM, said:

I took my Catapult K2 added a 275 LFE, a AC-10+4 tons of ammo, a HPPC,and 3 MLs. I also tried exchanging an ERPPC and a LPPC for the HPPC. It seemed to work better with the HPPC.

7 games( 4 losses,3 wins) 5 Kills,2 Kill Most Damage Done, 31 Kill Assists, 8 Component Destructions, and 1409 points of damage. My warhorn went off I swear I should have another kill. I think the fusion of Ash's & Baradul's designs works well.

If you feel you are having success with it, then go with it. Just save the design notes and change it if you ever feel the need. Remember it doesn't hurt (besides maybe your stats) to experiment. So experiment and see what you find effective and don't be afraid to try something else.

Honestly, I like the sounds of your build. Seems solid to me.

[jss78]

Yeah, on surface it's a perfectly reasonable combination of mid-range weapons. (Not really a bracket build at all.)

I might look at the amount of AC/10 ammo, I usually run only two tons (40 shots and possibly a few from the skill tree node). Might you have room for another heat sink instead?

[Spare Parts Bin]

jss78, on 28 February 2018 - 03:20 AM, said:

Yeah, on surface it's a perfectly reasonable combination of mid-range weapons. (Not really a bracket build at all.)

I might look at the amount of AC/10 ammo, I usually run only two tons (40 shots and possibly a few from the skill tree node). Might you have room for another heat sink instead?

Why so little ammo? I get nervous unless I have 60-100 rounds.