349 replies to this topic

[Skipmagnet]

Quick play oniy, I think, but it has both solo and group Q.

[Scurro]

Appogee, on 02 November 2017 - 12:50 AM, said:

Yes, I read that. "Progress is the progress a pilot has made this season compared to their average match score."

But I still couldn't work out what exactly it was supposed to mean.

Is it calculated as: "Progress" = "New Avg Match Score" / "Previous Avg Match Score" - 100

(Latest season average match score / average match score) * 100 -100

[Kin3ticX]

eyeballs, on 02 November 2017 - 03:02 AM, said:

Quick play oniy, I think, but it has both solo and group Q.

Most of mine are solo queue. Stomping group queue gets a nice w/l and a nice kdr but a lower matchscore because there is less meat to go around.

[chucklesMuch]

Are you wanting keep alts separate, combine or don't care?

[Scurro]

chucklesMuch, on 02 November 2017 - 01:37 PM, said:

Are you wanting keep alts separate, combine or don't care?

Alts can stay separate. Upon request from the player, I will (if I am not busy) edit the data from players who had a rename to combine their data to the new name.

Edited by Scurro, 02 November 2017 - 01:46 PM.

[Scurro]

Ranks are now determined by adjusted match score for players who have been active in the last three months.
Players who have stopped playing are still tracked but are now placed into "retirement". You can still search retired player stats. Retired players will be placed back into ranks if they return but their latest performance will have significant effect on their placement due to season weights.
Percentile rank also takes this into effect. Only active players are measured in rankings.
You can view the total amount of active and retired players at the bottom of the page.

Enjoy.

https://leaderboard.isengrim.org

[Jay Leon Hart]

Well gosh darn it, now I've dropped 1%

[Bombast]

Jay Leon Hart, on 05 November 2017 - 11:22 AM, said:

Well gosh darn it, now I've dropped 1%

I don't seem to have moved? I guess that's good.

[Shifty McSwift]

Lol the smurf alts favourite tool for self congratulation and e-peen measurements. <o

[Bombast]

Shifty McSwift, on 05 November 2017 - 11:30 AM, said:

Lol the smurf alts favourite tool for self congratulation and e-peen measurements. <o

How dare people attempt to track their performance. The fiends.

[Shifty McSwift]

Bombast, on 05 November 2017 - 11:34 AM, said:

How dare people attempt to track their performance. The fiends.

Yeah because that's what I said right?

How dare you say that about my mother!?!?!112

[Nightbird]

Could you dump the data from the last three seasons into separate excel/csv/googledoc files? If so I can cunch the numbers with my stats software to see what the matchmaker is doing wrong. It would be a lot more accurate if I had tier and mech information as well but it would be a good start.

Edited by Nightbird, 14 November 2017 - 08:23 AM.

[Lily from animove]

Nice list

also lol kaffeeangst 100% of his games in assaults. Such addiction to a classs

And then

MonPax 21228

yeah 21k games duuuude seriously, how?

Edited by Lily from animove, 14 November 2017 - 08:39 AM.

[Davegt27]

if Match maker is not using fire power in its analysis then that's a big problem

the old saying don't bring a knife to a gun fight comes to mind

I all so submit if fire power is not important then why does PGI spend so much time nerfing stuff

[Scurro]

Nightbird, on 14 November 2017 - 08:22 AM, said:

Could you dump the data from the last three seasons into separate excel/csv/googledoc files? If so I can cunch the numbers with my stats software to see what the matchmaker is doing wrong. It would be a lot more accurate if I had tier and mech information as well but it would be a good start.

Here is the last three leaderboard seasons pulled straight from the database. It will include some other data that is used in calculating scores.

Tier and mech information is not published by PGI and I am unable to pull that data.

Edited by Scurro, 14 November 2017 - 08:53 AM.

[Jackal Noble]

So, what am I supposed to do then go with one single mech all the fricking time. Ya no thanks lol. Nope, 6441... really?
I guess I should be more selfish instead of trying to be a team player.
Assault - camp in the back and farm.
Heavy - again, farm and not push.
Medium - idk
Light - not actually scout for team (TBH, that usually bites me in the @$$, will have spotted the entire enemy team on Crimson Straits twice over and team just gets rick-rolled, feels like a waste of time and no actual reward for it anyways.)

Obviously doing this game wrong.

Would be really nice if this data was parsed into solo-play vs group-play vs faction performance

Edited by JackalBeast, 14 November 2017 - 08:55 AM.

[Methanoid]

sceii, on 02 November 2017 - 02:00 AM, said:

Does this leaderboard include FP?

I would be amazed if the FP stats showed up, moreso if they showed that anyone actually played it outside of events in any number of worth, made even more ridiculous by the fact they keep pumping a lot of time and effort into something that is absolutely hated by the vast majority of whats left of the playerbase while making those changes constantly unpopular with no effort to change this.

[MischiefSC]

Nightbird, on 14 November 2017 - 08:22 AM, said:

Could you dump the data from the last three seasons into separate excel/csv/googledoc files? If so I can cunch the numbers with my stats software to see what the matchmaker is doing wrong. It would be a lot more accurate if I had tier and mech information as well but it would be a good start.

Mech info is sorta lost in the shuffle without an accurate 'tier' or more pointedly Elo style system to give you actual data on mech chassis performance.

Tier.... honestly it's an invisible factor. In reality what we have is 'Perpetually terribad/people with under X matches' group, a 'consistently just under mediocre/people within middle range Y matches' group and 'consistently mediocre or better/people with over Z matches' group. The matchmaker just tries to usually keep them out of games with each other.

I would be way more interested in seeing what the minute by minute population numbers were for matchmaking. So when it goes to make a match over a 90 second period, how many players does it have to choose from?

THAT would give me some valuable data to establish just how precise a player ranking system would need to be. If it's only got 60 players available and it needs to make 5 matches in that time.... well, honestly? A more precise matchmaker ranking would be useless. We should be happy we get approximately equal tonnage.

[Nightbird]

MischiefSC, on 14 November 2017 - 12:26 PM, said:

Mech info is sorta lost in the shuffle without an accurate 'tier' or more pointedly Elo style system to give you actual data on mech chassis performance.

Tier.... honestly it's an invisible factor. In reality what we have is 'Perpetually terribad/people with under X matches' group, a 'consistently just under mediocre/people within middle range Y matches' group and 'consistently mediocre or better/people with over Z matches' group. The matchmaker just tries to usually keep them out of games with each other.

I would be way more interested in seeing what the minute by minute population numbers were for matchmaking. So when it goes to make a match over a 90 second period, how many players does it have to choose from?

THAT would give me some valuable data to establish just how precise a player ranking system would need to be. If it's only got 60 players available and it needs to make 5 matches in that time.... well, honestly? A more precise matchmaker ranking would be useless. We should be happy we get approximately equal tonnage.

Given any group of 24 players, it should be possible to create teams with close to 50% win loss chance, but more data means more predictors.

While Tiers is invisible, what it should do is act as a counter weight to player skill. The better a player's skill, the higher their chance of winning, but the better the tier, the worse (via matchmaker) their chance of winning. Today Tiers work in that there is clearly a difference between Tier 1 and 5 players. But, it doesn't work in that within a tier, people can get win/loss ratios significantly different from 1.

The ideal data I need is separation of GQ & SQ data, and also a player's stats seperate by mech chassis, and even by weapons equipped. With that, I can calculate weights for everything, from a player's performance to the mech chassis selected to the loadout. Failing to have all of that, the data is lower quality as will be the predictions from it.

What I'll try to do is use one season to generate weights for stats we have, like KDR, Kills per match, avg match score, and see if I can predict the W/L ratio better than the MM. Remember, the MM always predicts 1, so I want to be on the right side of 1 and closer to the true W/L ratio. What will this accomplish? It will mean I can generate a better number to represent a player's Tier than the current PSR.

This better number can be used in place of the PSR formula, to create better matches.

[MischiefSC]

Nightbird, on 14 November 2017 - 01:03 PM, said:

Given any group of 24 players, it should be possible to create teams with close to 50% win loss chance, but more data means more predictors.

While Tiers is invisible, what it should do is act as a counter weight to player skill. The better a player's skill, the higher their chance of winning, but the better the tier, the worse (via matchmaker) their chance of winning. Today Tiers work in that there is clearly a difference between Tier 1 and 5 players. But, it doesn't work in that within a tier, people can get win/loss ratios significantly different from 1.

The ideal data I need is separation of GQ & SQ data, and also a player's stats seperate by mech chassis, and even by weapons equipped. With that, I can calculate weights for everything, from a player's performance to the mech chassis selected to the loadout. Failing to have all of that, the data is lower quality as will be the predictions from it.

What I'll try to do is use one season to generate weights for stats we have, like KDR, Kills per match, avg match score, and see if I can predict the W/L ratio better than the MM. Remember, the MM always predicts 1, so I want to be on the right side of 1 and closer to the true W/L ratio. What will this accomplish? It will mean I can generate a better number to represent a player's Tier than the current PSR.

This better number can be used in place of the PSR formula, to create better matches.

Tonnage.

It's not an equal or even consistent distribution.

Also the MM is not there to give a 50/50 chance; it's there to test predictions. You really want it to be off by as much as 8% to start, usually though about 4% assuming it has the confidence of 120+ matches per player involved. You want the variance to be off by a bit more than estimated margin of error. At least until you get the players to a point where your confidence in their score is 98-99%, but that's a lot of matches.

Oddly the more I think about it the more I think you sorta DO want tiers - just based of confidence in the players ranking. So the higher the confidence level the tighter you can try to make the balance for matchmaking and still be making accurate adjustments from the results.

The MM *wants* to predict 1, but remember it also has release valves so it can throw a team together within a timeframe.

KDR/Damage/Match Score are associated metrics, not predictors. At the highest tiers of performance you'll very likely see a tight correlation because everyone already has a comparable skillset. Out side of that top few % (probably 7 or 8%, on the skill curve that's the point where it turns into more of a cliff than a curve) W/L will have way less viable relation to other metrics.

Remember, a matchmaker that only works for comp tier players isn't going to work outside of comp queue.

However I'm a big fan of crunching numbers just to get newly crunched numbers. Refining telemetry gives you more useful numbers and different data points to then take on to new stuff. For example, I would love to know if on average KDR is more important or less important than damage for w/l. Also how high someone can get in w/l while consistently keeping less than a 1.5 KDR. Conversely how good can someones KDR get while still keeping barely a 1.2 or less w/l?

There's a lot I'd like to get out of that sort of data.