349 replies to this topic

[ForceUser]

Ollollo, on 02 March 2018 - 11:53 AM, said:

ForcedUser, thanks for the detailed and enlightning reply!

I'm sorry I wasn't very clear in my post. What I meant to ask was this: It seems my scores are too high! My overall score is higher than any of the seasons that make up the score. How can that be?

For instance, the overall score for the Lars Laser account linked above is 92 % but I have never scored above 90 % in any of the seasons.

There's a lot of reasons and they all kind of add to the disparity. It's because of how the algorithm takes all the various components and creates the overall ranking and how you compare to other people and how it treats older seasons' stats.

I think another big reason is that who is active tends to change from season to season. You might have a core of people who play a couple hundred matches every single season and that have relatively consistent results but there might be a relatively large number of people that might play a lot one month and then not so much the next, or have wider range of performance. Good players sometimes goes inactive so they get taken out of the overall ranking. So although in one season you ranked lower because a bunch of better players than you placed higher, since they are now considered inactive it doesn't change THAT seasons numbers but it does change the overall since it's based on the stats of current active players.

Other things like doing well consistently, hitting the 500 match mark (it's part of his algorithm), having the last 500 matches be better than your older matches, the weight distribution of the mechs you play and probably a lot of more obscure things (I'm not a statistician nor mathematician) all contribute to the disparity.

You can see the same thing with my overall ranking being higher than any one season.

Edited by ForceUser, 02 March 2018 - 12:27 PM.

[Grus]

it just sucks if you've had a few bad seasons way back in the day (talking single digit seasons) sandbagging you it effects your overall big time.

[Scurro]

Grus, on 02 March 2018 - 12:45 PM, said:

it just sucks if you've had a few bad seasons way back in the day (talking single digit seasons) sandbagging you it effects your overall big time.

Season weights help solve this problem. Each new season has a higher impact on changing your score. This can be both good and bad.

[Jackal Noble]

Scurro, on 02 March 2018 - 12:48 PM, said:

Season weights help solve this problem. Each new season has a higher impact on changing your score. This can be both good and bad.

How is it you are retired?

Grus, on 02 March 2018 - 12:45 PM, said:

it just sucks if you've had a few bad seasons way back in the day (talking single digit seasons) sandbagging you it effects your overall big time.

that and having ~7k games with about 6K of those being solo QP. My biggest irk is the group queue try-hards that think they are hot stuff. In reality, you are standing on the wings of others.

[Scurro]

JackalBeast, on 02 March 2018 - 12:56 PM, said:

How is it you are retired?

Was busy for a few months and played more than 10 games the 28th of February. Turns out someone at PGI messed up the timing for the season 20 and it ended the evening of the 28th and all my games counted for the next month

[Grus]

Scurro, on 02 March 2018 - 12:48 PM, said:

Season weights help solve this problem. Each new season has a higher impact on changing your score. This can be both good and bad.

[color=#000000]


The Jarl's List - Search Results
[/color]
[color=#000000]

Home Compare players Isengrim Leaderboard

Click headers to sort. Stats are updated on the first of the month.
Current Season: 20 Overview for: Grus Season Rank Percentile Games Played W/L Ratio Survive Rate K/D Ratio Average Score Adjusted Score Progress Light Medium Heavy Assault Overall: 7906 85% 1838 1.03 39% 1.47 241 249.4 22% 18% 22% 25% 33% Season 5 31759 10% 46 0.70 24% 0.63 130 121.3 4% 13% 15% 67% Season 6 24805 40% 70 0.79 37% 0.84 169 164.8 36% 20% 0% 47% 23% Season 7 19428 49% 74 0.92 39% 1.20 192 181.8 10% 0% 14% 66% 18% Season 8 18828 46% 134 0.68 34% 1.16 186 176.2 -3% 0% 61% 7% 30% Season 9 16405 51% 140 0.74 31% 0.99 195 184.7 5% 0% 31% 34% 28% Season 10 13376 60% 158 0.67 27% 1.18 210 201.3 9% 11% 28% 32% 28% Season 11 12187 64% 82 1.10 45% 1.73 240 231.3 15% 18% 0% 32% 43% Season 12 7820 75% 133 1.15 42% 1.75 268 255.8 11% 13% 21% 22% 44% Season 13 8411 75% 155 0.94 33% 1.31 270 257.2 1% 12% 0% 35% 47% Season 14 3530 89% 114 1.43 46% 1.66 303 293.2 14% 11% 58% 15% 16% Season 15 4442 85% 114 1.40 46% 2.03 293 282.3 -4% 11% 45% 21% 23% Season 16 13748 52% 101 1.46 43% 1.69 238 225.2 -20% 0% 29% 35% 30% Season 17 7162 74% 170 1.25 51% 1.99 247 258.2 15% 71% 0% 15% 13% Season 18 8214 77% 165 1.13 36% 1.63 276 262.6 2% 16% 15% 0% 66% Season 19 10202 70% 98 1.09 47% 1.73 248 246.4 -6% 41% 0% 19% 39% Season 20 3005 90% 84 1.80 45% 1.96 296 293.4 19% 32% 13% 39%
15% [/color]

the first 9 seasons hurt.

[Scurro]

Grus, on 02 March 2018 - 01:49 PM, said:

the first 9 seasons hurt.

Hurt might be too strong of a word.

You are still 85% percentile for overall even though your average percentile for your first 9 seasons was 52%

[Jackal Noble]

Scurro, on 02 March 2018 - 02:26 PM, said:

Hurt might be too strong of a word.

You are still 85% percentile for overall even though your average percentile for your first 9 seasons was 52%

exactly. try having your first 4 seasons (prolly more tbh) on a potatotop.

Edited by JackalBeast, 02 March 2018 - 02:46 PM.

[Grus]

Scurro, on 02 March 2018 - 02:26 PM, said:

Hurt might be too strong of a word.

You are still 85% percentile for overall even though your average percentile for your first 9 seasons was 52%

so maybe not so much this


more like this

[ocular tb]

One thing I noticed when looking at my own stats is that my first 11 seasons (1-4 were with a different name) were pretty mediocre with scores typically in the mid 200s, but starting with season 12 my scores got noticeably better. This coincides with the new skill tree dropping in May 2017 (season 11) and June being the first full season (12) under the new skill tree. I believe the extra survival points are what made the difference considering that I play lights a lot of the time and I noticed that some lights that had been hard to survive and do well in (Jenner) had now become serviceable and fun again. While the skill tree does have problems I think it's done some good things, especially for lights. Just a thought.

[Y E O N N E]

It did some good things for some Lights. It did very bad things for other Lights (e.g. Locust).

[Jackal Noble]

Yeonne Greene, on 02 March 2018 - 03:35 PM, said:

It did some good things for some Lights. It did very bad things for other Lights (e.g. Locust).

Not as annoying as the commando, but still surprisingly survivable imo.

[Y E O N N E]

JackalBeast, on 02 March 2018 - 03:40 PM, said:

Not as annoying as the commando, but still surprisingly survivable imo.

Not really talking about the survivability so much as how they nerfed the crap out of the weapons on the variants nobody played anyway, e.g. LCT-3S, LCT-1M, LCT-1V.

They even removed the range buff on the LCT-3V. The 3V! The on Light that's probably even worse than the Spider 5V. It's just...WTF?

[Jackal Noble]

Yeonne Greene, on 02 March 2018 - 03:52 PM, said:

Not really talking about the survivability so much as how they nerfed the crap out of the weapons on the variants nobody played anyway, e.g. LCT-3S, LCT-1M, LCT-1V.

They even removed the range buff on the LCT-3V. The 3V! The on Light that's probably even worse than the Spider 5V. It's just...WTF?

Obviously gearing up for the Flea.

[Kubernetes]

JackalBeast, on 02 March 2018 - 02:46 PM, said:

exactly. try having your first 4 seasons (prolly more tbh) on a potatotop.

Playing on decent hardware definitely makes a difference. I can see which season I got a new comp because my average MS went up considerably. Turns out my early performance wasn't indicative of an average pilot, but that of an above average pilot on a really old laptop.

[ForceUser]

JackalBeast, on 02 March 2018 - 12:58 PM, said:

How is it you are retired?



that and having ~7k games with about 6K of those being solo QP. My biggest irk is the group queue try-hards that think they are hot stuff. In reality, you are standing on the wings of others.

Have a gander at my stats on the list. My last season is 100% solo Q and done 100% in an Uziel. I think your irk is misplaced

[O L L O]

Scurro, on 02 March 2018 - 12:27 PM, said:

Maybe I am misunderstanding the statement, but your overall score is lower than multiple seasons you have had, not higher. Are you talking about percentile? Percentile for seasons is ranked among other players that played that season while as your overall percentile is ranked with all active players in the last three months (players who stop playing for three months are dropped out of the ranked pool).

Yes, I meant to say percentile, sorry about the confusion. The percentile values for the last three seasons are lower than the overall value for Ollollo. For my alts Lars Laser and Magnum Bonum, none of the seasons are as high as the overall value.


I just thought it was odd that the mean value is higher than any of its parts (sorry about the inaccurate mathematical terms here), but I realize that there are some mysteries in how percentile works.

Thanks for the reply!

[O L L O]

ForceUser, on 02 March 2018 - 12:27 PM, said:

There's a lot of reasons and they all kind of add to the disparity. It's because of how the algorithm takes all the various components and creates the overall ranking and how you compare to other people and how it treats older seasons' stats.

I think another big reason is that who is active tends to change from season to season. You might have a core of people who play a couple hundred matches every single season and that have relatively consistent results but there might be a relatively large number of people that might play a lot one month and then not so much the next, or have wider range of performance. Good players sometimes goes inactive so they get taken out of the overall ranking. So although in one season you ranked lower because a bunch of better players than you placed higher, since they are now considered inactive it doesn't change THAT seasons numbers but it does change the overall since it's based on the stats of current active players.

Other things like doing well consistently, hitting the 500 match mark (it's part of his algorithm), having the last 500 matches be better than your older matches, the weight distribution of the mechs you play and probably a lot of more obscure things (I'm not a statistician nor mathematician) all contribute to the disparity.

You can see the same thing with my overall ranking being higher than any one season.

Thanks for this insightful post! I think I now have a better grasp of how it works. Cheers!

[Dee Eight]

Kubernetes, on 02 March 2018 - 04:23 PM, said:

Playing on decent hardware definitely makes a difference. I can see which season I got a new comp because my average MS went up considerably. Turns out my early performance wasn't indicative of an average pilot, but that of an above average pilot on a really old laptop.

Same, season 12 I jumped into the mid 200s for average score, which lines up with when I started my computer upgrades. However the past nine months I've also spent significantly more time on my alts than on my main account but that's largely to do with that bankload of HSP leaving me little need to grind new mechs on the main account.

[Grus]

JackalBeast, on 02 March 2018 - 03:40 PM, said:

Not as annoying as the commando, but still surprisingly survivable imo.

SSSHHHHH!!! dont tell anybody!