#FF.Pickup All-time Records added

[9u-9u-9u]

Most Kills in a Round (Offense)

could we add record slots for most kills on offense for each class?

[KubeDawg]

Yaay now I know just how shitty a pickup player I am. woooo

[ddm999]

Originally Posted by 9u-9u-9u

could we add record slots for most kills on offense for each class?

Well, the only classes really played on offence is Scout and Medic, and they aren't played on defence really. :P

I suppose a defence Medic could sort of mess it up a bit.

[squeek.]

Originally Posted by ddm999

What are the requirements for Average Class Choice Frequency's?

For example, I have played Scout on Offence for 99.81% (up 7.02% of the record), and Engineer on Defence for 90.54% (up 10.08% of the record), yet I don't have these records presumably due to me not playing enough games. :P

I'm also tied with Juice for average kills as Scout, so I'd assume that has a requirement too.

There's no set-in-stone requirement. It uses a Bayesian average to determine the highest record and then displays the real average. For the Bayesian average, I basically just add an arbitrary amount to the total (I'm using a C of 7 for the most part) before calculating the average, which means that the lower the total, the lower the Bayesian average will be. Let's say you play scout for 9 out of 10 pickups (real average: .90; with a C of 7, that gives a Bayesian average of 9/17 or .529), and another person plays scout 40/50 pickups (real average: .80; with a C of 7, that gives a Bayesian average of 40/57 or .702). The 40/50 pickups is then set as the record because we are much more confident that it's a real statistic rather than a fluke that would change if more pickups were played.

Now, given that, for some of these records it might make sense to not use a Bayesian average or to lower the C. Class choice could be one of those, but if it gets lowered too much it could open up the possibility of one person playing like 5 pickups, only playing scout on O, and then disappearing, making the record 100% and therefore impossible to beat.

Originally Posted by 9u-9u-9u

could we add record slots for most kills on offense for each class?

Could do. I'm assuming you'd like to see stats for O demo/soli/engi?

[squeek.]

Made it so that if you hover over the values of the personal best and all-time record on a player's page it will show the Bayesian average.

[aleXtric]

im looking at them now and i see records that have dropped dramatically. like most teamkills "9" for defense. Is it only tracking daily rather than all time?

[squeek.]

Originally Posted by aleXtric

im looking at them now and i see records that have dropped dramatically. like most teamkills "9" for defense. Is it only tracking daily rather than all time?

I broke it temporarily; it was replacing all the records each round with only the stats from that round. Should all be back to normal now.

[9u-9u-9u]

Average Kills (Offense) 5.97

could we make it only calculate average O kills for all offensive classes that are not scout since scout isn't a DM class?


could we make a stat for improvements in average kills or just "greatest improvement"? since new people will start with a lower average so will always have one no matter how much they improve

[squeek.]

Originally Posted by 9u-9u-9u

could we make it only calculate average O kills for all offensive classes that are not scout since scout isn't a DM class?


could we make a stat for improvements in average kills or just "greatest improvement"? since new people will start with a lower average so will always have one no matter how much they improve

I think I'll be able to track changes in records and personal bests once I finish up what I'm doing. Once I get the foundational stuff done, I'll fix up any wonky records (SG upgrade efficiency in a round needs to be reworked as well since many people have 100%) and add new ones.

You're right about average O kills, didn't think about that. Maybe I should just get rid of all the average/most kills by class and switch them to average/most kills (offense) by class and average/most kills (defense) by class.

[squeek.]

I think I got leaderboards for each record working. Let me know if you find anything weird:

http://parser.ffpickup.com/v2/records/

Example:
http://parser.ffpickup.com/v2/record...yclass_demoman

[9u-9u-9u]

Originally Posted by squeek.

Maybe I should just get rid of all the average/most kills by class and switch them to average/most kills (offense) by class and average/most kills (defense) by class.

ah that's even better i like that idea

then we could also see the top DM scout

[ddm999]

Kinda sucks that to get a record you really need to have played a lot of pickups to get the time to overrule the Bayesian average (e.g. I've played 5000 time (seconds?) in pickups (on offence), I've got 99.8% played as Scout but 'cause of I've only played 5000 time, it gets shoved way down to 45.29% :L), but I guess it makes sense. Doubt I'm gonna get Average Class Choice: Offence Scout when the two people in first have 170,000 time played just on offence...

But seen as the highest time played on offence is 2.6 times that, that's not that high, really.

[squeek.]

Originally Posted by ddm999

Kinda sucks that to get a record you really need to have played a lot of pickups to get the time to overrule the Bayesian average (e.g. I've played 5000 time (seconds?) in pickups (on offence), I've got 99.8% played as Scout but 'cause of I've only played 5000 time, it gets shoved way down to 45.29% :L), but I guess it makes sense. Doubt I'm gonna get Average Class Choice: Offence Scout when the two people in first have 170,000 time played just on offence...

But seen as the highest time played on offence is 2.6 times that, that's not that high, really.

You're right, I might need to rethink the Bayesian average. I realized that I'm not even calculating the Bayesian average correctly, so it might work better if I fix that. You're meant to do (global average of time played on offense * global average of scout as offensive class choice + sum of time played as scout) / (global average of time played on offense + total time played on offense), which might change things, and make each Bayesian average proportional to the dataset rather than fairly arbitrary.

[ddm999]

The Bayesian average seems to be unusually low on Engineer records. :P

I've not played a lot of pickup games, but my average SG uptime is reduced from 49% to 20%, meaning it must be about 10%, which can't be right, can it?
Also:

Average SG kills: My record is 20, with average is 8, must be about 4,
Average SG upgrade efficiency: My record is 35%, with average is 28%, must be about 20%,
Average Engineer kills: My record is 30, with average is 12, must be about 6,
Average kills on D: My record is 26, with average is 12, must be about 4?

All these records seem unusually low to say the Bayesian average is meant to make it more, well, average...

How do you get less than 4 kills on defence? Is that even possible?

[squeek.]

Originally Posted by ddm999

The Bayesian average seems to be unusually low on Engineer records. :P

I've not played a lot of pickup games, but my average SG uptime is reduced from 49% to 20%, meaning it must be about 10%, which can't be right, can it?
Also:

Average SG kills: My record is 20, with average is 8, must be about 4,
Average SG upgrade efficiency: My record is 35%, with average is 28%, must be about 20%,
Average Engineer kills: My record is 30, with average is 12, must be about 6,
Average kills on D: My record is 26, with average is 12, must be about 4?

All these records seem unusually low to say the Bayesian average is meant to make it more, well, average...

How do you get less than 4 kills on defence? Is that even possible?

I haven't fixed the records to use a real Bayesian average yet, and I'm not totally sure I will. I lowered the constant of the Class Choice Percent records and am sort of regretting it; Wicked_Clown now has the offensive scout class choice record with 100% chosen in only ~21 rounds of offense played, making it a nearly unbeatable record if Wicked_Clown stops playing at this point. What I'm using the Bayesian average for is not to make it more average, it is to make outliers or unsustainable records not able to be the record holders in average categories. It's meant to make the records more consistent or meaningful, not more average.

You've only played the time equivalent of 5 rounds as engineer. The purpose of sorting by Bayesian average is to make it so even if you have 100% SG uptime in those 5 rounds, the system assumes that record is unsustainable. If the records were sorted purely by the real average and you held the record with 100% SG uptime and 5 pickups played, to keep that record you'd actually be encouraged to stop playing. The record couldn't realistically be improved, and the only thing that playing more would do is drive that record down. With the current system of Bayesian averaging, playing more actually increases the value of the record even if you aren't increasing your real average (in fact, you could even be decreasing it).

The current constant I'm using might be rewarding time played a bit too heavily, but I don't think by very much. Just so you know, the constant (C) I'm using right now is 7. The formula for each record is essentially:

(sum or count of relevant values) / (sum or count of relevant total + C)

I scale the C depending on the scale of the value/total as well, like for things that use sum of time played, I use 7*900 (900 seconds = 15 minutes, or one round).

[Sabbath]

cut that nerd talk out

[WiFiDi]

Originally Posted by Sabbath

cut that nerd talk out

[beees]

Originally Posted by squeek.

I haven't fixed the records to use a real Bayesian average yet, and I'm not totally sure I will. I lowered the constant of the Class Choice Percent records and am sort of regretting it; Wicked_Clown now has the offensive scout class choice record with 100% chosen in only ~21 rounds of offense played, making it a nearly unbeatable record if Wicked_Clown stops playing at this point. What I'm using the Bayesian average for is not to make it more average, it is to make outliers or unsustainable records not able to be the record holders in average categories. It's meant to make the records more consistent or meaningful, not more average.

You've only played the time equivalent of 5 rounds as engineer. The purpose of sorting by Bayesian average is to make it so even if you have 100% SG uptime in those 5 rounds, the system assumes that record is unsustainable. If the records were sorted purely by the real average and you held the record with 100% SG uptime and 5 pickups played, to keep that record you'd actually be encouraged to stop playing. The record couldn't realistically be improved, and the only thing that playing more would do is drive that record down. With the current system of Bayesian averaging, playing more actually increases the value of the record even if you aren't increasing your real average (in fact, you could even be decreasing it).

The current constant I'm using might be rewarding time played a bit too heavily, but I don't think by very much. Just so you know, the constant (C) I'm using right now is 7. The formula for each record is essentially:

(sum or count of relevant values) / (sum or count of relevant total + C)

I scale the C depending on the scale of the value/total as well, like for things that use sum of time played, I use 7*900 (900 seconds = 15 minutes, or one round).

Why aren't you summing an estimated average*C into the numerator
Isn't the formula bavg= (values + approxavg*C)/(C+numval)?

[Bayesian man]

White men have been using the name of our people for centuries in their mathematics without regard to the heritage, culture, or attainments of that which they steal from.

My Great-great-great Grandfather was one of the most respected men of our tribe in SouthEast Bayesia, he united the Ssengubar, Lutgeinkt, and Qw'sayloo clans in an effort to what led up to the last assualt of our race on the invading white man in the battle of Isandlwana.

That was the pinnacle of our battle with the white race and soon ended, because of succumbing in the later skirmishes to superior technology. The naming system your race often chooses to use is an unfortunate tactic in history, as the reason the Bayesians were defeated was the use of "white man's magic" or hence "his mathematic" of which the Bayesian forumulas were used in the artillery bombardments of 1882.

My people were honorable warriors, please respect them and their memory when using our name.

[WiFiDi]

^ i think its a spambot kill it