League of Legends: The New Thread
- PJ.
-
PJ. Hell in a Cell
- PJ.
- Hell in a Cell
- Hell in a Cell
- Posts: 4601
- Joined: January 5, 2007
- Location: somewhere better than you =*
- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
okay, tell me if this is a comparatively huge waste of time:
a tool that takes a player's champion pool (list ranging from 1 to many champions), and recommends a new champion for that champion pool based on matchup statistics or other statistics (like teamcomp variation)
like suppose you have a midlane champion pool of two champs
they complement each other in a lot of ways (doing well in lane matchups that the other doesn't)
but there are also a lot of match-ups neither champion in the pool covers really well
the program use this info to recommend champions based on how well they round out your pool
eh? eh?- SleepyKrew
-
SleepyKrew he/himSnark Attack
- SleepyKrew
he/him- Snark Attack
- Snark Attack
- Posts: 15746
- Joined: April 27, 2011
- Pronoun: he/him
- Location: quack
- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
yah i know the data's already there and i use it all the time, but i feel like there are certain things i'm doing over and over again that coding would make less of a time sink
and besides, i'm doing it in a very haphazard way, not really crunching the numbers so much as eyeballing them
i suppose this'll just be a personal project- Maestro
-
Maestro They/Them/AnyJack of All Trades
- Maestro
They/Them/Any- Jack of All Trades
- Jack of All Trades
- Posts: 8366
- Joined: April 18, 2012
- Pronoun: They/Them/Any
- Location: Chronically Online (EST)
- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
- PJ.
-
PJ. Hell in a Cell
- PJ.
- Hell in a Cell
- Hell in a Cell
- Posts: 4601
- Joined: January 5, 2007
- Location: somewhere better than you =*
- Cabd
-
Cabd QT Sniper
- Cabd
- QT Sniper
- QT Sniper
- Posts: 15501
- Joined: February 3, 2013
IIRC there's something that does that; was released on reddit like.. a year ago?ShowHave retired for good; Life is too busy to have time or energy for mafia. It was fun~
And then, a Miracle, a Dance Game and a flight of fancy struck, one more game into the abyss
- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
i have the data and am just thinking through functions on it atm.
now, thinking it through, if you just have matchup statistics to make your decision off of, there are two clearly different situations where your champion pool matters: when you pick before your opponent in the same role and when you pick after
when before, you usually pick more defensively, preparing for an adversarial pick by your opponent. This means you want to make a pick that isn't easily neutered (for example, aurelion sol, who suffers against assassins like fizz/diana). One who's worst matchups aren't bad.
when after, you can exploit your knowledge of the opponent's pick, and so can pick more aggressively. This means you make the pick that dominates the enemy's pick the most.
And all the while, the thing that gates your capacity to do these things is your champion pool - the set of champions you can and are willing to play at a level at/above your current mmr. Of course a lot more goes into whether a champion pool is good (synergy w teammates' picks, for example), but with just matchup statistics, these concerns seem core.- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
one straightforward/hyersimple way to profile a champion's safety is to look at their winrate against their worst matchup
if it's really low, that suggests that the pick can be countered really easily by an opponent simply by making that matchup happen - it also suggests that the champion has other weak matchups
if it's really high, that suggests that no matter what champion an opponent picks, the matchup will be pretty even
here are worst matchup winrates for different champs separated by roles: https://docs.google.com/spreadsheets/d/ ... sp=sharing
Outcomes aren't very surprising, but especially for low elo types could inform good decisions. For example, a novice might look at op.gg, see Karthus's high midlane winrate (53.29 right now, lower than only 3 other champs) and conclude he's worth picking whenever possible, failing to acknowledge that the pick can be countered very robustly. Similarly, someone maybe a bit more experienced but still unlearned might look at Aurelion Sol's kit and conclude that he's substantially more readily countered than other midlane picks; in reality, his worst matchup isn't as bad as the average midlaner's worst matchup, probably because his E enables him to roam even when his laning opponent is threatening.
I know that this kind of reasoning isn't foolproof given the data I've used - after all, Sol might be situational for reasons that extend beyond the specific laner he's opposed against. But - well, you get it.- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
The above can be extended to evaluate champion pools under the response case, too. In general, first for a champion pool we find its matchup winrates if the player always selects the best performing champion in that matchup. For example, if the opponent has chosen Orianna and the player's champion pool includes Ahri and Lissandra, we store the winrate for Ahri vs Orianna since Ahri has a higher winrate against Orianna than Lissandra does.
We thus have a set of winrates reflecting the champion pool's performance when we make picks optimally. To measure how well this champion pool covers its bases - ie how flexible it is, how reliably it exploits knowing the opponent's pick, etc - we can find look at the minimum winrate within this set. If it's low (like 47%), then there are some picks the opponent can make that this champion pool just can't effectively answer. If it's high (like 53%), then the converse is true: no matter what champion the opponent picks, you have a counterpick ready to make the game a more likely victory. This is just the champion pool's worst winrates.
However, on the other hand, preparing for the worst might not be the most effective strategy when building your champion pool. Maybe you're willing to tolerate significant disadvantages in some games in exchange for significant advantages in other games. As a more concrete example, maybe your champion pool struggles when the enemy picks Orianna, but wins so reliably otherwise that sculpting your pool to prepare for Orianna actually reduces your overall performance. I can come up with a 'response score' for your champion pool to measure how well your pool compares to others overall, not just in the worst case. We can take the same pooled set of winrates I described above as part of the computation of the worst case score above, and instead of finding the lowest winrate in the set, take an average of the winrates, weighting this mean by champion popularity (so that your winrate against Ahri, for example, matters more than your winrate against Kayle). With a percentile, we can even locate this score on a distribution of possible response scores for all champion pools of the same size as the one input, appropriately contextualizing it.
With these scoring functions ready, it becomes easy to compute instant answers to questions relevant to champion pool building:
- which champion could i add to improve my champion pool the most?
- of all these champions i like, which few of them would it be best to focus on?
- geez why am i typing all of this what a solipsistic fuck
yeahh
all the code's ready now, except the part where i make it usable by other human beings- RadiantCowbells
-
RadiantCowbells He/himSmooth Criminal
- RadiantCowbells
He/him- Smooth Criminal
- Smooth Criminal
- Posts: 70855
- Joined: February 24, 2013
- Pronoun: He/him
- Contact:
- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
- PJ.
-
PJ. Hell in a Cell
- PJ.
- Hell in a Cell
- Hell in a Cell
- Posts: 4601
- Joined: January 5, 2007
- Location: somewhere better than you =*
Spoken like a true one trick.In post 3161, RadiantCowbells wrote:I think you're undervaluing individual talent with a champion and overvaluing average ability with a championSometimes a sandwich is just a sandwich.- RayFrost
-
RayFrost Survivor
- RayFrost
- Survivor
- Survivor
- Posts: 10769
- Joined: August 2, 2009
- Location: Japan
- Contact:
- Untrod Tripod
-
Untrod Tripod Fat and Sassy
- Untrod Tripod
- Fat and Sassy
- Fat and Sassy
- Posts: 11652
- Joined: September 1, 2003
- SleepyKrew
-
SleepyKrew he/himSnark Attack
- SleepyKrew
he/him- Snark Attack
- Snark Attack
- Posts: 15746
- Joined: April 27, 2011
- Pronoun: he/him
- Location: quack
- PJ.
-
PJ. Hell in a Cell
- PJ.
- Hell in a Cell
- Hell in a Cell
- Posts: 4601
- Joined: January 5, 2007
- Location: somewhere better than you =*
In post 3166, SleepyKrew wrote:me except sleepo's internet might be shitty again so maybe I don't play anymoreSometimes a sandwich is just a sandwich.- Maestro
-
Maestro They/Them/AnyJack of All Trades
- Maestro
They/Them/Any- Jack of All Trades
- Jack of All Trades
- Posts: 8366
- Joined: April 18, 2012
- Pronoun: They/Them/Any
- Location: Chronically Online (EST)
- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
- PJ.
-
PJ. Hell in a Cell
- PJ.
- Hell in a Cell
- Hell in a Cell
- Posts: 4601
- Joined: January 5, 2007
- Location: somewhere better than you =*
- RayFrost
-
RayFrost Survivor
- RayFrost
- Survivor
- Survivor
- Posts: 10769
- Joined: August 2, 2009
- Location: Japan
- Contact:
Support: Janna, zyra, lulu, leona, thresh, nautilus, sona, sorakaIn post 3170, Psyche wrote:anyone want to give me their champion pool for a role and get some scores/recommendations generated?
Top: Renekton, nautilus, darius, malphite gragas, pantheon, irelia, kled
Mid: Karthus, anivia, lux, kassadin
Jungle: Amumu, j4
ADC: Lucian, Jhin, Ezrealdon't you feel silly now?- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
In post 3171, PJ. wrote:Ashe, jinx for ad
Warwick, vi, jarvan for jungle
Tryndamere jarvan top lane.
Annie only mid laneSpoiler: ad
Spoiler:
Spoiler:
Spoiler:
Ascending order.
Performance score is normalized for pickrate; deviations from actual winrate result from fact that picking a champion within the pool affects which other champions will be picked.- Psyche
-
Psyche he/theySurvivor
- Psyche
he/they- Survivor
- Survivor
- Posts: 10698
- Joined: April 28, 2011
- Pronoun: he/they
In post 3172, RayFrost wrote:
Support: Janna, zyra, lulu, leona, thresh, nautilus, sona, sorakaIn post 3170, Psyche wrote:anyone want to give me their champion pool for a role and get some scores/recommendations generated?
Top: Renekton, nautilus, darius, malphite gragas, pantheon, irelia, kled
Mid: Karthus, anivia, lux, kassadin
Jungle: Amumu, j4
ADC: Lucian, Jhin, EzrealSpoiler: support
Spoiler:
Spoiler:
Spoiler:
Spoiler:
If anyone has ideas for better or additional measures, I'm game. IMO the most useful part of all this is the last part, where I make recommendations based on worst matchup score. It's the one that really suggests a champion for your pool based on how it rounds the thing out. - Psyche
Copyright © MafiaScum. All rights reserved.
- Psyche
- RayFrost
- PJ.
- Psyche
- Psyche
- Maestro
- PJ.
- SleepyKrew
- Untrod Tripod
- RayFrost
- PJ.
- Psyche
- RadiantCowbells
- Psyche
- Psyche
- Psyche
- Cabd
- PJ.
- Psyche
- Maestro
- Psyche
- SleepyKrew
- Psyche
- PJ.