I made this: https://counterplay.xyz/lane/mid

not sure how useful it will actually be, but it's a fun project. I'm currently waiting for riot to give my app the OK so I can scan faster and get a much bigger samplesize.

Very nice. I enjoyed the GTO solution to blind picking an ad carry. Always pick Ashe.

One stat I like to have for solo queue pick and ban is basically pick rate multiplied by win rate over 50%. So I'd ban out 52% win rate Irelia with a 10% pick rate which would be a score of 20 over the 55% win rate Galio with a 1% win rate which would be a score of 5. To take it a step further you'd have to consider the delta in win rate between that and the next pick available. e.g. Spending a ban on 52.44% Sivir kind of sucks if it just pushes them to 52.43% Jinx. But a ban on 54.56% Wukong that gets them on 51.95% Riven seems good.

Another interesting subject is finding the champions underrated by win rate. e.g. Gragas is an unimpressive 50% win rate, but is drawing a huge ban rate. I suspect it would turn out that the good performances on that champion are far more impactful than a good Maokai jungle game which has a slightly higher win rate, but entirely unpopular. While you can see this in ban rates to a degree, it just tells you what picks the community feels that way about, and they can definitely be wrong.

You might want to add some coloring to the percentages of matchups.

added coloring to matchup tables.

i added (role winrate - 50) * (role pickrate) to the bans page as a "OP EV" and used it to sort the "consider banning these champs" table on the bans page (one motivation for this site was i wanted one page to load if i wanted to know what to ban and i hadn't played league in a whiel)

some thoughts on "OP EV"
- i used role winrate and role pickrate to compute the stat, because global winrate is usually diluted compared to role winrate. however, i'm not entirely happy with this, because this discounts the effect of "double whammy" bans (banning annie bans her from mid lane and support lane).
- "OP EV" has an inverse relationship with ban. because banned characters aren't picked, there will always be someone at the top of "OP EV" list that isn't banned. this actually makes it a great way to figure out 'maybe i should ban this person instead', but it makes it a bad way to figure out how this person stacks up against the currently accepted bans.

so, maybe i can fix it by making it (winrate - 50) * (pick + ban). unfortunately, i can't add "role pick" and "ban" safely. so i may just cheat and do ("role pick" * 2 + ban). or maybe make the whole thing global? i don't know. i'll play around with some of the other options later i guess.

maybe i'm not thinking clearly and it just needs to all be global. i just like role based stats better :/

great work so far

Annie is a particularly interesting one, because she has the number one winrate mid, but a sub-50% winrate as a support. So banning Annie out actually prevents the bad supports on the other team from taking her. Which lessons the value of the ban for me. Whereas Hecarim has above 50% winrate in multiple roles so it becomes a better ban.

I would go global with the ban stuff, solo queue is not a situation where you know the other team is going to Annie mid (which you don't want) or Annie support (which you probably do), you have to consider the champ as a whole by the averages. The role rates are useful in picking counter matchups which is a lot of the rest of the charts.

Rather than doing straight pick + ban or 2*pick + ban, I would just calculate an estimated pick rate. e.g. Sej is picked 7.3%, banned 80.33%. So I'd call that 7.3/(100-80.33) = 37.13% pick rate if not banned.

thanks karak

riot approved my site! so i'm now allowed to make requests at 300 times / second instead of 0.8 times / second. of course, i'm not really equipped to scan at that scale (parallelism, concurrency, ..) so that will be some work to figure out.

good call on the estimated pick rate. i'll try that next.

if you guys have any other ideas for analysis that can be done, post em here. i feel like teamcomp is one area that needs work. and i'll eventually have to make per-champion pages, but i don't want to just recreate champion.gg.

maybe how often the team with most counterpicks wins the game?

if that is possible

maybe win% by gold lead

and win% by cs lead

i mean that the stats show a couple time intervals like 10-20-30 minutes if possible

I've spent an entire four seconds thinking about this, so pardon that it's not fleshed out at all. Is there a way to show positive or negative impact of having Champion A on a team with Champion B? Like, Ori/Malph combo or like, Teemo and Eve (no tanks on team, basically)?

Team with most cc win% would be very interesting too

Definitely, champion.gg has bot lane synergy stats, so you can see while normally Ashe has a winrate of 55%, if you pair her with support Lux you will drop to 48%. But while there are only 414 bot lane combos with standard ADC and supports, there are 15.5k combos of two champions together, 2 million combos of three champions together, 232 million combos of four champions together, and 28 billion combos of full teams.

I got 50 active threads working, so I can scan about 3 games per second now (259k/day). on the downside, still a lot of the games I'm finding now are 5.10 games, and I don't know how long the stat crunching will take (champion.gg's faq says it takes them 15 hours near the end of the patch)

found a guy in plat who played ranked against 2 silvers, wtf: http://www.lolking.net/summoner/na/2...hes/1852577278

The Silver II Warwick was 8-3 just out of placements with an MMR of around 1450. The Silver I Thresh was duo with the Gold V Zed.

The Platinum V Fiddlesticks has a similar MMR to the Silver II Warwick about 1450. Looking at his ranked match history, he's gone 18-36 (33% winrate) over the last month to drop from around 1700 MMR to where he is now.

pushed an update with 25k 5.11 games, kinda a small samplesize. shyvana top lane not doing that well, but jungle is doing surprisingly well.

also, check out the bans page (and read the tooltips)

The bans page currently rates a champion as a must ban the further they are from 50%. So a champion with a 40% winrate is more of a must ban than a 55% winrate champion.

yeah, but only if they are bad and popular. the idea being you don't want someone on your own team to pick the popular, bad champ. also, the 'inverse bans' are discounted to 80% of the (distance * estimated pickrate)

I don't get it, there are more players on the other team that can pick the bad champ, so leaving him open should always hurt the other teams win percentage more than it does yours

Are you literally not seeing any Olaf games in these polling samples?

This.

Plus if you're like me and going through pick and ban with the option to dodge, you can always dodge out if your team ends up with the particularly bad popular champion.

not sure if I buy these arguments against inverse bans. if i plan to go top lane and ban out leblanc before picking my toplaner, my team has 4 left to pick and theirs has 5. yet, each of us only have 1 mid laner on our team, so the chance that either team picks leblanc is mostly the same. the only time this matters is when you would inverse ban a champion that plays in your lane. (this makes me think my 80% discount factor is wrong, too)

mitigating factor: you have the names of your teammates and can look up their tendencies or talk to them.

for the dodge option: what's different between banning out OPs and dodging if your team picks UPs vs banning out UPs and dodging if their team picks OPs? is it the perception that the OPs are picked more often? but that should be accounted for in the "op/up_ev"...

however, I agree that inverse bans don't make a lot of intuitive sense, so maybe there's something there that hasn't been articulated yet. i don't know. i moved them down to the bottom of the default sort.

here's something else that's weird: a lot of the inverse bans my system is recommending are considered 'op' by the community (ryze, gragas, possibly LB). what is this all about? is this the 'gragas open who want' effect where newcomers flock to the flavor of the month champ and play it badly, driving down its winrate?
compare the experience distribution between ahri and gragas:
http://champion.gg/champion/Gragas/Jungle
http://champion.gg/champion/Ahri/Middle

something to think about.

right now i'm filtering the tables to "must have above 1% role pickrate". i will eventually have a way to search for champions..

Wow, so Olaf looks like literally the least played champion in the game then? I don't notice any other exclusions on your list aside from him.

here are the least played: http://i.imgur.com/jhfUO9U.png
olaf's 2 percent match pick rate splits into (0.4% jungle / 0.6% top) lane+role pickrate, so he doesn't make the 1% cut. lane+role pickrate is half the match pick rate because it's team based.

so, I am pretty damn sure I have scanned every 5.11 NA plat+ game but I only have 98k games. At the time champion.gg was reporting over 1 million games scanned. I know he scans other regions, but still, that doesn't make sense. I sent him a message on reddit and he changed the header text from "Matches Analyzed" to "Champions Analyzed" (aka, his number was off by a factor of 10!)