the idea is great, also the database holds a looooooooooot of data.
but the website itself sucks sooooooooo hard.
that its nearly impossible to get to the right match without
typing "check" ingame.
normally i end up, counting up the numbers till i find the matches i am looking for
by bruteforce.
I wanted to know the win rates of alien-marines for the NSL Invitational
You're walking down a road that won't make you happy. You can't get statistically useful win-rates from a tournament with only around 100 matches.
Hmmm... nice sample size, even distribution of maps, all competitive teams with no glaringly huge skill gaps (ex. Archaea playing a brand new team, etc).
100 samples gives you a margin of error of about 10%, i.e. if the results were 50% marine wins then the "true" marine win rate could be anywhere between 40% and 60%.
I wanted to know the win rates of alien-marines for the NSL Invitational
You're walking down a road that won't make you happy. You can't get statistically useful win-rates from a tournament with only around 100 matches.
Hmmm... nice sample size, even distribution of maps, all competitive teams with no glaringly huge skill gaps (ex. Archaea playing a brand new team, etc).
Ya, you're right this data is worthless.
In no way is ~100 matches a "nice" sample size statistically speaking. Even if you include every match in Season 1, you're still not approaching a good sample size for balance.
Not every draw was alien ties. Both of Damage Networks draws were marine ties (we used our onii very poorly).
I would be curious to see these broken out less for balance and more for following the "sport" like stats on the backs of baseball cards, it would be interesting to see player and team stats like average time to first fade or wins with first upgrade choice... I am not sure what value they provide but then again I am not sure what value ERA technically provides in baseball... really just a fun way to compare effectiveness of teams and players. It would be cool if NS2Stats could provide a suite of reports following these tourneys. Then when WasabiOne is shout casting we can have stupid irrelevant statistics to talk about just like in baseball "This guy has a batting average of .338 when batting against left handed pitchers in the 7th inning of games in august on Tuesdays"
@Golden I watched your Dn vs OP. Solid marine play from both teams (until OP were neglecting jetpacks). I think it was you bilebombing the researching proto + AA? *facepalmed when it was pulled off* Top 2 qualify right? gratz and goodluck.
100 trials is decent but not useless. I wouldn't want to include season 1 matches though. 200 trials would reduce a good amount of error.
i dont really see how stats will prove anything because w/l can change when:
1. teams are not even skilled (like 70% of all matches srsly)
2. maps are not balanced for both sides
3. ping (not a big issue but still)
A couple of (important) points which cause the Invitational qualifiers' matches to not represent balance well:
1. A lot, lot, lot of the matches were a default win due to a forfeit, and thus all those games were 50-50 marine-alien wins, which skews things in favor of the marines (with the assumption that aliens are statistically stronger).
2. All groups had both strong teams and new, weaker teams, causing huge skill gaps between teams in some matches (whereas, for example, the skill differences in thr ENSL regular season divisions are much smaller), and thus again one team winning both marines and aliens in an uneven match tips the balance towards 50% with no justification.
As anecdotal evidence goes, out of the 8 rounds we actually played (not counting forfeit victories, which were half of all of our games), 7 rounds went to the aliens. I'm pretty sure the overall statistics would be in favor of the aliens still, though I cannot check this without going through every single match which I'm not going to do.
Comments
the idea is great, also the database holds a looooooooooot of data.
but the website itself sucks sooooooooo hard.
that its nearly impossible to get to the right match without
typing "check" ingame.
normally i end up, counting up the numbers till i find the matches i am looking for
by bruteforce.
that makes me so mad
You're walking down a road that won't make you happy. You can't get statistically useful win-rates from a tournament with only around 100 matches.
Hmmm... nice sample size, even distribution of maps, all competitive teams with no glaringly huge skill gaps (ex. Archaea playing a brand new team, etc).
Ya, you're right this data is worthless.
In no way is ~100 matches a "nice" sample size statistically speaking. Even if you include every match in Season 1, you're still not approaching a good sample size for balance.
Not every draw was alien ties. Both of Damage Networks draws were marine ties (we used our onii very poorly).
100 trials is decent but not useless. I wouldn't want to include season 1 matches though. 200 trials would reduce a good amount of error.
1. teams are not even skilled (like 70% of all matches srsly)
2. maps are not balanced for both sides
3. ping (not a big issue but still)
1. A lot, lot, lot of the matches were a default win due to a forfeit, and thus all those games were 50-50 marine-alien wins, which skews things in favor of the marines (with the assumption that aliens are statistically stronger).
2. All groups had both strong teams and new, weaker teams, causing huge skill gaps between teams in some matches (whereas, for example, the skill differences in thr ENSL regular season divisions are much smaller), and thus again one team winning both marines and aliens in an uneven match tips the balance towards 50% with no justification.
As anecdotal evidence goes, out of the 8 rounds we actually played (not counting forfeit victories, which were half of all of our games), 7 rounds went to the aliens. I'm pretty sure the overall statistics would be in favor of the aliens still, though I cannot check this without going through every single match which I'm not going to do.