Thats really cool man. I've actually thought about trying to do something similar myself (just general data mining). I'm not sure if I have the time to contribute right now, but would you accept coding help if offered?
I absolutely would! We've got a google group of contributors I can add you to if you can send me your email address (or check the site's about section, my email is on there).
Mmh nice work but Undyings winrate I just looked at is never going over 46,5-47%, most of the time it's lower but his winrate is actually better (according to Dotabuff it's 48%). It might be due to old games but even all-time statistics he is at 48%, some patches ago it was 47,34% (6.78). In the gametime up to 45 minutes it's not even at 44% on average in your statistics. That's why I'm not sure about how correct the rest of these statistics are. It might be for some reason correct for these 229.161 matches but it would be strange. Nethertheless, the graphics seem to go the right way in general.
Ah crap, I just realized I linked directly to the team builder instead of the new winrate stats:
I feel like I saw a thread about this awhile ago, if that was you nice to see the progress!
When u pick cancer theres no KOTL suggestion tho.
nice webapp.
where did get the stars for heroes ? I noticed they are slightly different to those in the dota2 game.
I like what you have done with this idea. The time-data graphs are great. If I fill out fully green bars on all categories in the team designer section the team will be a solid team.
You may want to go over some of the stars assigned to heroes again. For instance Pudge is listed as a jungler. It's only 1 star but I don't think he deserves even 1. On the other hand Batrider, one of the very best junglers has only 1 star. Or SF who has no initiating abilities (unless you count a blink-bkb ult doing a minor slow...but you really should not do this to start the fight someone needs to stun first)...anyway SF has 2 stars for initiator, but Slardar who's primary role is initiator and has an aoe stun has only 1 star.
I can't find anyone who, at 1:15:00, is higher than Anti-Mage. That's crazy to me.
Where are you accessing all of this data anyway? I'm a graduate student getting a masters in Statistics and Industrial Psychology, and if you ever need any assistance in any capacity I'd love to help. Do you utilize anything like SPSS, SAS, or R for these sorts of analyses?
@Crossfiyah
it's not crazy at all; he can push lanes very quickly, and most heroes caught alone will be blinked upon, abyssaled, right clicked by him and 2 images, and mana voided for a relatively quick death
His short cd blink also gives high mobility so it's hard to kill him without someone stunning him first and a bunch of heroes whacking him...so he can get away with more risky plays late game...unlike PA/Troll/etc.
What did seem strange is Drow Ranger's trend, having 80% win rate at 15 minutes in and slowly converging to where Riki's win rate is by 40 minutes. So at least for pubs, this suggests pushing a bunch of towers and farming your items in the first 20 min or so for a quick win is arguably a more valid method than playing her like a hard carry and dragging it out. I'm not sure what bracket these matches are played in but I assume even normal/high bracket players could get away with not knowing what to do and still win with Drow easily.
Drow snowballs like crazy with a little early game boost. Fed Drow is game over very quickly usually. Aura bonus to all ranged heroes too on team. That slow too. Most noobs don't realise that early game your job is to shut down drow as she is squishy and not let her get any last hits.
this is awesome, you should note which hero is for what color in the timedata graph
but we love it :D
I'm not surprised that drow loses late game at all. Drow stops scaling well after level 16. And ultimately she is a single target carry with low mobility. Drow can never match the top hard carries in team dps output. Put equal late game items on a Luna, Gyro, Medusa and they output 3x more dps because off the multi target factor even though drows "attack dmg" number will be higher.
Look at the Medusa graph. There is never a game long enough for Medusa to max out. Medusa is still getting stronger at 75 min.
It's also cuz in games with non-retards, someone will almost always be within 400 aoe during teamfights unless the Drow player is excellent at positioning. I'm guilty of ab(using) drow and the only way to catch those heroes (Luna/Gyro/Medusa/etc.) off guard and outcarry them is to position super well, something the vast majority of the player base cannot do. If you're out of position for 2 seconds you're pretty much dead (whereas Medusa can just click r and live through teamfights).
I wasn't surprised by Drow levelling off but the unrivaled 80% win rate early on suggests the remake of Marksmanship from a constant +agi ability to one that is predicated on your positioning is boosting her early game presence by a huge factor. +60 agi at level 11 is a free butterfly minus evasion. It's almost as if the AOE caveat doesn't apply in most games, which from my experience is the case (luckily for me).
PL drop pretty hard after 1 hour. Why is that ? i mean not a lot of carries can outcarry PL.
My guess is after an hour, even non-carries are starting to get enough gold to be dangerous (as a whole or even individually in some cases). Once roles 3-5 start getting tp boots, necrobook 3, shivas, sheep stick, etc, PL (or any carry) has more to worry about than just the enemy carry(s).
PL drops for much the same reason that Ara mentioned for drow and for the same reason naga drops off. Pubs are bad. PL never needs to take risks and can out split push any other hero late game for the win. But this requires patience and micro skills and map awareness that pubs just don't have.
I consider every PL death to be a failure. He is a carry that should be impossible to attack because he should be impossible to find with proper image micro. The only time you should know where the real PL is... is when you are purged and about to die in less than 2 seconds while your team is retreating. Since images can take towers and rax at zero risk, PL should never lose late game. But people fail to execute it, mostly they get impatient or want to dive for kills. Trying to get kills instead of just sending images to towers is how PL losses. He wins the game with zero kills, but its not fun for many people so they don't do it.
PL does not need kills, let your supports get the kills when you can.
My idea of excellent PL games (my average KDA is over 10, mostly because of nearly zero deaths)
http://dotabuff.com/matches/148751674 __________ 0-0-10 plenty of assists #1 tower dmg, at no risk
http://dotabuff.com/matches/212465670 ___________ 4-0-22 tons of assists #1 tower dmg, at no risk
http://dotabuff.com/matches/135305859 ___________ 5-1-15 tons of assists #1 tower dmg, only 1 death
http://dotabuff.com/matches/77819699 ___________ 6-0-8 maybe too many kills, #1 tower dmg, zero risk
http://dotabuff.com/matches/53398345 ___________ 7-0-9 maybe too many kills, #1 tower dmg, zero risk
http://dotabuff.com/matches/46772558 ___________ 5-0-12 plenty of assists, #1 tower dmg, zero risk
Mistakes were made!
http://dotabuff.com/matches/46752420 ___________ 12-3-13, not #1 tower dmg, 3 deaths too many
http://dotabuff.com/matches/39792564 ___________ 10-4-22, plenty of assists, #1 tower dmg, 4 deaths too many
If PL never dies and keeps images always going he is going to win the game. I don't play PL much because its boring. Its like a solved puzzle. I'm only going to lose if my team feeds or I play like crap.
^Relentless. you play PL as a pusher. Thats cool.
In pub game poeple expect u to carry them when u pick/random pl. you do need to get kills and ks.
PL is fun like the first 3 times when you go WOW! This guy is amazing! Then you realize every game is exactly the fucking same if you play correctly.
Будь ласка, увійдіть до системи для написання коментарів.
http://dotateam.me/
I tried to make it as easy to use as possible. Next goals are to take a look at farm dependency and kill dependency.