clarity VS butterfly

General * All data has been acquired from OpenDota, Stratz, and parsing replays using SkadiStats Clarity 2 * The numbers calculated here are based on each player's last 50 pro matches, exponentially weighted to give much more importance to recent performances * I have used 10 Cores, 5 Mids, and 10 Supports as the cut-off for what constitutes a top scoring player * I kept in the statistics for all players in addition to just the top players for your reference

Then you can create features yourself using various api (like opendota) or look for parsing github (i never used parsing tool myself, but i saved the link to this one : https://github.com/skadistats/clarity , no idea how good it is). Since smurf / cheaters are supposed to be a small fraction of all player, you'll end up with unbalanced dataset, so you'll prolly have to do semi supervised machine learning or at the very least supervised with weight.

So I actually use a couple different APIs. The reason for this is that I couldn't find a fantasy-related endpoint for the offical Dota API, and the match endpoint doesn't have all the data required for calculating fantasy points. To actually get all the required data, the replay needs to be downloaded and parsed (using tools like Clarity or Butterfly). OpenDota does this, and exposes all the relevant data in their match API endpoint.

The replays you download are that "raw data". You can parse it with something like Clarity.

So I actually use a couple different APIs. The reason for this is that I couldn't find a fantasy-related endpoint for the offical Dota API, and the match endpoint doesn't have all the data required for calculating fantasy points. To actually get all the required data, the replay needs to be downloaded and parsed (using tools like Clarity or Butterfly). OpenDota does this, and exposes all the relevant data in their match API endpoint.