Комментарии:
No comments 15 mins bro fell off
ОтветитьThat PokeHive data is pretty sick
ОтветитьPokehive data is cool. I think a comparison between those and trainerhill data would be good to show general trends rather than strict numbers
ОтветитьI'm personally a huge fan of the Pokehive data. Trainer Hill is also a great resource, but I get really granular with my parameters (using the last 10-14 days to include the most recent innovations, 100+ player tournaments, and top 50/25% placements). I'd be curious to see if there are any major differences between the data, and if that tells us anything about matchup spreads or techs. Love these videos!!
ОтветитьI'm gonna dissent from the other comments and say you should absolutely use TrainerHill data if the Pokehive data is that thin. We're looking at only a few games per matchup. We know that in all but the most extreme matchups, the true distribution is somewhere between 40/60 and 60/40, but if we only have 5 games per matchup, we'll get a ton of times where a matchup shows up as 80/20 or more extreme, or where a matchup that's actually favored one way shows up as 40/60 the other way. The percentages are going to swing wildly week-to-week, and that random variation totally overwhelms the signal when the true data is so close together.
Compare the Drago chart for Dortmund + Joinville to the same chart for Louisville, for example. Same format, pretty similar lists, so I'd expect few changes in the results. On Pokehive, 2 weeks earlier, Drago had a 28% win rate into BoltPon; this weekend, it was 52%. It had been 53% into Lugia; this weekend it was 25%. It went from 58% into Snorlax Stall to 42%. It went from 61% into Dragapult to 83%. From 56% into Terapagos to 72%. For the most part, it's not list changes causing this; it's just random variation. Charts with 15 or fewer matches going into the data for each deck pair just won't be refined enough to get the true distribution.
If you flipped a fair coin 15 times, you'd get a 7-8 split 39% of the time; the other 61% of the time, you've come away with data that says heads or tails is a 60/40 favorite or greater. Over 30% of the time, you come away saying it's 67-33 or more extreme. This is something we know is 50/50, and our data just doesn't have the power to tell us that.
useful and helpful as always, PokeHive data looks interesting, why not report both.
Ответитьpokehive data is very cool and nice to look at. thank you for these videos, you're amazing!
Ответить