News Shop
Events Forums

More Yomi Data


Your Oni’s solid tho. Hone the blade, CG. Hone the blade.


I’m just trying to get on the Mad King/Migo Quince level.


I’m pretty surprised by how highly rated my Quince is. And not at all surprised that the model says I should probably just mono-Setsuki. (High skill Gwen though, yay?)


I think it’s more telling you that your Setsuki is good-not-great, and that your Quince is actually better (though less sure about it). I I were looking at that I’d think you should mono-Quince my man.


I agree. Quince is definitely your top character. The question is really whether it’s enough dominant to make up for any particularly bad matchups (or maybe you just have a pocket Gwen/Setsuki for that).


Keep in mind: these charts can only show you where you are and where you have been; they cannot tell you where you should be going.


Sure, maybe I am overrating Setsuki and underrating Quince as pure characters - I was assuming that the difference in power levels is more than my skill difference can make up, but maybe I’m wrong!


Narrator voice: He was not, in fact, wrong.

I’m just saying that there might be a reason that there is no Fivec-but-for-Quince. I’m not saying that improving your Quince is not worth it.


I’m actually actively trying to convert the skill charts onto a y-axis that is more comprehensible (in particular, it’s going to be “win-chance against the top skill player/character combo, assuming even Elo and a 5-5 matchup” (i.e. converting the current [0, -4] scale via the inverse logit function)). That will make your relative level with a character more comparable to the matchup numbers.


I guess I just want to make sure the lesson people are learning from this data is not “I should focus exclusively on [weak character that I am the best at]” when “I should invest some more time in [strong character that I have an interested in]” or “I should use the characters I have the most experience with to counterpick my opponents as effectively as I can” are both probably more beneficial takeaways. If your goal is strictly to win sets while enjoying yourself, I mean.


I realized last night before playing @Hydro that I’d created an excellent competitive analysis tool. Namely:

  1. I looked him up
  2. found that his skill tournament picks were Arg and Troq
  3. decided I liked Ven against him
  4. lost soundly for two games
  5. and then switched back to Sets because she’s my crutch character

If I were actually competent with Ven, I probably could have stopped at step 3.


Notice how Vendetta only counterpicks Argagarg when you use the charts where you factor in skill level of players. I don’t suggest anybody who is not a Vendetta master counterpick Argagarg with Ven.


That’s definitely a fair point. I got rolled pretty hard. I’d love to hear your tips on how to play that MU better (in another thread).


That’s an interesting theory. I can’t wait to try it out for myself.


The problem with Yomi is that once you tell people how you win the people playing against you can just play against THAT. So would you rather I PMed you my thoughts or made a public topic?


I think I’m more interested in more high level play in the world than I am in winning particular MUs due to secret tech, so I’m ok with public. A project I’ve been thinking about is trying to capture MU specific advice in Yomibase, so having more specific MU thoughts is good.


By the way VengfulPickle, is everyone with tournament records in the analysis tools database, or just the people above?


Everyone is in the model. I’ve only generated images for the people above, which is all people that got to at least 1700 Elo, or played between Jan 2018 and whenever I last pulled the dataset from @mysticjuicer’s spreadsheet. I think you might have played a handful of games since I did the dump, which could be why you aren’t in the list above (assuming that was your question).


I didn’t expect to be in it. Where is the complete chart?

#140 has all of the historical Elo data. has all the historical match data that the Elo data is based on.

The rest of this is all just on my local machine, right now. Although I have pushed older versions of it up to Github.