Smogon Forums

...avoid this problem in your new system! I'm 100% certain you realize how outrageous it is to favor the small samples like this, from a probability theory point of view.

2. You said: "Easily the simplest rating system is the win-loss record, but that has the problem of rewarding players who only play poor players and punishing those who seek out challenging opponents." The system is primitive, and while it may...

...be unwanted to use WIN-LOSS (win percentage, with sample size requirements, of course) as the main ranking system, I think the numbers should at least be included on the ladder screens, which they currently aren't. The reasons are; it offers an unrivaled level of objectivity; the find battle button very strongly(!!!) lessens the issue of poor opponent/great opponent; it is just plain important, giving you...

...actual insight into the player's performance--I can't tell you how many times I've manually looked up another players rank (which includes win-loss, unlike the ladder screens), and been genuinely surprised by how impressive or unimpressive their (large sample) win percentages are compared to the rank they currently hold--after such a check, I feel like I know much more about a player's ability/performance than...

...I could ever hope to retrieve from a score based on any inherently arbitrary ranking system. Why not include WIN-LOSS? It also shows how many battles the score was drawn from.

3. Since all ranking systems are subjective either way, and people disagree on which one is right, why not just use them all at once (incl. win-loss) by making the ladder screen CATEGORIES SORTABLE? (~Choose Your Own Ranking System!~)...

...Just run one at default, and do suspect testing reqs or whatever from that one, no problems using several. Could also split the ladder into several ladders based on number of battles fought to allow more fair comparisons, but now I'm going beyond the promised three points, and I apologize for that.

I'd like to know what you think, but I've learned not to expect anything in these type of situations.

First off, next time try sending a PM. No need to flood my wall with six posts. Second, the system I'm developing is designed to easily facilitate determining reqs for suspect tests, with the idea that reqs are determined by win rate and by # of battles.

So if you have a perfect win record, you may only need 40 battles to get reqs, but if you go 3-1, that number may be more like 80.

The trick here is that, in a perfect world, every player regardless of skill would have a 50% win rate due to how PS does matchmaking. But this isn't super hard to account for.

Bottom line is that I've got this shit covered. As for sorting by other metrics, that's a PS issue, and it's up to Zarel whether he wants to implement it.

Yeah, wanted to PM, didn't find the option. When you say reqs do you mean reqs for voting on banning stuff, or reqs to get on the scoreboard? The latter is what I addressed, since small samples hog the top spots, when favorised over larger samples. I feel that this was a huge problem with the previous ranking system, and I wrote you basically to ask if this is something you're bent on trying to fix...

...with the new system. Should I interpret your statement of having shit covered as confirmation on this?

Second point is that I think W-L in the ladder screens (and not only via /ranksearch) is a much needed feature, so I wouldn't miss the opportunity to suggest it now when the system is being changed. I tried to argue why it is desirable, but this is up to you, I can only hope. You made no comment on this...