MP Ratings / Challenge Matches Open Thread


#21

What is the difference between Matchplay Rating and IFPA Rating? My (limited) understanding of them makes me wonder why we’d have both since it seems they essentially do the same thing.


#22

This one is meaningful. IFPA rating is meaningless.

Matchplay.events takes into consideration your actual head to head finishes against other players.

IFPA rating is a simulation based on your finishing position in a tournament. On top of this, it only looks at the 32 players above and below you.


#23

To my eyes, IFPA Rating is decent. Better than Ranking. Probably better than Eff %. My only gripe is that it decays too quickly, and changes too quickly (one result can send you up or down too much).


#24

But yes, good point about it actually looking at head to head matches. Wondering if the other calculations are mostly the same.


#25

I say it’s meaningless because it doesn’t mean anything. It’s just a number.

Against any player who plays in a tournament with you, you get either 1 win, 1 loss, or 1 tie against that player. It didn’t matter if you played them or not. It didn’t matter how many times you played them and what your record was. It also meant that you received more wins by playing in a larger tournament.

This was heavily exploited in 2015 when SuperLeague participants were being awarded hundreds of wins every month against players who were not actively participating in the league. The results were being recorded such that Joe NewYork had played against 200 unique players a month as if he had played 200 different chess matches against them and had won all of them. Even a player who finished in 10th place would have had a 95% win rate at which point Glicko gets extremely confident that you need to be placed a standard deviation or two higher than average players if beating them is so trivial that you can do it 95% of the time.

In addition, you should be able to use rating to calculate the win probability in a head to head match. Using IFPA rating, you cannot do this, because the rating takes into account 0 actual head-to-head matches until challenge matches are factored in.

The “fix” for this was that you only get wins, losses, and ties against up to 64 players in every tournament.


#26

I don’t have the exact list with me, but it’s filtering out:

  • Tournaments from before Jan. 1, 2016
  • Non-direct play tournaments (best game, pingolf etc.)
  • Tournaments marked as test tournaments
  • Tournaments where the start time and the end time are more than 7 days apart
  • Maybe something else, but I forget.

#27

This is my issue! But I do not know how to change End Time. i was not clicking End Tournament at the end of my tournaments for a while.


#28

What matters for you is whether or not you have “claimed” that player. The Timmy Sexton player never got attached to your user account. Because I’m not super smart, it’s impossible for you to claim a player unless there’s an in-progress tournament with that player participating.

There are two options:

  • Have the tournament organizers add your IFPA number to the Timmy Sexton players. This will autoclaim them.
  • Shoot me an email (your report counts) and I will do the claiming for you

Either way you then have to wait until I re-calculate the ratings from scratch to see old tournaments being folded into your rating. I plan on doing those on a weekly basis, but really I’m making that part up as I go along. :slight_smile:


#29

Tsk tsk tsk :slight_smile:

Send me a list of links to your tournaments. Going forward you’ll have to do the clicking quicker. Preferably the same day the tournament ended!


#30

They’re both Glicko calculations so they should be the same. That said, some of the constants are slightly different.

You can see the MP glicko values at the bottom: https://matchplay.events/live/ratings/info

You can see the IFPA glicko values here: https://www.ifpapinball.com/menu/ranking-info/#rating


#31

It’s like you looked at my roadmap document… Soon!


#32

Andreas - for Strikes tournaments where the bottom two get strikes are you including those individual game results?

It’s a wildly popular format and I don’t know about anyone else but since there’s no difference between 1st and 2nd we often have players intentionally draining as soon as they pass the 3rd place player so we can move the tournament forward.

Is that an issue for anyone else running this format?


#33

Group knockout tournaments have special handling. Let’s do the example where you have groups of four with bottom two getting a strike:

  • No result is recorded between 1st and 2nd place
  • No result is recorded between 3rd and 4th place

There are similar affordances made for other variants. E.g. if only 1 strike is awarded, only results between last place and the rest of the field are recorded.

In addition groups of 3 and 4 have a reduced effect on rating changes, just like they did in PARS (you get one guess as to who helped me with the math). For group knockout tournaments this means the reduced effect is determined both by the group size and by the number of strikes awarded.


#34

Awesome . . . because there’s a chance I didn’t record who came in 1st or 2nd accurately even when told who won. And by chance, I mean 100% of the time I randomly just clicked on the two “winners” for that group and moved on :slight_smile:


#35


#36

Is there any effort to get some of the data from Brackelope in there too? I know locally we have quite a few weeklies using Brackelope for their knockout and there is plenty of good data to help build the database.


#37

I reached out a couple of months ago but never heard back. I will work with any tournament app out there. The steps to integrate are pretty simple (generate a data file and put it on the internet and place a link in your app to an MP Ratings submission page).

I think the best way to get other software creators excited is for them to hear from players who want their results submitted.


#38

I ma not sure Issac has put a lot of time on Brackelop of late, let me reach out and talk to him :slight_smile:


#39

There’s now a “compare” button whenever you view a rating so you can compare yourself to that person. Shows all-time win-loss record along with the win probability.

I also added a “compare” to every game when you are playing a tournament so you can compare all players in a four-player group with a single click.

Haven’t added it to tournaments with multiple games per round because I haven’t found a good place to put it yet :frowning:

Example: https://matchplay.events/live/ratings/compare/377,394,119,11408


#40

sounds interesting, though I am not too sure how it works:

we are tied at 49-49 but there is two tables, with 53%/47%, even though it’s 49-49 it isn’t 50/50%? How does the % works since it does not seem directly related W-L score?

Having the two table is a bit confusing

in that example I won 29-22 against Lonnie or the opposite?