Quick OPR Question

[Ether]

Originally Posted by Ether

There are at least three ways you could deal with this...

Eugene Fang and I have been PM'ing about this. He came up with a clever and very effective method. See his discussion here.

[Ed Law]

Originally Posted by Ether

Eugene Fang and I have been PM'ing about this. He came up with a clever and very effective method. See his discussion here.

The method looks good but I am not sure if it is necessary. Here is my position on how to use OPR. Current event OPR should not be used until you have at least 4 or 5 matches depending on number of teams. Before that, I use max OPR from prior events. So in a way, it is similar to what Eugene Fang did. The match "prediction" using OPR is just a check. Since you are at the event, you should always use scouted data instead of OPR.
The work Eugene Fang did was great in showing how long it takes to converge.

[Caleb Sykes]

Originally Posted by Ether

Eugene Fang and I have been PM'ing about this. He came up with a clever and very effective method. See his discussion here.

Are the xOPR and ixOPR calculations only possible if every attending team has competed at least once? If so, that severely limits the usefulness of this effort, because this criterion is not met for a large majority of events.

[Caleb Sykes]

Originally Posted by Caleb Sykes

Are the xOPR and ixOPR calculations only possible if every attending team has competed at least once? If so, that severely limits the usefulness of this effort, because this criterion is not met for a large majority of events.

Obviously xOPR and ixOPR are possible without this constraint, since Eugene posted convergence graphs for Palmetto. What is the "other source for OPR initialization" that is used for cases like this?

[Ether]

Originally Posted by Caleb Sykes

Obviously xOPR and ixOPR are possible without this constraint, since Eugene posted convergence graphs for Palmetto. What is the "other source for OPR initialization" that is used for cases like this?

I don't know what Eugene used.

I haven't done any numerical experiments yet to confirm this, but if there is no prior OPR value for a given team at an event, I think you can just pick a reasonable number to seed the calculation, and let the 2x iteration take care of things from there.

[Ether]

Here is the basis/motivation for the ixOPR algorithm. Note how many matches must be played before the OPR stabilizes.

[Eugene Fang]

Originally Posted by Caleb Sykes

Obviously xOPR and ixOPR are possible without this constraint, since Eugene posted convergence graphs for Palmetto. What is the "other source for OPR initialization" that is used for cases like this?

Originally Posted by Ether

I think you can just pick a reasonable number to seed the calculation, and let the 2x iteration take care of things from there.

What Ether said shares my view on the initialization.

In the graphs, I used 0 as the initialization, which I knew was a bad idea at the time. I've been trying out using the average match score for the event (up to the currently played match) as the initialization for teams without past events with better* results.

* There's a lot of "fuzzy math" going on here and I know OPR is a pretty poor metric for estimating a team's performance, whether there are few matches played or many matches played. I'm just curious to see how accurate one can expect match predictions to be using no other source of scouting data.