Wrestlestat’s mysterious algorithm

[Tigerfan]

I was curious how Elam was ranked so low after his lone loss to Surber, so I used the compare feature to see the prediction for matchups between him and everyone ranked higher. Sure enough, the comparison feature predicted an Elam win over every other 197. That internal incongruity seems problematic. I’m curious if anyone else has looked at other scenarios and found anything similar?

Seems to me if your algorithm predicts that a wrestler wins every possible match, he should probably be ranked #1, unless there is a nemesis who repeatedly upsets him but is inconsistent and loses to multiple other wrestlers. Just another quirky factoid that is wrestlestat. 

[ionel]

[MPhillips]

Most of us just stopped trying to figure it out. Best of luck.

[1032004]

OK I think I may have found the most mysterious, and it involves 197.

Max Dean still #1 which is silly but seems in line with possibly the biggest problem with their algorithm that has been discussed in the past that they place too much emphasis on prior year's results (sometimes including several years back).

But Silas Allred, 27-7 in his first year as a starter and R12 finisher, is #2.  

The undefeated national champion and 2x finalist comes in all the way down at #5, with everyone above him having at least 3 losses.  Basically as a result of Dean being artificially high, all the guys that beat him this season are also artificially high.   But Nino doesn't get as much boost for beating those guys even though he didn't beat Dean H2H.

[jross]

It had Spencer Lee beating Gable Steveson by a major decision in the comparison match.  The algos are good, not great.

[ionel]

1 hour ago, jross said:

It had Spencer Lee beating Gable Steveson by a major decision in the comparison match.  The algos are good, not great.

Yeah but you can't prove it was wrong.