Who is ProfessorFBR???

Watson wrote:So... who is the mystery ranker???

fourfourtwo wrote:

Watson wrote:So... who is the mystery ranker???

I vote Hobbit. Or Watson. Damn sho isn't M - he's too arrogant to be anonymous.

And correction - bwgophers is the grandson of FBR, not the father.

deepthoughts wrote:I don't think it matters much who the Professor really is. It will simply come down to what his or her rankings look like. FBR should usually succeed in predicting outcomes between teams ranked 7 - 8+ slots apart from each other. FBR does not really predict much between head-to-head matchups of #1 vs #2 or #14 vs #15.

The suspicions of fixing rankings will never go away because most of that "noise" is usually between teams ranked very close to each other, something bwgophers and 02dad have both agreed is beyond the capability of the software.

deepthoughts wrote:
FBR can be played around with by a team as well, not just by the person doing the ranking. All one has to do is play competition ranked far above your capability, avoid games against weaker competition, and use many guest players to win a game or two against tough competition. You will vault up the rankings. Aggressive recruitment of guests will get you some results and FBR only looks at results.

deepthoughts wrote:
All this adds up to FBR being imperfect but that's OK -- because it offers the soccer forums its version of the crazy parent reality TV entertainment. It also means that the good Professor can hide in anonymity and watch the show. If the rankings are not right, it will become obvious to all when the outcomes no longer match the 7-8 apart slot rule that FBR does predict well. Stop all the worry about our new mystery man and lets enjoy the sideshow.

fourfourtwo wrote:

deepthoughts wrote:I don't think it matters much who the Professor really is. It will simply come down to what his or her rankings look like. FBR should usually succeed in predicting outcomes between teams ranked 7 - 8+ slots apart from each other. FBR does not really predict much between head-to-head matchups of #1 vs #2 or #14 vs #15.

You'd have to go to vegas before you'll find any system, human or otherwise, that can reliably predict the outcome between two closely matched teams. There is almost nothing at any price that will be better at ranking NTX girls soccer than FBR combined with a group of solid, knowledgeable human pollsters.

"7-8+ slots apart?" Not sure where you got that from. FBR groups "Tiers" of teams and reliably predicts how the tiers fare against each other. There is also a way to configure FBR such that the points gaps between teams should tell you something about their relative strength, but this "7-8 slots" you've come up with is completely arbitrary and not how FBR works.

If team #11 is on par with team #3 and they are in the same tier with similar point totals, FBR is not telling you that team #3 will always win a game between the two. On the other hand - suppose team #9 is in Tier 1 and Team #10 is in Tier 2, with a massive points gap between the two teams and a clear stratification between the two Tiers. In that case FBR is telling you team #9 is a much stronger team than team #10.

bwgophers wrote:

A little perspective from someone who spent way too much time doing this stuff for the ‘01’s…

NO ranking system is “perfect”. Not FBR, not Power Rankings, not human polls. They all have strengths, weaknesses, and different points of emphasis. Even when you combine all of these into a “BCS” style ranking, it still won’t be perfect.

Personally, I think the “accuracy” of FBR, and any of the other systems for that matter, is somewhere between deepthoughts’ “rusty machete” and his “laser scalpel”. The FBR ranking tiers are the “rusty machete”. However, there is still some further accuracy within the tiers as well.

We had tons of game data from the ‘01’s, and my guess is that game data from the ‘02’s would bear this out as well. What you will see is that IN GENERAL:

1) Games between teams within 5 ranking spots of one another will tend to be very competitive games (2 goal margin or less), and the lower ranked team coming out with a Win is not uncommon.

2) Games between teams ranked 5-10 spots from one another will tend to be competitive (again 2 goal margin or less in most cases), however, the higher ranked team will come out with a Win in a large majority of these games (>70%).

3) It will be very rare that a team ranked >10 spots below another will so much as manage a draw against the higher ranked opponent, and most games will have the higher ranked team winning by a margin >2 goals.

The biggest value of these rankings in my mind is for coaches to find leagues/tournaments that have an appropriate level of competition for their team. Also, for LDs and TDs to set up leagues/tourneys with an appropriate competitive balance. Most logical people don’t see any benefit in a team getting smoked by 4-5 goals every game, or a team smoking everyone else by 4-5 goals/game.

By the way, we did an evaluation of how each of the ’01 ranking systems (FBR, Power Rankings, Human Polls, and consolidated BCS) did at predicting the results of Lake Highlands QT. The results were a total wash. Each system got 8 out 11 1st weekend D1 qualifiers correct. Each system got 17 or 18 out of the 20 D1 teams correct. Each system got 27 or 28 of the 30 D1 + D3 qualifiers correct. So each of the systems were basically 90% accurate.

The bottom line is that just like deepthoughts said... if you are arguing or getting upset that a team is ranked a couple of spots too high or a couple of spots too low, or that Team A is ranked above or below Team B… You are wasting your time and energy.

deepthoughts wrote:
That's where I got my 7-8 slots thinking from, but I may have generalized too much. How many teams fall into a tier clearly matters too. Unfortunately, that is not easily seen or appreciated when you look at the FBR summary when posted.

fourfourtwo wrote:

deepthoughts wrote:
That's where I got my 7-8 slots thinking from, but I may have generalized too much. How many teams fall into a tier clearly matters too. Unfortunately, that is not easily seen or appreciated when you look at the FBR summary when posted.

Gophers was reporting trends he found in '01 data. Those trends held true at one point in time in the '01 build up before QT.

They very well could, but there is nothing in FBR that guarantees those same trends materialize for another age group, and I wonder whether those trends still hold for the top 35 '01 teams for games they've played since QT.

But you're right - it's all about the tiers, and they're a critical factor in whether FBR is configured effectively. The number of teams in a tier is based on results. Could be 4 teams in a tier, could be 14.

The ideal weekly FBR release IMO would post a few key pieces of information in addition to the final rankings:
1) the rules for how teams enter a tier and get relegated
2) the listing/ordering of teams within each tier
3) which teams have changed tiers as a result of last week's games
4) the tier matrix which shows how well the tiers are doing against each other

That takes extra time, and when you add in the inevitable line of questions/complaints at the front door each week...it's not for the faint of heart. Someone would have to genuinely enjoy this stuff to keep doing it week in, week out.

bwgophers wrote:By the way, interesting trend I spotted...

I went through my archives and counted 25 '01 teams that played in multiple leagues last Spring, including at least 50% of the consensus top 10 and top 25 teams.

By my count, in the '02's this Spring, there are 10 teams total playing in multiple leagues, with only 1 top 10 (FCD Renfro) and 4 Top 25 (FCD Renfro, TFC Elite, Solar Jones, and DT Vigil).

Just found the discrepancy kind of curious.