First up, killing the Supreme Court. Again. But still with numbers and statistics, because that's the best way to do things. Assume the Senate decides to stop being dumb. Then, Merrick Garland gets a hearing and since he's basically fine, he gets a seat on the Supreme Court. Since my least favorite justice is dead.
Wednesday, 30 March 2016
Sunday, 27 March 2016
Final final four
Today's also the last day I update the sports stuff for this year. Here's the table for the rest of the tournament:
#Bracket | N_R1 | PP_R1 | Nwrong_R1 | P_R1 | S_R1 | N_R2 | PP_R2 | Nwrong_R1 | P_R2 | S_R2 |
Mine | 32 | 1 | 6 | 26 | .995 | 16 | 2 | 4 | 50 | .998 |
Heart-of-the-cards | 32 | 1 | 10 | 22 | .656 | 16 | 2 | 8 | 38 | .320 |
Julie | 32 | 1 | 10 | 22 | .656 | 16 | 2 | 6 | 42 | .738 |
BHO | 32 | 1 | 9 | 23 | .823 | 16 | 2 | 6 | 43 | .820 |
538 | 32 | 1 | 8 | 24 | .928 | 16 | 2 | 7 | 42 | .738 |
Rank | 32 | 1 | 13 | 19 | .129 | 16 | 2 | 6 | 39 | .424 |
#Bracket | N_R3 | PP_R3 | Nwrong_R3 | P_R3 | S_R3 | N_R4 | PP_R4 | Nwrong_R4 | P_R4 |
Mine | 8 | 4 | 3 | 70 | .998903 | 4 | 8 | 3 | 78 |
Heart-of-the-cards | 8 | 4 | 6 | 46 | .044 | 4 | 8 | 4 | 46 |
Julie | 8 | 4 | 4 | 58 | .610 | 4 | 8 | 4 | 58 |
BHO | 8 | 4 | 4 | 59 | .674 | 4 | 8 | 3 | 67 |
538 | 8 | 4 | 2 | 66 | .955 | 4 | 8 | 2 | 82 |
Rank | 8 | 4 | 2 | 63 | .875 | 4 | 8 | 3 | 71 |
#Bracket | N_R3 | PP_R3 | Nwrong_R3 | P_R3 | N_R4 | PP_R4 | Nwrong_R4 | P_R4 |
Mine | 2 | 16 | 2 | 78 | 1 | 32 | 1 | 78 |
Heart-of-the-cards | 2 | 16 | 2 | 46 | 1 | 32 | 1 | 46 |
Julie | 2 | 16 | 2 | 58 | 1 | 32 | 1 | 58 |
BHO | 2 | 16 | 1+ | 67+ | 1 | 32 | 1 | 67+ |
538 | 2 | 16 | 2 | 82 | 1 | 32 | 1 | 82 |
Rank | 2 | 16 | 2 | 71 | 1 | 32 | 1 | 71 |
If the President gets his pick correct in the next round, then he'll win with an 83. Otherwise, 538 wins based on only getting two wrong in round 4. Everything else is locked in now, so there's nothing really to update anymore.
Friday, 25 March 2016
Round 3
Since it's the weekend, it's sports time. First up, my picks for this round of things:
Texas A&M:
This now has the added columns of S_RX. These are my simulated CDF values based on the Yahoo selection pick fractions given for each team. This is another piece of kind-of garbage code that I threw together earlier in the week. I think it's doing everything correctly, but I don't see any simulated results that get a total score above 83, and yahoo does list some in their leader list. Maybe 1e6 simulations isn't sufficient to fully probe things? Maybe I'm truncating or rounding something odd? The main idea behind this calculation is to see how well a given set of picks should rank.
One that I was doomed to get wrong. |
And the other doomed one. But a new mistake! |
Texas A&M:
29.687500 14.062500 3.125000 3 6 3 Texas A&M
28.125000 18.750000 21.875000 3 8 2 Oklahoma
First up, I think my analysis notes have been wrong on the previous posts. The file I'm pulling these numbers from is in 2016/2015/2014/group/game/rank/name format, not 2014/2015/2016 format. This changes the analysis for some of my previous mistakes, but I'm too lazy to go correct those. In any case, using this new, correct information, it looks like I thought (from the 2016 ratings) that Texas A&M should be slightly better than Oklahoma. Folding in previous years could have potentially altered that choice.
I was thinking a bit about adding some score-based information in as well. The idea being that each team scores a given median number of points across all their games, and have a given median number of points scored against them. By comparing how well a given score ranks in all their games, and against their opponent's, it should be possible to construct offense and defense ratings. This might be useful to say, "Team X is generally better, but they only are a +1 in offense, and they're playing a +4 defense, so they might not win." The other benefit would be to add two new metrics, which could then be used across the full multi-year dual-gender score set to determine which relative weights each should be assigned to a more complete prediction model.
I think the first step that I should do, though, is to dump all of that data into a database, instead of using horrible fixed-width formatted files to manage things. That's largely a consequence of not really caring a lot about the project.
In any case, here's the comparison table for round three:
#Bracket | N_R1 | PP_R1 | Nwrong_R1 | P_R1 | S_R1 | N_R2 | PP_R2 | Nwrong_R1 | P_R2 | S_R2 |
Mine | 32 | 1 | 6 | 26 | .995 | 16 | 2 | 4 | 50 | .998 |
Heart-of-the-cards | 32 | 1 | 10 | 22 | .656 | 16 | 2 | 8 | 38 | .320 |
Julie | 32 | 1 | 10 | 22 | .656 | 16 | 2 | 6 | 42 | .738 |
BHO | 32 | 1 | 9 | 23 | .823 | 16 | 2 | 6 | 43 | .820 |
538 | 32 | 1 | 8 | 24 | .928 | 16 | 2 | 7 | 42 | .738 |
Rank | 32 | 1 | 13 | 19 | .129 | 16 | 2 | 6 | 39 | .424 |
#Bracket | N_R3 | PP_R3 | Nwrong_R3 | P_R3 | S_R3 | N_R4 | PP_R4 | Nwrong_R4 | P_R4 | S_R4 |
Mine | 8 | 4 | 3 | 70 | .998903 | 4 | 8 | |||
Heart-of-the-cards | 8 | 4 | 6 | 46 | .044 | 4 | 8 | |||
Julie | 8 | 4 | 4 | 58 | .610 | 4 | 8 | |||
BHO | 8 | 4 | 4 | 59 | .674 | 4 | 8 | |||
538 | 8 | 4 | 2 | 66 | .955 | 4 | 8 | |||
Rank | 8 | 4 | 2 | 63 | .875 | 4 | 8 | |||
This now has the added columns of S_RX. These are my simulated CDF values based on the Yahoo selection pick fractions given for each team. This is another piece of kind-of garbage code that I threw together earlier in the week. I think it's doing everything correctly, but I don't see any simulated results that get a total score above 83, and yahoo does list some in their leader list. Maybe 1e6 simulations isn't sufficient to fully probe things? Maybe I'm truncating or rounding something odd? The main idea behind this calculation is to see how well a given set of picks should rank.
Sunday, 20 March 2016
Round 2
today was the end of round two of the sports thing. I also need to go back and update posts with the new label I've decided is probably useful, "sports". So I updated everything before the final game was over, and then had to double check nothing went wrong:
Again, three of my four mistakes this time around were caused by my winning choice being eliminated in the previous round. For the last one:
Xavier:
12.500000 42.187500 29.687500 2 7 7 Wisconsin
Why did I choose Xavier? Did I get confused and use the 2014 rankings instead of the 2016 ones? This looks like me being dumb. Maybe I took the #2 ranking too seriously? I should probably write down logic notes next time, so I can point to the error directly.
What does the scoring comparison look like?
Again, three of my four mistakes this time around were caused by my winning choice being eliminated in the previous round. For the last one:
Xavier:
12.500000 42.187500 29.687500 2 7 7 Wisconsin
34.375000 12.500000 14.062500 2 8 2 Xavier
What does the scoring comparison look like?
#Bracket | N_R1 | PP_R1 | Nwrong_R1 | P_R1 | N_R2 | PP_R2 | Nwrong_R1 | P_R2 |
Mine | 32 | 1 | 6 | 26 | 16 | 2 | 4 | 50 |
Heart-of-the-cards | 32 | 1 | 10 | 22 | 16 | 2 | 8 | 38 |
Julie | 32 | 1 | 10 | 22 | 16 | 2 | 6 | 42 |
BHO | 32 | 1 | 9 | 23 | 16 | 2 | 6 | 43 |
538 | 32 | 1 | 8 | 24 | 16 | 2 | 7 | 42 |
Rank | 32 | 1 | 13 | 19 | 16 | 2 | 6 | 39 |
Again the "rank" method is garbage, and shouldn't be used. Nate Silver had a tweet earlier about how this is apparently because it's based on RPI too much. Looking at wikipedia, it looks like RPI is an incomplete version of my LAM method. ¯\_(ツ)_/¯ This also shows the point where HotC totally falls apart, becoming the worst method. Everyone else is pretty well clumped together. I'm a bit surprised that 538 isn't doing better, given the "we included scores, and at-home values, and distances to the games, and the number of cats each player owns, and the SAT scores of each player."
This also makes me think I should have actually entered my selections into some pool. Maybe I should hone the method a bit more, and see how it works over a few more years. Or, alternatively, I could do the reasonably easy thing and apply the method to the historical data, and see if this consistently matches reality. Maybe next weekend, since I think it's a long one. This will also make me fix my master Makefile to put things into logical directories, and not just dump the outputs into a common directory.
Friday, 18 March 2016
Round one
Statistics results.
Ok, that West Virginia loss is going to hit the later rounds. |
As is Purdue. Not as bad as Michigan State, obviously. |
Let's look at the comparison table:
#Bracket | N_R1 | PP_R1 | Nwrong_R1 | P_R1 |
Mine | 32 | 1 | 6 | 26 |
Heart-of-the-cards | 32 | 1 | 10 | 22 |
Julie | 32 | 1 | 10 | 22 |
BHO | 32 | 1 | 9 | 23 |
538 | 32 | 1 | 8 | 24 |
Rank | 32 | 1 | 13 | 19 |
The columns are the bracket identifier, the number of games in the round, the points per correct selection in the round, the number wrong, and the total points. The brackets are mine above, the "Heart of the Cards" bracket taken by simply selecting teams based on the 2016 ranking I calculated, Julie's bracket, President Obama's, the 538 bracket taken by assuming constant composite rankings from their pre-tournament predictions, and a dummy bracket constructed by selecting teams based solely on their "sport rank" thing. That's actually working out a lot better than I expected. I was correct in shaking up the straight HotC numbers with a bit of historical data. Looking at the mistakes:
Arizona:
26.562500 43.750000 40.625000 1 5 6 Arizona
25.000000 37.500000 53.125000 5 1 11 Wichita St
I didn't believe the numbers, given the #11 ranking. From above, I should ignore the ranking in the future, because it's pretty crappy. The problem is that my numbers suggest that Wichita State is the best team in the entire thing, which doesn't seem like it's right.
West Virginia:
28.125000 21.875000 1.562500 2 6 3 West Virginia
34.375000 39.062500 45.312500 2 6 14 SF Austin
Ditto. My numbers predict that SF Austin is the second best team. I guess if either of them come out winning, I can say that I predicted it, and then tossed it in the trash.
Baylor:
17.187500 23.437500 20.312500 3 3 5 Baylor
25.000000 18.750000 7.812500 3 3 12 Yale
No clue, but it sounds like everyone was surprised by this one.
Purdue:
29.687500 14.062500 -3.125000 4 3 5 Purdue
37.500000 -7.812500 -3.125000 4 3 12 Ark Little Rock
My numbers say they both suck, so I went with last year's numbers to break the tie. I could have added in the 2014 values, but this was a #12 ranking, and I didn't believe those.
Dayton:
28.125000 26.562500 20.312500 4 7 7 Dayton
9.375000 7.812500 34.375000 4 7 10 Syracuse
This one I should have gotten right. I folded the two previous years in, and that said that I should trust consistency over a sudden jump. Maybe Syracuse has some new great player.
Michigan State:
35.937500 18.750000 28.125000 4 8 2 Michigan St
23.437500 3.125000 23.437500 4 8 15 MTSU
Again, this one seemed like it was a surprise to everyone. There are only three values of my ranking between these two values, so that kind of suggests they're within ~5% of each other in terms of skill. Oh well.
Tuesday, 15 March 2016
I didn't really do any of the improvements I discussed two years ago.
Basketball
Basically my solution this time was:
Basically my solution this time was:
- Check with Julie that no team went undefeated this year. That was my big problem last time.
- Run the 2016, 2015, and 2014 game solutions to determine the relative rankings for all of the teams in each of those years.
- Rank things based on the 2016 solution, letting 2015 solutions break ties. Also use this information (and the 2014) for solutions to:
- Anytime a #12 sport-ranked team is ranked substantially above a 1-4 sport-ranked team, assume something is off with the model, because I have a lot of #12 ranked teams ranked really high for some reason.
So the result table is:
#2014.score 2015.score 2016.score group game sport-rank 2016.score Team.name
23.437500 28.125000 40.625000 1 1 1 40.625000 Kansas
-9.375000 -21.875000 1.562500 1 1 16 1.562500 Austin Peay
18.750000 -3.125000 17.187500 1 2 8 17.187500 Colorado
29.687500 7.812500 20.312500 1 2 9 20.312500 Connecticut
3.125000 32.812500 26.562500 1 3 5 26.562500 Maryland
9.375000 20.312500 29.687500 1 3 12 29.687500 S Dakota St
10.937500 4.687500 20.312500 1 4 4 20.312500 California
14.062500 14.062500 34.375000 1 4 13 34.375000 Hawaii
40.625000 43.750000 26.562500 1 5 6 26.562500 Arizona
NAN NAN NAN 1 5 11 NAN VAN/WICH
1.562500 18.750000 28.125000 1 6 3 28.125000 Miami FL
14.062500 21.875000 9.375000 1 6 14 9.375000 Buffalo
12.500000 15.625000 17.187500 1 7 7 17.187500 Iowa
-20.312500 23.437500 15.625000 1 7 10 15.625000 Temple
37.500000 46.875000 37.500000 1 8 2 37.500000 Villanova
3.125000 -1.562500 17.187500 1 8 15 17.187500 UNC Asheville
21.875000 20.312500 34.375000 2 1 1 34.375000 North Carolina
NAN NAN NAN 2 1 16 NAN FGCU/FDU
-15.625000 -12.500000 14.062500 2 2 8 14.062500 USC
18.750000 17.187500 20.312500 2 2 9 20.312500 Providence
3.125000 10.937500 28.125000 2 3 5 28.125000 Indiana
4.687500 18.750000 37.500000 2 3 12 37.500000 Chattanooga
20.312500 53.125000 26.562500 2 4 4 26.562500 Kentucky
18.750000 17.187500 31.250000 2 4 13 31.250000 Stony Brook
-3.125000 37.500000 15.625000 2 5 6 15.625000 Notre Dame
NAN NAN NAN 2 5 11 NAN MICH/TULSA
1.562500 21.875000 28.125000 2 6 3 28.125000 West Virginia
45.312500 39.062500 34.375000 2 6 14 34.375000 SF Austin
29.687500 42.187500 12.500000 2 7 7 12.500000 Wisconsin
25.000000 6.250000 15.625000 2 7 10 15.625000 Pittsburgh
14.062500 12.500000 34.375000 2 8 2 34.375000 Xavier
12.500000 -6.250000 26.562500 2 8 15 26.562500 Weber St
21.875000 25.000000 34.375000 3 1 1 34.375000 Oregon
NAN NAN NAN 3 1 16 NAN HC/SOUTH
23.437500 -7.812500 29.687500 3 2 8 29.687500 St Joseph's PA
32.812500 18.750000 18.750000 3 2 9 18.750000 Cincinnati
20.312500 23.437500 17.187500 3 3 5 17.187500 Baylor
7.812500 18.750000 25.000000 3 3 12 25.000000 Yale
28.125000 40.625000 20.312500 3 4 4 20.312500 Duke
-21.875000 6.250000 28.125000 3 4 13 28.125000 UNC Wilmington
20.312500 10.937500 12.500000 3 5 6 12.500000 Texas
1.562500 42.187500 15.625000 3 5 11 15.625000 Northern Iowa
3.125000 14.062500 29.687500 3 6 3 29.687500 Texas A&M
26.562500 23.437500 17.187500 3 6 14 17.187500 WI Green Bay
0.000000 4.687500 10.937500 3 7 7 10.937500 Oregon St
28.125000 26.562500 23.437500 3 7 10 23.437500 VA Commonwealth
21.875000 18.750000 28.125000 3 8 2 28.125000 Oklahoma
-9.375000 -7.812500 25.000000 3 8 15 25.000000 CS Bakersfield
34.375000 40.625000 29.687500 4 1 1 29.687500 Virginia
7.812500 -1.562500 17.187500 4 1 16 17.187500 Hampton
-6.250000 -9.375000 10.937500 4 2 8 10.937500 Texas Tech
-4.687500 18.750000 17.187500 4 2 9 17.187500 Butler
-3.125000 14.062500 29.687500 4 3 5 29.687500 Purdue
-3.125000 -7.812500 37.500000 4 3 12 37.500000 Ark Little Rock
29.687500 26.562500 15.625000 4 4 4 15.625000 Iowa St
17.187500 26.562500 18.750000 4 4 13 18.750000 Iona
0.000000 1.562500 26.562500 4 5 6 26.562500 Seton Hall
34.375000 46.875000 29.687500 4 5 11 29.687500 Gonzaga
14.062500 25.000000 28.125000 4 6 3 28.125000 Utah
4.687500 -3.125000 25.000000 4 6 14 25.000000 Fresno St
20.312500 26.562500 28.125000 4 7 7 28.125000 Dayton
34.375000 7.812500 9.375000 4 7 10 9.375000 Syracuse
28.125000 18.750000 35.937500 4 8 2 35.937500 Michigan St
23.437500 3.125000 23.437500 4 8 15 23.437500 MTSU
-1.562500 10.937500 9.375000 5 1 11 9.375000 Vanderbilt
53.125000 37.500000 25.000000 5 1 11 25.000000 Wichita St
14.062500 17.187500 10.937500 5 2 16 10.937500 FL Gulf Coast
-17.187500 -20.312500 6.250000 5 2 16 6.250000 F Dickinson
26.562500 0.000000 15.625000 5 3 11 15.625000 Michigan
14.062500 18.750000 14.062500 5 3 11 14.062500 Tulsa
9.375000 -3.125000 -7.812500 5 4 16 -7.812500 Holy Cross
9.375000 1.562500 15.625000 5 4 16 15.625000 Southern Univ
So, using this, I can answer the following questions I saw while doing the research of "figuring out what FLGU means".
- I saw a thing asking if Holy Cross was underrated. My analysis says "no," and concludes with a "holy crap, no."
- Kansas is probably going to win it all.
- Julie was right, Michigan State should have been ranked higher than Virginia.
- I've already scored the two group 5 games that have played correctly.
Using the espn clicky thing to use these rules (I bent rule #4 to also apply to 13-ranked teams as well):
Group 1 and 2. |
Group 3 and 4. |
Final stuff. I don't know how to call the score thing. They're separated by ~4 points in the scores, or about 10%. So maybe 10 points, since basketball is a "log10(score) ~ 2" kind of game? |
For the remaining pre-game things, I have Michigan and Southern University winning those (in addition to the correctly called Wichita State and Florida Gulf Coast).
Subscribe to:
Posts (Atom)