Took me longer to return to this series than I expected. I think from here, this series will move along faster. I think my confidence in actually creating a useful metric is waning a bit. I think, perhaps, that I'm considering too many factors for the metric. Although, in the end, if I don't like the results, I can play with the factors until I do.
With that in mind, I think I need to make the bill hit number and State hits factor have a larger effect. To accomplish this, I need to lower the minimum for each. I'll be lowering the bill hit number minimum to 0.45 and the State hit minimum to 0.70. Using that change, here's how the coolness stacks up for my hits in the first three weeks of the year:
| Bill Hit # | Hit Location | Coolness | # |
|---|---|---|---|
| 3rd | OR | 46.69 | x1 |
| 2nd | Italy | 26.83 | x1 |
| 3rd | CA | 24.98 | x1 |
| 2nd | NV | 21.49 | x1 |
| 1st | MI | 16.82 | x1 |
| 1st | NJ | 16.25 | x2 |
| 1st | VA | 15.90 | x1 |
| 1st | PA | 15.82 | x1 |
| 1st | FL | 15.01 | x2 |
| 1st | AZ | 14.96 | x2 |
| 1st | WA | 14.63 | x2 |
| 1st | TX | 14.42 | x2 |
| 1st | HI | 13.60 | x4 |
| 1st | NV | 13.42 | x1 |
| 2nd | CA | 6.24 | x10 |
| 1st | CA | 0.00 | x55 |
The only hits to switch spots in coolness rank are the Florida and Arizona hits, and they switched because I now have more hits in Arizona than Florida.
After "rebalancing" the bill hit number and State hit factors, its time to add a third factor. The most logical place to go from here is county hit (after county, my plan is to add city and then zip code). The big question to answer here is which should be the bigger factor: State hit or county hit? The way I look at this is "would a hit in a new county in California be close in coolness to a hit in a State with just a few hits?" My gut reaction is "no", and as such, the State factor will be the larger one. As a starting point, I'll leave the minimum for State at 0.70 and make county's minimum 0.95. This makes the overall county/State combo minimum 0.665. I think that's ok for now, but this might need to be shifted when city and zip are added.
After running the program using the new factors, I think the most interesting data points are those for the first hit on bills in States other than CA. This data can be seen in the below table. The parathesis numbers indicate hits in county and hits in State.
| Bill Hit # | Hit Location | Coolness | # |
|---|---|---|---|
| 1st | Kalkaska MI (1:8) | 18.69 | x1 |
| 1st | Mercer NJ (1:13) | 18.13 | x1 |
| 1st | Essex NJ (3:13) | 17.71 | x1 |
| 1st | Prince William VA (2:17) | 17.46 | x1 |
| 1st | York PA (2:18) | 17.38 | x1 |
| 1st | Volusia FL (1:31) | 16.92 | x1 |
| 1st | Miami-Dade FL (6:31) | 16.36 | x1 |
| 1st | Cameron TX (1:44) | 16.34 | x1 |
| 1st | Pima AZ (7:32) | 16.27 | x2 |
| 1st | Pierce WA (2:39) | 16.22 | x1 |
| 1st | Travis TX (3:44) | 15.93 | x1 |
| 1st | King WA (21:39) | 15.66 | x1 |
| 1st | Kauai HI (16:67) | 14.75 | x1 |
| 1st | Clark NV (23:73) | 14.47 | x1 |
| 1st | Honolulu HI (45:67) | 14.41 | x3 |
Looking closely at the "out of order" coolness factors based on State hits, the first hit in Cameron county Texas is cooler than the 7th hit in Pima Arizona and the 2nd in Pierce Washington. Also the third hit in Travis Texas is cooler than the 21st in King Washington. Finally the 23rd hit in Clark Nevada is cooler than the 46th in Honolulu Hawaii.
I can buy into each of those. However, I'm not sure that I'd agree that the 7th hit in Pima Arizona would be cooler than the 2nd in Pierce Washinston. To fix this, I adjusted the exponential part of the county hit coolness equations. Meaning that the equation started as:
And then was adjusted to:
Recalculating for the hits above brought about this new coolness order:
| Bill Hit # | Hit Location | Coolness | # |
|---|---|---|---|
| 1st | Kalkaska MI (1:8) | 18.69 | x1 |
| 1st | Mercer NJ (1:13) | 18.13 | x1 |
| 1st | Essex NJ (3:13) | 17.20 | x1 |
| 1st | Prince William VA (2:17) | 16.98 | x1 |
| 1st | Volusia FL (1:31) | 16.92 | x1 |
| 1st | York PA (2:18) | 16.91 | x1 |
| 1st | Cameron TX (1:44) | 16.34 | x1 |
| 1st | Miami-Dade FL (6:31) | 15.83 | x1 |
| 1st | Pierce WA (2:39) | 15.76 | x1 |
| 1st | Pima AZ (7:32) | 15.74 | x2 |
| 1st | Travis TX (3:44) | 15.43 | x1 |
| 1st | King WA (21:39) | 15.11 | x1 |
| 1st | Kauai HI (16:67) | 14.21 | x1 |
| 1st | Clark NV (23:73) | 13.93 | x1 |
| 1st | Honolulu HI (45:67) | 13.89 | x3 |
Ok, so the Pierce Washington hit is now slightly cooler than the Pima Arizona hit. Good. Additionally, the Volusia Florida hit passed the York Pennsyvania hit slightly (I'm ok with this) and the Cameron Texas hit passed the Miami-Dade Florida hit (I'm also ok with this). So far so good.
To get some additional data points, I ran the program for all hits over the past three weeks. There were some more "interesting" hits in this time span, so I'll provide the table for all of the data except for the California first and second hits:
| Bill Hit # | Hit Location | Coolness | # |
|---|---|---|---|
| 3rd | Italy | 56.40 | x1 |
| 4th | Contra Costa CA (75:2222) | 54.03 | x1 |
| 3rd | Multnomah OR (11:53) | 45.97 | x1 |
| 2nd | Shelby TN (1:3) | 28.12 | x1 |
| 2nd | Saint Martin LA (1:4) | 27.81 | x1 |
| 2nd | Richmond GA (1:12) | 26.50 | x1 |
| 2nd | Chaffee CO (1:19) | 25.84 | x1 |
| 2nd | Weber UT (2:11) | 25.69 | x1 |
| 2nd | Catawba NC (2:15) | 25.27 | x1 |
| 2nd | Mason IL (1:32) | 24.97 | x1 |
| 3rd | Fresno CA (49:2222) | 24.66 | x1 |
| 2nd | Suffolk MA (5:30) | 23.89 | x1 |
| 2nd | Pima AZ (7:32) | 23.67 | x1 |
| 1st | Grant NM (1:7) | 18.83 | x1 |
| 1st | Dona Ana NM (1:7) | 18.83 | x1 |
| 1st | Washington AR (1:9) | 18.56 | x1 |
| 1st | Bonner ID (1:9) | 18.56 | x1 |
| 1st | Salt Lake UT (5:11) | 17.26 | x2 |
| 1st | Columbia FL (1:31) | 16.92 | x1 |
| 1st | Boulder CO (2:19) | 16.84 | x1 |
| 1st | Fulton NY (1:40) | 16.51 | x1 |
| 1st | Will IL (2:32) | 16.08 | x1 |
| 1st | Maricopa AZ (23:32) | 15.40 | x1 |
| 1st | Kauai HI (16:67) | 14.21 | x1 |
| 1st | Clark NV (23:73) | 13.93 | x1 |
| 1st | Honolulu HI (45:67) | 13.89 | x1 |
There are a few interesting hits of note in this data. The 3rd hit in Italy is cooler than the 4th hit in California but the CA hit is still cooler than the 3rd hit in Oregon. This is in line with my thinking that a 3rd hit is cooler than a 4th only if the 3rd hit State has 1/1000 of the hits of the 4th hit State. The 2nd hit in Weber UT being less cool than the 2nd hit in Chaffee CO is questionable but acceptable. The 2nd hit in Mason IL being cooler than the 3rd hit in Fresno CA breaks the 1/100 ratio for 2nd to 3rd hits, but with the Mason IL being the only hit in the county, that seems right. The Salt Lake UT hit being cooler than the Columbia FL hit is also questionable, but I think the order is correct.
Three factors in, I think I'm off to a good start. I wouldn't be surprised if I have to revisit some of the magnitudes along the way. Things will start to get interesting once I get past the rest of the "hit location" factors.
No comments:
Post a Comment