CFPOOL Rules
2007 marks the pool's twenty-second season. CFPOOL is the oldest tradition in rec.sport.football.college (and its predecessors) and may be one of the older participatory events on USEnet.
The pool is run for fun and "net bragging rights" only; no money is or ever has been involved. It is run as a "community service," on volunteered time. The pool is open to anyone able to send in an entry. Entries from computer prediction programs are especially encouraged.
The pool is a "confidence value" pool. This means that entrants are are asked to pick outright winners of games (without a "point spread"). Varying point values are assigned to each pick in the list (highest at the top, lowest at the bottom); the entrant sorts their predictions to assign point values to each predicted winner. There is a pool each week of the regular season (as long as there are enough games) and a final doubled-score pool for bowl games.
The pool is described in more detail in the following sections of this document.
The regular season of the pool will include games every Saturday, as long as there are enough games. Early or late in the season, when many teams are inactive, games from multiple weekends or multiple days may be combined into a single entry. The last entry of the regular season usually includes games from the Saturday before Thanksgiving through the conference championship games. The final entry for the season includes all of the bowl games.
Entries for the pool usually contain 18 games. They will be posted about a week in advance of the first game to be picked on the www.cfpool.com website. Entries are due by shortly before kickoff of the first game in the list (the official cutoff time is listed on the entry form).
The preferred method of submitting entries is via the Online Entry Forms. Entries can also be submitted by email if it is not possible to use the web-based entry form.
A perfect score is 1,000 points every week, regardless of the number of games on the list. Your score is the total of the "confidence values" for each line where you picked the winning team. The lines are numbered from the top (nearest the '$'); those nearest the top are worth the most.
For an 18 game week, confidence values are (by line number):
| 1) 110.83 | 4) 83.28 | 7) 62.57 | 10) 47.01 | 13) 35.32 | 16) 26.53 |
| 2) 100.77 | 5) 75.71 | 8) 56.88 | 11) 42.73 | 14) 32.11 | 17) 24.12 |
| 3) 91.61 | 6) 68.83 | 9) 51.71 | 12) 38.85 | 15) 29.19 | 18) 21.93 |
If your most confident pick (the one you placed first in the list of teams in your entry) wins, you get 110.83 points added to your score for the week. If that team loses, you get nothing for the pick. Similarly, if your bottom team (the pick in which you had least confidence) wins, you get only 21.93 points for that team. On the other hand, if that team loses, you've only lost out on the chance to score 21.93 points. In an 18 game week, getting your top pick correct is worth more points than getting your bottom four picks correct combined.
The actual values of each slot will vary depending on the total number of games included in the week. For more information the meaning of these numbers and their strategic implications, see the strategy section. For details on how the numbers are generated, see the FAQ section.
There is a 40-point bonus for having the highest score for the week. In the event of a tie for first place (which is rare), the bonus is divided up evenly. Weekly results and year-to-date totals will usually be posted Sunday (the day after the games).
The main part of the text version of the results posting looks like this:
+---------+------------+----------+----------+--------------+---------------+ | Score This Week | Year To Date | Entrant | | score | percentage |by thirds | score | percentage | Name | +---------+------------+----------+----------+--------------+---------------+ | 1000.00 | 18/18 100% | 06-06-06 | 3000.00 | 053/053 100% | PERFECT SCORE | | 801.71 | 13/18 72% | 06-04-03 | 2214.75 | 036/053 67% | Chris Stassen | | . . . . . . . . | | . . . . . . . . | | . . . . . . . . | +---------+------------+----------+----------+--------------+---------------+
How to read the table: in the above example, I managed a decent score (801.71) for the week, by predicting 13 of 18 games correctly. I scored six of six of the top one-third of confidence values, four of six of the middle one-third, and three of six of the bottom one-third. My total score for the all the weeks of the pool to date is 2,214.75. Overall, about 2/3 (67%) of my picks were correct.
The "bowl week" pool is worth double (2,000 points), and the bonus for the highest score is 100 points. The game list will be posted shortly after the conference championship games; entries will be due by kickoff of the first bowl game (usually a couple ofweeks later).
The year-total score for each entrant is the cumulative score for all the regular season plus bowls, with the lowest two scores dropped. For the sake of determining which weeks to drop, the bowl games entry is counted as two weeks, each with the same score. Your score is zero for any week that you don't enter.
Note that this is a change from the method that was used through 1991, where the highest outright total won the pool. The new system was introduced when previous entrants requested a scoring system that allows someone to miss a week (or two) without completely blowing their shot at the year-total championship.
The more certain you are about the outcome of a game, the higher you should place it when you sort your picks. I try to include fairly "close" games in the pool but preference is also given to televised games, and there will usually be at least some games where you can be pretty sure of the winner.
Note that for the confidence values listed in the scoring section you will receive more points for your top confidence pick than you would for the bottom four picks put together -- 110.83 versus 101.77 (29.19 + 26.53 + 24.12 + 21.93). Getting all of the bottom four picks wrong is not as costly as getting the top one wrong.
It is helpful to excel at picking game winners, but the real key to performing well in the pool is: coming up with a good ordering of your picks (i.e., assigning confidence values). You could be an average winner-picker (about 60%-70% right) and still win many weeks (and the year-total by a large margin) -- as long as your incorrect picks are consistently the games that you have placed at the lowest confidence levels.
By the way: there have been a handful of "perfect" entries in the pool's history, most notably four perfect scores in Week 2 of 2001. However, perfect entries are rather rare, considering that roughly 20,000 entries have been submitted over that span.
Don't take all of this too seriously -- just enter and have a good time. If you want to think deep meta-thoughts about the pool, perhaps this section will give you some ideas...
If you are familiar with all of the teams, you can probably guess how the majority of entrants will pick most of the games -- both who will win and which games will be high or low in most lists.
If you submit an entry too much like the majority one, there will be so many entries similar to yours that it will be difficult for you to score much above average. On the other hand, if you pick against the majority on nearly every game, you are bound to get a relatively low score. The consensus picks (denoted by the name ~Consensus in the the standings) usually finish near the top of the standings for the season.
There are two main ways to "break out of the pack":
If you move a team lower on your list than the majority, you stand to lose a lot less if there is an upset -- but the intervening teams, each of which moves up one notch, will cost you slightly more if they fail to win.
If you move a team higher on your list, you stand to lose a lot more points if they are upset -- but the intervening teams, each of which moves down one notch, will cost you slightly less if they fail to win.
I do not recommend picking more than a few each week. I also do not recommend picking an upset at a very high confidence level. If the upset occurs, you will break out of the pack even if the game is at a low confidence level -- but if the upset does not occur (which is more likely than not), you won't be hurt as badly if the team is near the bottom of your list.
A multiple of three is chosen because it makes the "by thirds" section of the scoring report work well. (For non-multiples, there aren't the same number of games in each "third.")
The fewer games there are in the pool, the smaller the total set of outcomes, and the more likely entrants will be to tie each other. The more games there are in the pool, the more painful it is to sort the entire list by confidence.
The range of 15-20 has been found to be a decent compromise -- it is about as many games as people can stand to sort, and there are relatively few ties (see next item).
The rule for generating 'confidence' values is: each higher one is 1.1 times the previous, rounded to two decimal places. The lowest confidence value (for a list with 18 games) is 21.93. 21.93 * 1.1 = 24.123. Rounded to two decimal places (24.12), you have the second confidence value. (And so on...)
That one rule -- combined with the requirement that all the values total 1000 and a known number of games -- uniquely determines the set of confidence values. The object in using such a rule to generate numbers is to decrease the number of tie scores.
Suppose I had instead used integers for confidence values (1 point for lowest, 18 for highest), as is commonly done in office pools that use confidence ratings. There would be only 172 different total scores -- integers zero [no picks right] through 171 [1+2+...+18]. There would be many different ways to get the same score (18, 17+1, 15+2+1, 7+5+3+2+1, etc.), and this would result in many tie scores.
But for the confidence values used in this pool, there are about 60,000 different week-total scores possible (when there are 18 games). Since there are 2^18 (about 260,000) different sets of confidence values that could be added up, there are still multiple ways to end up with the many of the scores. But the overlap is relatively minor -- even when there are 200 entrants in the pool, tie scores anywhere in the result list are rare. If there were only 172 possible scores, there would probably be about 100 tied entrants out of 200 entries.
Another benefit of this system is even when I am forced to include a different number of games in the list, each pool is worth the same number of points. This helps the 'drop lowest scores' algorithm (for the year-total championship) to work.
This section is here mainly for historical reasons. There are no longer tie games in Division-I football. If the score is tied at the end of regulation, a special tiebreaker is played.
A tie game usually scores half of the associated confidence value. For example: if the fourth team after the '$' in your entry (for a pool with 18 games) tied their opponent, you would be awarded 41.64 points (which is half of 83.28, the fourth confidence value).
Exception: if the tied team is listed last on your entry, you score the full confidence value. (If there were two ties in the list of games, you get full credit for placing a tied team in either of the bottom two spots of your list; the bottom three spots if there were three ties... and so on.)
The intent is to keep 1000 as the perfect score. A side effect is that you are rewarded for not being very confident about a game that ends up tied.
It depends on how serious the error is:
The pool takes about 5-10 hours per week to run. Most of the time goes to fixing entries that won't parse, notifying people by mail of entry problems, and selecting "close" games for the coming week. Much of the repetitive work -- such as making recaps or scoring entries -- requires relatively little operator intervention.
The pool is being run currently by Steve Schmadeke and Nick Lonsky of Talient, Inc. Steve is a Husker fan, making the 1000-mile round trip from the Denver area to Lincoln for all of the home games. Steve makes all the selections for the game list each week, so any perceived bias can be laid at his doorstep.
We inherited operation of the pool from cjohn mace, who ran the pool from 1995 through 1999 and established the cfpool.com domain. Prior to cjohn, the pool was run, with two interruptions, by Chris Stassen who established the pool in 1986. Mannie Lowe ran the pool in 1989 and 1990. "johndoe" ran it in 1993.
There has been a strong tradition in the past for computer-based ratings systems to participate in the CFPOOL. Computer rankings are recognizable by the asterisks that demark their names in the standings (e.g. *gfsys2*). If you're a current participant whose entry is based on your computer system, let us know and we'll update the database to include the asterisks.
Also in the past, there was a volunteer network to submit entries based on other well-known ratings such as the Sagarin System and the Vegas Line. We would like to rebuild that network to cover the complete set of all the computer systems in the BCS, both current and recently departed if possible.
Email us if you would like to "own" one of these entries. We will forward you the password to submit the entry for that system each week. If you feel you would volunteer but are afraid other commitments would prevent you from submitting the entry for that system before the deadline each and every week, we could still use your help. You can send in the picks retroactively for those weeks you couldn't process the games in time. Of course, the entries for these systems will be differentiated from the rest of the entries (computer-generated or otherwise) and won't be eligible for the weekly bonus.
Finally, we are considering making a web service available for submitting entries directly from your computer ranking systems, allowing you to automate the process of fetching the game list, computing the confidence order and submitting the resulting entry. Let us know if you are interested in such a system.
Back to RSFC CFPOOL Home Page.
Comments/Questions: cfpool