/* ]]> */
Oct 112009
 

Office football pools tend to payout in any of these schemes: 1) at season end, 2) weekly, and 3) mixed, a combination of season end and weekly payouts. Did you know the best strategy for each payout scheme is different? Below I discuss the best strategies for season end payouts and weekly payouts. I will discuss strategies for mixed payouts in a future post.

Strategies may vary by whether your picks are confidence weighted.

How to win the season ending prize: PointMax
Your best chance at winning the season ending prize is the Point Maximization (PointMax) strategy. It works against overconfident participants because they’ll ruin their season total by picking too many upsets. It works against the weak because they give up on PointMax if it’s not in first place after Week N. You just pick the favorites each week and never flinch.

Confidence scoring
In a pool with confidence scoring, sort the favorites by magnitude of Win Probability in descending order. I don’t charge for WinProbabilities or PointMax picks because they’re an open secret. I offer them as a convenience, but if for some reason you don’t like WinProbability, substitute Vegas point spreads, Sagarin Pure Points ratings (on USA Today), AccuScore predictions (Wall Street Journal), or Advanced NFL Stats Win Chance (on NYT’s Fifth Down column). The sorts won’t be too different most of the time. All of these are widely available free of charge.

Regular scoring
If your pool doesn’t have confidence scoring, just pick the favorites. Look at this site or USA Today to see their gambling lines page in the sports section. Now both the Wall Street Journal (AccuScore) and New York Times (Fifth Down, on Fridays) also publish such predictions.

How to win the most weekly payouts: Weekly Payout Maximizer
As good as it is over the season as a whole, PointMax is the worst strategy for winning weekly payouts! It’s too conservative. Weekly Payout Maximizer (WPM) will greatly increase your chances of winning weekly payouts.

Here’s the free advice that in the future will cost money:
For pools without confidence scoring, the list below is organized according to the number of participants in your pool. (Confidence pool players will have to wait for WPM to be tuned to allow for all those permutations).

In a regular scoring (no confidence weighting) pool, pick all the favorites except:

If your pool has 10 participants: pick against Indianapolis
If your pool has 11 to 30 participants: pick against Indianapolis, New York Jets
If your pool has 31 or more participants: pick against Indianapolis, New York Jets, Dallas, and Jacksonville

Below I discuss the philosophy and some simulation results behind WPM. My wife strongly encourages you to stop reading right here. However if you’re interested understanding details on the strategy behind WPM, read on.

Pool Entrants are Biased toward Favorites
Football pool contestants have learned not to pick too many upsets. As a result, each week they overbet some favorite teams. You have an excellent chance at winning more weekly payouts if you recognize such opportunities and make calculated picks when favorites are overbet.

Weekly Payout Maximizer: Identifying When to Bet Against Overbet Favorites
Given each week’s Win Probabilities, WPM identifies which, if any, underdogs to pick. WPM doesn’t predict which teams are likely to upset, it identifies whether any favorites are overbet.

The number of upsets it recommends will vary from week to week, and it will vary according to the number of participants in your pool. The number of upsets depends on the distribution of Win Probabilities, the distribution of office pool selections, and the number of participants in your office pool. If your pool has only 10 participants, for 2009 Week 5 it recommends picking only one upset – Indianapolis – but if your pool has 40 or more participants this week it recommends picking four: Indianapolis, NY Jets, Dallas, and Jacksonville.

This is not some tout’s “upset special;” it is an actuarially determined strategy for winning more weekly payouts than the other participants in your pool. The strategy depends not on expecting the Colts to lose, but on recognizing that if they do lose (they have roughly a 40% chance of losing this week) then your pool participants’ picks will fare very poorly. And if they win, so many of your pool participants will have picked them it won’t help anyone who picked them very much.

WPM is a high risk strategy. In order to reap the most weekly wins you need to risk having some bad weeks.

Do not use WPM if you hope to win the season total prize!

Monte Carlo Simulation Analysis: WPM Tailors Recommendations according to #Participants
16,000 Simulations Each for Various Pool Sizes, (Nonconfidence scoring)
Taking NFL 2009 Week 5 as an example of how this works, here are the WPM results from thousands of simulations using the Win Probabilities and the distributions of online office pool selections. The optimal number of upsets gets smaller depending on the number of contestants in your pool.

Scenarios: 10, 20, 30, 40, 50, 75 Participants
In each scenario there are 1 WPM, 1 Point Maximization Participants, and the rest are Average Joes. To avoid depending on unrealistic assumptions, I randomly apply the actual pick distributions of online office pool Participants to the Average Joes.

Percentage of Weekly Payouts Won by WPM vs. Average Joe and PointMax
10 Players
WPM recommendation: Pick Indianapolis to be upset
Percentage of weekly payouts won by
PointMax/Average Joe/WPM
9.3 / 9.4 / 16.2

Percentage of Weekly Payouts Won by WPM vs. Average Joe and PointMax
20 Participants
WPM recommendation: Pick Indianapolis and NYJ to be upset
Percentage of weekly payouts won by
PointMax/Average Joe/WPM
3.3 / 4.9 / 8.7

Percentage of Weekly Payouts Won by WPM vs. Average Joe and PointMax
30 Participants
WPM recommendation: Pick Indianapolis and NYJ to be upset
Percentage of weekly payouts won by
PointMax/Average Joe/WPM
1.6 / 3.3 / 6.7

Percentage of Weekly Payouts Won by WPM vs. Average Joe and PointMax
40 Participants
WPM recommendation: Pick Indianapolis, NYJ, Dallas and Jacksonville to be upset
Percentage of weekly payouts won by
PointMax/Average Joe/WPM
1.0 / 2.5 / 5.2

Percentage of Weekly Payouts Won by WPM vs. Average Joe and PointMax
50 Participants
WPM recommendation: Pick Indianapolis, NYJ, Dallas and Jacksonville to be upset
Percentage of weekly payouts won by
PointMax/Average Joe/WPM
0.7 / 2.0 / 4.4

Percentage of Weekly Payouts Won by WPM vs. Average Joe and PointMax
75 Participants
WPM recommendation: Pick Indianapolis, NYJ, Dallas and Jacksonville to be upset
Percentage of weekly payouts won by
PointMax/Average Joe/WPM
0.4 / 1.3 / 3.5

This means that in a very small pool – one with 10 participants, if every week is like Week 5 the participant applying WPM will win 16.2% of the weekly payouts. In a 20 participant pool, he’ll pick two upsets and win 8.7% of the weeks. In a 75 participant pool, he will pick 4 upsets and will still win 3.5% of the weeks. That compares to the Average Joe who wins 1/75 of the time. PWM will probably not win the Season Total, but he’ll win the most money over the course of the season.

Notice a few aspects about these simulation results:
1) The Average Joe’s chance remains consistently 1/#Players
2) PointMax’s chance is better than Average Joe only in very small pools
3) The more participants in the pool, the better the improvement for WPM over Average

PointMax’s consistency is your opportunity
Although it does a lousy job at winning any single week, PointMax still wins the most points over the 16,000 simulations for each scenario. But its consistency give risk to opportunity for you to win more weekly payouts by using WPM.

 Posted by on October 11, 2009 at 11:12 am