So -- we agree that fading the squares is a legitimate betting strategy. The question is -- how can we find out which teams the squares like each week and how shall we weigh the data?
The challenge of using a "fading" system is establishing consistency. Sure, you can ask the drunk guy sitting at the bar next to you who he likes every week and bet the other way. But some weeks, that guy sitting next to you may be a decent handicapper. Same thing with the people inside the sports books. If you are fading a single person, how do you know the bettor won't get frustrated and start fading his own picks? It happens. Perhaps the bad handicapper will read something and get wiser about betting. Maybe he'll just give up. There are several ways to get burned by relying on a single person or a random sampling that is too small. We have to overcome that with a large enough sample to truly indicate public sentiment.
The Goldman Experiment solves all of these sticky issues. It is a consensus-based system that uses a very large sampling. It easily identifies who the squares like each week. It gives us multiple games. In short, it's powerful. Here's how it goes.
About 12 years ago, a good friend of mine named S. Goldman (friends call him "Goldie") entered a local football handicapping contest. Goldie lives in New York. The contest is a collection of office pools and regular people who pick NFL games against the point spread. Last year, there were 228 people entered into the contest. Everyone in the contest pays an entry fee and the entrants who pick the most games correctly win back a portion of the prize money.
The contest is based on weekly lines that appear in the New York Daily News (the lines in that newspaper are pretty accurate for a mainstream sports daily). Each Friday, contestants must pick SIX games against the spread and mail them to the coordinator. No totals. Just sides. Then, the plays are entered into a spreadsheet. In Goldie's own words, here's how he gets his plays each week:
I apply WEIGHTED scores to the picks, and then compare the WEIGHTED results. It is NOT enough for me to bet a team just because lots of squares are on it. I look for the combination of EVERYONE on one side, and NO ONE on the other. The games I bet are the ones with the LARGEST DIFFERENTIAL on the weighted scores. The weighting is simple. Since every player picks six games, I assign points in descending order from 6. So, the top pick is worth 6 points, the next is worth 5, and so forth.
So, if 30 players select the GIANTS as their TOP pick, I assign the GIANTS 180 points (30 x 6). Then, if 22 players pick the GIANTS as their second best pick, I assign 5 points for those picks, or 110 points (22 x 5). If 12 people pick the GIANTS as their third best pick, I assign 4 points or (12 x 4 = 48 points). And, so on. Then, I add up the points for each position (180+110+48...etc.) to arrive at the WEIGHTED SCORE for each team. Finally, I match up the teams, compare their weighted scores, and bet the games with the LARGEST DISCREPANCY.
For example, if the RAVENS are playing the JETS, the RAVENS may get weighted score of 434 -- the highest of any team that week. But, if the JETS weighted score is 330, the difference is only 104 (434 - 330 =104). On the other hand, if the STEELERS are playing the COLTS, and the weighted scores are STEELERS: 280 and COLTS: 72, the difference is 208 (280 - 72 = 208). In this example, the COLTS would be the bet, but the JETS would not, despite the fact that so many people liked the RAVENS. Enough interest in the JETS cancels the play. In the COLTS/STEELERS game however, the combination of the number of players on the STEELERS, COMBINED with the fact that virtually NO ONE is on the COLTS makes that the attractive game.
Finally, because the pool is New York-based, I "discount" the GIANTS and JETS weighted scores by 20 percent. This is a random decision on my part....however there is NO DOUBT that a statistically significant number of players select either the JETS or the GIANTS or both every week because they are fans of the teams. I'm sure that the same local bias is found in MIAMI, DALLAS, GREEN BAY, etc. So, the weighted scores on the local teams are higher than they would be if the sample was national. The 20 percent discount is the random number I have applied in an attempt to even out this effect. It is subject to adjustment.
For the past two NFL seasons, Goldie has privately e-mailed me the weekly picks so I could monitor the results (Note: Up until this point, I have never used the information other than as a curiosity. These picks have never influenced my selections). For Goldie, fading the contest picks has been a consistently positive method of picking winners.
Last year -- out of 228 entrants, only 53 entrants (less than a quarter) picked more than 50 percent winners. 166 contestants picked less than 50 percent. 9 players hit exactly 50 percent. Since there was no vig on these games, the results should have been about half, or 114 contestants on the positive side of 50 percent. Since a coin flip would give you an even 50-50 split (with no vig), it goes to show that this handicapping contest is filled with squares.
Now, here's what's REALLY astounding. For all 228 players, the average success rate of picking ATS winners was just 41.5 percent! A stunning number of players picked games at less than 40 percent. A significant number of contestants picked 35 percent (or less) winners. But NO ONE came close to picking the opposite (65 percent or over). In fact, in three particular weeks last season, not a single player was able to go 6-0 for the week. Yet, in those same weeks, 12 players went 0-6 !!! It is much more common in this pool for players to go 0-6 for the week than 6-0. It wouldn't seem possible, but that happens every season. And remember, they only have to pick six games....not the whole card.
Last season, the top differential bets won 66 percent of the time. The final season record was 28-14. The prior season I was 25-15. The season before that, a remarkable 32-13. But, that was probably making up for the season before that one. In 1998, the fans had a "good" season. Goldie went 35-33 and lost money that year (because of the juice). So, that's just one average season for the fans out of the last five. The overall record of The Goldman System speaks for itself -- 85 wins and 42 losses for an astounding 67 percent win rate over the past three seasons.
The best thing about The Goldman System is that there is no handicapping required. Once the plays are determined, I will post the plays here each week -- most likely on Saturdays. The "top differential" bets means that we get about 3 plays per week with this system. The system uses a large enough sample to represent the betting public (squares). It is also consistent since it puts us in the position of playing the role of the sportsbook and fading the public action. It is sort of the anti-thesis of the Hilton Football Contest, where we take the consensus of the public and go the other way.
Starting with NFL Week 1 of the regular season, I will have a special section in my daily report which gives readers the top picks of Goldman's contest. That means, we will bet the other way. This will be kept as a separate section of the daily report. NOTE: The Goldman polyester plays will not in any way effect my analysis or recommendations -- which will be independently researched.
The challenge of using a "fading" system is establishing consistency. Sure, you can ask the drunk guy sitting at the bar next to you who he likes every week and bet the other way. But some weeks, that guy sitting next to you may be a decent handicapper. Same thing with the people inside the sports books. If you are fading a single person, how do you know the bettor won't get frustrated and start fading his own picks? It happens. Perhaps the bad handicapper will read something and get wiser about betting. Maybe he'll just give up. There are several ways to get burned by relying on a single person or a random sampling that is too small. We have to overcome that with a large enough sample to truly indicate public sentiment.
The Goldman Experiment solves all of these sticky issues. It is a consensus-based system that uses a very large sampling. It easily identifies who the squares like each week. It gives us multiple games. In short, it's powerful. Here's how it goes.
About 12 years ago, a good friend of mine named S. Goldman (friends call him "Goldie") entered a local football handicapping contest. Goldie lives in New York. The contest is a collection of office pools and regular people who pick NFL games against the point spread. Last year, there were 228 people entered into the contest. Everyone in the contest pays an entry fee and the entrants who pick the most games correctly win back a portion of the prize money.
The contest is based on weekly lines that appear in the New York Daily News (the lines in that newspaper are pretty accurate for a mainstream sports daily). Each Friday, contestants must pick SIX games against the spread and mail them to the coordinator. No totals. Just sides. Then, the plays are entered into a spreadsheet. In Goldie's own words, here's how he gets his plays each week:
I apply WEIGHTED scores to the picks, and then compare the WEIGHTED results. It is NOT enough for me to bet a team just because lots of squares are on it. I look for the combination of EVERYONE on one side, and NO ONE on the other. The games I bet are the ones with the LARGEST DIFFERENTIAL on the weighted scores. The weighting is simple. Since every player picks six games, I assign points in descending order from 6. So, the top pick is worth 6 points, the next is worth 5, and so forth.
So, if 30 players select the GIANTS as their TOP pick, I assign the GIANTS 180 points (30 x 6). Then, if 22 players pick the GIANTS as their second best pick, I assign 5 points for those picks, or 110 points (22 x 5). If 12 people pick the GIANTS as their third best pick, I assign 4 points or (12 x 4 = 48 points). And, so on. Then, I add up the points for each position (180+110+48...etc.) to arrive at the WEIGHTED SCORE for each team. Finally, I match up the teams, compare their weighted scores, and bet the games with the LARGEST DISCREPANCY.
For example, if the RAVENS are playing the JETS, the RAVENS may get weighted score of 434 -- the highest of any team that week. But, if the JETS weighted score is 330, the difference is only 104 (434 - 330 =104). On the other hand, if the STEELERS are playing the COLTS, and the weighted scores are STEELERS: 280 and COLTS: 72, the difference is 208 (280 - 72 = 208). In this example, the COLTS would be the bet, but the JETS would not, despite the fact that so many people liked the RAVENS. Enough interest in the JETS cancels the play. In the COLTS/STEELERS game however, the combination of the number of players on the STEELERS, COMBINED with the fact that virtually NO ONE is on the COLTS makes that the attractive game.
Finally, because the pool is New York-based, I "discount" the GIANTS and JETS weighted scores by 20 percent. This is a random decision on my part....however there is NO DOUBT that a statistically significant number of players select either the JETS or the GIANTS or both every week because they are fans of the teams. I'm sure that the same local bias is found in MIAMI, DALLAS, GREEN BAY, etc. So, the weighted scores on the local teams are higher than they would be if the sample was national. The 20 percent discount is the random number I have applied in an attempt to even out this effect. It is subject to adjustment.
For the past two NFL seasons, Goldie has privately e-mailed me the weekly picks so I could monitor the results (Note: Up until this point, I have never used the information other than as a curiosity. These picks have never influenced my selections). For Goldie, fading the contest picks has been a consistently positive method of picking winners.
Last year -- out of 228 entrants, only 53 entrants (less than a quarter) picked more than 50 percent winners. 166 contestants picked less than 50 percent. 9 players hit exactly 50 percent. Since there was no vig on these games, the results should have been about half, or 114 contestants on the positive side of 50 percent. Since a coin flip would give you an even 50-50 split (with no vig), it goes to show that this handicapping contest is filled with squares.
Now, here's what's REALLY astounding. For all 228 players, the average success rate of picking ATS winners was just 41.5 percent! A stunning number of players picked games at less than 40 percent. A significant number of contestants picked 35 percent (or less) winners. But NO ONE came close to picking the opposite (65 percent or over). In fact, in three particular weeks last season, not a single player was able to go 6-0 for the week. Yet, in those same weeks, 12 players went 0-6 !!! It is much more common in this pool for players to go 0-6 for the week than 6-0. It wouldn't seem possible, but that happens every season. And remember, they only have to pick six games....not the whole card.
Last season, the top differential bets won 66 percent of the time. The final season record was 28-14. The prior season I was 25-15. The season before that, a remarkable 32-13. But, that was probably making up for the season before that one. In 1998, the fans had a "good" season. Goldie went 35-33 and lost money that year (because of the juice). So, that's just one average season for the fans out of the last five. The overall record of The Goldman System speaks for itself -- 85 wins and 42 losses for an astounding 67 percent win rate over the past three seasons.
The best thing about The Goldman System is that there is no handicapping required. Once the plays are determined, I will post the plays here each week -- most likely on Saturdays. The "top differential" bets means that we get about 3 plays per week with this system. The system uses a large enough sample to represent the betting public (squares). It is also consistent since it puts us in the position of playing the role of the sportsbook and fading the public action. It is sort of the anti-thesis of the Hilton Football Contest, where we take the consensus of the public and go the other way.
Starting with NFL Week 1 of the regular season, I will have a special section in my daily report which gives readers the top picks of Goldman's contest. That means, we will bet the other way. This will be kept as a separate section of the daily report. NOTE: The Goldman polyester plays will not in any way effect my analysis or recommendations -- which will be independently researched.
